Ecological Modelling 176 (2004) 291–312 A simple tropical ecosystem model of carbon, water and energy fluxes Silvia N. Monteiro Santos, Marcos Heil Costa∗ Department of Agricultural Engineering, Federal University of Viçosa, Av. P. H. Rolfs, s/n, Viçosa MG 36571-000, Brazil Received 13 August 2002; received in revised form 14 October 2003; accepted 29 October 2003 Abstract A simple tropical ecosystem model (SITE) was developed to study the response of tropical ecosystems to environmental conditions. SITE fills the niche of an ecosystem model of intermediate complexity, sophisticated enough to be used to study the fast dynamics of tropical ecosystems, while simple enough to be used to introduce the ecosystem modelling concepts to students and inexperienced modellers. SITE is a dynamic model that incorporates several processes: canopy infrared radiation balance, solar radiation balance, aerodynamic processes, canopy physiology and transpiration, balance of water intercepted by the canopy, transport of mass and energy in the atmosphere, soil heat flux, soil water flux and carbon balance. It is structured with one canopy layer and two soil layers, and is forced by hourly data of temperature, radiation balance, precipitation, humidity and wind, and simulates the fluxes of CO2 , water and energy, as well as the dynamics of carbon in the ecosystem. For calibration and validation, we used fluxes of CO2 , water vapour and sensible heat, measured at a primary evergreen forest site in Eastern Amazonia, Brazil. Even though SITE is considerably less complex than other models of similar goals, it reproduces well the hourly variability of the fluxes of CO2 and water vapour, and simulates the seasonal scale balance of those elements properly. SITE is available as a 1200-line FORTRAN code, and as a computer spreadsheet. We believe the model will be useful to help train the next generation of tropical ecologists in the use of ecosystem models. © 2004 Elsevier B.V. All rights reserved. Keywords: Tropical forest; Ecosystem model; Carbon flux; Water vapour flux; Biosphere–atmosphere interaction 1. Introduction The interannual and interdecadal variability in climate, and other changes in the environment, like rising atmospheric CO2 concentration and large-scale changes in land cover, have motivated several studies about the behaviour of ecosystems in a changing environment. Such studies lead to the development of ∗ Corresponding author. Tel.: +55-31-3899-1899; fax: +55-31-3899-2735. E-mail addresses: [email protected], [email protected] (M.H. Costa). several numerical models to understand the effects of these changes on the carbon, water and energy fluxes between the ecosystems and the atmosphere. These fluxes are strongly coupled, so the integrated representation of the carbon, water and energy balance components is fundamental to the understanding of the functioning of an ecosystem. Hurtt et al. (1998) grouped terrestrial ecosystem models developed to study global changes according to their objectives: biogeochemical models (type I) like CENTURY (Parton et al., 1988), BGC (Running and Gower, 1991), and TEM (Raich et al., 1991; Melillo et al., 1993); biophysical models (type II) like BATS 0304-3800/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2003.10.032 292 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 (Dickinson et al., 1984), SiB (Sellers et al., 1986) and LSX (Pollard and Thompson, 1995); and biogeographical models (type III) like DOLY (Woodward, 1987), BIOME (Prentice et al., 1992; Haxeltine and Prentice, 1996), and MAPSS (Neilson, 1995). However, there are integrated models that fit in more than one category, like IBIS (Foley et al., 1996), that would be classified as type I + II + III, or LSM (Bonan, 1996), type I + II. Models whose objectives spread through more than one category allow a deeper study of the interactions between biosphere and atmosphere, in particular the several feedback mechanisms involved. The development of these models let several authors to discuss the potential effects of the changes of the vegetation cover on the regional and global climate, concentrating on the dynamical evolution of the ecological, biophysical, biogeochemical and biogeographical processes that happen in different time scales (Hurtt et al., 1998). Process-based models, when used in conjunction with data collected at micrometeorological sites, are really useful