Grant Agreement No: 265307 Project start date: 1st January 2011 Project end date: 30rd June 2014 BROWSE Bystanders, Residents, Operators and WorkerS Exposure models for plant protection products SEVENTH FRAMEWORK PROGRAMME Theme: Environment (including climate change) Technical report WP1.4: Operator exposure: Boom spraying, Mixing/loading, Orchard spraying and Hand Held Applications Title of Deliverable Deliverable number Related WP and Tasks Type of Document Authors Compiled by Date Technical report WP1: Operator exposure: boom spraying, mixing/loading , orchard spraying and hand held applications 1.4 WP1 Public Report Partners WP1 Henk Goede, Suzanne Spaan, Thies Oosterwijk, Neleida Marrufo, Agathi Charistou, Victoria Roelofs, Clare Butler-Ellis, Richard Glass, Kiki Machera, Marc Kennedy, Helen Owen, Dav Stott, Mark Fee, Andy Hart, Rianda Gerritsen-Ebben 31/10/2014 Project coordinator Dr Andy Hart Fera, United Kingdom E-mail: [email protected] http://www.browseproject.eu 1 Table of Contents Abbreviations .......................................................................................................................................... 4 1 Introduction .................................................................................................................................... 5 2 Scope and Aim ................................................................................................................................. 6 3 BROWSE WP1 operator models ...................................................................................................... 7 3.1 User interface .................................................................................................................................. 7 3.2 Exposure models ............................................................................................................................. 7 3.2.1 Boom spraying inhalation model .................................................................................. 12 3.2.2 Boom spraying dermal model ....................................................................................... 26 3.2.3 Mixing and loading inhalation model ........................................................................... 37 3.2.4 Mixing and loading dermal models ............................................................................. 44 3.2.5 Orchard spraying inhalation model .............................................................................. 71 3.2.6 Orchard spraying dermal models.................................................................................. 79 3.2.7 Handheld Application inhalation model ....................................................................... 90 3.2.8 Handheld Application Dermal model ............................................................................ 95 3.3 Aggregated exposure .................................................................................................................. 130 3.3.1 External exposure ....................................................................................................... 130 3.3.2 Internal exposure ........................................................................................................ 133 3.4 Quality assurance and testing procedures.................................................................................. 137 4 Comparison with existing models ............................................................................................... 140 5 Model outcome interpretation and level of conservatism ......................................................... 141 5.1 Exposure outcomes ..................................................................................................................... 141 5.2 Routes and sources of exposure included .................................................................................. 141 5.3 Exposure scenarios included ....................................................................................................... 141 5.4 Population ................................................................................................................................... 142 5.5 Representativeness of model outcomes versus “real world”..................................................... 142 2 5.6 Level of uncertainty and variability ............................................................................................. 142 5.7 Model comparison as indication for level of conservatism ........................................................ 143 5.8 Overall level of conservatism ...................................................................................................... 143 6 Conclusion ................................................................................................................................... 144 7 Recommendations for future development and refinement ..................................................... 146 8 References................................................................................................................................... 148 3 Abbreviations a.s. Active substance ASE Airborne spray emission AR Volume spray rate BS Boom spraying CTS Closed Transfer Systems HHA Handheld application LAI Leaf Index Area ML Mixing & loading MTD Mechanical Transfer Devices PPE Personal protective Equipment PPP Plant Protection Product PPPs Plant Protection Products RA Risk Assessment RPE Respiratory Protective Equipment SA Sensitivity analysis 4 1 Introduction This document provides an overview of the operator models (WP 1) developed as part of the Bystander, Resident, Operator and WorkerS (BROWSE) project. BROWSE is an EU 7th Framework Programme and is scheduled from 2011 to mid 2014. The main objectives of the project are to: • Review, improve and extend the models currently used in risk assessment (RA) of Plant Protection Products (PPPs) to evaluate the exposure of operators • Development of new models for operator exposure scenarios • Involve all relevant stakeholders and end users and take account of gender issues and regional differences The expected impact of the project is to contribute to the implementation of Regulation (EC) 1107/2009 and Dir. 2009/128/EC on the Sustainable Use of PPPs. A Deliverable 1.1 was compiled to review existing models and data for operator exposure (https://secure.fera.defra.gov.uk/browse/project/deliverables-&-reports.cfm). A common finding for PPPs models is that they are based on outdated empirical data (non GLP and GLP) and only cover limited scenarios. PPP models also derive surrogate exposure values using different percentiles and using different normalisation parameters. Generally speaking, the models estimate external inhalation as well as external dermal (potential and/or actual) exposures. A large variation in the use of defaults for personal protective equipment is applied. Biocide and industrial chemicals models are an exception, where probabilistic task-based models such as BEAT and ART are used. It was concluded that more and reliable exposure data are required for modeling purposes. The collation of the exposure data found in the open (public), grey literature and in data(bases) of existing and emerging models, revealed a significant amount of papers (or reports) with quantitative exposure data. However, although most of the collated exposure data-sources identified are publicly available, only about one third of the sources have raw data (detailed measurement data) that is readily accessible and only about half of these consist of data for plant protection products (PPPs). It became clear at an early stage of the BROWSE database development that the available data does not provide the desired level of contextual information required for modeling purposes. In order to develop probabilistic models, new and good quality data is required. Unfortunately the BROWSE consortia did not get access to new/recent and contextually-rich data. 5 Nevertheless, within the BROWSE framework the following options were considered for the further development of a model for operator exposure: • Improvement of the existing empirical models with more data • Combination of an empirical and mechanistic modeling approach • Development of a mechanistic model • Inclusion of probabilistic modeling, taking into account probability distributions and uncertainty in exposure The aim and scope of WP1 was formulated after a comprehensive review of the currently available PPPs, biocide and other exposure models, and considering the scenarios that are covered in current PPP models. 2 Scope and Aim The following aims are proposed for the operator models (WP1): • Develop a single, new and improved modeling framework, integrating all available exposure data • Incorporate recently-generated, high-quality exposure data if available • Take explicit account of key factors and mechanisms influencing exposure, account for nonlinearity, strengthen estimation where data are limited and to produce probability distributions of exposure • Use data on operator practices in different Member States to take account of controls & protective equipment • Take account of regional & gender differences where possible The following scenarios were selected for WP1 operator model development: • boom spraying • mixing & loading (liquids and solids) • orchard spraying • hand-held spraying 6 3 BROWSE WP1 operator models 3.1 User interface See relevant software documentation. The relevant user inputs required for the user-interface are presented for each model in the respective sections of this report. 3.2 Exposure models Conceptual model A generic conceptual model has been developed (Figure 1) to support the development of models for each scenario. It describes the transport of a PPP from the source to the receptor (operator). This model considers three main routes of personal exposure, i.e. (i) inhalation (via respiratory tract), (ii) dermal (via skin) and (iii) ingestion (via mouth). The model consists of different compartments through which the pesticide can pass. These compartments can broadly be divided into three steps: • Emission from source (source, local source zone) • Transport between source and operator (air-, surface- and bulk- zone) • Transport at the operator (respiratory protective equipment (RPE), outer clothing contaminant layer, inner clothing contaminant layer, skin contaminant layer, (pre-)oral /mouth) The transport of PPPs occur through 4 mechanisms, (i) separation from gas/vapour or solid particles from parent material (source strength), (ii) transport between compartments, (iii) loss of pesticide from compartments due to sinks, and (iv) uptake by the receptor (Tielemans et al, 2008). The latter mechanism (uptake) is only included as an estimate after modeling the different routes of exposure. The inhalation exposure models accounts for the most important determinants that are involved in the airborne transfer of PPPs from the formulation. The dermal exposure is the most complex route of exposure and occurs through three potential pathways, i.e. (i) deposition from the air, (ii) contacts between surfaces and the body, and (iii) bulk transfer through splashes or dripping (from liquids) and impaction (from solids) (Schneider et al, 1999). It should be noted that the three potential dermal pathways of exposure are scenario-specific and not always relevant (and therefore not included in all the models). And lastly, a generic estimate is made of ingestion exposure that considers contamination of the hands or gloves and the probability that it is transferred into the mouth and oral route. 7 Figure 1: Conceptual model for operator exposure to PPPs Surface zone Local source zone Bulk zone Air zone Source Deposition Splashes/impaction Surface contact Respiratory protective equipment Outer clothing contaminant layer Inner clothing contaminant layer Respiratory tract Oral Inhalation Ingestion Skin contaminant layer (different body parts) Dermal Modeling approach For each scenario and exposure route, the most important determinants that influence the transport of a PPP from the source to the receptor were identified and, where appropriate, included in the models. Their effect sizes and correlations were evaluated and determined by means of data analyses, literature reviews and expert opinion. Subsequently, mechanistic algorithms were developed using the available evidence to underpin the models. In general, the following modeling procedure was followed for each scenario: • Collation of available evidence from the literature • Statistical (regression and correlation) analyses of the available exposure data • Identifying exposure determinants and allocating effect sizes and distributions (incl. information from EFSA and BROWSE survey data, e.g. Glass et al, 2012) • Adopt experimental data as input for the models (where available and/or appropriate) 8 • Developing mechanistic algorithms • Coding the datasets with the (missing) model input parameters, and introducing imputed data where appropriate using statistical techniques (e.g. bootstrapping) • Sensitivity analysis (SA) of the model algorithms • Comparison and testing of model outputs with data (no calibration or fitting/transforming was performed due to data quality issues) • Comparing the model outputs with existing model outputs Depending on the available evidence of exposure determinants of a given scenario, available experimental data and existing models, the most useful type of model was selected. This resulted in different levels of complexity of the respective mechanistic algorithms. Examples include: • Using models of other BROWSE work packages (input from WP3 for the BS inhalation model) • Using experimental data (surface contamination and hand/body exposure levels for the BS hands, ML hands and ML body models) • Using existing models (Advanced REACH tool (ART) for ML inhalation) • Using extrapolation (BS body extrapolated from BS hands model) Table 1 gives an overview of the principal inputs of each model and the data used for testing. Table 1: Key model inputs and data used Model Examples of key inputs Spray volume rate WP3 estimate (airborne fraction BS estimate) inhalation Vehicle-sprayer type (dispersion estimate) Surface contamination levels (% of BS hands total spray volume) Contact event frequency Extrapolation BS hands model BS body Contact event frequency *ML ART model inhalation *ML hands Container contamination levels Hand contamination levels Derived from Exposure data used for testing User input Experimental spray drift data & WP3 model Literature review BROWSE database Inhalation data Vehicle-sprayer contamination studies User input Literature review User input Existing calibrated model (based on literature review and expert elicitation procedure) Experimental studies Experimental studies 9 BROWSE database Bare hand & Protected hand BROWSE database Whole body (excl. hands) BROWSE database Inhalation data BROWSE database Bare hand & Model Examples of key inputs Contact event frequency Body contamination levels *ML body Contact event frequency Spray volume rate Orchard Fraction of overspray spraying Row spacing Inhalation Vehicle/ forward speed Wind speed Orchard Surface contamination levels (% of spraying total spray volume) hands Contact event frequency Orchard spraying Deposition estimation (based on total cockpit contamination) body HandSpray volume rate held models Surface contamination level Derived from User input Experimental studies User input User input Experimental studies User input User input User input Vehicle-sprayer contamination studies User input Vehicle-sprayer contamination studies User input Application efficiency Studies on Surface Contamination of the Equipment when ML Air Concentration Literature review Exposure data used for testing Protected hand BROWSE database Whole body (excl. hands) BROWSE database Inhalation data BROWSE database Bare hand & Protected hand BROWSE database Whole body (excl. hands) BROWSE database Hands and Whole body (excluding hands) Existing calibrated model (based on literature review and expert elicitation procedure) *Note a distinction is made between manual and mechanical mixing and loading (ML) models (not indicated in this Table) Sensitivity analysis Sensitivity analysis (SA) was performed for the model algorithms of each scenario. Using the input ranges of the model algorithm a Latin hypercube sampling space of 400 points are generated across the inputs. Due to several discrete inputs the samples are converted from a discrete uniform distribution. Preferably, separate sensitivity analyses (SA) are required for different combinations of categorical variables in the models. Model testing with data and imputation of data For each of the models described in this report we have some empirical data available which consists of information on the inputs for the models (e.g. concentration of active substance, wind speed etc.) 10 along with the measured output (e.g. potential dermal exposure on the hands and / or body etc.) from the field study. For each of these datasets we run the relevant model and compare the model outputs with the measured output. However, several datasets are not complete and contain missing pieces of information. Where this was the case the missing input or inputs were imputed by empirical sampling from existing values for that input in the available data. As a result, this report distinguishes between ‘true datasets’ and ‘imputed datasets’. To summarise the output we calculate the geometric mean, median, 2.5th percentile and 97.5th percentile. We then plot the geometric mean of the output against the true measured values that we have from the available data and plot the 95% confidence intervals, medians and geometric means of the output against the true values (all on log10 scale) for comparison. Except where indicated otherwise we use 10,000 iterations of each model. Where relevant, a clear distinction is made between the measured values or results for the model runs when the data set used was complete and when input values needed to be imputed in the respective plots. To test the models, two types of plots / graphs are presented: • The measured potential exposure under specific conditions (y-axis) against the geometric mean of the model outputs (x-axis) under the same conditions (i.e. same model inputs) • The individual measured potential exposure (dots) and estimated model outputs (grey lines) under the same conditions (i.e. same model inputs) plotted with the exposure level ( y-axis). We present two of these plots with the results sorted in two different ways. The first plot is sorted on the median model output and the second is sorted on the measured exposure values. The models were only compared or tested with available exposure data. Thus no calibration or transforming/fitting was performed in any of the models. Reasons for not calibrating or fitting the models were: (1) Considerable variation in both the data and model outputs, and (2) Sub-standard data quality of the BROWSE database (e.g. limited or missing contextual data) Regression analyses will not be robust when based on small datasets (the regression will automatically remove any datasets with missing values). Although imputation is used to replace the missing values, it’s not clear what effect this will have on the outputs. 11 Dermal sampling techniques For potential dermal exposures, a preliminary analysis of the data showed no conclusive differences in hand exposure data using different sampling techniques, i.e. bare hands and protected gloves. ‘Bare hands’ data refer to the measured amount of potential dermal exposure on unprotected hands (no protective gloves worn) or on sampling (non-protective) gloves, e.g. cotton or Tyvek gloves. ‘Protective gloves’ data refer to the measured amount of potential dermal exposure on protective gloves. When testing the hand models, the data used for hand exposure were based on both these sampling techniques (depending on the data available). So combining the sampling techniques in the data used for model testing may result in increased or unknown variability and outputs. Background calculations Various background calculations are performed in the software in order to calculate model inputs. Typical examples include an estimate of the spraying time (to estimate the spray volume rate), the ML periods and the number of containers used (Appendix L). 3.2.1 Boom spraying inhalation model (a) Description The boom spraying scenario represents field sprayers using hydraulic nozzles, but excludes airassisted operations. Operator inhalation exposure is assumed to originate from the spraying equipment that releases airborne droplets during boom spraying (BS) activities. The inhalation model focuses on droplet spray and predicts the air concentration of droplets in the breathing zone of the operator. The dispersion process behind a moving vehicle is complex, and the available mechanistic models that are used in drift prediction (e.g. individual drop trajectory and Gaussian models) are not suitable to predict transfer from the boom to the vehicle. However, these models may be useful to estimate the amount (or fraction) of airborne spray at the boom. To estimate inhalation exposure of operators during BS, the transfer mechanisms associated with wake and plume are considered (see Appendix A). Proposed model: The model consists of three parts that represents the three consecutive phases of the sourcereceptor model: (i) Local airborne spray emission (ASE) 12 (ii) Dispersion from boom to vehicle (to estimate Cvehicle) (iii) Dispersion from vehicle to operator using cabin efficiencies (If) More details of input variables of the model are presented in Tables 2 and 3. (i) Local airborne spray emission (ASE) The amount of airborne spray around the vehicle is divided into (and the sum of) two local airborne spray emission estimates (ASEwakein & ASEplumein). The spray emission rate (in L/s) implicitly takes account of the spraying time and assumes that the resulting air concentration is steady over time. It describes the dispersion from the boom to the vehicle via wakes (ASEwakein) (boom nozzles directly behind the vehicle-sprayer) and via a plume (ASEplumein) (boom nozzles located on outer boom part). Mass balance inputs to estimate the ASE is the spray volume rate (l/s), derived from the total volume applied and the spraying time. The fraction of spray volume that remains airborne in close proximity of the boom is expressed as a factor (fairborne spray). This estimate is obtained from the droplet trajectory model (Browse WP3 model) and used to estimate the fairborne spray for different droplet qualities, i.e. fine, medium, course and very course (Appendix A). The output of the model is based on fixed variables for distance (at 2 meters downwind of the boom), at a worst case 90 degrees wind angle. The model also includes factors such as wind speed, boom height (above crop), crop height and vehicle speed. The WP3 output is a factor that represents the estimated fraction of airborne droplets. It is derived from the WP3 emulator by dividing the value by 2 and then adjusted these values to be in line with the intended fairborne spray estimate. (ii) Dispersion from boom to vehicle (to estimate Cvehicle) To estimate the dispersion of spray from the boom to the operator, the previously described ASE estimates are used as input. Both the mass balance inputs of ASEwakein and ASEplumein inputs are affected by the presence of boom shields (bs), Table 7. The dispersion of droplet spray by plume (ASEplumein) from the outer part of boom (not directly affected by wake effects) is determined by a plume factor (pf) based on an estimate of the probability of plume reaching the vehicle based on different wind speeds, wind directions, driving directions and the fraction expected to reach the vehicle-sprayer (Appendix A). Both the ASEwakein and ASEplumein estimates are assumed to be relevant as droplet spray around the vehicle-sprayer. The dispersion of droplet spray around the vehicle-sprayer is determined by its size (height, width, 13 length) and the vehicle speed. The rationale behind this is that the area around the vehicle-sprayer is continuously dispersed by the vehicle speed (vvehicle) through a surface area (i.e. the width and height of the vehicle-sprayer, Avehicle) that is perpendicular to the driving direction (Tables 4 and 5). The length of the vehicle-sprayer (d), distinguished as vehicle-mounted/self-propelled and trailermounted rigs, represents the distance between the vehicle and the boom (Table 6). All these parameters are used to estimate the concentration of droplet spray around the vehicle Cvehicle (see algorithms). (iii) Dispersion from vehicle to operator using cabin efficiencies (If) Personal enclosures on vehicles (e.g. cabins) may affect the transfer of spray to the operator. The categories proposed for cabins are presented in Table 8 (Appendix D). (b) Algorithms Ipde = Cvehicle * If * Ccoi Ipde Potential inhalation exposure concentration to the active substance (in droplet spray) (µg/m3) Ccoi Concentration of a.s. in spray volume (µg/l) Cvehicle Concentration of droplet spray at the vehicle (l/m3) If Cabin efficiency (factor) where Cvehicle = (ASEwakein + (ASEplumein ∗ 𝑝𝑓)) ∗ 𝑏𝑠 ∗𝑑 Avehicle ∗ vvehicle ASEwakein flux of spray from boom nozzles in close proximity of vehicle affected by wake (l/s) ASEplumein flux of spray from nozzles further along the boom affected by a plume effect (l/s) pf fraction of plume that reaches the vehicle-sprayer bs effect of boom shielding Avehicle surface area perpendicular to the vehicle’s driving direction where air exchange occurs (m2) based on w (width of vehicle-sprayer) and h (height of vehicle-sprayer) 14 vvehicle vehicle speed (m/s) d vehicle-sprayer length categories based on vehicle-mounted and self-propelled (VM) and trailer-mounted (TM) rigs With ASEwakein = AR ∗ fairborne spray ∗ And ASEplumein = AR ∗ fairborne spray ∗ w lb 𝑙𝑏 − 𝑤 2 ∗ 𝑙𝑏 AR spray volume rate (L/s) fairborne spray fraction of spray volume that remain airborne in close proximity of the boom immediately after spray emission (factor) [from WP3 model] where: # • airChild = (emulatorValue / 2 ) * vehicle speed (m/s)/ nozzle flow rate (ml/s) / number of nozzles # • airAdult = (emulatorValue / 2 ) * 2 *vehicle speed (m/s)/ nozzle flow rate (ml/s) / number of nozzles • fAirborneSpray = airAdult – airChild w width of vehicle-sprayer (= 2 times vehicle/sprayer width, m) lb total boom width (m) Note: if w>lb, then w=lb where the boom width (lb) is assumed true (it is therefore reasonably assumed that the boom width is at least the width of the virtual space (2x vehicle-sprayer width) – in order to avoid using negative values) # a fixed default value is applied assuming a 90 degree wind angle, equivalent to following wind for the operator, divided by 2 on the assumption that the wind will only be following for half of the time as the tractor will either have a head wind or tail wind. (c) Sensitivity analysis Sensitivity analysis (SA) was carried out on the output representing potential dermal exposure on the body, for the boom spraying and mixing/loading model. SA is designed to show which inputs or 15 groups of inputs has the greatest impact on the output, given their individual ranges and variability. SA can also show the nature of the input/output relationships and provided a useful check on the model behavior. For this reason, it has been used as part of our testing process to assess whether the input/output relationship is realistic for individual modules. An example of this is given below in Figure 2. The inputs which have the most influence are wind speed (m/s), concentration of a.s. in spray volume (g/l), spray volume rate (l/s) and cabin factor. As the first three increase, the log(exposure) generally increases. For cabin factor, as the cabin factor increases from no cabin to a cabin with pressurized/filtered ventilation, the log(exposure) decreases. For this example, the partitioning of output variance is also presented in Table 9. This partitioning shows how each of the variability distributions, assigned to individual model inputs, contributes to the induced total output variation. SA contributes to the testing of the model, in addition to the more detailed testing described below. The details of the SA method with results from the latest Browse model and some conclusions are given in Appendix M. (d) Model testing Please refer to the type of plots/graphs and explanation under item 3.2. For the testing of the BS inhalation model, the value of boom shielding (bs) was fixed at 1, assuming the absence of boom shielding in the data. Figure 3 shows the model estimates plotted against all the available data (including all imputed datasets). Using a different plot type, Figure 4 indicates whether the data concerns vehicle-(rear-mounted) and self-propelled sprayers (circles) or trailermounted sprayers (squares). The output suggests that the model, in some cases, under-predicts for trailer-mounted sprayers and over-predicts for vehicle/self-propelled sprayers. However, most of these outputs are based on imputed data (blue) and this assumption may therefore not be conclusive. Figures 5 and 6 show the performance of the model using complete datasets only (with only boom shielding imputed with a fixed value), which indicates that the model outputs fit better with the complete datasets than the imputed datasets. 16 (e) Tables Table 2: Boom spraying inhalation model - model inputs* No 1 Inputs Concentration of a.s. in spray volume (g/l) Spray volume rate (l/s) Cabin factor (3 categories) Lower 0.05 Upper 50 Type selected value Based on BROWSE data 0.01 0.0001 1.5 1 selected value Normal 2.4 32 selected value 0.01 1 Uniform 6 7 8 9 Vehicle sprayer type height*width (8 categories) Vehicle or trailer mounted – length/distance (3 categories) Boom Height above crop (m) Boom width (m) Wind speed (m/s) Plume factor 0.2 2 0.5 NA 1.2 45 10 NA selected value selected value selected value fixed at 0.01 10 11 Forward speed (km/h) Boom shielding# 4 NA 25 NA selected value fixed at 0.7 if boom shielding present and 1 if absent BROWSE data Literature review (Appendix D) Different sources (Appendix B) DEFRA SID5 report (Appendix B) WP3 & BROWSE data WP3 & BROWSE data WP3 & BROWSE data Expert estimate (Appendix A) WP3 & BROWSE data Literature review (Appendix C) 2 3 4 5 * # separate models developed using (i) different spray qualities (fine, medium, course, very course) and (ii) different crop heights (0.1, 0.5, 1.0, 1.5m) fixed at 1 for the testing in the absence of contextual data Table 3: Boom spraying inhalation model - input variables Input variables Concentration of a.s. in spray volume Cabin factor Unit g/l User input User input Model input User input From/for Mass balance input n/a (i) No cabin (ii) Cabin without pressurized/filtere d ventilation 1 Sample from U(0.1, 0.25) with probability 0.25, from U(0.25, 0.5) with probability 0.5 and from U(0.5, 0.9) with probability 0.25 Sample from U(0.0001, 0.05) with probability 0.25, from U(0.05, 0.15) with probability 0.5 and from U(0.15, 0.3) with probability 0.25 User input converted for use in WP3 model Literature review Table 7 Appendix D (iii) Cabin with pressurized/filtere d ventilation Wind speed km/h From WP3: 1-20 km/h 17 WP3 model (to estimate fairborne spray) Input variables Boom height above crop Unit M User input From WP3: 0-1.2 m range for users to choose from From WP3: User can choose from 0.1, 0.5, 1.0, 1.5 Estimated from total area sprayed, dose, final mixing & loading concentration Estimated from total area sprayed, boom width and vehicle speed Model input User input Crop Height M Volume applied L Spraying time Min Vehicle-sprayer width M Categories: S, M, L, U, sSP, mSP, lSP, O Estimated from total area sprayed, boom width and vehicle speed (((areaSprayedHa*10000)/(totalB oomWidthM*dForward_SpeedK MH*16.67)*1.25)) Sampled from respective categories Vehicle-sprayer height M Categories: S, M, L, U, sSP, mSP, lSP, O Sample from respective categories Vehicle-sprayer length/distance M Sampled from respective categories Boom width (m) M Categories: vehiclemounted and selfpropelled (VM), trailermounted (TM), unknown (O) User input Forward speed km/h 4-25 km/h User input Droplet Size n/a Use selected droplet size Wind Angle ° (i) Fine (ii) Medium (iii) Coarse (iv) Very coarse 90° Fixed Number of nozzles - None Use selected value Estimated from total area sprayed, dose, final mixing & loading concentration User input Emulator value will be divided by 2 for adult and child before taking the difference For WP3 estimate: Total number of nozzles calculated with nozzle spacing and number of passes. Estimated from rounding down (boom width*2) 18 From/for WP3 model (to estimate fairborne spray) WP3 model (to estimate fairborne spray) As input for spray volume rate (AR) As input for spray volume rate (AR) Literature review Table 4 Appendix B Literature review Table 5 Appendix B Literature review Table 6 Appendix B To estimate ASEwakein and ASEplumein (see algorithm) WP3 model and dispersion estimate WP3 model (to estimate fairborne spray) WP3 model (to estimate fairborne spray) WP3 model (to estimate fairborne spray) Input variables Distance used to estimate fairborne spray) Plume factor Unit M User input None Model input Fixed at 2m downwind for WP3 estimate - None Fixed at 0.01 Boom shielding - User input Shielding absent (1) Shielding present (0.7) From/for WP3 model (to estimate fairborne spray) Expert estimate Appendix A Literature review Table 7 Appendix C Table 4: Vehicle widths categories Vehicle type* Codes Width of vehiclesprayer relevant for wake (m)2 3 Range (m) S Width of the vehicle-sprayer (m)1 1.3 Small Average, Unimog, small&medium self-propelled Large, large self-propelled M, U, sSP, mSP 2.3 5 4.0-6.0 L, lSP 3.3 7 6.0-8.0 Unknown O - - 2-8 2.0-3.0 1 proposed width of vehicle ~2 times vehicle width (to take account of area around sprayer affected by wake effects) * The following vehicle-sprayers are included in the model (Appendix B): S = quad, golf cart, mini tractor (e.g. John Deere compact 1 series) M = average-sized tractor (e.g. John Deere 2 series) L = large-sized tractor, 4WD (e.g. John Deere 9R/RT series) U = unimog or similar sSP = small self-propelled (≤1000L tank) mSP = average sized self-propelled (>1000-3000L tank) lSP = large self-propelled (≥3000L tank) O = unknown 2 Table 5: Height dimensions of vehicles used for boom spraying Vehicle type Codes Height (m) Range (m) Small S 1.5 1.2-2.0 Average, Unimog, Small&medium self-propelled M, U, sSP, mSP 2.5 2.0-3.0 Large, large self-propelled L, lSP 3.5 3.0-4.0 Unknown O - 1.2-4.0 19 Table 6: Length / distance categories based on vehicle and spraying rig configurations1 Type Code Range (factor) * Length, distance^ (range, m) Lower Upper Front-fitted booms FB 0 Fixed 1.0 - Vehicle-mounted and self-propelled VM 1-6 0.1 0.9 Trailer-mounted TM 4-12 0.01 0.25 O 0-12 0.01 1.0 Unknown 1 based on literature review (Appendix B) vehicle-sprayer configurations with front-fitted booms (>0m) are currently assigned with a fixed value of 1, assuming no reduced effect compared with drawn booms ^ Table 7: Proposal of multipliers/factors for boom shields1 Category Description % Drift reduction A No shields 0 n/a 1 B Boom shields (any type) 30 10 – 70 0.7 1 Note: Range Multiplier based on literature review (Appendix C) For testing, the current dataset used for model development has no information on boom shields. For now, a factor of 1 are assigned to all data entries Table 8: Proposal cabin categories and associated multipliers/factors1 Category Description A B C 1 * No cabin Cabin without pressurized/filtered ventilation, OR with ventilation but without filtration Cabin with pressurized / filtered ventilation* % Cabin efficiency 1 60 n/a 10 - 90 Range (factor) Lower Upper Fixed 1 0.1 0.9 70 - >99 0.0001 Range (%) 90 based on literature review (Appendix D) criteria will be used in the user interface to verify ventilation & filtration compliances (ASAE S-525) 20 0.3 Table 9: Effect of variance explained by the model inputs (for fine nozzle, crop height 0,1m only)* Number Input Total effect of variance individually (%) 18.70 Total effect of variance including joint effects (%) 20.37 1 Concentration of a.s. in spray volume (g/l) 2 Spray volume rate (l/s) 17.14 19.29 3 Cabin factor 13.46 13.92 4 Boom Height above crop (m) 6.25 7.49 5 Boom width (m) 0.29 0.35 6 Wind speed (m/s) 30.49 32.70 7 Vehicle sprayer 4.56 6.12 8 Vehicle or trailer mounted 0.06 0.07 9 Plume factor 0.88 1.01 10 Forward speed (km/h) 0.91 1.10 11 Boom shielding 2.26 3.31 * Please note that GEM-SA may be outdated. The SA is intended for continuous inputs, so re-runs required for each categorical input 21 (f) Figures Figure 2: Example of SA output: Plot of main effects for crop height 0.1m and fine nozzle type W C A F Wind speed Concentration of a.s. in spray volume Spray volume rate (l/s) Cabin factor F W F 7 F W C A F F C A W A C F W C A C A C A C A W F 6 A C F W F A C F 5 W F A C W A C 4 main effect W W W 0.0 0.2 0.4 0.6 0.8 1.0 standardised input value Mean effect 95% probability interval Least important mean effects 22 Figure 3: Inhalation exposure: Plot of log10(all data*) (measured potential inhalation exposure all data*) and log10(geometric mean(model outputs)) *including results based on datasets where some inputs were imputed datasets (blue) and results where all inputs were known (apart from boom shielding) (red). 10000 1000 100 10 1 0.1 0.01 log 10 (Measured potential inhalation exposure(µg/m3)) 100000 0.001 0.001 0.01 0.1 1 10 100 1000 3 log10 (Geometric mean(Model outputs(µg/m ))) 23 10000 100000 Figure 4: Inhalation exposure: Plot of 95% CIs (grey lines), geometric means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10(measured potential inhalation exposure) [blue=where some inputs were imputed; red=complete set of inputs to the model except for fixed boom shielding imputation; circles=vehicle-mounted and selfpropelled; squares=trailer-mounted]. The index refers to the ordering of the model outputs and measured values when sorted on the medians of the model outputs. 100000 10 log (Potential inhalation exposure (µg/m3)) 10000 1000 100 10 1 0.1 0.01 0.001 0.0001 0 20 40 60 24 80 100 Index 120 140 160 180 Figure 5: Inhalation exposure: Plot of 95% CIs (grey lines), geometric means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10 (measured potential inhalation exposure [red dots represent complete sets of model inputs except for fixed boom shielding imputation]). The index refers to the number assigned to the model outputs and measured inhalation exposure when they are sorted based on the measured inhalation exposure. 100 10 1 0.1 0.01 10 log (Potential inhalation exposure (µg/m3)) 1000 0.001 0.0001 0 10 20 30 Index 25 40 50 60 Figure 6: Inhalation exposure: Plot of 95% CIs (grey lines), means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10 (measured potential inhalation exposure for datasets which had a complete set of inputs except for fixed boom shielding imputation = red dots) The index refers to the sorting on the medians of the model output. 100 10 1 0.1 0.01 10 log (Potential inhalation exposure (µg/m3)) 1000 0.001 0.0001 0 10 20 30 Index 40 50 60 3.2.2 Boom spraying dermal model (a) Description The boom spraying dermal models consist of a hand model (Dph) and whole body model (Dpw). The following activities of an operator during the boom spraying scenario are included: • Operating the vehicle (vehicle cockpit with & without cabin) • Stepping into/out of vehicle (contact with the vehicle interior & exterior) • Incidental activities with the spraying rig, e.g. mounting/dismounting the rig, manual (un-) folding of boom, maintenance of nozzles. The models consider important determinants of dermal exposure associated with field sprayer operations, e.g. type of spraying equipment (trailer, rear-mounted), number of unplugged nozzles and number of application tasks (Lebailly et al, 2009). As high dermal exposures are associated with 26 incidental activities with contaminated spraying rigs and nozzles (as opposed to the vehicle cockpit), the model includes both the vehicle cockpit and spraying rig as sources of exposure. The cleaning of vehicles and spraying equipment are not included as a parameter in the models because the effect of this activity on exposure is not known. Hand model The hand model considers two routes of exposure, i.e. surface contacts and deposition. An estimate of exposure from surface contacts is based on two key parameters as described by Gorman et al (2012), i.e. frequency of contacts and surface contamination levels. The estimate of deposition onto the hands is based on the inhalation estimate (Ipde). The input parameters and ranges of the models are presented in Table 10. (i) Frequency of surface contacts The hand model applies event frequencies to provide an indication of surface contact frequencies. The following event frequencies are included in the hand model (see Table 10): • Surface contacts with the vehicle cockpit is represented by events of stepping into/out of the vehicle, i.e. the number of mixing & loading periods (Fmlp) and number of times leaving the vehicle during trouble-shooting or nozzle maintenance (Fnr) • Surface contacts with the delivery system/rig is represented by (i) trouble-shooting and nozzle maintenance events (Fnr) and (ii) incidental contacts such as (de-)mounting and boom (un)folding prior to and after spraying (Frig) The mixing & loading periods (Fmlp) are estimated from the volume applied and the tank volume. The volume applied parameter is estimated from user inputs and are based on the total area sprayed (Ha), dose (kg/Ha) and final mixing & loading concentration (g/l). To determine the frequency of nozzle maintenance (Fnr), information on the frequency of trouble shooting/nozzle maintenance during boom spraying (e.g. Lebailly et al, 2009; Bell & Lloyd, 1988) was consulted. For the BS dermal hand model, a default of 3 nozzle maintenance events during spraying operations is proposed. When testing the model with data (with no information of nozzle maintenance frequencies), it is assumed that contact events with the nozzles during spraying (Fnr) are between 0 and a worst case of 15, with a 90% probability of between 0 and 5 and 10% probability of between 6 and 15. 27 Other events that may occur prior to or after spraying activities include nozzle replacement, (de-) mounting of the rig and manual boom (un)folding. A fixed event frequency with the rig and nozzles before and after spraying (Frig) is set at 1. Frig are not included as user input assuming that this is not a known input. These values may be updated with evidence of frequencies of different activities during boom spraying with different vehicle-sprayer types (e.g. Lebailly et al, 2009). (ii) Surface contamination levels The surface contamination level (µg/cm2 of a.s.) on field sprayers are highly variable and limited evidence is available on the determinants influencing the deposition on vehicle and sprayer surfaces of field sprayers. Ideally the surface contamination levels should be derived from data for different vehicle-sprayer configurations, different spray qualities, etc. However, this information is only available in a few specific settings (e.g. Balsari & Marucco, 2003). As result, it was decided to use a broad range of contamination levels as input for the model (Table 11 & Appendix E). For this purpose, contamination levels are based on three ‘Volume applied’ categories based on the percentage of the total spray volume that deposits on the vehicle and delivery system. Other parameters included in the model algorithm are spray quality, wind speed, sprayer type (or distance/length) and cabins. Spray quality is considered to distinguish between flat fan low drift nozzles (VMD=220um) and flat fan conventional nozzle (VMD=165um) (Balsari & Marucco, 2003). For wind speed, a distinction is made between wind speeds of <1m/s and >1m/s, assuming an increase in external contamination with higher wind speed (van de Zande et al, 2007). Self–propelled sprayers are known to have significantly higher external contamination levels compared to vehicle-mounted and trailer-mounted spraying rigs (Ramwell, et al, 2004). Other field trials indicate that dermal exposure during boom spraying applications are higher among farmers using rear-mounted sprayers compared to trailer-mounted sprayers (Lebailly et al, 2009). In the absence of conclusive evidence, it is assumed that the distance between the boom and operator affects dispersion and deposition. For this purpose, the model distinguishes between vehiclemounted and self-propelled sprayers (VM) and trailer-mounted sprayers (TM). As the surface contamination level estimate on the vehicle cockpit concerns the vehicle exterior, a cabin efficiency factor is also included. The effect sizes proposed for distance (vehicle- & trailer-mounted) and cabin 28 efficiency in the boom spraying inhalation model are adopted for this model (Tables 6 and 8 respectively). In addition, the transfer efficiency from surfaces to body parts (Cf) (Appendix H), body surfaces areas (BP) and the affected surface area of body parts (lcn/BP) (Appendix I) are included in the algorithm. The surface-to-hand transfer efficiency (Cf) is corrected to account for multiple contacts, for both contacts with the cockpit (Cftot_cockpit) and rig (Cftot_Rig) (Appendix J). (iii) Deposition Deposition on the hands is estimated by using the inhalation estimate as input (Ipde), while considering the droplet settling velocity (vdep) and application time (t) (see table 10). One side of both hands is considered relevant as the affected surface area of the hands (50%). Whole body (excl. hands) model To estimate the whole body dermal exposure, the hand model is used with an extrapolation based on a review of body part distributions during boom spraying activities (Appendix F). The extrapolation distinguishes between the presence (Fnr >0) or absence (Fnr = 0) of hand exposure due to trouble shooting and nozzle maintenance. The extrapolation assumes that in the absence of trouble-shooting and nozzle maintenance (Fnr = 0), the hands are less contaminated compared to the whole body (see algorithm). (b) Algorithms I Hand exposure model (Dph) Dph= ((Sccockpit * Ccoi * Sq * Ws * d * If) * (Cftot_cockpit * BPhands*lcn/BPpalms) + (Scrig * Ccoi * Sq * ws) * (Cftot_rig * BPhands*lcn/BPpalms) + Ddep Where 𝐷𝑑𝑒𝑝 = 𝐼𝑝𝑑𝑒 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑡𝑜𝑝 And 29 𝐶𝑓𝑡𝑜𝑡_𝑐𝑜𝑐𝑘𝑝𝑖𝑡 = 𝐶𝑓𝑡𝑜𝑡_𝑟𝑖𝑔 = 𝐹𝑚𝑙𝑝+ 𝐹𝑛𝑟 � �𝐶𝑓 0.5(𝑘−1) � 𝑘=1 𝐹𝑛𝑟+ 𝐹𝑟𝑖𝑔 � �𝐶𝑓 0.5(𝑘−1) � 𝑘=1 Dph potential dermal exposure to a.s. in µg on the hands Dph_cockpit potential dermal exposure to a.s. in µg on the hands (from vehicle/cockpit surfaces) Dph_rig potential dermal exposure to a.s. in µg on the hands (from rig, boom, nozzles) Sccockpit surface contamination levels (l/m2) of vehicle/cockpit Scrig surface contamination levels (l/m2) of rig Ccoi concentration of a.s. in spraying volume (µg/l) Sq droplet quality (SqFine = 2.8; SqMediumCourseVeryCourse = 1) ws ws ≤ 1m/s (factor 0,5), ws > 1m/s (factor 1) Fmlp number of mixing & loading periods (= tankfulls applied) (assuming that the operator stepped in/out of the vehicle during M&L periods) Fnr number of nozzle maintenance events (indicating both (i) frequency of stepping i/o of vehicle (contacts with cockpit), and (ii) contacts with boom/nozzles) Frig frequency of incidental contacts with the rig/boom excl. nozzle maintenance (assuming a default of 1 contact event during (de-)mounting or boom folding) prior to and after spraying d distance effect (mounted/self-propelled (VM) and trailer mounted (TM) sprayers) If cabin factor BPhands total surface area of hands (m2) lcn/BPpalms affected surface area of hands (m2 of BP) (palms) Cf transfer efficiency for the grasping of smooth surfaces = triangular distribution with lower value 0.23, mode 0.45 and upper 0.68 Cftot_cockpit corrected transfer efficiency for multiple contacts (maximum: 0.89 with Cf = 0.45; Fmlp + Fnr = 6) (Appendix H) Cftot_Rig corrected transfer efficiency for multiple contacts (maximum: 0.89 with Cf = 0.45; Fnr + Frig = 6) (Appendix H) Ddep Potential dermal hand exposure to droplets via deposition (µg) 30 Ipde Inhalation exposure estimate from inhalation model - a.i. in air (µg/m3) t Total application time (s) vdep droplet settling velocity average deposition speed of droplets in the air (m/s) lcn/ BPhands_top Affected surface area of hands (m2 of BP) Rationale (Fmlp + Fnr) represent the frequency of surface contact events with the vehicle cockpit (Sccockpit) when stepping into/out of the vehicle during trouble-shooting and mixing & loading periods (Fnr + Frig) represent the frequency of surface contact events with the delivery system/rig (Scrig) during trouble-shooting and incidental contacts ((de-)mounting, boom folding) prior to and after spraying II Whole body (excl. hands) exposure model (Dpw) The following equation is used to estimate the potential dermal exposure (Dpw) of the whole body in µg: Dpw = Dph * Ew Dpw potential dermal exposure to a.s. in µg on the whole body (excl. hands) Dph potential dermal exposure to a.s. in µg on the hands (from hand exposure model) Ew extrapolation from hand exposure model (ew-nozzle or ew-nonozzle) to body exposure ew-nozzle hand exposure incl. nozzle maintenance extrapolated to whole body (=0,2), where Fnr >0; assuming 80% of total dermal exposure is distributed on the hands and 20% on the whole body ew-nonozzle hand exposure excl. nozzle maintenance extrapolated to whole body (=0,35), where Fnr = 0; assuming 65% of total dermal exposure is distributed on the hands and 35% on the whole body Note: for testing purposes, Ew, is estimated using the Fnr as indicated in Table 10 (90% probability between 0-5, and 10% probability between 6-15). Software: default 3 (c) Model testing Please refer to the type of plots/graphs and explanation under item 3.2 31 I Hand exposure model (Dph) The available data is not suitable for testing the BS hand model, because of data gaps of important parameters. For example, the number of contact events with the nozzles is not known in the data. Assuming a frequency of nozzle maintenance (Fnr) of 90% probability between 0 and 5 and 10% probability between 6 and 15, the model predicts a conservative exposure compared to the broad range of data (Figure 7). A default nozzle maintenance frequency is set at 3 in the software. Unfortunately, inputting a distribution for nozzle events will not be feasible as the probabilistic output of the model (using iterations) will not be able to produce a single estimate with repeated runs of the model. II Whole body (excl. hands) exposure model (Dpw) The whole body exposure model was compared with the dataset using the extrapolation of the hand model (Figures 1). (d) Tables Table 10: Boom spraying dermal models - main user and model inputs* No User inputs Model inputs Lower Upper 1 Concentration of a.s. Concentration of a.s. 0.05 50 in spray volume (g/l) in spray volume, Ccoi (g/l) 2 Volume applied Volume applied (l), 3 50 25000 (estimated from total categories area sprayed, dose, final mixing&loading concentration) 3 n/a Cockpit surface 1E-4 5E-2 contamination levels, Sccockpit (l/m2) 4 n/a Rig surface 1E-2 2.5E-1 contamination levels, Scrig (l/m2) 5 Number of M&L Number of M&L 1 20 periods (estimated periods (tankfulls with volume applied applied) (Fmlp) and tank size) 6 Frequency of nozzle Frequency of nozzle 0 15 maintenance (Fnr); maintenance (Fnr); 32 Type selected value Based on BROWSE data selected value BROWSE data uniform distributions Field trial data Table 11 Appendix E Field trial data Table 11 Appendix E BROWSE data uniform distributions selected value skew (90% probability Field studies No User inputs user selected: 0 - 5; >5; default in software = 3 Model inputs (With 90% probability between 0-5) (for testing with data only) Lower Upper 7 n/a Fixed at 1 - 8 Cabin factor (3 categories) Frequency of contacts with rig prior to/after spraying (Frig) Cabin factor (3 categories with ranges) 0.0001 9 Vehicle or trailer mounted (3 categories – VM, TM, Ounknown)) 10 Spray quality (sq) (4 categories) 11 Wind speed (ws) 12 n/a 13 Gender (male, female) Vehicle-sprayer length (d) Vehicle-mounted (0.1 – 0.9) Trailer-mounted (0.01 - 0.25) Front-fitted booms fixed at 1 Spray quality (sq)(4 categories); SqFine = 2.8; SqMedium / Coarse / VeryCoarse = 1 Wind speed (ws) Ws ≤ 1m/s (factor 0,5), Ws > 1m/s (factor 1) Transfer efficiency from surface to the hands (Cf); e.g. 1st contact = 0.45; 2nd-6th contact (*0.5); see text Surface area of hands (BP) (m2) 14 n/a 15 From BS inhalation Fraction affected surface area of hands (palms) (lcn/BPpalms) Inhalation exposure / 33 Type between 0 and 5 and 10% probability between 6 and 15) Fixed Based on 1 see inhalation model 0.01 1 uniform distributions Literature review Table 8 Appendix D Field trial data Table 6 Appendix B 1 2.8 selected value Field trials 0.5 1.0 selected value Field trials 0.23 0.68 triangular distribution between 0.23 and 0.68 with a mode at 0.45 IOM database Literature review Appendix H & J - - Appendix I Fixed 0.5 - lognormal distributions (Males: log(SA) ~ Normal (-2.2319, 0.109922) Females: log(SA) ~ Normal(2.4313, 0.10362)) Fixed Expert opinion Field studies Literature review Appendix I See previous No User inputs model 16 n/a Model inputs air concentration, Ipde (µg/m3) Time (s), t (spraying time) Lower Upper Type Fixed at user input 17 n/a Droplet setting velocity, vdep 0.078 18 n/a Fixed 0.5 19 n/a Affected surface area of hands (back of hands), (lcn/BPhands_top) Body model (extrapolation from hands); with nozzle maintenance (Fnr>0) and without nozzle maintenance (Fnr=0) 0.2 0.313 Uniform 0.35 fixed to 0.2 if Fnr > 0 or 0.35 if Fnr = 0 (based on 2 Fnr situations) Based on section Derived from background calculations (Appendix….) Example: Nuyttens et al. 2009 Literature review Appendix I Literature review Appendix F * Models distinguish between Volumes applied and contamination levels of different surfaces Table 11: Surface contamination levels using Volume Applied categories Category 1 2 3 * + ^ Volume applied Lower <50 1000 5000 Upper 1000 5000 >10000 Contamination level cockpit, Sccockpit (l/m2)*+ Lower Upper -4 1.0E 5.0E-3 1.0E-4 2.50E-2 1.0E-4 5.0E-2 Contamination level delivery system/rig, Scrig (l/m2)*^ Lower Upper -2 1.0E 2.5E-1 1.0E-2 2.5E-1 1.0E-2 2.5E-1 broad categories extrapolated from field sprayer and orchard sprayer contamination studies using the % of total spray volume depositing on the vehicle and delivery system (e.g. Balsari & Marucco, 2003, Ramwell et al, 2005, Michielsen et al, 2012). for Sccockpit, a lower estimate of category 2 is adopted for categories 1 and 3 to account for a minimum contamination level at the start of the operation for Scrig, categories 2 is used as reference for categories 1&3 to account for surface loading on the rig and surface loading over time. A worst case highly contaminated delivery system/rig contamination (incl. potential dripping from nozzles) is applied as an upper range for all categories 34 (e) Figures Figure 1: Hand exposure: Plot of log10(measured potential dermal exposure (hands)*) and log10(geometric mean(model outputs)) (assuming a 90% probability of 0-5 nozzle maintenance events) *including data sets which have model inputs that were imputed 35 Figure 2: Whole body exposure: Plot of log10(measured potential dermal exposure (body) *) and log10(geometric mean(model outputs)) * including data sets which have model inputs that were imputed 36 3.2.3 Mixing and loading inhalation model (a) Using the ART model Although often considered to be negligible, inhalation exposure to pesticides during the mixing & loading phase cannot be ruled out, since pure formulation is handled. Therefore, inhalation exposure will also be modeled for the mixing/loading scenario. Until now only a limited amount of specific measurement data is available for this activity. However, the activities as performed during mixing/loading are considered to be comparable as the ones performed with chemicals in general, for which an exposure model for the estimation of inhalation exposure exists, namely the Advanced REACH Tool (ART) (www.advancedreachtool.com), which is developed based on a mechanistic model for inhalation exposure. Therefore it was decided to use this model to estimate inhalation exposure during mixing/loading activities with PPPs. The ART mechanistic model is based on a conceptual framework that adopts a source receptor approach, which describes the transport of a contaminant from the source to the receptor and defines seven independent principal modifying factors: substance emission potential, activity emission potential, localized controls, segregation, personal enclosure, surface contamination, and dispersion. ART currently differentiates between three different exposure types: vapours, mists, and dust (fumes, fibres, and gases are presently excluded). Various sources were used to assign numerical values to the multipliers to each modifying factor. The evidence used to underpin this assessment procedure was based on chemical and physical laws. In addition, empirical data obtained from literature were used. Where this was not possible, expert elicitation was applied for the assessment procedure. Multipliers for all modifying factors were peer reviewed by leading experts from industry, research institutes, and public authorities across the globe. In addition, several workshops with experts were organized to discuss the proposed exposure multipliers (Fransman et al., 2011). 