tools to understand the functioning of an ecosystem. However, most comprehensive ecosystem models are extremely sophisticated, theoretically and mathematically, and are usually seen as “black boxes” by the users. In this work we develop a model of intermediate complexity to estimate carbon, water vapour and energy fluxes between a tropical forest ecosystem and the atmosphere. In order to appropriately represent the functioning of the ecosystem, the model considers the main physical, chemical and biological processes involved in the mentioned fluxes. The model fills the niche of an ecosystem model of intermediate complexity, sophisticated enough to be used to study the fast dynamics of tropical ecosystems, while simple enough to be used to introduce the ecosystem modelling concepts to students and inexperienced modellers. 2. Model description The main purpose of Simple Tropical Ecosystem Model (SITE) is to reliably simulate mass and energy fluxes between the ecosystem and the atmosphere in a simple way. According to the model classification of Hurtt et al. (1998), SITE is classified as I + II type, the same category of LSM (Bonan, 1996). Users of SITE include researchers that want quick modelling analy- ses but, due to its relative simplicity, it is a fine model to be used in intermediary-level modelling courses of the atmosphere-biosphere interaction. The model reconciles simplicity with a rigorous treatment of the several physical, chemical and biological processes involved: canopy solar and infrared radiation balances, aerodynamic processes, canopy physiology and transpiration, canopy water balance, mass and energy transport in the atmosphere, soil heat flux, soil water flux and carbon balance. SITE is based on previously developed models, mainly LSX (Pollard and Thompson, 1995), LSM (Bonan, 1996), IBIS (Foley et al., 1996), and SiB2 (Sellers et al., 1996), being much simpler, though. The strategy for simplification included the following steps: (a) a careful study of these models and of the most recent modelling literature; (b) among the several modelling alternatives considered, selection of the simplest methodology used to simulate each process; (c) elimination of unnecessary model elements; and finally (d) simplification of the mathematics and numerical methods used. Steps (c) and (d) were repeated several times, advancing and retroceding in the simplification process to evaluate the quality of the simulation, compared against observed data, the numerical stability of the solution, and the respect to the physical principles. It was also assumed that the model would be used in heavily instrumented areas like micrometeorological sites, where there is a good availability of data that could be used for model calibration and validation. Some intended characteristics and uses of the model defined some important simplifications. For example, the intended restricted use of the model for tropical regions allowed the elimination of snow and ice thermodynamics. The assumption that the model will be used in instrumented areas, where the albedo is known, permitted important simplifications in the solar radiation equations. SITE is a dynamical point model that uses an integration time step (dt) of one hour, representing a point of land totally covered by an evergreen broadleaf forest. Small modifications may be necessary for the representation of other tropical ecosystems. The model is structured with one layer of canopy and two layers of soil (Fig. 1). It is forced by eight hourly data measured above the canopy: air temperature (T) and specific humidity (q), horizontal wind S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 56 that depends on the leaf and stem density. The components of the infrared radiation balance are shown in Fig. 1. Iu , Is and Ig are the net infrared radiation flux absorbed by leaves, stems and soil surface (W m−2 ); I↑ and I↓ are the upward and downward infrared radiation flux below tree level per overall area (W m−2 ); Ia is the downward infrared radiation flux from the atmosphere (W m−2 ) and I is the net long wave radiation measured above the canopy. The solutions for the calculation of infrared radiation balance for each component of the canopy are similar to those used in the LSX model (Pollard and Thompson, 1995, Appendix 4, Eqs. (A18)–(A25)). za 50 Ia z1 height (m) 40 z12 30 layers u and s z2 Iu Is I I 20 10 0 293 Ig zg layer g zd layer d 2.2. Canopy solar radiation balance Fig. 1. Schematic representation of the model and the infrared radiation balance: u, s, g, and d, refer to leaves, stems, ground soil layer and deep soil layer, respectively. speed (u), incident short wave radiation (S), net long wave radiation (I), albedo (α), precipitation (P) and atmospheric pressure (Patm ). The model also uses parameters of the biophysical characteristics of the vegetation (Table 1). The main output variables of the model are latent heat flux, sensible heat flux, water vapour flux and net ecosystem production. The next sections present a detailed description of each module of the model. 