37 The model consists of one algorithm to estimate the contribution from near-field (NF) [equation 1] and one for estimating the contribution from far-field (FF) sources [equation 2]. Personal exposure from a near-field source (Cnf) is a multiplicative function of substance emission potential (E), activity emission potential (H), (primary) localized control (LC1), secondary localized control (LC2; in case two localized controls are used simultaneously, and dispersion (D). The algorithm for a far-field source (Cff) also includes segregation (Seg) and personal enclosure/separation (Sep). The level of surface contamination (Su) for each activity depends on the location of the source, i.e. whether there is (i) a near-field source only [equation 3], (ii) a far-field source only [equation 4], or (iii) both near- and far-field sources [in which case the surface contamination in the near-field is assumed to dominate that of the far-field, see equation 3]. 38 Subsequently, the overall exposure is estimated by algorithm equation (5). The algorithm considers multiple activities [and exposure time (texposure)] within an 8-h work shift (ttotal) and also allows periods with assumingly zero exposure (tnon-exposure). With regard to dispersion, the following assumptions are made: • In case of outdoor mixing/loading, it is assumed that this takes places close to buildings • In case of mixing/loading under a shelter/covering/roof, it is assumed that these conditions are comparable to ‘outdoors – close to buildings’ • In case of indoor mixing, only good natural ventilation is assumed For use within BROWSE, the above algorithms are included in the total algorithm for the estimation of exposure during mixing loading. To do so, the relevant modifying factors within the ART mechanistic model were considered relevant for the mixing & loading scenario. If possible, defaults were chosen, and to do so, the relevant modifying factors per exposure scenario were identified. It is assumed that far field exposure is not relevant for mixing & loading PPPs, since in practice one operator is performing all the activities, and thus no secondary exposure is assumed. Also the distance of the operator to the source is assumed to be <1 meter. For the estimation of inhalation exposure with the ART model, it is necessary to take into account the duration of the various activities that together form the mixing & loading scenario. To be able to do so, assumptions are made with regard to the relative contribution of the various activities (transport, opening/closing, decanting) to the total duration of a mixing & loading period, both for solids and liquids (Table 14). (b) Inhalation exposure estimates and testing The inhalation model was not tested using the exposure dataset. However, the model has been calibrated as described in Schinkel et al. (2011). The mechanistic model output provides a (dimensionless) relative score for the GM exposure of a scenario and is fitted to exposure 39 measurements to ‘translate’ these scores to a quantitative exposure estimate in mg/m3 with the following equation: Yijk is the exposure level for the kth measurement within the jth company in the ith scenario. Xijk is the ln-transformed exposure level; ln(α) is the intercept (natural logarithm of the slope on the natural scale); δi represents the random effect of the ith scenario, cij represents the random effect of the jth company in the ith scenario and εijk is the residual error term. It is assumed that δi, cij and εijk values are normally distributed with mean equal to zero and variances representing the between-scenario, between-company (or work-site), and within-company components of variance. The companies are nested within scenarios. With this method the relative ART mechanistic model scores are still proportional to actual exposure levels and importantly the effects of individual MFs are preserved. The intercept (ln(α)) represents the estimated exposure if the ART model score is 1. The calibration results presented in Table 15 are used to estimate inhalation exposure for mixing & loading activities based on the relative ART-scores for specific scenarios (based on broad model). The calibration provided insight into the uncertainty of the estimated GM for specific scenarios. This uncertainty is expressed as an UF, and this UF is used to calculate confidence limits around the estimated GM exposure. The analyses indicate that the model could estimate with 90% confidence GM exposure levels within a factor between two and six of the measured GM exposure levels (depending on the form of exposure). The total percentage of explained variance is 61% for the abrasion exposure form (not stated here), 64% for dust (non-abrasive), 60% for vapours (not stated here) and 30% for mist. Two models were developed and presented below, i.e. ML liquids and ML solids. 3.2.3.1 Mixing and loading liquids Main model and user inputs are presented in Tables 12 and 13. It is assumed that only low-volatile substances are used in liquid formulations (in ART framework cut off point: vapour pressure <10 Pa), and thus only exposure to mists is taken into account. In case of mechanical application, inhalation exposure during the mixing of the diluted formulation in the tank of the spray equipment is considered negligible, since it is assumed that this occurs in the 40 far field, with low-volatile substances, with relatively undisturbed surfaces (no aerosol formation) and a limited open surface area (opening tank). In case of liquid formulations it is assumed that a liter of liquid formulation weighs one kg (comparable to water), to be able to use the concentration of active substance in the formulation (in g/L) as the weight fraction (g/kg). 3.2.3.2 Mixing and loading solids Main model and user inputs are presented in Tables 16. Dustiness classifications in ART were adapted for pesticide formulations as shown in Table 17. (a) Tables Table 12: Mixing & loading liquids inhalation model - main user and model inputs* No 1 Model inputs Concentration of a.s. of formulation in container/packaging, C (µg/l) Lower 5 Upper 1000 Type selected value Based on * 2 3 Viscosity (fixed medium) Partial vapour pressure Fixed 0.3 - - * * - Fixed fixed 10Pa/30000Pa Fixed - Fixed Fixed * * - selected value * 2.7 selected value * 1 selected value * 0.001 selected value * - Fixed * 1 selected value * fcoi 4 M&L activities (3 categories) – (i) transport; (ii) opening & closing; (iii) emptying & pouring 5 Splash loading (emptying) Fixed 3 2 6 Contaminated surface area (0,3-1m ) Fixed 0.001 (transport, opening&closing) 7 Duration of activities (based on number of containers used * default time) 8 (i) Indoor (natural ventilation, large room) or 0.75 outdoors (ii) Indoor (natural ventilation, small room) (iii) Outdoors (only close to building or under shelter) 9 Localized controls ((i) extraction (canopy 0.5 hood) (0.5); (ii) extraction (other) (1); (iii) none (emptying only) (1), default = 1 10 Use rate (l/min) ((based on total number of 0.01 containers used * container size (l) / duration of activities (min)) (emptying only) 11 Level of contamination (transport & Fixed 0.1 opening/closing) (default of <10% of surface) 12 Level of containment (emptying only) 0.3 * Fransman et al 2011; Schinkel et al, 2011; Tielemans et al 2008 Table 13: Default values used for M&L (liquids) inhalation (ART) model for three activities 41 * Parameter Description Viscosity Low Transport Opening & dosing Emptying / pouring 1 1 1 0,3 0,3 0,3 10-100 l/min - - 0,01 1-101 l/min - - 0,003 0,1-1 l/min - - 0,001 Open process - - 1 Product-to-air interface - - 0,3 Splash loading - - 3 2 0,001 0,001 - medium Use rate Level of containment Type of application Contaminated surface area 0,3-1m Level of contamination 10-90% of surface 0,3 0,3 - <10% of surface 0,1 0,1 - None 1 1 - Local ventilation - - 0,5 Indoor large room 0,9 0,9 0,9 Indoor small room 2,7 2,7 2,7 Outdoor close to buildings 0,75 0,75 0,75 Housekeeping practices 0,01 0,01 0,01 Localized controls Dispersion Surface contamination Table 14: Use of duration in the M&L (liquids) inhalation model Scenario/ activity Activity class ART Total mix / load period Transport Opening / closing Handling of contaminated objects Handling of contaminated objects Pouring / emptying Transfer of liquid products – falling liquids 1 min per event 0.5 min opening, 0.5 min closing 1.5 min per event Assumed duration (in min) 1 min per container 1 min per container 1.5 min per event Table 15: Calibration results of the ART model 2 bs (95% CI) a 2 bc σ 2 residual Dust 3.01 0.81 (0.25-1.36) 0.38 (0.12-0.64) 2.29 (2.03-2.55) 3.48 Mist 10.23 1.14 (0.17-2.10) 1.65 (0.94-2.36) 2.62 (2.06-3.18) 5.41 between-scenario component of variance, CI = confidence interval between-company component of variance c residual error component of variance d total variance 42 (95% CI) 2 d total σ b σ c ln(α) a (95% CI) b Exposure form σ Table 16 Mixing & loading solids inhalation model - main user and model inputs*(see Appendix O for detailed information) No 1 2 Model inputs Concentration of a.s. of formulation in container/packaging, C (µg/kg) Lower Sample data Upper Sample data Type selected value - - 0.01 Fixed 1 Fixed 0.03 1.0 - selected value fixed fixed Fixed 1 Fixed 3 Fixed 1 - - fixed fixed fixed selected value 0.75 2.7 selected value 0.5 1 selected value 1 3 selected value Fixed 0.01 - fixed fcoi M&L activities (3 categories) – (i) transport; (ii) opening & closing; (iii) emptying & pouring 3 Dustiness (see Table 17) 4 Moisture content 5 Level of contamination - handling of slightly contaminated (layers of less than few grams) objects (transport, opening & closing) 6 Carefulness of handling (transport, emptying); routine/normal 7 Dropping height (drop height ≥ 0.5 m); emptying only 8 Level of containment (open) 9 Duration of activities (based on number of containers used * default time) (see Table 3) 10 (i) Indoor (natural ventilation, large room) or outdoors (0.8991) (ii) Indoor (natural ventilation, small room) (2.7) (iii) Outdoors (only close to building or under shelter) (0.75) 11 Localized controls ((i) extraction (canopy hood)(0.5); (ii) extraction (other) (1) ; (iii) none (emptying only), default = none (1) 12 Use rate (l/min) ((based on total number of containers used * container size (l) / duration of activities (min)) (emptying only) (i) Transferring 10-100 kg/min (3) (ii) Transferring 1-10 kg/min (1) 13 Surface contamination (housekeeping practices) * Fransman et al 2011; Schinkel et al, 2011; Tielemans et al 2008 Table 17 fixed Dustiness classification for M&L (solids) inhalation model Description (i) Firm granules, flakes or pellets Classification (data, user interface) Macro granule (GG) (ii) Granules, flakes or pellets Fine granule (FG) Wettable granule (WG)* Water soluble granule (SG)* (iii) Coarse dust (iv) Fine/course dust or powder (v) Extremely fine and light powder Wettable powder (WP)* (Water-soluble) powder for dry seed treatment (DS, SS) Flo-dust (GP) Dustable powder (DP) NA # transport; opening & closing; emptying / dumping * fine/course dust is considered reasonable worst case, to be applied as default 43 All activities# 0,01 0,03 0,1 0,3 (default)* 1 3.2.4 Mixing and loading dermal models 3.2.4.1 (a) Mixing and loading liquids Description This scenario includes mixing and loading (ML) of liquids during the ground boom, broadcast, handheld spraying and aerial spraying operations. Based on the different M&L procedures and the available information and data, the following models were developed: • Open pour to tank top or induction bowl - Hands model - Whole body (excl. hands) model • Transfer with a mechanical transfer/coupling devices (MTD) fitted on induction bowls, incl. closed transfer systems (CTS) - Total body model (hands & body) The following activities of an operator during the mixing & loading scenario were considered in the model development: • Transport / collection of containers • Opening and closing of containers • Emptying and decanting • Using premixing containers and measuring jugs The models emphasize the principle route of exposure as surface contacts with containers used. Deposition of droplets during ML scenarios is assumed negligible. The models use quantitative values derived from experimental data as mass balance inputs, i.e. (i) surface contamination levels and/or (ii) body part contamination levels. These inputs differentiate between ML methods, container sizes and probability of spillages for single-event ML activities. To develop a mechanistic algorithm, other important parameters are included to estimate operator dermal exposure, e.g. concentration of a.s. in the formulation, frequency of surface contacts, transfer efficiency from surfaces to body and affected body parts. 44 Model 1a: Open pour to tank top or induction bowl I Hand exposure model The algorithm used to estimate dermal exposure during mixing and loading is based on two key parameters associated with surface contacts as described by Gorman et al (2012), i.e. frequency of contacts and surface contamination levels. The input parameters and ranges of the models are presented in Table 16. In order to incorporate exposure through splashes and dripping, the model algorithm includes experimental hand contamination data of a standardized emptying procedure of the respective container sizes. Dermal hand exposure is first estimated per ML period (Dph_mlp) in µg/hands, which is assumed a summation of hand-to-surface exposure to the container exterior (Dph_container-ext, see item ii) and direct emission from splashes or dripping (Dph_es, see item iii). Subsequently, if more than one ML period is relevant, the model estimates a total hand exposure for multiple ML periods, expressed as (see algorithms). For this purpose, the frequency of hand contacts with containers are used (item i). (i) Frequency of surface contacts The hand model applies event frequencies to provide an indication of surface contact frequencies. The following event frequencies are included in the hand model: • Surface contacts based on the total number of containers used incl. re-use (Ftnc) • Number of ML periods (Fmlp) (assuming a worst case where the 1st container used during each ML period is being re-used, unsealed and/or contaminated) Ftnc is based on the number of containers used per ML period and gives an indication of the contact events. The number of ML periods (Fmlp) is included in the algorithm to estimate the total number of containers used per ML period. The total number of containers used per ML period (Ftnc/Fmlp) is considered representative of contact events, with a distribution derived from the data of ≈1-15. Ftnc and Fmlp is estimated with the tank size (l), final ML concentration of a.s. in spray volume (g/l), container size (l), formulation concentration (g/l) and total spray volume applied (l). The hand model algorithm includes the transfer efficiency from surface to the hands (Appendix H), as well as the affected surface area of the hands involved in contacts (Appendix I). The surface-to-hand transfer efficiency (Cf) is corrected to account for multiple contacts, expressed as Cftot-container-ext (Appendix J). 45 (ii) Surface contamination levels Surface contamination levels were derived from various experimental studies (Glass et al, 2009; Gilbert et al, 2000; Mathers et al, 1999). The data distinguish between two ML methods (tank-top and induction bowl pour) and four container sizes (1L, 5L, 10L, 20L). Surface contamination levels (lower and upper range, Sccontainer-ext) were derived for the different container sizes and expressed in different categories, each with a probability. The probabilities are based on the probability frequencies in the experimental data (Table 17). The derived values are therefore indicative and only provide information about the relative performance of the different container sizes when used with a particular filling technique (Appendix G). (iii) Spillages In order to incorporate exposure through splashes and dripping, the model algorithm includes experimental hand contamination data of a standardized emptying procedure. Hand contamination levels (es) were derived from the available experimental studies assuming that a significant amount of contamination occurred through spillages (using the standard CSL container pouring test) (Table 18). These values were derived for tank-top and induction bowl ML methods using the four container size categories. Probability distributions of splashes occurring on the hands during the use of different container sizes and ML methods are also available and included in the hand contamination estimate. The derived values are based on limited experimental data and are only indicative input of hand contamination levels in practice. Model 1b: Open pour to tank top or induction bowl II Whole body exposure (excl. hands) model Whole body contamination levels (eopen pour) were derived from experimental studies for tank-top pour and induction bowl pour ML methods using four container size categories (Table 19). These values were derived from whole body contamination (coveralls) based on the standard CSL container pouring test protocol (for the pouring into a simulated induction hopper). In the absence of data for tank-top pouring, the latter values were adopted from induction bowl pouring. The derived values are based on limited experimental data and are only indicative input of whole body contamination levels in practice. 46 Model 2: Mechanical transfer devices (MTD) and closed systems I Total body (incl. hands) exposure model Limited data is available for mechanical transfer devices (MTD) and closed transfer systems (CTS). These ML methods are also more complicated due to the wide range of systems used and protocols followed for each system. The experimental data available distinguish between six systems, where a number of the systems had a common design of valve connection to the pesticide kegs and dry break coupling connecting the system to a mock sprayer. In the absence of consistent experimental data for different body parts, it was decided to derive the total operator contamination from all the available systems sampled, giving a rough estimate of operator contamination (Table 19). (b) Algorithms Model 1: Open pour to tank top or induction bowl I Hand exposure model (Dph) Per M&L period Dph_mlp = Dph_container-ext + Dph_es Where Dph_container-ext = Cfcoi * (Sccontainer-ext * Cftot_container-ext* BPhands * lcn/BPpalms)) Dph_es = Cfcoi * (es * (Ftnc/Fmlp)) 𝐹𝑡𝑛𝑐 𝐹𝑚𝑙𝑝 𝐶𝑓𝑡𝑜𝑡_𝑐𝑜𝑛𝑡𝑎𝑖𝑛𝑒𝑟_𝑒𝑥𝑡 = � 𝑘=1 𝐹𝑡𝑛𝑐 𝐹𝑚𝑙𝑝 𝐹 �� 𝑡𝑛𝑐 �−1� �𝐶𝑓 0.5 𝐹𝑚𝑙𝑝 � and 𝐹𝑡𝑛𝑐 𝐹𝑚𝑙𝑝 ≥ 1 is always an integer ≥ 1 as Ftnc is an integer * Fmlp Dph_container-ext Hand exposure from hand contacts with container exteriors ((µg/hands) Dph_es Hand exposure from direct emissions such as splashes and dripping from containers (µg/hands) 47 For combined ML periods Dph = Dph_mlp * Fmlp Dph_mlp dermal hand exposure per ML period (µg/hands) Sccontainer-ext surface contamination level for different container sizes and M&L methods (l/cm2) Cfcoi concentration of a.s. in formulation / packaging (µg/l) Ftnc total number of containers used incl. repeated use (Ftnc) Fmlp number of ML periods es hand contamination representing dripping/splashes after standard emptying operation (separate for tank-top (es-tanktop) and induction bowl (es-induction bowl) BP total surface area of hands (cm2) lcn/BP affected surface area of hands (cm2 of BP) Cf transfer efficiency from surface to the hands Cftot-container-ext corrected transfer coefficient for multiple contacts (maximum: 0.89 with Cf = 0.45; 𝐹𝑡𝑛𝑐 𝐹𝑚𝑙𝑝 = 6) (Appendix H) Model 1: Open pour to tank top or induction bowl II Whole body (excl. hands) exposure model (Dpw) Dpw = Cfcoi * eopen pour * Ftnc Dpw potential dermal exposure to a.s. in µg on the whole body (excl. hands) Cfcoi concentration of a.s. of formulation in container/packaging (µg/l) eopen pour whole body exposure during a standard emptying operation / per container used (l/whole body) as per container size and ML method Ftnc total number of containers used incl. re-use of un-emptied containers Model 2: Mechanical transfer devices (MTD) using induction hoppers and closed transfer systems (CTS) I Total body (incl. hands) exposure model (Dpw) Dpw = Cfcoi * emtd * Fmlp 48 Dpw potential dermal exposure to a.s. in µg on the total body (incl. hands) Cfcoi concentration of a.s. of formulation in container/packaging (µg/l) emtd indicative dermal exposure during the use of MTS and CTS per ML period (l/whole body incl. hands) Fmlp (c) frequency of contacts as number of ML periods Model testing Please refer to the type of plots/graphs and explanation under item 3.2. Also, the number of data available for single and multiple M&L periods vary and are different in the respective plots. Model 1: Open pour to tank top or induction bowl I Hand exposure model Only ML data was used (excl. application) to test the hand model. As shown in the algorithms, the model first estimates hand exposure for a single ML period (Figure 13 & 14). As shown in Figure 14, model estimates tend to over-estimate for the ML of boom sprayers and under-estimate for handheld spraying operations. However, this may be due to the imputed data used. The ML model then estimates the exposures of all the ML periods combined (Figure 15 & 16), which includes ML data with both single and multiple ML periods. Mixing & loading during hand-held spraying operations, in particular, shows a high variability and broad range in exposures (also see Appendix K). The final model estimates of all the ML periods combined (Figure 16) appear to be conservative but do not capture the large variation in the exposure data. However, many of the outliers may be ascribed to less reliable imputed datasets. Model 1: Open pour to tank top or induction bowl II Whole body exposure model Testing of the whole body model with exposure data is plotted in Figures 15 and 16. Model 2: Mechanical transfer devices (MTD) and closed systems I Total body exposure model Too few datasets are available to compare model outputs with data. 49 (d) Tables Table 16: Mixing & loading dermal models - main user and model inputs* No 1 User inputs Concentration of a.s. of formulation in container/packaging (g/l) 2 4 Container size – 4 categories: ≤2,5l; >2,5l - <7,5l; 7,5l <15l; >15l M&L method (3 categories): - Tank-top (open) pour - Induction bowl (open) pour - Mechanical transfer/coupling devices (MTD) and closed transfer systems (CTS) n/a 5 n/a 6 (Volume applied/tank volume) roundup 7 8 (Volume applied, tank size, in-use concentration (Ccoi), container size, formulation concentration (Cfcoi)) roundup n/a 9 n/a 10 n/a 3 Model inputs Concentration of a.s. of formulation in container/packaging, Cfcoi (g/l) Container size – 4 categories based on experimental data (1, 5, 10 & 20L) M&L method (3 categories) Lower 5 Upper 1000 Type selected value Based on BROWSE data 0.5 25 selected value Experimental studies Appendix G - - selected value Experimental studies Appendix G Container contamination levels, Sccontainer-ext 2 (l/cm ), incl. frequency distributions (Table 19) Hand exposure after standard emptying operation, es (l/hands), as per container size and M&L method, incl. frequency distributions (Table 20) Frequency of contacts as number of M&L periods (Fmlp) Frequency of contacts as total number of containers used incl. repeated use (Ftnc) 2.0E -10 9.0E -7 uniform distributions Experimental studies Table 17 Appendix G 1.0E -6 2.0E -4 uniform distributions Experimental studies Table 18 Appendix G Dermal exposure during the use of MTSs and CTS (l/total body), emtd, Body exposure (eopen pour) during standard operation per container (L/body) as per container size and M&L method (Table 21) Transfer efficiency from surface to the hands (Cf); 1E -7 1.4E 1E -7 2.0E 50 1 15 selected value BROWSE data 1 45 selected value BROWSE data -3 uniform -4 uniform Experimental studies Table 19 Appendix G Experimental studies Table 19 Appendix G 0.16 0.47 triangular distribution IOM database Literature review No User inputs Model inputs 1st contact = 0.31; 2nd6th contact (*0.5) 11 Gender (male/female) Surface area of hands, 2 cm (BP) (testing: fixed male) 12 n/a Lower Fraction affected surface Fixed area of hands (palms) 0.42 (lcn/BPpalms) * models distinguish between container sizes and M&L methods 51 Upper - - Type between 0.16 and 0.47 with a mode of 0.31 lognormal distribution – see BS dermal model selected value Based on Appendix H&J Appendix I Appendix I Expert opinion Table 17: Surface contamination levels (Sccontainer-ext,L/cm2)^ and associated frequency distributions (%) derived for (manual) open pour M&L methods* for tank-top pour and induction bowl pour using four container size categories Tank-top category 1 lower Upper 0,001 0,01 1 2,00E-09 2,00E-08 5 6,00E-10 10 20 ml/container category 2 P* category 3 lower > upper < P* lower upper 0,01 1 70 2,00E-08 2,00E-06 30 6,00E-09 35 6,00E-09 6,00E-07 4,00E-10 4,00E-09 5 4,00E-09 2,00E-10 2,00E-09 5 2,00E-09 P* 1 1,5 – 4,5 40 6,00E-07 9,00E-07 25 4,00E-07 60 4,00E-07 5,00E-07 35 4,00E-07 50 4,00E-07 9,00E-07 45 lower upper P* 1 1,5 – 4,5 Container size (L) Induction bowl category 1 lower Upper 0,001 0,01 1 2,00E-09 2,00E-08 5 6,00E-10 10 20 ml/container category 2 P* category 3 lower > upper < P* 0,01 1 70 2,00E-08 2,00E-06 30 6,00E-09 25 6,00E-09 6,00E-07 70 6,00E-07 9,00E-07 5 4,00E-10 4,00E-09 35 4,00E-09 4,00E-07 60 4,00E-07 5,00E-07 5 2,00E-10 2,00E-09 10 2,00E-09 4,00E-07 55 4,00E-07 9,00E-07 35 Container size (L) * Frequency distributions 2 Container surface area (defaults, cm ) ^ 1 600 5 1754 10 2785 20 4421 Indicative ranges were derived from an analysis of surface contamination levels of different container sizes after a standard pouring procedure (Mathers et al 2000, Gilbert et al, 1999; Gilbert et al, 2000, Glass et al, 2009). Indicative frequency probabilities were derived from Glass et al (2009) 52 Table 18: Hand contamination levels (es, L/hands) and frequency distributions (%) derived for (manual) open pour M&L methods using four container size categories ~, tank-top and induction bowl respectively Tank-top category 1 category 2 lower Upper P* lower upper P* 1 1,00E-06 1,00E-05 99 1,00E-05 2,00E-04 1 5 1,00E-06 1,00E-05 95 1,00E-05 2,00E-04 5 10 1,00E-06 1,00E-05 70 1,00E-05 2,00E-04 30 20 1,00E-06 1,00E-05 60 1,00E-05 2,00E-04 40 Container size (L) Induction bowl category 1 category 2 lower Upper P* lower upper P* 1 1,00E-06 1,00E-05 99 1,00E-05 2,00E-04 1 5 1,00E-06 1,00E-05 99 1,00E-05 2,00E-04 1 10 1,00E-06 1,00E-05 80 1,00E-05 2,00E-04 20 20 1,00E-06 1,00E-05 70 1,00E-05 2,00E-04 30 Container size (L) ^ Indicative ranges were derived from an analysis of hand contamination levels during a standard pouring procedure (Mathers et al 2000, Gilbert et al, 1999; Gilbert et al, 2000, Glass et al, 2009). Indicative frequency distribution were derived from Glass et al (2009) 53 Table 19: Whole body contamination levels (eopen pour) derived for (manual) open pour M&L methods using four container size categories, and total body exposure during use of mechanical transfer devices (emtd) Category Container size (L) Lower 1 2 ^ Note Upper Whole body contamination (excl. 1 hands) after open pour , eopen pour (l/whole body)^ Mean Lower Upper 1.0E -8 1.0E -6 1.5E -7 1.5E 1 0.5 2.5 1.0E -7 2 >2.5 7.5 1.5E -6 5.0E -7 2.0E -6 3 >7.5 15 5.0E -6 4 >15 25 2.0E -5 5 25 - - Total body contamination after using 2 mechanical transfer devices (MTD) , emtd (l/total body)^ Mean* - Lower - Upper - -5 - - - 5.0E -5 - - - 2.0E -4 - - - - 7E -5 1.0E -7 1.5E -3 derived from whole body contamination (coveralls) after a standard pouring procedure (induction bowl). These values were also adopted for tank-top pour. Note: If raw data was not available for all combinations of methods and container sizes, values were adopted from other similar-sized container sizes operator contamination based on different body parts depending on type of loading systems used. Measurements taken after one protocol of six different mechanical transfer devices including both induction hoppers and closed transfer systems (including volumes of 5l-25l) Lower and upper ranges were derived using the mean values with a factor 0,1 and 10 respectively (as standard deviations could not be calculated from the studies) Derived from: eopen pour (Mathers et al 2000, Gilbert et al, 1999, Gilbert et al, 2000; Glass et al, 2009) and emtd (Glass et al, 2009) 54 (e) Figures Figure 13: Hand exposure: Plot of log10(potential dermal exposure (hands)*) and log10(geometric mean(model outputs)) (single M&L period) *including data sets which have model inputs that were imputed log 10 (Measured potential dermal exposure(µg/hands)) 1e+08 1e+07 1e+06 100000 10000 1000 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1000 10000 100000 1e+06 1e+07 1e+08 log10 (Geometric mean(Model outputs(µg/hands))) 55 Figure 14: Hand exposure: Plot of 95% CIs (grey lines), geometric means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10(measured potential dermal exposure (hands)*) (purple squares=boom sprayer, orange crosses=hand-held sprayers) (single M&L period) *including data sets which have model inputs that were imputed 1e+06 100000 10000 1000 100 10 log (Potential dermal exposure (µg/hands)) 1e+07 10 1 0 10 20 30 Index 56 40 50 60 Figure 15: Hand exposure: Plot of log10(measured potential dermal exposure (hands)*) and log10(geometric mean(model outputs)) (all M&L periods combined) * including data sets which have model inputs that were imputed log 10 (Measured potential dermal exposure(µg/hands)) 1e+08 1e+07 1e+06 100000 10000 1000 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1000 10000 100000 1e+06 1e+07 1e+08 log10 (Geometric mean(Model outputs(µg/hands))) 57 Figure 16: Hand exposure: Plot of 95% CIs (grey lines), geometric means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10(measured potential dermal exposure (hands)*) (blue circles=aerial, purple squares=boom sprayer, red diamonds=broadcast sprayer, orange crosses=hand-held sprayers) (all M&L periods combined) * including data sets which have model inputs that were imputed 10 log (Potential dermal exposure (µg/hands)) 1e+07 1e+06 100000 10000 1000 100 10 1 0.1 0.01 0 20 40 80 60 Index 58 100 120 140 Figure 17: Whole body exposure: Plot of log10(measured potential dermal exposure (body)) and log10(geometric mean(model outputs)) (Red circles= datasets where no model inputs needed to be imputed; blue circles= data sets which have model inputs that were imputed) (all M&L periods combined) log 10 (Measured potential dermal exposure(µg/body)) 1e+07 1e+06 100000 10000 1000 100 10 1 0.1 0.01 0.01 0.1 1 10 100 1000 10000 100000 1e+06 1e+07 log10 (Geometric mean (Model outputs(µg/body))) 59 Figure 18: Whole body exposure: Plot of 95% CIs (grey lines), geometric means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10(measured potential dermal exposure (body)*) (blue circles=aerial, purple squares=boom sprayer, red diamonds=broadcast sprayer, orange crosses=hand-held sprayers) (all M&L periods combined) * including data sets which have model inputs that were imputed 1e+07 10 log (Potential dermal exposure (µg/body)) 1e+06 100000 10000 1000 100 10 1 0.1 0.01 0.001 0 20 40 60 80 Index 60 100 120 140 160 3.2.4.2 Mixing and loading solids (a) Description The mixing and loading (ML) of solids scenario includes all operations where PPPs with a solid formulation (e.g. granules or powders) are mixed and loaded. These mixing/loading operations may occur in combination with for instance ground boom, broadcast, hand-held and aerial spraying operations, but are considered a separate activity in view of estimating exposure. The following models were developed: • Manual handling and emptying solid formulations from packaging (≤25kg), which includes collection, carrying, opening/closing, weighing and scooping: - Hands model - Whole body (excl. hands) model The models apply quantitative values derived from experimental data as mass balance inputs, i.e. body part contamination levels. It is assumed that the most important routes of exposure associated with dermal exposure are represented in the derived data, i.e. deposition on the skin and surface-tohand contacts. Hand and body exposure models The algorithm used to estimate dermal exposure during mixing and loading of solids (see paragraph ‘algorithms’) is based on experimental data (body part contamination values) in combination with model input parameters such as the concentration of a.s. in the formulation, number of packaging (e.g. bags/boxes) handled used per M&L period, amount of formulation used, dustiness of the product, the use of local ventilation, wind speed and the body part surface areas (see Table 20). (i) Body part contamination levels Indicative body part contamination levels were derived from an experimental study (Veldhof et al, 2006) with separate estimates for the hands (eh_w&e), arms (ea_w&e), torso (et_w&e) and legs (el_w&e). It is assumed that the sub-activity of emptying (or dumping) of solids is the most important activity in terms of dust emissions. As such, the experimental data used to derive body contamination levels are based on a single emptying event of a given amount of product, but it also include a sequence of other sub-activities such as collection, carrying, opening/closing of bags/boxes, and weighing and scooping. The data distinguish between two key determinants of exposure, i.e. (1) the amount of 61 formulation used and (2) the dustiness of the formulation during single events (emptying). Based on regression analysis, these two parameters were found to have the most significant effect on dermal exposure (Veldhof et al, 2006). The experimental data include the following amounts used: 200g, 1kg, 5kg, 10kg and 15kg. Considering the detailed analysis of the data (Veldhof et al, 2006), four categories were proposed for this exposure model (<1kg, 1,0-7,5kg, 7,5-12,5kg, >12,5-25kg). The dustiness of the formulation type is categorized as granule, powder and fine powder. The dustiness classification of the experimental test product was translated to a corresponding classification of PPP products. The contamination levels for hands, arms, torso and legs are presented in Table 20. The derived values are indicative values based on limited data only. More details on the experimental set-up and data derivation is presented in Appendix P. (ii) Frequency of events Both the hand and whole body models apply event frequencies that provide an indication of the frequency that a single emptying event will occur. The following event frequencies are included in the hand model: • Total number of packaging used excl. re-use (Fnc) • Number of ML periods (Fmlp) Since the body part contamination estimates are based on experimental data during a single event of emptying (a given amount of) solids, these contamination estimates are multiplied with the number of these events in the model. To derive a body part contamination estimate based on the ‘amount used’ during a single event, a background calculation is used to estimate the amount of product used per M&L period (and per packaging if multiple bags/boxes are used) (see background calculations, Appendix L). For example, the hand contamination estimate (eh_w&e) is multiplied with the number of packaging used per M&L period (Fnc/Fmlp). No evidence is available on how body part contamination during a single event will increase with multiple M&L periods or use of multiple packaging. In line with experimental skin loading studies (Appendix J), a skin loading constant (fdl) is applied to simulate a gradual increase with multiple M&L periods (see algorithms). (iii) Other factors Although the experimental data used to derive hand contamination levels (eh_w&e) are corrected for amounts used (<1kg – 25kg) and formulation type (granule, powder, fine powder), it does not take 62 account of potentially important parameters such as wind speed / air velocity (and wind direction) and local exhaust ventilation (indoors). These parameters are included in the algorithm as shown in Table 20. Local exhaust ventilation was not used in the experimental study. Although extraction ventilation is not commonly used during the mixing and loading of powders in the agricultural sector, the use of LEV cannot not be excluded either. A distribution of efficiency values for local exhaust capturing hoods (indoors) is proposed (i.e. efficacy values of 0.01-1) based on dumping activities in industry (from approximately 55% effectiveness) (Marquart et al, 2003). The experimental data represent exposure conditions of an indoor setting with relatively little airflow and unknown wind direction. To account for uncertainty regarding dispersion that is caused by wind speed and wind direction in diverse workplace settings (both indoors and outdoors), a broad range of effects are proposed (both increasing as decreasing) depending on the wind direction. Considering the limited information that will be available as model inputs, a broad distinction is made between low wind speeds (≤1m/s; efficacy values of 0.3-3) and higher wind speeds of greater than 1m/s (efficacy values of 0.1-10). Distributions of body surfaces areas (BP, cm2) for hands, arms, torso/trunk and legs are included in the algorithm, which differentiate between male and female distributions (see Appendix I). These body surface areas roughly correspond to the body surface areas used in the experimental study on which the body contamination levels are based. Based on the experimental data, it is assumed that the exposure to the rest of the body is negligible (Appendix P). (b) I Algorithms Hand exposure model (Dph) Per M&L period Dph = Cfcoi * ((eh_w&e * BPhands* Fnc/Fmlp) * vlev * ws) (1) M&L periods combined Dpht = Dph * fdl (2) where constant in fdl = 0.1 with each consecutive Fnc/Fmlp event (see previous runs with fdl) 63 Dph dermal hand exposure per M&L period (µg/hands) Dpht total dermal hand exposure of all M&L periods combined (µg/hands) Cfcoi concentration of a.s. of formulation in packaging (µg/kg) eh_w&e hand contamination during handling and emptying activities (kg/cm2) based on the amount used per emptying event (kg/event) Fnc total number of containers used Fmlp number of M&L periods (if Fnc/Fmlp <1 then equal to 1) vlev local exhaust extraction (canopy) hood used (indoors) ws wind speed, air velocity BPhands surface area of the hands (cm2) II Whole body exposure model (Dpb) Per M&L period Dpb = Cfcoi * (((ea_w&e * BParm + el_w&e * BPlegs + et_w&e * BPtorso) * Fnc/Fmlp) * vlev * ws) (1) M&L periods combined Dpbt = Dpb * fdl (2) where constant in fdl = 0.1 with each consecutive Fnc/Fmlp event Dph dermal whole body exposure per M&L period (µg/whole body) (assuming that only the arms, torso and legs are potentially exposed) Dpbt total dermal whole body exposure of all M&L periods combined (µg/whole body) Cfcoi concentration of a.s. of formulation in packaging (µg/kg) ea_w&e arm contamination during handling and emptying activities (kg/cm2) based on the amount used per emptying event (kg/event) el_w&e leg contamination during handling and emptying activities (kg/cm2) based on the amount used per emptying event (kg/event) et_w&e torso contamination during handling and emptying activities (kg/cm2) based on the amount used per emptying event (kg/event) Fnc total number of containers used 64 Fmlp number of M&L periods (if Fnc/Fmlp <1 then equal to 1) vlev local exhaust extraction (canopy) hood used (indoors) ws wind speed, air velocity (indoors & outdoors) BParm surface area of the arms (cm2) BPlegs surface area of the legs (cm2) BPtorso surface area of the torso / trunk (cm2) (c) Model testing Please refer to the type of plots/graphs and explanation under item 3.2. Note that the number of data available for single and multiple M&L periods vary and are different in the respective plots. I Hand dermal exposure model Only ML data was used (excl. application) to test the hand model. As shown in the algorithms, the model first estimates hand exposure for a single ML period, however, the data was plotted with model outputs with all ML combined, as shown in the algorithms. As indicated in Figures 19 and 20 (hand models) and Figures 21 and 22 (whole body models), very little data is available to evaluate the model outputs. Overall, it is clear that too little data is available to perform a proper testing to evaluate the variance explained by the models. A few issues that may determine the accuracy of the models are the sub-activities that are part of the exposure scenario of the experimental data (weighing, scooping), that are not always relevant in the dataset with PPP exposure data as tested. (d) Tables Table 20 No 1 2 3 Main user and model inputs * (also see Appendix P) User inputs Concentration of a.s. of formulation in container / packaging (mg/kg) Amount used per packaging <=1kg >1kg - <7.5kg 7.5kg - <12.5kg >12.5kg – 25kg Input required to Model inputs Concentration of a.s. of formulation in packaging, Cfcoi (µg/kg) Lower from data Upper from data Type selected value Based on BROWSE data Container size – 4 categories based on experimental data (200g, 1kg-5kg, 10kg, 15kg) 0.2 25 selected value Experimental data (Appendix G) Hand exposure after single # # Uniform Experimental 65 4 5 6 7 8 9 10 11 calculate: (i) amount used, (ii) product type^ Input required to calculate: (i) amount used, (ii) product type^ Input required to calculate: (i) amount used, (ii) product type^ Input required to calculate: (i) amount used, (ii) product type^ (see parameters required in Appendix L) (see parameters required in Appendix L) LEV used (indoors only) Wind speed, direction (indoors & outdoors) Gender (male/female) scooping & emptying operation^, eh_w&e (µg/cm2) Arms exposure after scooping & emptying operation^, ea_w&e (µg/cm2) data, Table 2a Appendix G Appendix L Experimental data, Table 2 Appendix G Appendix L Experimental data, Table 2 Appendix G Appendix L Experimental data, Table 2 Appendix G Appendix L BROWSE data Appendix L # # Uniform Legs exposure after scooping & emptying operation^, el_w&e (µg/cm2) # # uniform Torso exposure after scooping & emptying operation^, et_w&e (µg/cm2) # # uniform Total number of containers used (Fnc) 1 selected value Number of M&L periods (Fmlp) (Fnc/Fmlp max 6 events) from data from data / calcs from data selected value BROWSE data Extraction capturing hood used (indoors only) (yes=0.45-0.01; no=1) If ws ≤ 1m/s → 0.3 – 3.0; If ws > 1m/s → 0.1 – 10 0.01 1.0 normal Marquart et al, 2003 0.1 10 uniform Expert opinion Surface area of body part, cm2 (hands=BPhands; arms=BParms; legs=BPlegs; torso/trunk=BPtorso); (testing: fixed male) * * lognormal distributi on Appendix I* # see Table 2a/b, dependent on formulation type, etc. ^ includes dustiness of product (granule, powder, fine powder) 66 (e) Figures Figure 19: Hand exposure: Plot of log10(measured potential dermal exposure (hands)*) and log10(geometric mean(model outputs)) (all M&L periods combined) *including data sets which have model inputs that were ‘complete’ (red) and imputed (blue) 67 Figure 20: Hand exposure: Plot of 95% CIs (grey lines), geometric means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10(measured potential dermal exposure (hands)*) (all M&L periods combined) *including data sets which have model inputs that were ‘complete’ (red) and imputed (blue) 68 Figure 21: Whole body exposure: Plot of log10(measured potential dermal exposure (whole body)*) and log10(geometric mean(model outputs)) (all M&L periods combined) *including data sets which have model inputs that were ‘complete’ (red) and imputed (blue) 69 Figure 22: Whole body exposure: Plot of 95% CIs (grey lines), geometric means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10(measured potential dermal exposure (whole body)*) (all M&L periods combined) *including data sets which have model inputs that were ‘complete’ (red) and imputed (blue) 70 3.2.5 Orchard spraying inhalation model (a) Description Operator inhalation exposure is assumed to primarily originate from the spraying equipment that releases airborne droplets during orchard spraying (OS) activities. The inhalation model focuses on droplet spray and predicts the air concentration of droplets in the breathing zone of the operator much like the Boom spraying model. This air concentration is converted to the air concentration of the given active substance (a.s.) based on the concentration of a.s. in the spraying volume. To estimate inhalation exposure of operators during Orchard spraying, the following three mechanisms are considered (please see appendix A for more information): • The airborne spray: The overspray factor is defined as the fractions of spray that remains airborne between the tree rows of the orchard and to which the operator could get exposed. Cross and colleagues 1 2 3 found that the fraction of overspray is dependent on the spray quality but not on the height of the trees, the application rate or the air flow. Based on the available evidence the model predicts the total amount of overspray based on the spray quality and the application rate only. • The refreshment of air around the vehicle: As the vehicle moves forward it constantly drives into fresh air. This effect is modeled by g refreshing the air around the vehicle at the rate of the driving speed and a using a factor for the wind speed. The area of fresh air into which the vehicle is moving is described by the vehicle height and the tree row spacing. • The wake and dispersion toward the operator: The turbulence around the vehicle and spraying equipment (in particular relevant for spray from nozzles close to the vehicle) will have an 1 Cross, J. V., P. J. Walklate, R. A. Murray, and G. M. Richardson. 2003. Spray deposits and losses in different sized apple trees from an axial fan orchard sprayer: 3. effects of air volumetric flow rate. Crop Protection 22 (2): 381-94 2 Cross, J. V., P. J. Walklate, R. A. Murray, and G. M. Richardson. 2001. Spray deposits and losses in different sized apple trees from an axial fan orchard sprayer: 1. effects of spray liquid flow rate. Crop Protection 20 (1): 13-30. 3 Cross, J. V., P. J. Walklate, R. A. Murray, and G. M. Richardson. 2001. Spray deposits and losses in different sized apple trees from an axial fan orchard sprayer: 2. effects of spray quality. Crop Protection 20 (4): 333-43 71 influence on the dispersion of spray from the spray equipment to the operator. Analysis of the data showed that the wake and dispersion is well explained by using the vehicle size and vehicle speed to describe the spray dispersion towards the operator. Proposed model The model consists of three parts that represents the three consecutive phases of the sourcereceptor model: (i) The airborne spray (ii) Dispersion around the vehicle (Wake and air Refreshment) (iii) Dispersion from spray around the vehicle to operator using cabin efficiencies (If) Input variables of the model are represented in Tables 1 and 2. (i) The airborne spray The amount of spray emitted at the spray boom that potentially remains airborne around the vehicle (that is not deposited on the trees or past the tree row as drift) is defined as the overspray. Based on the spray quality, the fraction of spray that deposits on the ground between trees in relation to the total volume sprayed rate. This fraction is called the overspray fraction. Together with the total volume spray rate it provided the model input for the total amount of spray airborne around the vehicle. For determination of the total overspray: - The spray volume rate (SPR) is used as input - The fraction of spray that remains airborne between the tree rows (foverspray) is based on experimental data. The foverspray value is derived for 3 spray qualities (medium, fine, very fine) and is based on the fraction of spray that deposits on the ground in relation to the total airborne spray. (Appendix A and Table 3). (ii) Dispersion around the vehicle (Wake and air Refreshment) The dispersion around the vehicle is described by two processes: Firstly, the reduction of the air concentration of spray by air refreshment due to the driving of the vehicle into fresh air and by the wind. The total amount of overspray is divided by the area over which are refreshment takes place (perpendicular to the driving direction) multiplied by the vehicle 72 speed and 85% of the wind speed. The area is defined by the area made up by the vehicle height and the row spacing of the trees. The 85 percentage of wind speed is assumed to describe the wind direction that adds to the air refreshment (only the 15% more or less equal to the driving direction is excluded). The dispersion of spray from the spray equipment area to the rest of the vehicle is described by the wake factor (wf) (a factor for the turbulence) and the dilution of spray over a volume described by the vehicle distance category (d) (Table 5) and the vehicle speed (vvehicle). The vehicle speed is used to describe the imaginable increased distance from the operator to the spray equipment due to displacement of the operator away from the equipment during spraying. (iii) Dispersion from spray around the vehicle to operator using cabin efficiencies (If) Personal enclosures on vehicles (e.g. cabins) may affect the transfer of spray to the operator. The categories proposed for cabins are presented in Table 7. (b) Algorithms Ipde = Ipde Cvehicle * If * Ccoi Potential inhalation exposure concentration to the active substance (in droplet spray) (g/m3) Ccoi Concentration of a.s. in spray volume (g/l) Cvehicle Concentration of droplet spray at the vehicle (l/m3) If Cabin efficiency (factor) where Cvehicle = 𝐴𝑅 ∗ �𝑓𝑜𝑣𝑒𝑟𝑠𝑝𝑟𝑎𝑦 � 𝑤𝑓 ∗ 𝐻𝑉𝑒ℎ𝑖𝑐𝑙𝑒 ∗ 𝑑𝑟𝑜𝑤 ∗ (vvehicle + 0.85 ∗ 𝑣𝑤𝑖𝑛𝑑 ) (𝑑 + vvehicle )3 AR Application rate (L/s) fairborne spray overspray fraction: fraction of spray volume that remains airborne between the tree rows 73 Hvehicle Vehicle height (m) drow Row spacing (m) vvehicle vehicle speed (m/s) vwind wind speed (m/s) d vehicle distance category based on vehicle-mounted and self-propelled (VM) and trailer-mounted (TM) rigs wf (c) wake factor Testing and calibration Model Testing Please refer to the type of plots/graphs and explanation under item 3.2. for the testing of the Orchard Spraying inhalation model. Since no data points were complete for all the required inputs all data are imputed. Imputed data were based on expected values compared to all other datapoints. Figure 1 shows the model estimates plotted against all the available data. Figure 2 shows the model estimated against the data points with confidence intervals. Due to the imputed data used for testing, no conclusions can be drawn from the correlation found between the data and model outputs, except that the models predict exposures in the overall expected range. Figures 3 shows the same results but with increasing data values. 74 (d) Tables Table 1: Orchard spraying inhalation model - input variables Abbr Cc.o.i. If Input variables Unit Concentration of g/l a.s. in spray liquid Cabin efficiency n/a (factor) User input User input Model input User input From/for Mass balance input (iv)No cabin (v) Cabin without pressurized/filtered ventilation 1 Sample from U(0.1, 0.25) with probability 0.25, from U(0.25, 0.5) with probability 0.5 and from U(0.5, 0.9) with probability 0.25 Sample from U(0.0001, 0.05) with probability 0.25, from U(0.05, 0.15) with probability 0.5 and from U(0.15, 0.3) with probability 0.25 User input Sample from U(0.010.04) Sample from U(0.020.08) Sample from U(0.030.22) Sample from U)0.010.22) lognormal right skewed (mean at 0.05) Sample from respective categories User input User input User input Sampled from respective categories Literature review Table 7 Appendix D (vi)Cabin with pressurized/filtered ventilation AR foverspra l/s n/a y Application rate Overspray fraction Hvehicle Vehicle height m drow vvehilce vwind d Row spacing Vehicle speed Wind speed Vehicle distance category m m/s m/s m User input Medium spray quality Fine spray quality Very fine spray quality Unknown Categories: S, M, L, U, sSP, mSP, lSP, O Check in data 1-8 m/s 1-7 m/s Categories: vehiclemounted and selfpropelled (VM), trailermounted (TM), unknown (O) 75 Literature review Table …… Appendix ….. Browse data Literature review Table 6 Appendix B Table 2: Proposal cabin categories and associated multipliers/factors % Cabin Cabin Type A No cabin 1 n/a Fixed 1 B Cabin without pressurized/filtered 60 10 - 90 0.1 0.9 90 70 - >99 0.0001 0.3 efficiency Range (%) Range (factor) Category Lower Upper ventilation, OR with ventilation but without filtration C Cabin with pressurized/filtered ventilation* 1 based on literature review (Appendix D) * criteria will be used in the user interface to verify ventilation & filtration compliances (ASAE S-525) Table 3: Overspray fraction Spray quality1 Code Overspray factor Medium M Fine Range (factor) * Lower Upper 0.025 0.01 0.04 F 0.05 0.02 0.08 Extra fine XF 0.125 0.03 0.22 Unknown 0 0.05 0.01 0.22 1 based on literature review (Appendix …..) Table 4: Vehicle Height Type Code No Cabin Length, distance^ Range (factor) * (range, m) Lower Upper A 1.5 1 2 Cabin B,C 2.5 2 3 Unknown O 2 1 3 76 Table 5: Length / distance categories based on vehicle and spraying rig configurations1 Type Code Length, distance^ Range (factor) * (range, m) Lower Upper VM 3.5 1 6 Trailer-mounted TM 8 4 12 Unknown O 6.5 1 12 Vehicle-mounted and self-propelled 1 based on literature review (Appendix B) (e) Figures Figure 1: Inhalation exposure: Plot of the (fitted) log10(data) and log10(mean(model outputs)) 100 10 0.1 0.01 log 10 (Data(µg/l)) 1 0.001 0.0001 1e-05 1e-06 1e-06 1e-05 1 0.1 0.01 0.0001 0.001 log10 (Mean(Model outputs(µg/l))) 10 100 Figure 2: Inhalation exposure: Plot for the lo10(fitted model outputs) with the log10(data) for the complete datasets (with increasing model values). (Plot of 95% CIs (grey lines), means (red lines), medians (black lines), and imputed datasets (blue dots). 77 10 0.1 0.01 0.001 10 log (Model outputs (µg/l)) 1 0.0001 1e-05 1e-06 0 10 20 30 40 60 50 Data index 70 80 90 100 figure 3: Inhalation exposure: Plot for the lo10(fitted model outputs) with the log10(data) for the complete datasets (with increasing data values). (Plot of 95% CIs (grey lines), means (red lines), medians (black lines), and imputed datasets (blue dots). 10 0.1 0.01 0.001 10 log (Model outputs (µg/l)) 1 0.0001 1e-05 1e-06 0 10 20 30 40 50 60 78 70 80 90 100 3.2.6 Orchard spraying dermal models Description The Orchard spraying dermal models consist of a hand model (Dph) and whole body model (Dpw). The following activities of an operator during the orchard spraying scenario are included: • Operating the vehicle (vehicle cockpit with & without cabin) • Stepping into/out of vehicle (contact with the vehicle interior & exterior) • Incidental activities with the spraying rig, adjustment of spraying rig and maintenance of nozzles. The models consider important determinants of dermal exposure associated with orchard sprayer operations, e.g. type of spraying equipment (trailer, rear-mounted), the size of the spraying activity and number of application tasks. The model predicts the exposure based on 2 routes of exposure: deposition from airborne spray and surface contacts with contaminated surfaces. As high dermal exposures are associated with incidental activities with contaminated spraying rigs and nozzles (as opposed to the vehicle cockpit), the model includes both the vehicle cockpit and spraying rig as sources of exposure for the surface contact route. The cleaning of vehicles and spraying equipment are not included as a parameter in the models because the effect of this activity on exposure is not known. Hand model Deposition The deposition of spray on the hand is modelled by using vehicle contamination data. It is assumed that the contamination of the control area (cabin or vehicle controls) of the vehicle occurs through spray deposition only. The amount of deposition on this part of the vehicle is assumed to deposition on the upper parts of the hands of the operator as well. (the palms will be more exposed to surfaces rather than spray deposition). Deposition of spray on the hand therefore is only assumed to occur on the top of the hands and their respective surface area (BPapplicationtop) (Appendix H). The level of exposure due to deposition is calculated directly from the relevant vehicle contamination levels (Table 7 and Appendix E) and the exposed surface area. The last term of the hand exposure model algorithm represents the deposition. 