2.1. Canopy infrared radiation balance The infrared radiation is treated as if each vegetation level is a semi-transparent plane with an emissivity The solutions for the calculation of the canopy solar radiation balance used in most models (e.g., Dickinson et al., 1984; Sellers et al., 1986, 1996; Pollard and Thompson, 1995; Bonan, 1996) are based on the individual representation of direct and diffuse fluxes, usually dividing calculations between visible and near infrared bands. These solutions are mathematically very complex, and are beyond the intended uses of this model. Considering that one of the objectives of this work is the development of a simple model, that the model is designed to be used in experimental areas where the albedo is assumed to be known, and that there is a significant uncertainty on the canopy transmissivity values, we propose a simple mathematical solution of the problem. Assuming that all reflection occurs on the top of the canopy, which requires a dense evergreen forest, the equations for the canopy solar radiation balance are independent of the calculations of direct and diffuse solar radiation for the visible and near infrared bands. Despite this simplification, the Table 1 Vegetation biophysical parameters Variable Symbol Used value Source Specific leaf area (m2 leaf kg−1 C) Stem area index (m2 m−2 ) Top height of upper canopy (m) Bottom height of upper canopy (m) Atmospheric roughness length for the canopy (m) Atmospheric roughness length for the soil surface (m) Zero-plane displacement (m) Intermediate height of upper canopy (m) Sl SAI z1 z2 zoa zog d z12 13 1 40 30 2.35 0.005 30.0 32.35 Medina and Cuevas (1996); Roberts et al. (1996) Estimated Measured on site Measured on site Shuttleworth (1988) Shuttleworth et al. (1984) Carswell et al. (2002) d + zoa 294 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 S Sr z u za logarithmic profile of the wind – high z0 Su Ss u1 z1 z12 d z2 u12 u2 u2g Sg Fig. 2. Schematic representation of the solar radiation balance. proposed treatment conserves energy throughout the canopy. The components of the canopy solar radiation balance are shown in Fig. 2, where Su , Ss and Sg are the solar radiation flux absorbed by leaves, stems and the soil surface (W m−2 ), respectively, and are computed using Eqs. (1)–(3): Su = (1 − α − τu )S (1) Ss = τu S(1 − τs ) (2) Sg = τu τs S (3) where Sr is the solar radiation reflected by the canopy (Sr = αS), τ u and τ s are the leaf and stem transmissivities. This assumption implies that the radiation reflected by the ground is negligible, which usually requires that Sg S (τ u τ s ∼ 0). The more open the canopy is, and the higher the ground albedo is, the less valid Eqs. (1)–(3) are. However, this assumption may also be valid in ecosystems other than a dense tropical forest. For example, if the ground albedo is very low (<0.1, like in a wetland) and the canopy is relatively open (τ u τ s ∼ 0.1), the error in the radiation absorbed by the soil surface is still on the order of 1% of S. 2.3. Aerodynamic processes Following several authors (Pollard and Thompson, 1995; Sellers et al., 1996; Campbell and Norman, 1998), we assumed the vertical wind profile shown in Fig. 3. Between the levels za and z1 , airflow is as- z0a1 exponential profile of the wind logarithmic profile of the wind – low z0 z02g 0 u Fig. 3. Schematic representation of the wind profile. sumed to be predominantly turbulent with a high value of aerodynamic roughness, with the horizontal wind speed decreasing logarithmically from za to z1 . The wind profile decreases exponentially from z1 downwards to z2 , and a logarithmic wind profile is assumed between the levels z2 and the surface, with a zero-plane displacement equal to zero and a low value of