79 Surface contacts 𝑆𝑐𝑐𝑜𝑐𝑘𝑝𝑖𝑡 ∗ 𝐶𝑐𝑜𝑖 ∗ 𝐼𝑓 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 � � 𝐵𝑃𝑎𝑝𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛𝑇𝑜𝑝 The algorithm used to estimate dermal exposure during orchard spraying is based on two key parameters associated with surface contacts as described by Gorman et al (2012), i.e. frequency of contacts and surface contamination levels. The input parameters and ranges of the models are presented in Table 6. (iv) Frequency of surface contacts The hand model applies event frequencies to provide an indication of surface contact frequencies. The following event frequencies are included in the hand model (see Table 6): • Surface contacts with the vehicle cockpit exterior is represented by events of stepping into/out of the vehicle, i.e. the number of mixing & loading periods (Fmlp) and number of times leaving the vehicle during trouble-shooting or nozzle maintenance (Fnr) • Surface contacts with the delivery system/rig is represented by (i) trouble-shooting and nozzle maintenance events (Fnr) and (ii) incidental contacts such as (de-)mounting and boom (un)folding prior to and after spraying (Frig) The mixing & loading periods (Fmlp) are estimated from the volume applied and the tank volume. The volume applied parameter is estimated from user inputs and are based on the total area sprayed (Ha), dose (kg/Ha) and final mixing & loading concentration (g/l). Since no information on the frequency of nozzle maintenance and incidental contacts was available, defaults from boom spraying were adopted. These are between 0 and a worst case of 15, with a 90% probability of between 0 and 5 and 10% probability of between 6 and 15. Other events that may occur prior to or after spraying activities include nozzle replacement and boom adjustment. A fixed event frequency with the rig and nozzles before and after spraying (Frig) is set at 1. Frig are not included as user input assuming that this is not a known input. (v) Surface contamination levels The surface contamination level (µg/cm2 of a.s.) on field sprayers are highly variable and limited evidence is available on the determinants influencing the deposition on vehicle and sprayer surfaces 80 of field sprayers. Ideally the surface contamination levels should be derived from data for different vehicle-sprayer configurations, different spray qualities, etc. However, this information is only available in a few specific settings (e.g. Balsari & Marucco, 2003). As result, it was decided to use a broad range of contamination levels as input for the model (Table 7 & Appendix E). For this purpose, contamination levels are based on four ‘Volume applied’ categories based on the percentage of the total spray volume that deposits on the vehicle and delivery system. As the surface contamination level estimate on the vehicle cockpit concerns the vehicle exterior, a cabin efficiency factor is also included. The effect sizes proposed for cabin efficiency in the orchard and boom spraying inhalation model are adopted for this model (Error! Reference source not found.). In addition, the transfer efficiency from surfaces to body parts (Cf) (Appendix H), body surfaces areas (BP) and the affected surface area of body parts (lcn/BP) (Appendix I) are included in the algorithm. The surface-to-hand transfer efficiency (Cf) is corrected to account for multiple contacts, for both contacts with the cockpit (Cftot_cockpit) and rig (Cftot_Rig) (Appendix J). Surface contact are only assumed to occur with the palms of the hands (BPpalms) (Appendix H). Whole body (excl. hands) model To estimate the whole body dermal exposure, it is assumed that whole body exposure occurs only due to deposition during application. The mechanism for whole body deposition is similar as for the hand model except it is applied to the whole body surface area. Whole body exposure during incidental contacts might be relevant but where excluded from the models due to a lack of data. Moreover the relatively high air-concentration during application (compared to for example boom spraying) make other sources of exposure less significant. Whole body exposure during stepping in and out of the vehicle is also not modelled partly because exposure to the whole body seems less likely during this activity. Algorithms I Hand exposure model (Dph) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 𝐷𝑝ℎ = ��𝑆𝑐𝑐𝑜𝑐𝑘𝑝𝑖𝑡 ∗ 𝐶𝑐𝑜𝑖 ∗ 𝐶𝑓𝑡𝑜𝑡_𝑐𝑜𝑐𝑘𝑝𝑖𝑡 � + �𝑆𝑐𝑓𝑎𝑛 ∗ 𝐶𝑐𝑜𝑖 ∗ 𝐶𝑓𝑡𝑜𝑡_𝑟𝑖𝑔 �� ∗ � � 𝐵𝑃𝑝𝑎𝑙𝑚𝑠 𝑆𝑐𝑐𝑜𝑐𝑘𝑝𝑖𝑡 ∗ 𝐶𝑐𝑜𝑖 ∗ 𝐼𝑓 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 +� � 𝐵𝑃𝑎𝑝𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛𝑇𝑜𝑝 81 𝐶𝑓𝑡𝑜𝑡_𝑐𝑜𝑐𝑘𝑝𝑖𝑡 = 𝐶𝑓𝑡𝑜𝑡_𝑟𝑖𝑔 = 𝐹𝑚𝑙𝑝+ 𝐹𝑛𝑟 � �𝐶𝑓 0.5(𝑘−1) � 𝑘=1 𝐹𝑛𝑟+ 𝐹𝑟𝑖𝑔 � �𝐶𝑓 0.5(𝑘−1) � 𝑘=1 𝐹=6 𝐶𝑓𝑚𝑎𝑥 = ��𝐶𝑓 0.5(𝑘−1) � 𝑘=1 Dph potential dermal exposure to a.s. in µg on the hands Sccockpit surface contamination levels (l/m2) of cockpit Ccoi concentration of a.s. in spraying volume (ug/l) Cftot_cockppit corrected transfer coefficient for multiple contacts (maximum: 0.89 with Cf = 0.45; Fmlp + Fnr = 6) Scfan surface contamination levels (l/m2) of the fan Cftot_rig corrected transfer coefficient for multiple contacts (maximum: 0.89 with Cf = 0.45; Fnr + Frig = 6) BPhands total surface area of hands (m2) lcn/BPpalms affected surface area of hands (m2 of BP) (palms) If cabin factor Cfmax corrected transfer coefficient for a maximum amount of contacts (0.89 with Cf = 0.45: F= 6) lcn/BPApplicationTop affected surface area of hands (m2 of BP) (top of hands) Cf transfer coefficient from surface to the hands FmLp number of mixing & loading periods Fnr number of nozzle maintenance events Frig frequency of incidental contacts with the rig/boom Rationale (FmLp + Fnr) represent the frequency of surface contacts with the vehicle cockpit (Sccockpit) when stepping into/out of the vehicle during trouble-shooting and mixing & loading periods 82 (Fnr + Frig) represent the frequency of surface contacts with the delivery system/rig (Scrig) during trouble-shooting and incidental contacts ((de-)mounting, boom folding) prior to and after spraying II Whole body (excl. hands) exposure model (Dpw) The following equation is used to estimate the potential dermal exposure (Dpw) of the whole body in ug: 𝐷𝑝𝑤 = 𝑆𝑐𝑐𝑜𝑐𝑘𝑝𝑖𝑡 ∗ 𝐶𝑐𝑜𝑖 ∗ 𝐼𝑓 ∗ 𝐵𝑃𝑤ℎ𝑜𝑙𝑒 𝑏𝑜𝑑𝑦 Dpw potential dermal exposure to a.s. in µg on the whole body excluding hands Sccockpit surface contamination levels (l/m2) of cockpit Ccoi concentration of a.s. in spraying volume (ug/l) BPwhole body total surface area of the whole body (m2) Model testing Please refer to the type of plots/graphs and explanation under item 3.2 I Hand exposure model (Dph) The available data is not suitable for testing the Orchard spraying hand model, because of data gaps of important parameters. For example, the number of contact events with the nozzles is not known in the data. Assuming a frequency of nozzle maintenance (Fnr) of 90% probability between 0 and 5 and 10% probability between 6 and 15, the model predicts a conservative exposure compared to the broad range of data (figure 4 and figure 5). A default nozzle maintenance frequency is set at 3 in the software. Unfortunately, inputting a distribution for nozzle events will not be feasible as the probabilistic output of the model (using iterations) will not be able to produce a single estimate with repeated runs of the model. II Whole body (excl. hands) exposure model (Dpw) The whole body exposure model was compared with the dataset using the deposition of spray on the operator body (either inside or outside the cabin) only (figure 6 and figure 7). 83 (f) Tables Table 6: Orchard spraying dermal models - main user and model inputs* abbr User inputs Model inputs Lower Upper Type Based on Cc.o.i Concentration of a.s. Concentration of 0.05 50 selected value BROWSE data in spray volume (g/l) a.s. in spray 50 25000 selected value BROWSE data 0.001 0.25 uniform Field trial data distributions Table 11 volume (g/l) Volume applied Volume applied (l), (estimated from total 4 categories area sprayed, dose, final mixing&loading concentration) Sccockpit n/a Cockpit surface contamination levels (l/m2) Scrig n/a Appendix E Rig surface 0.036 0.25 contamination Uniform Field trial data Distributions Table 11 levels (l/m2) Fmlp Appendix E Number of M&L Number of M&L periods (estimated periods (tankfulls with volume applied applied) 1 20 selected value BROWSE data 0 5 Sample from Field studies and tank size) Fnr Frequency of nozzle Frequency of maintenance (Fnr); nozzle discrete uniform user selected: 0 - 5; maintenance; distribution >5; (With 90% between 0 - 5 default in software = 3 probability between 0-5) (for testing with data only) Frig n/a Frequency of Fixed contacts with rig at 1 84 - Fixed Expert opinion Field studies abbr User inputs Model inputs Lower Upper Type Based on 0.0001 1 see inhalation Literature model review prior to/after spraying Cabin factor (3 Cabin factor (3 categories) categories with ranges) Table 8 Appendix D Vehicle or trailer Vehicle-sprayer mounted (3 categories length (d) – VM, TM, O- Vehicle-mounted unknown)) (0.1 – 0.9) 0.01 1 uniform Field trial data distributions Table 6 Appendix B Trailer-mounted (0.01 - 0.25) Front-fitted booms fixed at 1 Cf n/a Transfer efficiency 0.23 0.68 triangular IOM database from surface to the distribution Literature hands (Cf); e.g. 1st between 0.23 review contact = 0.45; 2nd- and 0.68 with a Appendix H & 6th contact (*0.5); mode at 0.45 J lognormal Appendix I see text BPhands Gender (male, female) Surface area of hands (m2) - - distributions (Males: log(SA) ~ Normal (2.2319, 0.109922) Females: log(SA) ~ Normal(2.4313, 0.10362)) 85 abbr User inputs Model inputs Lower Upper Type Based on BPwholeb Gender (male, female) Surface area of the - - Lognormal Appendix I whole body (m2) ody distributions (Males: LCn/ n/a Fraction of Fixed BPapplicati affected surface 0.5 on top area of hands (top) lcn/ n/a BPpalms - Fixed Literature review Appendix I Fraction affected Fixed surface area of 0.5 - Fixed Literature review hands (palms) Appendix I * Models distinguish between Volumes applied and contamination levels of different surfaces Table 7: Surface contamination levels using Volume Applied categories Category Volume applied Contamination level cockpit, Contamination Sccockpit (l/m2)* * ** level delivery system/rig, Scrig (l/m2)* Lower Upper Lower Upper Lower Upper 1 <50 500 0.001 0.025 0.036 0.25** 2 500 1000 0.003 0.051 0.18 0.25** 3 1000 3000 0.006 0.152 0.25** 0.25** 4 3000 >10000 0.017 0.25** 0.25** 0.25** broad categories extrapolated from orchard sprayer contamination studies using the % of total spray volume depositing on the vehicle and delivery system (e.g. Balsari & Marucco, 2003, Ramwell et al, 2005, Michielsen et al, 2012). A maximum surface loading of 0.25 liters per square meter is assumed 86 (g) Figures Figure 4: Hand exposure: Plot of log10(measured potential dermal exposure (hands) and log10(geometric mean(model outputs)). Assuming a 90% probability of 0-5 nozzle maintenance events. All data sets included imputed values. 1e+06 100000 1000 100 log 10 (Data(µg/hands)) 10000 10 1 0.1 0.1 1 10000 1000 100 10 log10 (Mean(Model outputs(µg/hands))) 87 100000 1e+06 Figure 5: Hand exposure: Plot of 95% CIs (grey lines), geometric means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10(measured potential dermal exposure (hands). Assuming a 90% probability of 0-5 nozzle maintenance events. All data sets included imputed values. 1e+06 10 log (Model outputs (µg/hands)) 100000 10000 1000 100 10 1 0.1 0 20 40 60 80 Data index 100 88 120 140 Figure 6: Whole body exposure: Plot of log10(measured potential dermal exposure (body) and log10(geometric mean(model outputs). All data sets included imputed values. 1e+07 100000 10000 log 10 (Data(µg/body)) 1e+06 1000 100 10 10 100 1000 10000 100000 log10 (Mean (Model outputs(µg/body))) 89 1e+06 1e+07 Figure 7: Whole body exposure: Plot of 95% CIs (grey lines), geometric means (red lines) and medians (black dashed lines) for the log10(model outputs) with the log10(measured potential dermal exposure (body). All datasets included imputed values. 10 log (Model outputs (µg/body)) 1e+06 100000 10000 1000 100 0 20 40 60 80 100 120 Data index 140 160 180 200 3.2.7 Handheld Application inhalation model (a) Description The hand-held application (HHA) scenario represents an operator applying PPP by using a spray lance or gun either connected to a tank by a hose, or connected to a backpack The HHA model accounts for indoor (greenhouses) as well as outdoor (soft fruit trees, vineyards, vegetables crops, etc.) environments. Operator inhalation exposure is assumed to primarily originate from the spraying equipment that releases airborne droplets HHA activities. The inhalation model focuses on droplet spray and predicts the air concentration of droplets in the breathing zone of the operator much like the Boom and Orchard spraying models. This air concentration is converted to the air concentration of the given active substance (a.s.) based on the concentration of a.s. in the spraying volume. For the use of the ART model, it was assumed that far field exposure is not relevant for mixing & loading PPPs, 90 since in practice one operator is performing all the activities, and thus no secondary exposure is assumed. Also the distance of the operator to the source is assumed to be <1 meter. Furthermore, it is assumed that only low-volatile substances are used liquid formulations (in ART framework cut off point: vapour pressure <10 Pa), and thus only exposure to mists is taken into account. In case of liquid formulations it is assumed that a litter of liquid formulation weighs one kg (comparable to water), to be able to use the concentration of active substance in the formulation (in g/L) as the weight fraction (g/kg). With regard to dispersion, the following assumptions are made: • In case of outdoor application, it is assumed that this takes places close to buildings • In case of application under a shelter/covering/roof, it is assumed that these conditions are comparable to ‘outdoors – close to buildings’ • In case of application inside greenhouses only good natural ventilation is assumed Main ART model inputs (default values) are presented in Table 1. (b) Model testing The model was tested with a dataset containing empirical data from which the ART score was derived for each observation (e.g. weight fraction) along with the measured output (e.g. inhalation exposure). The dataset used to test the model contained more than 600 data points. Where this was the case, the missing inputs were imputed by empirical sampling from existing values in the available data. To summarize the output, the geometric mean, median, 2.5th percentile and 97.5th percentile were calculated. Thereafter, the geometric mean of the output against the true measured values with 95% confidence intervals, medians and geometric means of the output against the true values (all on log10 scale) were plotted for comparison. For each sub-scenario, 10,000 iterations were used. The models were only compared or tested with available exposure data. Thus no calibration or transforming/fitting was performed. Results from the model testing are presented in figures 1 to 4. For all the sub-scenario, The y-axis represents the measured potential inhalation exposure. The x-axis represents the geometric mean of 91 the model outputs. Blue dots indicate imputed values, red dots indicate that art scores were available. Figures 1 to 4 show the result of the model testing. Art default values for HHA air concentration (Table 1) were tested against measured data representative of HHA application activities. (c) Tables Table 1: Default values used for HHA Inhalation (ART) model Modifying factor Relevant parameter(s) Description Multiplier Substance emission potential (E) Viscosity Low viscosity (like water) 1.0 E = 10/30000 * 1.0 * Weight fraction % active substance in spray liquid Exactly Use rate High application rate (> 3 L/min) Moderate application rate (0.3-3 L/min) Low application rate (0.030.3 L/min) Very low application rate (<0.03 L/min) Any direction (including upward) 3 3 Default 1 Default No localized controls Spraying with no or low compressed air use No localized controls 1.0 Default Indoor: Room volume and air changes per hour Natural ventilation, large room volume (greenhouse) 0.8991 Outdoor: Distance of the source from buildings Far from buildings Close to buildings Housekeeping practices (Default level: no specific cleaning practices, no protective clothing that repel spills, process not fully enclosed) 0.2 0.75 0.01 weight fraction Activity emission potential (H) Spray application of liquids – surface spraying of liquids H= 1 * 3 * 1 Direction of application Spray technique Localized controls (LC) LC=1.0 Dispersion (D) Indoor: D=0.90 Outdoor: D=0.2 Surface contamination (Su) Su=0.01 92 1 Coupling with BROWSE user inputs Relate to formulation type (EC, SC) division into categories Default Is considered to be known (input), or can be derived from other known information like the formulation, the amount used, etc. Default 0.3 0.1 Default Default (d) Figures Figure 1. Inhalation exposure. Plot of the log10 (geometric mean of the concentration of a.i. in air*) against the log10 (measured inhalation exposure(both in ug/m3)). Blue dots indicate imputed values, red dots indicate that art scores were available. (Backpack Downwards scenario) *including a dataset which have art model inputs that were imputed 100000 10000 1000 log 10 (Data(µg/m3)) 100 10 1 0.1 0.01 0.001 0.0001 1e-05 1e-05 0.0001 0.001 1 0.1 0.01 100 10 1000 10000 100000 log10(GM of concentration in air (µg/m3)) Figure 2. Inhalation exposure. Plot of the log10 (geometric mean of the concentration of a.i. in air*) against the log10 (measured inhalation exposure(both in ug/m3)). Blue dots indicate imputed values, red dots indicate that art scores were available. (Backpack Up and Downwards scenario) *including a dataset which have art model inputs that were imputed 100000 10000 1000 log 10 (Data(µg/m3)) 100 10 1 0.1 0.01 0.001 0.0001 1e-05 1e-05 0.0001 0.001 0.01 0.1 1 10 100 1000 10000 100000 log10(GM of concentration in air (µg/m3)) 93 Figure 3. Inhalation exposure. Plot of the log10 (geometric mean of the concentration of a.i. in air*) against the log10 (measured inhalation exposure(both in ug/m3)). Blue dots indicate imputed values, red dots indicate that art scores were available. (Gun and Hose Downwards scenario) *including a dataset which have art model inputs that were imputed 100000 10000 1000 log 10 (Data(µg/m3)) 100 10 1 0.1 0.01 0.001 0.0001 1e-05 1e-05 0.0001 0.001 0.01 0.1 1 10 100 1000 10000 100000 log10(GM of concentration in air (µg/m3)) Figure 4.Inhalation exposure. Plot of the log10 (geometric mean of the concentration of a.i. in air*) against the log10 (measured inhalation exposure(both in ug/m3)). Blue dots indicate imputed values, red dots indicate that art scores were available. (Gun and Hose Up and Downwards scenario) *including a dataset which have art model inputs that were imputed 100000 10000 1000 log 10 (Data(µg/m3)) 100 10 1 0.1 0.01 0.001 0.0001 1e-05 1e-05 0.0001 0.001 0.01 0.1 1 10 100 1000 10000 100000 log10(GM of concentration in air (µg/m3)) 94 3.2.8 Handheld Application Dermal model (a) Description Previous studies suggest that the fraction of the applied volume available for deposition onto the worker, as well as the body surface area affected will be influenced by the spraying direction (Brouwer et al. (2001); Berger-Preiss et al. (2005) Hughes et al. 2008) Leona et al. 1992) Marquart, (2003)). Likewise, the equipment used will determine the degree of exposure as well as the distribution of contamination over the operator body. (Dosemeci (2002); Leonas et al. (1992); Machera (2003); Matuo & Matuo (1998); Nuyttens (2009); Sanjrani (1990); Wicke (1999)) Therefore, the HHA model describes four sub-scenarios based on the application direction and the equipment used, for which separate models for hand PDE and whole body (excluding hands) PDE are developed. The following sub-scenarios are distinguished: 1. Backpack – Downward spraying: The application method consists of a handheld spray lance or gun connected to a backpack or knapsack. The spraying direction is exclusively downwards because the crop is shorter than the average height of the hands above the floor (0.7m), or because the type of treatment does not require upwards application (e.g. herbicides). 2. Backpack – Up- and Downward spraying: This scenario includes crops higher than 0.7m, for which the application direction can be downwards, horizontal and/or upwards. The application method consists in a handheld spray lance or gun connected to a backpack or knapsack. 3. Gun and hose – Downward spraying: The application method consists in a handheld spraying lance or gun connected to a tank by a hose. The tank can be either fixed or tractor mounted. The spraying direction is exclusively downwards since the crop height or the type of treatment does not require upwards application. 4. Gun and hose – Up- and Downward spraying. This scenario involves handheld application on crops higher than 0.7m by using a spraying gun or lance connected to a fixed or mobile tank 95 by a hose. This scenario includes crops higher than 0.7m so that the application direction can be downwards, horizontal and/or upwards. Operator dermal exposure originates from the spraying equipment that releases airborne droplets during HHA activities. A certain fraction of the spray will deposit over the operator hands and body. Upon incidental contact with the crop a percentage of the volume applied will be transferred from the crop to the operator. Likewise, the fraction deposited on the equipment can be transferred to the operator hands or body by direct contact. Splashes produced when loading the backpack represent a source of contamination by transfer from the backpack to the operator back or hands. The bulk exposure pathway for handheld application activities includes three kinds of events, namely leakages from the equipment, splashes from the backpack and drifting from the treated crop. Even when the bulk pathway may have an important contribution in case an incidental event occurs, the mentioned unintended events can be avoided by following the best agricultural practices (FAO, 2013) (FAO, (2001a); (2001b)). It is assumed that operators take into account the Best Agricultural Practices, and follow the PPP label, therefore there is not need to estimate exposure due to misuse (BROWSE, 2011). Regarding bulk exposure from the treated crop, the model assumes that the optimal application rate is assessed and low drift equipment is used. In case that drifting occurs, the fate of the contaminant should be the area below the crop. Since the operator applies the product from a distance large enough in which he has freedom of mobility, this event can be neglected. It is known that even when the best agricultural practices are followed, some types of contacts cannot be avoided. Crops are usually designed to maximize productivity by increasing the yield per unit area. In that case, the operator may not be able to freely move between two adjacent crop canopies leading to frequent direct contact with the crop. This pathway of exposure represent a common event when applying PPP by using HHA techniques, therefore contacts with the treated crop has been modelled as part of the “Direct Contact” pathway Likewise, splashes are likely to occur during mixing and loading periods. Although the mixing and loading activity has been modelled in a different scenario, the operator can be exposed to the 96 splashes produced during this activity by direct contact with the equipment, and therefore this source of contamination has been included within the HHA model. Tables 1 and 2 offer an overview of the pathways included in each scenario for both models, Hands exposure and Whole Body (excluding hands) exposure Similar equations apply for the different subscenarios based on three key parameters associated with surface contacts as described by Gorman et al (2012) (frequency of contacts and surface contamination levels), Brouwer et al. (2001) and Cohen Hubal et al (2005) (transfer efficiencies after continuous contacts). The model algorithms include the transfer efficiency from the donor surface to the hands and body (Appendix H), as well as the affected surface area of the hands (Appendix I and Table 5) or body (Table 5) involved in contacts. The surface-to-hand transfer efficiency (Cf) is corrected to account for multiple contacts, expressed as Cfmax (maximum number of contact with the spray gun), Cfbackpack (transfer efficiency after multiple contacts with the backpack) or Cfcrop (transfer efficiency after multiple contacts with the treated crop) (Appendix J). Model inputs are different for each sub-scenario (Tables 2). HAND MODEL: The model consists of two pathways of exposure: (i) Deposition of airborne droplets (ii) 1. Contact with the crop 2. Contact with the equipment The input parameters and ranges of the models are presented in Table3. (i) Deposition of airborne droplets (Ddhapp) The air concentration (Cair) (derived from the ART model) together with the droplet setting velocity (vdep) and the application time (t) will determine the degree of exposure via deposition (𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 1). The Stoke law’s defines the droplet setting velocity as a factor directly proportional to the diameter of the droplet (m) multiplied by the gravitational acceleration and by the differences in 97 density between the droplet and the air (kg∙m-3). This velocity is inversely proportional to the dynamic viscosity of the air (Kg m-1∙s-1) (see equation 7). Droplets <100 µm are regarded as prone to drift to non-target surfaces (Arvidsson et al., (2011); Miller & Ellis, (2000); Salyani & Cromwell, (1992)). On the other hand, droplets <50 µm take a long time to deposit (Sundaram & Sundaram 1991). Moreover, they can evaporate in a time shorter than 1.8 seconds under certain climatic conditions (Hofman & Solseng, 2001) and therefore they were not considered relevant for the assessment of potential dermal exposure. Inputs parameter used to estimate the droplet setting velocity are presented in Table 4. 2 φ 2 δ g −δa ·g v= · · 9 4 η The proportion of droplets sizes relies among other factors on the nozzle type, application pressure and technique used (Nuyttens et al., 2007). Since the user is unlikely to be acquainted with this information, a uniform distribution of droplets sizes was considered. As mentioned before, the PPP spraying solutions density is assumed to be 1000 Kg/m3 (comparable to water). When considering standard conditions, the droplet setting velocity ranges from 0.078 to 0.303 m•s-1 (see Table 4). The droplet setting velocity assessed by using the mentioned equation coincides with the sedimentation velocity estimated by Nuyttens et al. 2009 with a similar approach. ii.1) Contact with the treated crop ( 𝑫𝒄𝒉𝒄𝒓𝒐𝒑 ): ii.1.1 The surface contamination level of the crop, D It is known that the deposition efficiency of PPP application (E) tends to be less than 100%. Spray deposition efficiency is dependent upon multiple factors like leaf morphology, droplet size (Smith (2000); Abdelbagi & Adams, (1987); Uk, (1977), crop height, (Abdelbagi & Adams, 1987), the equipment used (Juste et al. 1990; Sánchez-Hermosilla et al., 2011), volume applied (Uk, (1977); Sánchez-Hermosilla et al.,( 2011)) and the use of commercial adjuvants (van Zyl et al.,2010) etc. Additionally the PPP is not homogeneously distributed over the crop (Sánchez-Hermosilla et al., 2011). Moreover, the amount of spray reaching the inside zones is remarkably lower than that on the outside zones (Juste et al., 1990). It was decided to take a broad range of spray deposition 98 efficiency in which the worst case scenario (spray deposition efficiency of 100%) was included (L. R. Ahuja, (2000); Zabkiewicz, (2007)). The total leaf area determines the volume in which the product applied will distribute. The Leaf Area Index (LAI) is defined as the leaf surface area per unit of soil surface area, reflecting a close relationship between the deposition and distribution of the applications (Rincón, Páez, & Sánchez-Hermosilla, 2009). To determine this index, a model that estimates this index from the crop height was used (equation 8). The correlation between the canopy height (H) and the LAI provides a correlation coefficient of 0.91 when used to estimate the LAI of crops with a regular distribution of nodes and leaf development. LAI=0.99·(H)1.49 (Rincón et al., 2009) (Equation 8) The average PPP per leaf area can be assessed by relating the amount of active substance of PPP that reaches the target surface with the LAI by integrating the model by Rincón et al (2009) with the formula developed by Gil et al. (2007) to estimate the spray deposition efficiency ( equation 5.1). Ccoi is the concentration of active substance of PPP in the air (µg/m3). Ar is the Application rate (L/m2). 𝐷= 𝐸 ∗ 𝐶𝑐𝑜𝑖 ∗ 𝐴𝑟 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.1) 0.99 ∗ 𝐻1.49 ∗ 103 Once the crop surface contamination is known, it is also needed to consider the frequency of contacts, the transfer efficiency from the treated crop to the operator hands, as well as the affected surface area of the operator hands in order to estimate hands dermal exposure (see equation 5). ii.1.2 Surface contacts with the treated crop, (𝑪𝒇𝒄𝒓𝒐𝒑 ) This factor has been linked to the row space (Fr), where row space is defined as the space between the outer leaves of two opposite crop canopies (see Table 3). According to Brouwer (1999) and 99 Cohen Hubal, (2005, 2008), after the 6th contact with the donor surface the dermal loading does not increase with further contacts, therefore the upper limit of contacts per application episode is set at the 6th contact event (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.2). * Considering that the hands of the operator are unlikely to be in contact with crops shorter than 0.7m, dermal hand exposure by direct contact with the crop applies only to scenarios 2 and 4. ii.2 Contact with the equipment ii.2.1Surface contact with the spray gun 𝑫𝒄𝒉𝒈𝒖𝒏 As it was stated before, the equation to estimate dermal exposure by surface contact with the gun relies on the frequency of contacts, the transfer after continuous contacts from the gun to the operator hands, and the gun surface contamination (See equation 2) ii.2.1.1 Frequency of contact with the spray gun or lance (and with the hose that connects the spray gun with the tank in scenarios 1 and 2) (𝑪𝒇𝒎𝒂𝒙 ): The operator holds the spray lance or gun during the whole spray process. The model considers a new contact event each time that the operator changes the position in which he/she holds the spray lance or gun, namely after a mixing and loading period, and when changing the spraying row since the operator tends to pull the hose toward him/herself. Considering that a typical crop area contains at least 6 rows (expert judgement), the upper limit for number of contacts per application episode was taken in order to simplify the model (see equation 2.2) ii.2.1.2 Spray gun surface contamination level (𝑺𝒄𝒈𝒖𝒏 ): The HHA model considers that the main source of contamination is the deposition of airborne droplets from the air (See equation 2.1). Since the operator is assumed to follow the good agricultural practices, additional sources of contamination (leakages) as consequence of odd behavior are neglected. The deposition rate relies on the air concentration (𝐶𝑎𝑖𝑟 ) multiplied by the application time (t) and by the droplet setting velocity (𝑣𝑑𝑒𝑝 ). 100 ii.2.2 Surface contact with the backpack (𝐃𝐜𝐡𝐛𝐚𝐜𝐤𝐩𝐚𝐜𝐤 ) Hand exposure by direct contact with the backpack considers the frequency of contacts as well as the variation in the transfer efficiency from the backpack to the operator hands (See equation 3) ii.2.2.1 Frequency of contact with the backpack (𝑪𝒇𝒃𝒂𝒄𝒌𝒑𝒂𝒄𝒌 ): Represented by events of putting on/taking off the backpack before and after a mixing/loading event (𝐹𝑚&𝑙 ). The mixing & loading periods are either an user input, or estimated from the volume applied and the tank volume in case this value is unknown by the user. The volume applied parameter is estimated from user inputs and are based on the total area sprayed (Ha), dose (kg/Ha) and final mixing & loading concentration (g/l) (See equation 3.2). The surface-to-hand transfer efficiency (Cf) is corrected to account for multiple contacts, expressed as Cfbackpack. An upper limit of 6th contact events for application episode was taken. ii.2.2.2 Backpack surface contamination level (𝑺𝒄𝒃𝒂𝒄𝒌𝒑𝒂𝒄𝒌_𝑻𝒐𝒕 ): Represented by deposition of airborne droplets onto the backpack, as well as splashes produced when loading the backpack during a mixing and loading period. The contribution of the deposition pathway to the total tank surface contamination relies on the air concentration (𝐶𝑎𝑖𝑟 ) multiplied by the application time (t) and by the droplet setting velocity (𝑣𝑑𝑒𝑝 ) (See equation 3.1). Due to the lack of information on spraying tank surface contamination, an estimation of this was derived from experimental studies on surface contamination levels of induction hoppers (Glass et al, 2009; Gilbert et al, 2000; Mathers et al, 1999). The induction hopper is a receptacle similar in size to the backpack and the method of filling is similar for both containers. Even when the opening size of the induction bowl can be up to twice as much bigger as the opening of the spraying backpack is, field observations often showed that operators tend to be careful when loading the backpack, therefore both situations are comparable. The tank contamination due to splashes was estimated from the container size (L) and the concentration of active substances in the spray volume (µg/L). Container size varies from 0 to 7 liters, each with a probability. The probabilities are based on the probability frequencies in the experimental data (see Table 7). The derived values are therefore indicative and only provide information about the relative performance of the different container sizes when used with a particular filling technique (Appendix G). 101 Data was transformed from contamination levels per operation/container to contamination levels per square centimeter (L/cm2) for a given container size. For this purpose, surface contamination levels were estimated using surface areas of containers. First, values were derived for the container exterior and container upper (lid, cap) of different container sizes. The derived values are therefore indicative and only provide information about the relative performance of the different container sizes when used with a particular filling technique. * PDE by direct contact with the backpack applies only to scenarios 1 and 2 In addition, the transfer efficiency from surfaces to body parts (Cf) (Appendix H), body surfaces areas (BP) and the affected surface area of body parts (lcn/BP) (Appendix I) are included in the algorithms. The surface-to-hand transfer efficiency (Cf) is corrected to account for multiple contacts with an upper limit of six contacts. Whole body (excluding hands) model The whole body (excluding hands) model consists of two pathways of exposure: (i) Deposition of airborne droplets (ii) 1. Contact with the crop 2. Contact with the equipment Similarly as for the dermal hand model, the whole body (excl. hands) model considers that transfer from the bulk to the operator is not relevant (and therefore not included in the model) since operators are assumed to follow the best management practice in order to avoid the occurrence of incidental events such a splashes, leakages and drifting (Appendix N). To model the whole body (excl. hands) exposure, similar equations are used, nevertheless the pathways included in whole body exposure (see Table 2) as well as the model inputs vary from one to another model. Similarly, the body surface area is extended in this model. Table 4 summarizes the proportion of the whole body affected according to this approach. This model takes into account a 102 maximum affected body surface area of 0.94 considering that the hands and head are excluded from this model. The model assumes that the whole body (excl. hands) is exposed to deposition of droplets (equation 4).Whole body exposure by direct contact with the treated crop is linked to the crop height. (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.4). During a typical application performance by using a spraying knapsack or backpack, just the back and the hands of the operator are prompt to be exposed by contact with the equipment, therefore the equation to model this pathway considers that just the back body surface is relevant (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 6)* * PDE by direct contact with the backpack applies only to scenario 1and 2 (a) Model testing Please refer to section 3.2.7 for detailed information about the database used to test the model as well as the statistical analysis performed. Result of the model testing are presented in figures 1to 10. Figures 1 and 2 show the result from the sensitivity analysis performed to probe the suitability of the four sub-scenarios proposed for the hand and whole body dermal exposure models. Results of the hand and whole body models testing are presented in figures 3 to 10. The graphical representations for all the sub-scenarios depict the measured potential dermal exposure (y-axis) against the geometric mean of the model outputs (x-axis). The individual measured potential dermal exposure (dots) and estimated model outputs (grey lines) plotted with the exposure level ( y-axis). (b) Algorithms Scenario 1: Backpack – Downward spraying I. Hand exposure model: Dph = Ddhapp + Dchgun + Dchbackpack Dph Potential dermal hand exposure a.s (µg) Ddhapp Potential dermal hand exposure to droplets via deposition (µg) 103 Dch gun Potential dermal hand exposure via surface contacts with the gun (µg) Dch backpack Potential dermal hand exposure via surface contacts with the bacpack (µg) Deposition 𝐷𝑑ℎ𝑎𝑝𝑝 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 𝐼) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑡𝑜𝑝 Cair Concentration of a.i. in air (µg/m3) t Total application time (s) vdep Droplet settling velocity average deposition speed of droplets in the air (m/s) BPhands Total surface area of the hands (m2) lcn/ BPhands_top Affected surface area of hands (m2 of BP) Surface contacts with the gun 𝐷𝑐ℎ𝑔𝑢𝑛 = 𝑆𝑐𝑔𝑢𝑛 ∗ 𝐶𝑓𝑚𝑎𝑥 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑝𝑎𝑙𝑚𝑠 𝑆𝑐𝑔𝑢𝑛 = (𝑆𝑐𝑇𝑎𝑛𝑘 ∗ 𝐶𝑐𝑜𝑖 ) + (𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2.1) 𝐹=6 𝐶𝑓𝑚𝑎𝑥 = � 𝐶𝑓𝑔𝑟𝑎𝑠𝑝 ∗ 0.5(𝐾−1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2.2) 𝐾=1 Scgun = Surface contamination of the gun Cfmax Corrected transfer efficiency for maximum amount of contacts (maximum: 0.89 with Cf = 0.45; Fgrasp= 6) (Appendix H) 104 Cfgrasp Transfer efficiency for grasping of smooth surfaces (triangular distribution with lower value 0.23, mode 0.45 and upper 0.68) lcn/ BPhands_palms Affected surface area of hands (m2 of BP) Surface contact with the backpack 𝐷𝑐ℎ𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 = 𝑆𝑐𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘_𝑇𝑜𝑡 ∗ 𝐶𝑓𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 3) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑝𝑎𝑙𝑚𝑠 𝑆𝑐𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘_𝑇𝑜𝑡 = (𝑆𝑐𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 ∗ 𝐶𝑐𝑜𝑖 ) + (𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 3.1) 𝐹𝑚&𝑙 𝐶𝑓𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 = � 𝐶𝑓𝑔𝑟𝑎𝑠𝑝 ∗ 0.5(𝐹𝑚&𝑙 −1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 3.2) 𝐾=1 Scbackpack Surface contamination level of the backpack (l/cm2) Cfbackpack Corrected transfer efficiency for multiple contacts (maximum: 0.89 with Cf = 0.45; Fgrasp = 6) (Appendix H) Fm&l Frequency of M&L periods Ccoi Concentration of a.s in spraying volume (µg/L) II. Whole body exposure model: Dpw = Ddwapp + Dcwcrop + Dcwequipment Dpw Potential dermal whole body exposure a.s (µg) Ddwapp Potential dermal whole body exposure to droplets via deposition (µg) Dcwcrop Potential dermal whole body exposure via surface contacts with the crop (µg) Dcwbackpack Potential dermal whole body exposure via surface contacts with the backpack (µg) 105 Deposition 𝐷𝑑𝑤𝑎𝑝𝑝 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ∗ 𝐵𝑃𝑤ℎ𝑜𝑙𝑒 𝑏𝑜𝑑𝑦 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 4) Cair Concentration of a.i. in air (µg/m3) t Total application time (s) vdep Droplet settling velocity average deposition speed of droplets in the air (m/s) BPwhole body Total surface area of the body (m2) Lcn/BPbody-ch Proportion of body part affected according to crop height (m2 of BP) Surface contacts with the crop 𝐷𝑐𝑤𝑐𝑟𝑜𝑝 = 𝐷 ∗ 𝐶𝑓𝑐𝑟𝑜𝑝 ∗ 𝐵𝑃𝑤ℎ𝑜𝑙𝑒 𝑏𝑜𝑑𝑦 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5) 𝐵𝑃𝑏𝑜𝑑𝑦−𝑐ℎ 𝐷= 𝐸∗𝐶𝑐𝑜𝑖 ∗𝐴𝑟 0.99∗𝐻 1.49 ∗103 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.1) 𝐹𝑟 𝐶𝑓𝑐𝑟𝑜𝑝 = � 𝐶𝑓𝑠𝑚𝑢𝑑𝑔𝑒 ∗ 0.5(𝐹𝑟−1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.2) 𝐾=1 D Average of PPP deposited per unit of leaf area (µg/m2) E Spray efficiency (100 %) Ccoi Concentration of a.s in spraying volume (µg/L) Ar Application rate (L/m2) H Crop Height (m) Fr Frequency of contact according to row space (Table 5) Cfcrop Corrected transfer efficiency for maximum amount of contacts (maximum: 0.22 with Cf = 0.112; Fgsmudge= 6) (Appendix H) Cfsmudge Transfer efficiency for the smudging of smooth surfaces (triangular distribution with lower value 0.019, mode 0.112 and upper 0.45 (Appendix H) 106 Surface contact with the backpack 𝐷𝑐𝑤𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 = 𝑆𝑐𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 ∗ 𝐶𝑐𝑜𝑖 ∗ 𝐶𝑓𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 ∗ 𝐵𝑃𝑏𝑜𝑑𝑦 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 6) 𝐵𝑃𝑏𝑎𝑐𝑘 𝐹𝑚&𝑙 𝐶𝑓𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 = � 𝐶𝑓𝑔𝑟𝑎𝑠𝑝 ∗ 0.5(𝐹𝑚&𝑙 −1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 3.2) 𝐾=1 Scbackpack Surface contamination level of the backpack (l/cm2) Cfbackpack Corrected transfer efficiency for multiple contacts (maximum: 0.89 with Cf = 0.45; Fgrasp = 6) Cfgrasp Transfer efficiency for grasping of smooth surfaces (triangular distribution with lower value 0.23, mode 0.45 and upper 0.68) Lcn/BPback Affected surface area of the back (m2 of BP) Scenario 2: Backpack – Up- and Downward spraying I. Hand exposure model: Dph = Ddhapp + Dchcrop + Dchgun + Dchbackpack Dph Potential whole body dermal exposure a.s (µg) Ddhapp Potential dermal hand exposure to droplets via deposition (µg) Dchcrop Potential dermal hand exposure via surface contacts with the crop (µg) Dchgun Potential dermal hand exposure via surface contacts with the gun (µg) Dchbackpack Potential dermal hand exposure via surface contacts with the backpack (µg) 107 Deposition 𝐷𝑑ℎ𝑎𝑝𝑝 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 𝐼) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑡𝑜𝑝 Cair Concentration of a.i. in air (µg/m3) t Total application time (s) vdep Droplet settling velocity average deposition speed of droplets in the air (m/s) BPhands Total surface area of the hands (m2) lcn/ BPhands_top Affected surface area of hands (m2 of BP) Surface contact with the crop 𝐷𝑐ℎ𝑐𝑟𝑜𝑝 = 𝐷 ∗ 𝐶𝑓𝑐𝑟𝑜𝑝 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 𝐷= 𝐸∗𝐶𝑐𝑜𝑖 ∗𝐴𝑟 0.99∗𝐻 1.49 ∗103 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.1) 𝐹𝑟 𝐶𝑓𝑐𝑟𝑜𝑝 = � 𝐶𝑓𝑠𝑚𝑢𝑑𝑔𝑒 ∗ 0.5(𝐹𝑟−1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.2) 𝐾=1 D Average of PPP deposited per unit of leaf area (µg/m2) E Spray efficiency factor Ccoi Concentration of a.s in spraying volume (µg/L) Ar Application rate (L/m2) H Crop Height (m) Cfcrop Corrected transfer efficiency for maximum amount of contacts (maximum: 0.22 with Cf = 0.112; Fgsmudge= 6) (Appendix H) Cfsmudge Transfer efficiency for the smudging of smooth surfaces (triangular distribution with lower value 0.019, mode 0.112 and upper 0.45 (Appendix H) Fr Frequency of contact according to row space (Table 5) 108 Surface contacts with the gun 𝐷𝑐ℎ𝑔𝑢𝑛 = 𝑆𝑐𝑔𝑢𝑛 ∗ 𝐶𝑓𝑚𝑎𝑥 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑝𝑎𝑙𝑚𝑠 𝑆𝑐𝑔𝑢𝑛 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2.1) 𝐹=6 𝐶𝑓𝑚𝑎𝑥 = � 𝐶𝑓𝑔𝑟𝑎𝑠𝑝 ∗ 0.5(𝐹−1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2.2) 𝐾=1 Scgun = Surface contamination of the gun Cfmax Corrected transfer efficiency for maximum amount of contacts (maximum: 0.89 with Cf = 0.45; Fgrasp= 6) (Appendix H) Cfgrasp Transfer efficiency for grasping of smooth surfaces (triangular distribution with lower value 0.23, mode 0.45 and upper 0.68) lcn/ BPhands_palms Affected surface area of hands (m2 of BP) Surface contact with the backpack 𝐷𝑐ℎ𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 = 𝑆𝑐𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘_𝑇𝑜𝑡 ∗ 𝐶𝑓𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 3) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑝𝑎𝑙𝑚𝑠 𝑆𝑐𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘_𝑇𝑜𝑡 = (𝑆𝑐𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 ∗ 𝐶𝑐𝑜𝑖 ) + (𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 3.1) 𝐹𝑚&𝑙 Scbackpack Cfbackpack 𝐶𝑓𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 = � 𝐶𝑓𝑔𝑟𝑎𝑠𝑝 ∗ 0.5(𝐹𝑚&𝑙 −1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 3.2) 𝐾=1 Surface contamination level of the backpack (l/cm2) Corrected transfer efficiency for multiple contacts (maximum: 0.89 with Cf = 0.45; Fgrasp = 6) (Appendix H) Cfgrasp Transfer efficiency for grasping of smooth surfaces (triangular distribution with lower value 0.23, mode 0.45 and upper 0.68) 109 II. Fm&l Frequency of M&L periods Ccoi Concentration of a.s in spraying volume (µg/L) Whole body exposure model: Dpw = Ddwapp + Dcwcrop + Dcwbackpack Dpw Potential whole body dermal exposure a.s (µg) Ddwapp Potential dermal whole body exposure to droplets via deposition (µg) Dcwcrop Potential dermal whole body exposure via surface contacts with the crop (µg) Dcwbackpack Potential dermal whole body exposure via surface contacts with the backpack (µg) Deposition 𝐷𝑑𝑤𝑎𝑝𝑝 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ∗ 𝐵𝑃𝑤ℎ𝑜𝑙𝑒 𝑏𝑜𝑑𝑦 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 4) Cair Concentration of a.i. in air (µg/m3) t Total application time (s) vdep Droplet settling velocity average deposition speed of droplets in the air (m/s) BPwhole body Total surface area of the body (m2) Surface contacts with the crop 𝐷𝑐𝑤𝑐𝑟𝑜𝑝 = 𝐷= 𝐷 ∗ 𝐶𝑓𝑐𝑟𝑜𝑝 ∗ 𝐵𝑃𝑤ℎ𝑜𝑙𝑒 𝑏𝑜𝑑𝑦 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5) 𝐵𝑃𝑏𝑜𝑑𝑦−𝑐ℎ 𝐸∗𝐶𝑐𝑜𝑖 ∗𝐴𝑟 0.99∗𝐻 1.49 ∗103 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.1) 𝐹𝑟 𝐶𝑓𝑐𝑟𝑜𝑝 = � 𝐶𝑓𝑠𝑚𝑢𝑑𝑔𝑒 ∗ 0.5(𝐹𝑟−1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.2) 𝐾=1 110 D Average of PPP deposited per unit of leaf area (µg/m2) E Spray efficiency factor Ccoi Concentration of a.s in spraying volume (µg/L) Ar Application rate (L/m2) H Crop Height (m) Cfcrop Corrected transfer efficiency for maximum amount of contacts (maximum: 0.22 with Cf = 0.112; Fgsmudge= 6) (Appendix H) Cfsmudge Transfer efficiency for the smudging of smooth surfaces (triangular distribution with lower value 0.019, mode 0.112 and upper 0.45 (Appendix H) Fr Frequency of contact according to row space (Table 6) Lcn/BPbody-ch Affected surface area of the back (m2 of BP) Surface contacts with the backpack 𝐷𝑐𝑤𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 = 𝑆𝑐𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 ∗𝐶𝑐𝑜𝑖 ∗𝐶𝑓𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 ∗𝐵𝑃𝑏𝑜𝑑𝑦 ∗𝑙𝑐𝑛 𝐵𝑃𝑏𝑎𝑐𝑘 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 6) 𝐹𝑚&𝑙 𝐶𝑓𝑏𝑎𝑐𝑘𝑝𝑎𝑐𝑘 = � 𝐶𝑓𝑔𝑟𝑎𝑠𝑝 ∗ 0.5(𝐹𝑚&𝑙 −1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 3.2) 𝐾=1 Scbackpack Surface contamination level of the backpack (l/cm2) Cfbackpack Corrected transfer efficiency for multiple contacts (maximum: 0.89 with Cf = 0.45; Fgrasp = 6) (Appendix H) Cfgrasp Transfer efficiency for grasping of smooth surfaces (triangular distribution with lower value 0.23, mode 0.45 and upper 0.68) Fm&l Frequency of M&L periods Ccoi Concentration of a.s in spraying volume (µg/L) Lcn/BPback Affected surface area of the back (m2 of BP) 111 Scenario 3: Gun and Hose – Downward spraying I. Hand exposure model: Dph = Ddhapp + Dchgun Dph Potential hands dermal exposure a.s (µg) Ddhapp Potential dermal hand exposure to droplets via deposition (µg) Dchgun Potential dermal hand exposure via surface contacts with the gun (µg) Deposition 𝐷𝑑ℎ𝑎𝑝𝑝 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 𝐼) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑡𝑜𝑝 Cair Concentration of a.i. in air (µg/m3) t Total application time (s) vdep Droplet settling velocity average deposition speed of droplets in the air (m/s) BPhands Total surface area of the hands (m2) lcn/ BPhands_top Affected surface area of hands (m2 of BP) Surface contacts with the gun 𝐷𝑐ℎ𝑔𝑢𝑛 = 𝑆𝑐𝑔𝑢𝑛 ∗ 𝐶𝑓𝑚𝑎𝑥 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑝𝑎𝑙𝑚𝑠 𝑆𝑐𝑔𝑢𝑛 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2.1) 𝐹=6 𝐶𝑓𝑚𝑎𝑥 = � 𝐶𝑓𝑔𝑟𝑎𝑠𝑝 ∗ 0.5(𝐹−1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2.2) 𝐾=1 112 Scgun = Surface contamination of the gun Cfmax Corrected transfer efficiency for maximum amount of contacts (maximum: 0.89 with Cf = 0.45; Fgrasp= 6) (Appendix H) Cfgrasp Transfer efficiency for grasping of smooth surfaces (triangular distribution with lower value 0.23, mode 0.45 and upper 0.68) lcn/ BPhands_palms Affected surface area of hands (m2 of BP) II. Whole body exposure model: Dpw = Ddapp + Dccrop Dpw Potential whole body dermal exposure a.s (µg) Ddwapp Potential dermal whole body exposure to droplets via deposition (µg) Dcwcrop Potential dermal whole body exposure via surface contacts with the crop (µg) Deposition 𝐷𝑑𝑤𝑎𝑝𝑝 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ∗ 𝐵𝑃𝑤ℎ𝑜𝑙𝑒 𝑏𝑜𝑑𝑦 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 4) Cair Concentration of a.i. in air (µg/m3) t Total application time (s) vdep Droplet settling velocity average deposition speed of droplets in the air (m/s) BPwhole body Total surface area of the body (m2) Surface contacts with the crop 𝐷𝑐𝑤𝑐𝑟𝑜𝑝 = 𝐷 ∗ 𝐶𝑓𝑐𝑟𝑜𝑝 ∗ 𝐵𝑃𝑤ℎ𝑜𝑙𝑒 𝑏𝑜𝑑𝑦 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5) 𝐵𝑃𝑏𝑜𝑑𝑦−𝑐ℎ 113 𝐷= 𝐸∗𝐶𝑐𝑜𝑖 ∗𝐴𝑟 0.99∗𝐻 1.49 ∗103 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.1) 𝐹𝑟 𝐶𝑓𝑐𝑟𝑜𝑝 = � 𝐶𝑓𝑠𝑚𝑢𝑑𝑔𝑒 ∗ 0.5(𝐹𝑟−1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.2) 𝐾=1 D Average of PPP deposited per unit of leaf area (µg/m2) E Spray efficiency factor Ccoi Concentration of a.s in spraying volume (µg/L) Ar Application rate (L/m2) H Crop Height (m) Fr Frequency of contact according to row space (Table 6) Cfcrop Corrected transfer efficiency for maximum amount of contacts (maximum: 0.22 with Cf = 0.112; Fgsmudge= 6) (Appendix H) Cfsmudge Transfer efficiency for the smudging of smooth surfaces (triangular distribution with lower value 0.019, mode 0.112 and upper 0.45 (Appendix H) BPwhole body Total surface area of the body (m2) Lcn/BPbody-ch Proportion of body part affected according to crop height (m2 of BP) Scenario 4: Gun and Hose – Up- and Downward spraying I. Hand exposure model Dph = Ddhapp + Dchgun + Dchcrop Dph Potential hands dermal exposure a.s (µg) Ddhapp Potential dermal hand exposure to droplets via deposition (µg) Dchcrop Potential dermal hand exposure via surface contacts with the crop (µg) Dchgun Potential dermal hand exposure via surface contacts with the gun (µg) 114 Deposition 𝐷𝑑ℎ𝑎𝑝𝑝 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 𝐼) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑡𝑜𝑝 Cair Concentration of a.i. in air (µg/m3) t Total application time (s) vdep Droplet settling velocity average deposition speed of droplets in the air (m/s) BPhands Total surface area of the hands (m2) lcn/ BPhands_top Affected surface area of hands (m2 of BP) Surface contacts with the crop 𝐷𝑐ℎ𝑐𝑟𝑜𝑝 = 𝐷 ∗ 𝐶𝑓𝑐𝑟𝑜𝑝 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 𝐷= 𝐸∗𝐶𝑐𝑜𝑖 ∗𝐴𝑟 0.99∗𝐻 1.49 ∗103 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.1) 𝐹𝑟 𝐶𝑓𝑐𝑟𝑜𝑝 = � 𝐶𝑓𝑠𝑚𝑢𝑑𝑔𝑒 ∗ 0.5(𝐹𝑟−1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.2) 𝐾=1 D Average of PPP deposited per unit of leaf area (µg/m2) E Spray efficiency factor Ccoi Concentration of a.s in spraying volume (µg/L) Ar Application rate (L/m2) H Crop Height (m) Cfcrop Corrected transfer efficiency for maximum amount of contacts (maximum: 0.22 with Cf = 0.112; Fgsmudge= 6) (Appendix H) Cfsmudge Transfer efficiency for the smudging of smooth surfaces (triangular distribution with lower value 0.019, mode 0.112 and upper 0.45 (Appendix H) Fr Frequency of contact according to row space (Table 6) 115 Surface contacts with the gun 𝐷𝑐ℎ𝑔𝑢𝑛 = 𝑆𝑐𝑔𝑢𝑛 ∗ 𝐶𝑓𝑚𝑎𝑥 ∗ 𝐵𝑃ℎ𝑎𝑛𝑑𝑠 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2) 𝐵𝑃ℎ𝑎𝑛𝑑𝑠_𝑝𝑎𝑙𝑚𝑠 Scgun = Surface contamination of the gun Cfmax Corrected transfer efficiency for maximum amount of contacts (maximum: 0.89 with Cf = 0.45; Fgrasp= 6) (Appendix H) Cfgrasp Transfer efficiency for grasping of smooth surfaces (triangular distribution with lower value 0.23, mode 0.45 and upper 0.68) lcn/ BPhands_palms Affected surface area of hands (m2 of BP) II. Whole body exposure model: Dpw =Ddwapp + Dcwapp Dpw Potential whole body dermal exposure a.s (µg) Ddwapp Potential dermal whole body exposure to droplets via deposition (µg) Dcwcrop Potential dermal whole body exposure via surface contacts with the crop (µg) Deposition 𝐷𝑑𝑤𝑎𝑝𝑝 = 𝐶𝑎𝑖𝑟 ∗ 𝑡 ∗ 𝑣𝑑𝑒𝑝 ∗ 𝐵𝑃𝑤ℎ𝑜𝑙𝑒 𝑏𝑜𝑑𝑦 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 4) Cair Concentration of a.i. in air (µg/m3) t Total application time (s) vdep Droplet settling velocity average deposition speed of droplets in the air (m/s) BPwhole body Total surface area of the body (m2) 116 Surface contacts with the crop 𝐷𝑐𝑤𝑐𝑟𝑜𝑝 = 𝐷= 𝐷 ∗ 𝐶𝑓𝑐𝑟𝑜𝑝 ∗ 𝐵𝑃𝑤ℎ𝑜𝑙𝑒 𝑏𝑜𝑑𝑦 ∗ 𝑙𝑐𝑛 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5) 𝐵𝑃𝑏𝑜𝑑𝑦−𝑐ℎ 𝐸∗𝐶𝑐𝑜𝑖 ∗𝐴𝑟 0.99∗𝐻 1.49 ∗103 (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.1) 𝐹𝑟 𝐶𝑓𝑐𝑟𝑜𝑝 = � 𝐶𝑓𝑠𝑚𝑢𝑑𝑔𝑒 ∗ 0.5(𝐹𝑟−1) (𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5.2) 𝐾=1 D Average of PPP deposited per unit of leaf area (µg/m2) E Spray efficiency factor Ccoi Concentration of a.s in spraying volume (µg/L) Ar Application rate (L/m2) H Crop Height (m) Fr Frequency of contact according to row space (Table 6) Cfcrop Corrected transfer efficiency for maximum amount of contacts (maximum: 0.22 with Cf = 0.112; Fgsmudge= 6) (Appendix H) Cfsmudge Transfer efficiency for the smudging of smooth surfaces (triangular distribution with lower value 0.019, mode 0.112 and upper 0.45 (Appendix H) (c) BPwhole body Total surface area of the body (m2) Lcn/BPbody-ch Proportion of body part affected according to crop height (m2 of BP) Model testing Please refer to the type of plots/graphs and explanation under item 3.2.6 (f) 117 The whole model was run with a dataset of more than 600 data entries in which some inputs were imputed. There were at least 120 data entries per scenario. Imputation was done by empirical sampling from existing values in the data and 2,000 iterations were used. Mean, median, 2.5th percentile and 97.5th percentiles were calculated and plotted against the true values that we have from data and plot the 95% confidence intervals, medians and means of the output against the true values (all on log10 scale) for comparison Additionally the suitability of considering four different scenarios based on the spraying direction and the equipment used was tested. Figures 1 and 2 prove that the variation in exposure decreases by considering the above mentioned scenarios. As shown in figures 3 to 10, the whole model tends to underestimate exposure for hand exposure as well as for whole body exposure, nevertheless it can be due to the imputed data used. The model provides a high variability and broad range in exposure Scenario 1: Backpack Downward Spraying Testing of the hands exposure model in plotted in Figure 3 and whole body is plotted in Figure 4 Scenario 2: Backpack Up and Downward Spraying Testing of the hands exposure model in plotted in Figure 5 and whole body is plotted in Figure 6 Scenario 3: Gun and Hose Downward Spraying Testing of the hands exposure model in plotted in Figure 7 and whole body is plotted in Figure 8 Scenario 4: Gun and Hose Up and Downward Spraying Testing of the hands exposure model in plotted in Figure 9 and whole body is plotted in Figure 10 (d) Comment to Gun and Hose Up and Downwards spraying Within the Browse Reserve fund Report Field trials conducted in Greece data on operator exposures for hand held application of a gun and hose up and downward scenario were collected. 118 These data are considered to mimic worst case scenario’s. In these experiments the transfer from crop to operator was measured during spraying processes. This was done by first spraying pesticide A and later spraying pesticide B while measuring pesticide A. Crops were very dense and high with narrow row spacings. Comparisons of our model estimations of crop transfer and the experimental data collected in the reserve fund projecte are displayed in the table Operator Model output (% of total exposure contribution by the crop transfer route) Total Body Hands (e) 0 – 11 % 0 – 22.6 % Reserve fund output (worst case) (% of measured pesticide A transferred from crop during application of pesticide B directly after application of pesticide A) 10 – 16 % 9 – 21 % Tables Table 1: Hand Exposure Models, Scenarios and Pathways included Model Scenario Pathway Hands Exposure Backpack Downward Potential dermal hand exposure to droplets via deposition (Ddhapp) Potential dermal hand exposure via surface contacts with the gun (Ddhgun) Spraying Potential dermal hand exposure via surface contacts with the backpack (Ddhbackpack) Backpack Up and Downward Spraying Potential dermal hand exposure to droplets via deposition (Ddhapp) Potential dermal hand exposure via surface contacts with the crop (Ddhcrop) Potential dermal hand exposure via surface contacts with the gun (Ddhgun) Potential dermal hand exposure via surface contacts with the backpack (Ddhbackpack) Gun and Hose Downward Spraying Gun and Hose Up and Downward Spraying Potential dermal hand exposure to droplets via deposition (Ddhapp) Potential dermal hand exposure via surface contacts with the gun (Ddhgun) Potential dermal hand exposure to droplets via deposition (Ddhapp) Potential dermal hand exposure via surface contacts with the crop (Ddhcrop) Potential dermal hand exposure via surface contacts with the gun (Ddhgun) Table 2: Whole Body Exposure Models, Scenarios and Pathways included Model Scenario Pathway Whole Body (excluding Backpack Downward Potential dermal whole body exposure to droplets via deposition (Ddwapp) hands) Exposure Spraying Potential dermal whole body exposure via surface contacts with the crop (Dcwcrop ) Potential dermal whole body exposure via surface contacts with the backpack (Dcwbackpack ) Backpack Up and Downward Spraying Potential dermal whole body exposure to droplets via deposition (Ddwapp) Potential dermal whole body exposure via surface contacts with the crop (Dcwcrop ) Potential dermal whole body exposure via surface contacts with the backpack (Dcwbackpack ) 119 Gun and Hose Potential dermal whole body exposure to droplets via deposition (Ddwapp) Downward Spraying Potential dermal whole body exposure via surface contacts with the crop (Dcwcrop ) Gun and Hose Up and Potential dermal whole body exposure to droplets via deposition (Ddwapp) Downward Spraying Potential dermal whole body exposure via surface contacts with the crop (Dcwcrop ) Table 3: Handheld application model inputs abbr User input Model Inputs Lower Upper Type t Total application Application Time 0 1500 Selected value Concentration of a.s. in 0.05 50 selected value Based on time of spraying activity (s) Cc.o.i Concentration of BROWSE data a.s. in spray volume spray volume (g/l) (g/l) Cair n/a Concentration of A.I. in Fixed air (µg/m3) according to inhalation model Fmlp n/a Number of M&L periods 1 20 selected value BROWSE data 1 6 Uniform Table 6 (estimated with volume applied and tank size) Fr n/a Frequency of contact according to row space distribution according to row space H Crop height (m) Crop height (m) Selected value Ar Application Rate, Application Rate, (L/ha) Selected value (L/ha) E n/a Spray deposition Efficiency factor 0.6 1 Default is 1 (L. R. Ahuja, (2000); Zabkiewicz, (2007)). 120 abbr User input Model Inputs Lower Upper Type Based on drow Row space (m) Row space (m) 0.1 >1.5 Selected Value (Frenich et al., 2002), (Machera, 2003) Cfcrop n/a Corrected transfer - - Fixed Appendix H 0.019 0.45 triangular Appendix H efficiency for maximum amount of contacts (maximum: 0.22 with Cf = 0.112; Fgsmudge= 6) Cfsmudge n/a Transfer efficiency for the smudging of smooth distribution with surfaces lower value 0.019, mode 0.112 and upper 0.45 BPhands Gender (male, female) Surface area of hands - - 2 (m ) lognormal Appendix I distributions (Males: log(SA) ~ Normal (-2.2319, 0.109922) Females: log(SA) ~ Normal(2.4313, 0.10362)) BPwholeb Gender (male, Surface area of the ody female) whole body (m2) - - Lognormal Appendix I distributions (Males: Lcn/BPb ody-ch n/a Proportion of body part affected according to crop height 121 - - Table 5 abbr User input Model Inputs lcn/ n/a Affected surface area of - BPhands_t Lower Upper Type Based on - Table 5 top of hands op * Number of contacts with the treated crop is linked to the row space (See table 6) Table 4. Droplet setting velocity variables Symbol Description Units V Droplet Setting Velocity m∙s ø Droplet Diameter m δg Droplet density m-3 1,000 m-3 1.275 Air Density Kg∙ ἠ Dynamic viscosity of the air kg∙m ∙s g Gravitational constant m∙s -1 -1 -2 2 2 2 Total surface area of the body, m (BPwhole body) 2 -5 No No -4 5∙10 - 1∙10 Kg∙ δa Total surface area of hands, m (BPhands) Effect Range 0.078 – 0.303 - Table 5. Body surface area affected. Default values User No Input Variable input 1 Multiplier -1 -5 1.78 ∙10 9.81 Inputs Multiplier Effect Range Type distribution Male 0,107 0.09 – 0.131 US EPA (1985) Female 0.089 0.076 0.106 Mean 0.107 (5% 0.09, 95% 0.131) Mean 0.076 (5% 0.076, 95% 0.106) Male 1.95 1.62 – 2.41 US EPA (1985) Female 1.76 1.40 – 2.24 Mean 1.95 (5% 1.62, 95% 2.41) Mean 1.76 (5% 1.40, 95% 2.24) – of Based on US EPA (1985) US EPA (1985) 3 Total surface area of the back, m No 0.155 US EPA 2000 4 Lcn/Bphands_top No 0.5 ? 5 Lcn/Bphands-palms No 0.5 6 Lcn/BPback No 1 7 Icn/BPbody-ch Crop height(m) 0-0.19 US EPA 1996 Male 0.066 Female 0.066 122 0.2-0.59 Male 0.2 Female 0.19 Male 0.4 Female 0.39 1-1.29 Backpack model Male 0.66 Female 0.64 1-1.29 gun and hose model Male 0.86 Female 0.81 >1.3 Backpack model Male 0.74 Female 0.68 >1.3 gun and hose model Male 0.94 Female 0.86 0.6-0.99 Table 6 Frequency of contacts according to row space No 8 Input Variable User input Number of Contacts according to Row space no Number contacts 1-2 3-4 5-6 of Row (m) Space Base on >1.5 0.91 – 1.49 <90 (Brouwer et al., 2001) (Cohen Hubal et al., 2005) Table 7. Overview of induction bowl contamination after emptying a full container Mathers et al 2000 Glass et al 2009 Container Induction bowl Induction bowl Induction bowl Induction bowl Induction bowl Surface area size contamination contamination contamination contamination contamination induction bowl contamination (L) (L) (L) (L) (m ) SD 1 5 3.10E-06 1.13E- Glass et al 2009 (L) 2 Average surfac -2 (L∙m ) 1.00E-06 6.00E-06 0,25 4,00E-06 3.10E-06 7.00E-06 0,25 1,22E-05 3.80E-05 1.01E-04 0,25 8,76E-05 5.10E-05 8.60E-05 0,25 4,28E-04 05 10 5.8E-06 8.70E06 20 5.00E-05 7.00E05 123 2.20E-04 3.80E-04 Table 8. Default values for Transfer Coefficients. Cf Cfgrasp Description / specifics Reference N median AM SD min max Donor Surface = Spray tank Surface to glove, Smudge, >12 hours, Dry, Mass/mass, Liquid Ramwell et al., 2010 Decontamination of agricultural sprayers, HSE report RR792 6 48.5 47.8 23.2 21 79 11.21 12.7 1,9 45.5 Donor Surface= treated crop to 16 Cfsmudge hands, >12 hours, Wet and Dry, 8 Mass/mass. (f) Figures Figure 1. Comparison of geometric means against true values for all hand PDE models. Red represents the Backpack Downwards scenario, Blue represents the Gun and Hose Downwards scenario 2, Black is the Gun and Hose Up and Downwards scenario 3 and green is Backpack Up and Downwards scenario 4. 1e+07 1e+06 log 10 (Data(µg/hands)) 100000 10000 1000 100 10 1 0.1 0.01 0.001 0.001 0.01 0.1 1 10 100 1000 10000 100000 1e+06 1e+07 log10 (Mean(Model outputs(µg/hands))) 124 Figure 2. Comparison of geometric means against true values for all whole body excluding hand PDE models. Red represents the Backpack Downwards scenario, Blue represents the Gun and Hose Downwards scenario 2, Black is the Gun and Hose Up and Downwards scenario 3 and green is Backpack Up and Downwards scenario 4. 1e+09 1e+08 1e+06 100000 10000 log 10 (Data(µg/body)) 1e+07 1000 100 10 1 1 10 1000 10000 100000 1e+06 1e+07 1e+08 1e+09 100 log10 (Mean(Model outputs(µg/body))) Figure 3. Hands exposure. Plot of 95% CIs (grey lines), geometric means (red dots) and medians (black dashed lines) for the log10 (model outputs) with the log10 (measured potential dermal exposure (hands)*) (blue diamonds). (Backpack Downwards scenario). * Including data sets which have model inputs that were imputed 1e+07 100000 10000 1000 100 10 log (Model outputs (µg/hands)) 1e+06 10 1 0.1 0 20 40 60 80 Data index 125 100 120 140 160 Figure 4. Whole body (excluding hands) exposure. Plot of 95% CIs (grey lines), geometric means (red dots) and medians (black dashed lines) for the log10 (model outputs) with the log10 (measured potential dermal exposure (whole body excluding hands)*) (blue diamonds). (Backpack Downwards scenario).* Including data sets which have model inputs that were imputed 1e+08 1e+06 100000 10000 1000 10 log (Model outputs (µg/body)) 1e+07 100 10 1 0 20 40 60 80 100 Data index 120 140 160 180 Figure 5. Hands exposure. Plot of 95% CIs (grey lines), geometric means (red dots) and medians (black dashed lines) for the log10 (model outputs) with the log10 (measured potential dermal exposure (hands)*) (blue diamonds). (Backpack up and downwards scenario). * Including data sets which have model inputs that were imputed 1e+06 10 log (Model outputs (µg/hands)) 100000 10000 1000 100 10 1 0.1 0.01 0.001 0 10 20 30 40 50 60 Data index 126 70 80 90 100 Figure 6. Whole body (excluding hands) exposure. Plot of 95% CIs (grey lines), geometric means (red dots) and medians (black dashed lines) for the log10 (model outputs) with the log10 (measured potential dermal exposure (whole body excluding hands)*) (blue diamonds). (Backpack up and downwards scenario). * Including data sets which have model inputs that were imputed 1e+07 10 log (Model outputs (µg/body)) 1e+06 100000 10000 1000 100 10 1 0 20 40 60 Data index 127 80 100 120 Figure 7. Hand exposure. Plot of 95% CIs (grey lines), geometric means (red dots) and medians (black dashed lines) for the log10 (model outputs) with the log10 (measured potential dermal exposure (hands)*) (blue diamonds). (Gun and hose downwards scenario). *Including data sets which have model inputs that were imputed 10 log (Model outputs (µg/hands)) 100000 10000 1000 100 10 0 10 20 30 40 50 60 Data index 70 80 90 100 Figure 8. Whole body (excluding hand) exposure. Plot of 95% CIs (grey lines), geometric means (red dots) and medians (black dashed lines) for the log10 (model outputs) with the log10 (measured potential dermal exposure (whole body excluding hands)*) (blue diamonds). (Gun and hose downwards scenario). *Including data sets which have model inputs that were imputed 1e+07 10 log (Model outputs (µg/body)) 1e+06 100000 10000 1000 100 10 0 10 20 30 40 Data index 128 50 60 70 80 Figure 9. Hand exposure. Plot of 95% CIs (grey lines), geometric means (red dots) and medians (black dashed lines) for the log10 (model outputs) with the log10 (measured potential dermal exposure (hands)*). (Gun and hose up and downwards scenario). * Including data sets which have model inputs that were imputed) (blue diamonds) 100000 1000 100 10 10 log (Model outputs (µg/hands)) 10000 1 0.1 0 20 40 60 80 Data index 100 120 140 Figure 10. Whole body (excluding hands) exposure. Plot of 95% CIs (grey lines), geometric means (red dots) and medians (black dashed lines) for the log10 (model outputs) with the log10 (measured potential dermal exposure (whole body excluding hands)*) (Gun and hose up and downwards scenario) * including data sets which have model inputs that were imputed) (blue diamonds) 1e+09 1e+07 1e+06 100000 10000 10 log (Model outputs (µg/body)) 1e+08 1000 100 10 0 20 40 60 80 100 Data index 129 120 140 160 3.3 Aggregated exposure 3.3.1 External exposure As presented in the previous section, potential inhalation and dermal exposure models were developed for each scenario. The ingestion estimates are based on input from the dermal model and default estimates of hand-to-mouth events. For the inhalation and dermal exposure models, a distinction is made between potential and actual exposure estimates. For dermal exposure, model estimates distinguish between body surface areas of males and females (Appendix I). Potential exposure The ‘potential’ exposure models predict an operator’s exposure to PPP without taking into account (protective) clothing or respiratory protective equipment. The potential exposure level therefore reflects the exposure levels typically measured during field studies, e.g. an inhalation sample in the breathing zone of the operator (and not taking account of the effect of respirators), or the amount of PPP sampled on the exterior of the body (on gloves and clothing, but also on the skin if unprotected). These models can therefore be calibrated and validated using potential exposure data. Actual exposure The estimates of the potential exposure models are subsequently used for ‘actual’ exposure estimates. An actual estimate represents the predicted PPP exposure considering the reduction in exposure due to skin or respiratory protection. In some instances, the actual and potential exposure will be equal because no protection was used. For ingestion exposure the potential exposure is equal to the actual exposure. However, the effect of respiratory equipment (gloves, respirators) needs to be considered to estimate the (default) frequency of hand to mouth contacts. The ingestion exposure will also be estimated separately for each scenario (since hand contamination levels may vary). The actual estimates cannot always be validated because of a lack of data. Data on actual exposures are generally limited to measurements taken inside respirators, or skin samples taken underneath clothing or PPE. Extensive literature reviews and analyses of data on the effectiveness of PPE are applied as default reduction factors in the models (see separate report from reserve fund proposal on this). 130 For dermal exposure, the BROWSE models distinguish between hand exposure and body exposure. However, to include the use of PPE and work wear as risk management measures, the models distinguish between hand, body and head exposure. For this purpose, an estimate is made to differentiate between body and head exposure. Based on a literature review of body part distributions (Appendix F), the following distributions of head exposure compared to the total body exposures are presented in Table 21. 3.3.1.1 Inhalation exposure (Ia) The actual inhalation exposure (Ia) is estimated using the following equation: Ia = Ipde * Prpe Where, Ipde potential inhalation exposure to a.s. in droplet spray in µg/m3 Prpe protection factor of respiratory protective equipment (RPE) (Table 20) 3.3.1.2 Dermal exposure (Da) The actual dermal exposure (Da) is estimated using the following equation: Da = (Dph * Pppe) + (Dpw * Pppe) Where, Dph Potential dermal exposure to a.s. in µg on the hands Dpw Potential dermal exposure to a.s. in µg on the whole body (excl. hands) Pppe Protection factor of personal protective equipment (PPE) (Table 20) 3.3.1.3 Ingestion exposure (Ina) Ingestion of pesticides may occur as a result of hand-to-mouth contacts in the workplace, also referred to as inadvertent ingestion exposure. It has been estimated that 16% of the working population in the UK may be exposed to hazardous materials by the ingestion route (Cherrie et al, 2006). To estimate ingestion exposure, a similar approach was used as described by Gorman NG et al (2012). The actual ingestion exposure (Ina) is estimated as follows: 131 Ina = Dph * Am/Ah * SE * (N*t) Ina ingestion exposure, µg (per scenario) Dph potential dermal exposure to a.s. in µg on both hands (per scenario) Am area of one hand making contact with the mouth [cm2] (fixed 75cm2) (14% from EPA, Michaud) Ah area of both hands [cm2] (EPA’s value for male hands of 1070cm2) Am/Ah 0.07 SE skin to mouth transfer factor (0.43); see Table 23 N number of hand-to-mouth contacts (events/hr); Table 22 t exposure time of scenario in hrs (minutes/60; roundup=1). So N*t ≥1 roundup More information of each parameter is presented in Table 22. Evidence for the input parameters of ingestion exposure was obtained from various sources. The potential dermal exposure to a.s. on both hands (Dph) for a given per scenario is considered the mass balance input. The relative (contaminated) hand surface area (Am) that may contact the peri-oral or oral cavity is estimated using an EPA default value (14%), assuming that only the fingertips of one hands are involved in a hand-to-mouth contact event. A hand-to-mouth transfer factor (0.43) was calculated using EPA and IOM data (Table 23). The latter value is derived for hand-to-mouth (oral cavity) transfer, but does not include the transfer efficiency from the peri-oral area to the oral cavity Since hand-to-mouth contacts may primarily involve contacts with the peri-oral region (and much less contacts directly with the mouth), the value of 0.43 may be too conservative. A value of 10% (0.1) may also be considered for the transfer efficiency into the mouth (Cherrie et al, 2006). To derive values for the frequency of hand-to-mouth contacts, an observational study from Gorman NG & colleagues (under review) was used. They observed workers’ hand-to-mouth behaviors in different industries, worksites and tasks. The majority of the contacts occurred between the hands and peri-oral area (lips and within 2cm of the lips). Interestingly, the hand-to-mouth contact frequencies were significantly higher among workers in the high risk perception group. The high risk group is workers who are aware that they are using highly hazardous or toxic substances. This suggests that workers working with substances they consider hazardous may not necessarily reduce their hand-to-mouth contact behavior. It was found that the hand-to-mouth contact frequency (contacts/hr) was significantly higher between-tasks than during tasks, and significantly higher in 132 industrial worksites (AM=7.6; 6.5-26) compared with laboratories (AM=0.5; 0-4). Among workers who used personal protective equipment (PPE), the contact frequencies (contacts/hr) were lower when gloves (AM=1.2) and respirators (AM=0.1) were worn, than when they were not worn (gloves; AM=4.8; respirators; AM=5.3). To derive default values for the ‘number of hand-to-mouth contacts’ (N), we assumed that the operator exposure scenarios may be more comparable to the ‘laboratory/research’ work sectors. The reason for this assumption is that this sector adopted more stringent ‘best practices’ and respirator use compared with the ‘industrial sector’. Since the WP1 models assess task-based exposures, the ‘between tasks’ hand-to-mouth contacts are excluded. We decided to distinguish between the presence and absence of respirators or glove use. Since the effect of the combined use of both gloves and respirators were not presented in the study, a hand-to-mouth contact frequency (N) was derived for ‘use of gloves/respirators’. Considering the overall outcome of the study described earlier, it was decided to assign a uniform distribution for when gloves/respirators are worn (0.1; 1.0), and when not worn (0.5; 5.0). Existing PPP models do not take account of ingestion exposure for operators. Observational evidence is available that indicates that ingestion exposure is possible through hand-to-mouth contacts in different industries (Gorman Ng et al, 2013). However, limited evidence is available from epidemiological studies to substantiate actual ingestion exposure, e.g. using biomonitoring. With that in mind, and the evidence currently available, we feel that it is appropriate to include an ingestion estimate in the WP1 operator models. Considering the fact that the ingestion exposure route is new and not commonly used for regulatory purposes in the agricultural sector, the ingestion estimate is by default not activated in the software. However, it is up to the software user to activate the ingestion module and interpret the ingestion exposure predicted by the model. 3.3.2 Internal exposure Absorption values together with the bodyweight are applied to convert the external (actual) exposure to the internal exposure. This is done per exposure route in order to compare the exposure with the AOEL (given in mg/kg bw). For inhalation and ingestion exposure a default absorption of 100% is used. The absorption value used for dermal exposure is substance-specific and also specific for pure and diluted formulations (e.g. different for ML and boom spraying scenarios). 133 The EFSA guidelines are used to take account of the variations. The software therefore allows users to enter their own values. (g) Tables Table 20: Personal protective equipment (PPE) and work wear reduction factors PPE or work wear item Hand exposure Protective (chemical resistant) gloves Gloves unspecified Gloves unspecified Latex/PE/Vinyl/PVC gloves Latex/PE/Vinyl/PVC gloves Nitrile gloves Nitrile gloves Body exposure Long sleeved shirt and trousers (Working clothing) Unspecified work clothing Unspecified work clothing Protective coverall Protective certified coverall Protection factor (by which exposure in absence of protection should be multiplied) Specific exposure value affected Type or source of value Liquids Solids 5% Dermal exposure hands only – Constant-EFSA 3.95% M/L Liquids 0.85% Solids 4.52% 0.00-97.07% M/L Liquids 0.00-19.17% Solids 0.00-56.93% 5.51% M/L Liquids: 0.18% 0.00-70.85% M/L Liquids: 0.00-4.05% 3.14% M/L Liquids 0.85% Solids 4.15% 0.00-97.07% M/L Liquids 0.00-19.17% Solids 0.00-56.93% Dermal exposure hands only – 75th percentile- BROWSE Dermal exposure hands only – Distribution- BROWSE Dermal exposure hands only – 75th percentile- BROWSE Dermal exposure hands only – Distribution- BROWSE Dermal exposure hands only – 75th percentile- BROWSE Dermal exposure hands only – Distribution- BROWSE 10% Dermal exposure body only – Constant-EFSA 7.2% Dermal exposure body only Dermal exposure body only Dermal exposure body only Dermal exposure body only – 75 percentile- BROWSE – Distribution- BROWSE – Constant-EFSA – Constant-EFSA 10% 0.01-71.31% 10% 5% 134 th PPE or work wear item Cotton coverall Protection factor (by which exposure in absence of protection should be multiplied) 9.60% Cotton coverall 0.01-50.07% Polyester-cotton coverall Polyester-cotton coverall Polyester-cotton coverall for highintensity crops (handheld spraying only) Polyester-cotton coverall for highintensity crops (handheld spraying only) Uncertified rain suit 4.25% Certified PPE 0.1% Head exposure Hood and visor 5% Hood 50% RPE mask type Half and full face masks Filter type FP1, P1 and similar FFP2, P2 and similar Inhalation exposure RPE mask Filter type type Half and FP1, P1 full face and masks similar FFP2, P2 and similar Specific exposure value affected Type or source of value Dermal exposure body only Dermal exposure body only Dermal exposure body only Dermal exposure body only Dermal exposure body only – 75 percentile- BROWSE – Distribution- BROWSE – 75th percentile- BROWSE – Distribution- BROWSE – 75th percentile- BROWSE 0.16-52.92% Dermal exposure body only – Distribution- BROWSE 0.2% Dermal exposure body only Dermal exposure body only – AM- BROWSE (indicative value) AM- BROWSE (indicative value) Dermal exposure head only Dermal exposure head only – Constant-EFSA – Constant-EFSA 80% Dermal exposure head only – Constant-EFSA 80% Dermal exposure head only – Constant-EFSA 25% Inhalation exposure Constant-EFSA 10% Inhalation exposure Constant-EFSA 0.05-71.31% 29.08% 135 – th Table 21: Distribution of percentage of dermal head exposure compared to dermal total exposure* Scenario Distribution of % head exposure compared to total dermal exposure (head, body, hands) Boom spraying 0-6% Orchard spraying 0-17% Hand held spraying high level 0-12% Hand held spraying low level 0-1% Mixing and loading 0-1% *the upper percentage is currently proposed and programmed in the software Table 22: Ingestion exposure - main user and model inputs No 1 User inputs n/a Model inputs Potential dermal exposure to a.s. in µg on both hands (Boom spraying) (Dph_bs) Lower Full distribution Upper - Based on BS hand model output 2 n/a M&L hand model output n/a Full distribution 2 Fixed 75 cm - 3 - 4 n/a Potential dermal exposure to a.s. in µg on both hands (Mixing & loading) (Dph_ml) Area of one hand making contact with the 2 mouth (cm ) (Am) 2 Area of both hands [cm ] (Ah) - 5 n/a Skin to mouth transfer factor (SE) Fixed 1070 2 cm Fixed 0.43 (14% from EPA, Michaud) Appendix I 6 PPE inputs (relevant for both BS and ML) Number of hand-to-mouth contacts# (events/hr) (i) if gloves or respirators are used (events/scenario) if gloves or respirators are not used (events/scenario) Exposure time of scenario in hrs (minutes/60; roundup=1). So N*t ≥1 roundup uniform uniform 0.1 1.0 0.5 Scenariospecific 5.0 - 7 - EPA, IOM data (Table 23) Cherrie et al 2007; Gorman Ng (in press) Exposure duration (per scenario) *different values may be proposed for the frequency of hand-to-mouth contacts to distinguish between PPE use and no PPE use #assuming primarily peri-oral contact Table 23: Analysis outputs to derive a skin-to-mouth transfer factor (SE) N AM SD GM Min 50th percentile IOM 12 0.14 0.31 -* -0.14 0.05 EPA 27 0.48 0.13 0.46 0.22 0.50 Combined 39 0.38 0.26 -* -0.14 0.43~ 136 IOM 0.22 0.33 0.64 0.93 1.00 75th percentile 90th percentile 95th percentile 99th percentile Max ~ * EPA 0.59 0.62 0.66 0.70 0.71 Combined 0.57 0.61 0.67 0.89 1.00 th The 50 percentile was used for the skin to mouth transfer factor (0.43) Due to negative values 3.4 Quality assurance and testing procedures (h) Exposure data For the inclusion or non-inclusion of exposure data collected from literature a decision tree was used (see Figure 1) for uptake of these data in the exposure database used for modeling purposes. Studies should contain information on inhalation, dermal (potential and/or actual) or biomonitoring exposure data, preferably at the level of individual measurements (raw data), but at least with summary statistics with regard to measured exposure levels (minimum, maximum and AM/GM, SD/GSD). Figure 19: Decision tree acceptance exposure data Exposure data B,R,O,W? Adequate description study design and monitoring methods Analysing methodology adequate described Contextual information on exposure determinants available Inclusion in database No inclusion No inclusion No inclusion For included studies, the study design is adequately described (with general information like year of study, location, target persons, site of application, indoor/outdoor, environmental conditions, etc.), including the methods used to measure exposure (whole body, patch techniques, etc.) and the sampling devices/material used. In principle, only studies in which personal measurements are performed are included (i.e. stationary measurements will only be used for scenarios where no personal measurements are available for operators. The guidelines (e.g. GLP) under which the study 137 is carried out are described. The number of study subjects measured, test sites etc. are clearly described. All the conditions of the study are reported so a decision on representativeness and relevance to the models to be developed can be made. As a minimum the study reports for included studies contain an outline or reference to description of the analytical method. The analytical method used should be a validated method. Ideally, the report should state if the data have been corrected for either for field or laboratory recoveries. Also how data <LOD or <LOQ are being handled is described properly. Although the data from the EUROPOEM database was not checked (see paragraph 6.2), based on the information available in the executive summaries of the studies and information from persons involved in the EUROPOEM project itself, the analytical quality of the study and the representativeness of the study were combined in an overall quality score, which has been documented in the database, together with the underlying remarks about the analytical quality, representativeness and other remarks. This information was taken into account when selecting data for the use during the model development, for which a tiered approach was used, depending on the availability of data for certain scenarios. The inclusion of data from studies was discussed within the project team. (i) Development of the exposure database All the exposure data that are entered in the database by a project member have been checked on good entry by another person from the project team and errors have been corrected. The starting point of the BROWSE exposure database was the EUROPOEM database, consisting of already entered data, including quality check. The data from the original EUROPOEM database have not been checked again, in part because it was known that this was already done extensively within the EUROPOEM project, and in part because we had no access to the a large part of the original study reports. 138 (j) Model development For the development of the mechanistic models all relevant literature has been reviewed and summarized. This information was used as input for model parameters and the resulting model algorithms and inputs were developed in the statistical package R (by FERA statistician). The programmed models and model outputs were checked by TNO. All the relevant contextual information in the database was translated and coded according to relevant model parameters required to test the models. In case of missing information, empirical sampling was performed to obtain imputed values. All model runs and testing was critically evaluated, checked and re-run if required to exclude errors in the models. (k) Software The programmed model in R was checked by FERA and directly used to develop the software. All code was peer reviewed within the software team and then tested by partners to ensure it performed as intended, before release to the BROWSE Advisory Panel and stakeholders for further testing. 139 4 Comparison with existing models The comparison with existing models is describes in “Operator exposure: Comparison of BROWSE model with the existing models: Annex to the Final WP1 Technical Report” by Charistou et al, 2014. 140 5 Model outcome interpretation and level of conservatism 5.1 Exposure outcomes Potential exposure estimates of the WP1 models are expressed differently depending on the route of exposure, e.g. inhalation (µg/m3), dermal (µg/body part) and ingestion (µg). These values are converted to mg/kg bodyweight for each exposure route, distinguishing between potential and actual exposure after taking account of the effect of PPE and/or work clothing. Subsequently these estimates are translated to the absorbed amount (mg/kg body weight) and the proportion of the AOEL or AAOEL. The model outputs can therefore relate to longer-term or acute estimates related to the AOEL or AAOEL (75th and 95th percentiles respectively). 5.2 Routes and sources of exposure included Where appropriate, the different routes of exposure (inhalation, dermal and ingestion) are included in the respective models. The dermal exposure pathways that are often neglected in exposure modeling, like bulk/splashes, are also included if relevant, for instance in the M&L dermal model. Also, model estimates always include both hand and body exposures. The models only estimates primary exposure (direct exposure due to work performed by the operator him/herself), and thus not from possible secondary sources of exposure (e.g. other application sources in the vicinity). 5.3 Exposure scenarios included The operator exposure models are developed for the following scenarios: • Boom spraying • Mixing & loading (liquids and solids) • Orchard spraying • Hand-held spraying Depending on input parameters like vehicle sprayer type, mixing and loading method, sprayed area, tank volume, the presence of a cabin, wearing PPE etc. these scenarios are further subdivided in order to make a fit-for purpose exposure assessment for reasonable worst-case and typical scenarios. 141 5.4 Population The models are considered relevant for both males and females and all age groups, although no explicit account was made to include differences between these population groups. An exception is where a distinction is made between body surface areas of males and females (in the dermal models). Although most of the available information on operator exposure, especially the exposure data, is based on studies with males, the mechanistic models with their underlying parameters are considered also applicable for estimating exposure of female operators, although some difference in actual working practices may exist. The models intend to cover operators that perform their work according to Good Agricultural Practice (GAP). Misuse during handling of PPPs is not taken into account. 5.5 Representativeness of model outcomes versus “real world” Model inputs are mostly obtained from experimental studies or evidence from literature. These model inputs (distributions) represent a range (with lower and upper values) of the best available data on ‘actual’ conditions in practice. Model parameter distributions are sometimes based on small-sized populations (or number of operators), limited experimental data (and conditions of use) and may not always include differences found between field studies or workplace settings. This will obviously increase the level of uncertainty in the model parameter inputs and model outputs. 5.6 Level of uncertainty and variability All the WP1 models are probabilistic and dependent on iterations of (distributions) of model input parameters. In turn, the models are capable of producing an exposure distribution. The median, 75th percentile, and 95th percentile are presented to the user in the software. Variability and uncertainty were quantified using distributions for a large number of internal model parameters (see WP1 Technical Report), and in some of the options for user inputs (body weight, breathing rate and PPE). Therefore the percentiles output by the model represent the effect of variability and uncertainty in all these factors on the exposure estimate. This output variance has also been referred to in this report when carrying out sensitivity analysis (SA). One of the outputs from SA is a table in which the relative contributions to the total output variation can be quantified and ranked to show their relative impact on the model output. 142 5.7 Model comparison as indication for level of conservatism An indication of the level of conservatism can be seen from the model testing as presented in this report. The models produce exposure estimates that are in the overall range of the available exposure data. However, these estimates do not always represent the required level of conservatism expected for regulatory risk assessment. On the other hand, a considerable part of the data used for testing is relatively old, and therefore the exposure levels in the available test set may be higher and thus less representative of current agricultural practices and conditions. Based on these considerations, and the fact that only a limited amount of recent exposure data was available during model development, it is assumed that the level of conservatism of the models is applicable for current agricultural practices and conditions. 5.8 Overall level of conservatism The level of conservatism in the models and final distributions of model inputs and defaults is reviewed (in collaboration with other work packages) to ensure an acceptable level of conservatism. 143 6 (a) Conclusion Rationale for the modeling approach chosen The aim of WP1 was to develop a single, new and improved modeling framework for operator exposure, to integrate all available exposure data and take explicit account of key factors and mechanisms influencing exposure. To meet this objective, it was decided to use a mechanistic and probabilistic modeling approach. Unfortunately WP1 was not able to get access to new operator exposure data. However, an advantage of mechanistic models is that exposure data is not required as direct input for model development, and that the model inputs can be improved and updated on a regular basis using new evidence. WP1 was also fortunate to get access to experimental data that could provide indicative input distributions in some of the models. For the most part, the data that were collated in the BROWSE database and used for testing / checking of the WP1 operator models mostly consisted of existing and old data (often dating back to the 1990s and earlier). This data (mostly EUROPOEM) has previously been found outdated, of low quality and not representative in the rapidly (technological) advancing agricultural sector. It is also often referred to as low quality because of high variability seen in the data itself. Whether this is true or not, it is clear that the high variability in operator exposure data is also evident in new data. An example is the modeling approach described in the development of a new (empirical) agricultural operator exposure model (AOEM) (BfR, 2003; Großkopf et al, 2013). The AOEM is based on new empirical data in which the determinants influencing exposure are unraveled and explicitly taken into account. The latter publication emphasizes the importance of more detailed modeling approaches that are required for operator exposure models. (b) Status of models The models presented in this report have not been thoroughly validated. In order to do so, new and good quality data is required. However, outputs of the respective models were tested / checked with (mostly outdated) datasets extracted from the BROWSE database. As complete datasets were not always available for testing, imputed datasets were also used. Compared to existing models, the BROWSE operator models include more model parameters and distinguish between more exposure routes or exposure types (e.g. both hand and whole body, ingestion). In general, the models produce exposure estimates that are in the overall range of the available exposure data. However, when plotting the confidence intervals for each model output, it is clear 144 that the model outputs do not always capture all the variance in the highly variable exposure data. Suppose the data used for testing can be considered acceptable, these model outputs may under- or over-estimate exposure. However, it is evident from the model testing plots that imputed data dominate the outliers in the data versus model comparisons. As imputed data is commonly (and often exclusively) used in the model testing, it was not possible to estimate the actual variance explained by the respective models. It is, however, evident that the whole body dermal models explain more variance than the hand models. Due to the fact that few ‘true datasets’ were available for testing, the effect of between-study variances etc. were not investigated. Also, no conclusions can be drawn from the testing results in terms of the selected (sub-) scenarios, e.g. whether it is acceptable to develop a single model for mixing & loading for all applications (field sprayers, broadcast sprayers, hand-held, aerial). (c) Use of models in regulatory framework With the development of the operator exposure models and the integration into the software, the aim is to contribute to the implementation of Regulation (EC) 1107/2009 and Dir. 2009/128/EC on the Sustainable Use of PPPs. Much effort was put into the integration of current agricultural practices in the EU from publications and survey data (from the BROWSE and EFSA survey data, e.g. Glass et al, 2012). With the implementation of the models in a user-friendly tool with sufficient flexibility for the user to make use of reasonable worst case defaults given in the software or the use of own values, the tool can be used to perform risk assessments in the framework of Regulation (EC) 1107/2009. Next to this the tool can identify risk indicators (for example the effect of the use of a cabin on exposure, drift-reducing nozzles, etc.) meeting the requirements of Dir. 2009/128/EC on the Sustainable Use of PPPs. 145 7 Recommendations for future development and refinement In order to keep the models up to date and refine them where necessary after thorough user feedback, the following steps can be considered. • A thorough analysis of all the available dermal exposure data is proposed to determine the differences in dermal sampling techniques (e.g. interception and removal techniques) and the effect on the variability found in exposure data. This might give insight into the effect of different dermal sampling techniques on dermal exposure, and a possible introduction of a correction factor • In case new, more recent, exposure data becomes available, this can be used to check and if necessary refine the models. For example recently the AOEM data became available so it would be good for the future to use this data to validation and refinement of the current models. • More extensive final sensitivity analyses of the algorithms can be performed. This could include different categorical scenario inputs. • An additional validation of the models could be performed using new data that has not been used for model development. 146 (a) Disclaimer The model comparisons described in a separate annex (see reference in chapter 4) and calculations included in the report are performed with software developed by FERA which is at this moment still under development. Therefore no rights can be derived from the results presented in this report. 147 8 References Abdelbagi, H., & Adams, A. (1987). Influence of droplet size, air-assistance and electrostatic charge upon the distribution of ultra-low-volume sprays on tomatoes. Crop Protection, 6, 226–233. Retrieved from http://www.sciencedirect.com/science/article/pii/0261219487900433 Berger-Preiss, E., Boehncke, A., Könnecker, G., Mangelsdorf, I., Holthenrich, D., & Koch, W. (2005). Inhalational and dermal exposures during spray application of biocides. International Journal of Hygiene and Environmental Health, 208(5), 357–72. doi:10.1016/j.ijheh.2005.04.006 Brouwer, D., Semple, S., Marquart, J., & Cherrie, J. (2001). A dermal model for spray painters. Part I: Subjective exposure modelling of spray paint deposition. Annals of Occupational …, 45(1), 15– 23. Retrieved from http://annhyg.oxfordjournals.org/content/45/1/15.short BROWSE. (2011). Prioritisation of scenarios and overall model objectives for operator, worker, bystander & resident exposure modelling and identification of data required from WP4. Cohen Hubal, E. a, Suggs, J. C., Nishioka, M. G., & Ivancic, W. a. (2005). Characterizing residue transfer efficiencies using a fluorescent imaging technique. Journal of Exposure Analysis and Environmental Epidemiology, 15(3), 261–70. doi:10.1038/sj.jea.7500400 Dosemeci, M. (2002). A Quantitative Approach for Estimating Exposure to Pesticides in the Agricultural Health Study. Annals of Occupational Hygiene, 46(2), 245–260. doi:10.1093/annhyg/mef011 FAO. (2001a). Guidelines on minimun requirements for agricultural pesticide application equipment. Volume one. FAO. (2001b). Guidelines on standards for agricultural pesticide application equipment and related test procedures, vol. 1. FAO. (2013). International Code of Conduct on the Distribution and Use of Pesticides Annotated list of Technical Guidelines for the implementation of the International Code of Conduct on the Distribution and Use of Pesticides, (January). Gil, E., Escolà, a., Rosell, J. R., Planas, S., & Val, L. (2007). Variable rate application of plant protection products in vineyard using ultrasonic sensors. Crop Protection, 26(8), 1287–1297. doi:10.1016/j.cropro.2006.11.003 Hughes, E. a, Flores, A. P., Ramos, L. M., Zalts, A., Richard Glass, C., & Montserrat, J. M. (2008). Potential dermal exposure to deltamethrin and risk assessment for manual sprayers: influence of crop type. The Science of the Total Environment, 391(1), 34–40. doi:10.1016/j.scitotenv.2007.09.034 Juste, F., Sanchez, S., Ibanez, R., Val, L., & Garcia, C. (1990). Measurement of spray deposition and efficiency of pesticide application in citrus orchards. Journal of Agricultural …, (August 1988), 148 187–196. Retrieved from http://www.sciencedirect.com/science/article/pii/S0021863405801258 Leonas, K. K., Yu, X. K., & Hall, D. (1992). Deposition Patterns on Garments during Application of Lawn and Garden Chemicals: A Comparison of Six Equipment Types, 234, 230–234. Machera, K. (2003). Determination of Potential Dermal and Inhalation Operator Exposure to Malathion in Greenhouses with the Whole Body Dosimetry Method. Annals of Occupational Hygiene, 47(1), 61–70. doi:10.1093/annhyg/mef097 Marquart, J. (2003). Determinants of Dermal Exposure Relevant for Exposure Modelling in Regulatory Risk Assessment. Annals of Occupational Hygiene, 47(8), 599–607. doi:10.1093/annhyg/meg096 Matuo, T., & Matuo, Y. K. (1998). Efficiency of Safety Measures Applied to a Manual Knapsack Sprayer for Paraquat Application to Maize ( Zea mays L .), 701, 698–701. Nuyttens, D., Braekman, P., Windey, S., & Sonck, B. (2009). Potential dermal pesticide exposure affected by greenhouse spray application technique. Pest Management Science, 65(7), 781–90. doi:10.1002/ps.1755 Rincón, V., Páez, F., & Sánchez-Hermosilla, J. (2009). Estimation of Leaf-Area Index in greenhouse tomato crop to determine the volume rate to spray. cirg.ageng2012.org. Retrieved from http://www.cirg.ageng2012.org/images/fotosg/tabla_137_C0402.pdf Sánchez-Hermosilla, J., Rincón, V. J., Páez, F., Agüera, F., & Carvajal, F. (2011). Field evaluation of a self-propelled sprayer and effects of the application rate on spray deposition and losses to the ground in greenhouse tomato crops. Pest Management Science, 67(8), 942–947. doi:10.1002/ps.2135 Sanjrani, W. (1990). Effect of application volume and method on spray operator contamination by insecticide during cotton spraying, 9(October). Smith, D., & Askew, S. (2000). Droplet size and leaf morphology effects on pesticide spray deposition. Transactions of the …, 43(2), 255–259. Retrieved from http://www.prairieswine.com/pdf/2987.pdf Uk, S. (1977). Tracing insecticide spray droplets by sizes on natural surfaces. The state of the art and its value. Pesticide Science, 8(5), 501–509. doi:10.1002/ps.2780080512 Van Zyl, S. a., Brink, J.-C., Calitz, F. J., & Fourie, P. H. (2010). Effects of adjuvants on deposition efficiency of fenhexamid sprays applied to Chardonnay grapevine foliage. Crop Protection, 29(8), 843–852. doi:10.1016/j.cropro.2010.04.017 Wicke, H., Ba, G., & Frie, R. (1999). Comparison of spray operator exposure during orchard spraying with hand-held equipment fitted with standard and air injector nozzles, 18, 0–7. 149 Zabkiewicz, J. a. (2007). Spray formulation efficacy—holistic and futuristic perspectives. Crop Protection, 26(3), 312–319. doi:10.1016/j.cropro.2005.08.019 Abdelbagi, H., & Adams, A. (1987). Influence of droplet size, air-assistance and electrostatic charge upon the distribution of ultra-low-volume sprays on tomatoes. Crop Protection, 6, 226–233. Retrieved from http://www.sciencedirect.com/science/article/pii/0261219487900433 ASAE S525. Agricultural cabs – Engineering Control of Environmental Air Quality. 2003. Balsari P, Marucco P, Oggero G. 2006. External contamination of sprayers in vineyards. Aspects of Applied Biology. 77: 1-6 Balsari P, Marucco P. 2003. Sprayer cleaning: The importance of the sprayer adjustment on the external contamination. ASAE Meeting Presentation, Paper number 031097. Berger-Preiss, E., Boehncke, A., Könnecker, G., Mangelsdorf, I., Holthenrich, D., & Koch, W. (2005). Inhalational and dermal exposures during spray application of biocides. International journal of hygiene and environmental health, 208(5), 357–72. doi:10.1016/j.ijheh.2005.04.006 Bundesinstitut für Risikobewertung (BfR). Joint Development of a new Agricultural Operator Exposure Model. Project report, Berlin 2013. ISBN3-943963-03-8 CEN (2004) prEN 15051 Workplace atmospheres- measurement of dustiness of bulk materialsRequirements and test methods. CEN/TC 137 N330 Comité Européen de Normalisation, Brussels, Belgium. Brouwer, D., Semple, S., Marquart, J., & Cherrie, J. (2001). A dermal model for spray painters. Part I: Subjective exposure modelling of spray paint deposition. Annals of Occupational …, 45(1), 15–23. Retrieved from http://annhyg.oxfordjournals.org/content/45/1/15.short Cherrie JW, Semple S, Christopher Y, Saleem A, Hughson GW, Philips A. 2006. How important is inadvertent ingestion of hazardous substances at work? Ann. Occ. Hyg. 50 (7): 693-704 150 CropLife International. Catalogue of pesticide formulation types and international coding system. CropLife International, Technical monograph No. 2, 6th Edition, revision May 2008, 2008. Christopher Y. 2008. Inadvertent ingestion exposure to hazardous substances in the workplace. PhD Thesis. Aberdeen, UK: University of Aberdeen. Department for Environment Food and Rural Affairs (DEFRA). SID5 (Rev. 07/10). 2011. Field spray drift studies to mature winter cereal crops with modern application practices to inform policy on setting of buffer zones in the UK. Research Project Final Report. Project Code: PS2017 Dosemeci, M. (2002). A Quantitative Approach for Estimating Exposure to Pesticides in the Agricultural Health Study. Annals of Occupational Hygiene, 46(2), 245–260. doi:10.1093/annhyg/mef011 US EPA. Human Health Evaluation Manual. Part E: Supplemental Guidance for Dermal Risk Assessment, RAG Vol. 1, 1999. Fransman W, van Tongeren M, Cherrie JW, Tischer M, Schneider T, Schinkel J, Kromhout H, Warren N, Goede H, Tielemans E. Advances Reach Tool (ART): Development of the mechanistic model. Ann. Occup. Hyg. 2011; 55 (9): 957-979. Gerritsen-Ebben MG, Brouwer DH, van Hemmen JJ. Effective Personal Protective Equipment (PPE). Default setting of PPE for registration purposes of agrochemical and biocidal pesticides. TNO report V7333. Zeist, 2007 Gil, E., Escolà, a., Rosell, J. R., Planas, S., & Val, L. (2007). Variable rate application of plant protection products in vineyard using ultrasonic sensors. Crop Protection, 26(8), 1287–1297. doi:10.1016/j.cropro.2006.11.003 Gilbert AJ, Wild SA, Mathers JJ, Glass CR. Measurement of spillage and contamination arising from the progressive emptying of a 10 litre container and evaluation of the influence of pack size on hand contamination. York: Central Science laboratory; 1999. Report nr FD 98/91. 151 Gilbert, AJ, C R Glass, R Lewis, JJ Mathers, F Mazzi, S A Wild. 2000. Determination of operator and environmental contamination with simulated pouring studies into an induction bowl. Project Number PA1701, Report Number FD 99/61. Glass CR, Mathers JJ, Lewis RJ, Harrington PM, Gilbert AJ, Smith S. Understanding exposure to agricultural pesticide concentrates. York: Central Science Laboratory; 2009?. Report nr HSE Contract Ref: 4030/R51.193; DEFRA Contract Ref: PA1722. Glass R, Garthwaite D, Pote A, et al. Collection and assessment of data relevant for non-dietary cumulative exposure to pesticides and proposal for conceptual approaches for non-dietary cumulative exposure assessment. External scientific report CFT/EFSA/PPR/2010/04. Supporting publications 2012:EN-346. Gorman Ng M, Semple S, Cherrie JW, Christopher Y, Northage C, Tielemans E, Veroughstraete V, van Tongeren M. 2012. The relationship between inadvertent ingestion and dermal exposure pathways: A new integrated conceptual model and a database of dermal and oral transfer efficiencies. Ann. Occ. Hyg. Published Online First: 23 July 2012. doi:10.1093/annhyg/mes041 Großkopf C, Mielke H, Westphal D, Erdtmann-Vourliotis M, Hamey P, Bouneb F, Rautmann D, Stauber F, Wicke W, Maasfeld W, Salazar JD, Chester G, Martin S. A new model for the prediction of agricultural operator exposure during professional application of plant protection products in outdoor crops. Journal of Consumer Protection and Food Safety: Published online: DOI 10.1007/s00003-013-0836-x Hubal, Cohen, Suggs, J. C., Nishioka, M. G., & Ivancic, W. a. (2005). Characterizing residue transfer efficiencies using a fluorescent imaging technique. Journal of exposure analysis and environmental epidemiology, 15(3), 261–70. doi:10.1038/sj.jea.7500400 Hughes, E. a, Flores, A. P., Ramos, L. M., Zalts, A., Richard Glass, C., & Montserrat, J. M. (2008). Potential dermal exposure to deltamethrin and risk assessment for manual sprayers: influence of crop type. The Science of the total environment, 391(1), 34–40. doi:10.1016/j.scitotenv.2007.09.034 152 Institute of Occupational Medicine (IOM). Database of Dermal and Oral Transfer Efficiencies. (http://www.iom-world.org/research/research-expertise/exposure-assessment/database-of-dermaland-oral-transfer-efficiencies/) Juste, F., Sanchez, S., Ibanez, R., Val, L., & Garcia, C. (1990). Measurement of spray deposition and efficiency of pesticide application in citrus orchards. Journal of Agricultural …, (August 1988), 187– 196. Retrieved from http://www.sciencedirect.com/science/article/pii/S0021863405801258 Kennedy, M. C., Anderson, C. W., Conti S., and O’Hagan. (2006). Case Studies in Gaussian Process Modelling of Computer Codes. Reliability Engineering and System Safety. 91, 1301—1309. Lebailly P, Bouchart V, Baldi I, Lecluse Y, Heutte N, Gislard A, Malas JP. 2009. Exposure to pesticides in openfield farming in France. Ann Occ Hyg. 53(1): 69-81 Leonas, K. K., Yu, X. K., & Hall, D. (1992). Deposition Patterns on Garments during Application of Lawn and Garden Chemicals: A Comparison of Six Equipment Types, 234, 230–234. Lloyd GA, Bell GJ. Hydraulic nozzles: Comparative Spray Drift study. Ministry of Agriculture, Fisheries and Food (MAFF). United Kingdom. SC7704, 1988 Machera, K. (2003). Determination of Potential Dermal and Inhalation Operator Exposure to Malathion in Greenhouses with the Whole Body Dosimetry Method. Annals of Occupational Hygiene, 47(1), 61–70. doi:10.1093/annhyg/mef097 Marquart J, Lansink C, Engel R, van Hemmen J. 1999. Effectiveness of local exhaust ventilation during dumping of powders from bags. TNO report v99.267, Zeist, The Netherlands Marquart, J. (2003). Determinants of Dermal Exposure Relevant for Exposure Modelling in Regulatory Risk Assessment. Annals of doi:10.1093/annhyg/meg096 153 Occupational Hygiene, 47(8), 599–607. Mathers JJ, Glass CR, Gilbert AJ, Wild SA. Determination of operator and environmental contamination with simulated pouring studies into an induction bowl. York: Central Science Laboratory; 2000. Report nr FD 99/61. Matuo, T., & Matuo, Y. K. (1998). Efficiency of Safety Measures Applied to a Manual Knapsack Sprayer for Paraquat Application to Maize ( Zea mays L .), 701, 698–701. Michielsen JMGP, van de Zande JC, Wenneker H, Stallinga H, Van Velde P. 2012. External loading of an orchard sprayer with agrochemicals during application. Aspects of Applied Biology. 114: 151-157 Nuyttens, D., Braekman, P., Windey, S., & Sonck, B. (2009). Potential dermal pesticide exposure affected by greenhouse spray application technique. Pest management science, 65(7), 781–90. doi:10.1002/ps.1755 Ramwell CT, Johnson PD, Boxall ABA, Rimmer DA. 2004. Pesticide residues on the external surfaces of field-crop sprayers: environmental impact. Pest Management 60:795–802 (SciencePest Manag Sci (online: 2004) Ramwell CT, Johnson PD, Boxall ABA, Rimmer DA. 2005. Pesticide Residues on the External Surfaces of Field Crop Sprayers: Occupational Exposure. Ann. Occup. Hyg., Vol. 49, No. 4, pp. 345–350 Rincón, V., Páez, F., & Sánchez-Hermosilla, J. (2009). Estimation of Leaf-Area Index in greenhouse tomato crop to determine the volume rate to spray. cirg.ageng2012.org. Retrieved from http://www.cirg.ageng2012.org/images/fotosg/tabla_137_C0402.pdf Sánchez-Hermosilla, J., Rincón, V. J., Páez, F., Agüera, F., & Carvajal, F. (2011). Field evaluation of a self-propelled sprayer and effects of the application rate on spray deposition and losses to the ground in greenhouse tomato crops. Pest doi:10.1002/ps.2135 154 Management Science, 67(8), 942–947. Sanjrani, W. (1990). Effect of application volume and method on spray operator contamination by insecticide during cotton spraying, 9(October). Schinkel J, Warren N, Fransman W, van Tongeren M, McDonnell P, Voogd E, Cherrie JW, Tischer M, Kromhout H, Tielemans E. Calibration of the ART mechanistic model. J. Environ. Monit. 2011; 13: 1374–82. Schneider T, Vermeulen R, Brouwer DH, Cherrie JW, Kromhout H, Fogh CL. Conceptual model for assessment of dermal exposure. Occup Environ Med 1999;56:765–773 Smith, D., & Askew, S. (2000). Droplet size and leaf morphology effects on pesticide spray deposition. Transactions of the …, 43(2), 255–259. Retrieved from http://www.prairieswine.com/pdf/2987.pdf Wild SA, Mathers JJ, Glass CR. 2000. The potential for operator hand contamination to pesticides during the mixing and loading procedure. Aspects of Applied biology, 57: 179-183 Tielemans E, Schneider T, Goede H, Tischer M, Warren N, Kromhout H, van Tongeren M, van Hemmen JJ, Cherrie JW. Conceptual model for assessment of inhalation exposure: defining modifying factors. Ann. Occup. Hyg. 2008; 52 (7): 577-586. Van de Zande JC. Inventarisatie externe verontreiniging spuitapparatuur. Plant Research International BV, Wageningen. Nota 470. 2007. Van Zyl, S. a., Brink, J.-C., Calitz, F. J., & Fourie, P. H. (2010). Effects of adjuvants on deposition efficiency of fenhexamid sprays applied to Chardonnay grapevine foliage. Crop Protection, 29(8), 843–852. doi:10.1016/j.cropro.2010.04.017 Uk, S. (1977). Tracing insecticide spray droplets by sizes on natural surfaces. The state of the art and its value. Pesticide Science, 8(5), 501–509. doi:10.1002/ps.2780080512 155 United States Environmental Protection Agency (USEPA). Occupational and residential exposure test guidelines. United States Environmental Protection Agency (US-EPA); 1996. Report nr OPPTS 875.1100. United States Environmental Protection Agency (USEPA). Human Health Evaluation Manual. Part E: Supplemental Guidance for Dermal Risk Assessment, RAG Vol. 1, 1999. Veldhof RJG, Lurvink MWM, de Vreede SAF, Gijsbers JHJ, Tjoe Nij E, Brouwer DH. 2006. Dermal exposure to dust during manual transfer of powder; the effect of amount handled and dustiness. TNO report V5816, Zeist, The Netherlands Wicke, H., Ba, G., & Frie, R. (1999). Comparison of spray operator exposure during orchard spraying with hand-held equipment fitted with standard and air injector nozzles, 18, 0–7 Zabkiewicz, J. a. (2007). Spray formulation efficacy—holistic and futuristic perspectives. Crop Protection, 26(3), 312–319. doi:10.1016/j.cropro.2005.08.019 156
© Copyright 2026 Paperzz