aerodynamic roughness, characteristic of the soil surface (Table 1). These conditions imply a wind speed nearly constant between z2 and z = 0 levels. Using diabatic correction factors to account for thermally induced turbulence (Campbell and Norman, 1998, pp. 96–97), the wind speed at different levels can be calculated using Eqs. (4)–(8). In the regions where the wind profile is logarithmic, if the wind speed in the upper part of the profile, and the values of d and zo are known, the friction velocity of the profile (a1 or 2g) is determined by Eq. (4) or (8). Subsequently, wind speed at any height in the region (e.g., z1 ) can be calculated by Eq. (5). In the region where the wind profile is exponential, if the wind speed at the top of the region and the attenuation coefficient are known, the speed at any level can be calculated by Eqs. (6) and (7). Friction velocity in the profile a1 : 0.41u u∗a1 = [ln((za − d)/zoa ) + ψma1 ] (4) S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 Wind speed at level z1 : u∗ u1 = a1 [ln((z1 − d)/zoa ) + ψma1 ] 0.41 295 q12 (5) ru=1/su Wind speed at level z12 : z12 −1 u12 = u1 exp a z1 Wind speed at level z2 : z2 u2 = u1 exp a −1 z1 Friction velocity in the profile 2g : 0.41u2 u∗2g = [ln(z2 /zog )ψm2g ] (6) rs (7) qu (8) Fig. 4. Schematic representation of the leaf transpiration. 2.4. Canopy physiology and transpiration This module of SITE is based on the equations proposed by Farquhar et al. (1980), Farquhar and Sharkey (1982), Collatz et al. (1991, 1992), Sellers et al. (1996), Foley et al. (1996) and Campbell and Norman (1998). The transpiration per unit area of leaf is calculated dividing the difference of the air specific humidity by the resistance to the water vapour flux (Fig. 4). (qu − q12 )LAI (1/su ) + (1/gs ) qu – q12 rs + ru rs=1/gs where a is the wind coefficient for a forest and ψmi is diabatic correction factors for momentum, where i denote the a1 and 2g profiles. The turbulent transfer coefficient atmosphere– atmosphere (σ a , m s−1 ) is calculated integrating the flux between the levels z12 and za (Fig. 6) (Campbell and Norman, 1998, pp. 98). The soil–atmosphere turbulent transfer coefficient (σ g ) is calculated similarly. The diffusive transfer coefficients between leaves and the atmosphere (su , m s−1 ) and between stems and the atmosphere (ss , m s−1 ) are calculated using empirical approximations based on wind tunnels experiments (Campbell and Norman, 1998, pp. 101). Etu = Etu = (9) where Etu is the transpiration (kg H2 O m−2 s−1 ) and gs is the canopy conductance. In Fig. 4, the canopy resistance (rs = 1/gs ) is controlled by the degree of opening of the stomata, which is associated to several factors like CO2 concentration, temperature, humidity, and the whole physiological process in the plant. The canopy boundary layer resistance (ru = 1/su ) depends directly on the wind and on leaf characteristics, such as structure, size and shape, as described in Section 2.3. Etu assumes only positive values, i.e., Etu = 0 if q12 > qu . The stomata are parameterised by a connected photosynthesis–conductance scheme, which considers the diffusion of CO2 between leaf and air. The photosynthesis rate is simulated according to the Farquhar equations (Farquhar et al., 1980; Farquhar and Sharkey, 1982). CO2 diffusion from atmosphere to chloroplasts does not occur as a gas, but in liquid phase, as CO2 molecules are dissolved in water in the cell vicinity (Fig. 5). The CO2 diffusion through the stomata is described by the equations of Collatz et al. (1991). The net photosynthesis rate (An ) or dry matter production is the balance between the processes of absorption (Ag ) and release of CO2 (Ru ). The model considers that the gross photosynthesis (Ag ) is the minimum between two potential capacities (Je and Jc ): Ag = min(Je , Jc ) (10) The gross photosynthesis rate limited by light (mol CO2 m−2 s−1 ) is expressed as: CO2i − Γ∗ Je = α3 APAR (11) CO2i + 2Γ∗ 296 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 active radiation (mol photons m−2 s−1 ). The gross photosynthesis rate limited by the activity of the Rubisco enzyme (mol m−2 s−1 ) is given by: CO2a rb=1/gb Vm (CO2i − Γ∗ ) CO2i + Kc (1 + ([O2 ]/Ko )) Jc = An CO2s rs=1/gs 1.6 rs CO2i Fig. 5. Representation of the net CO2 flux through the plant stomata. where Γ∗ = [O2 ]/2τ is the compensation point for gross photosynthesis (mol mol−1 ); [O2 ] is the oxygen concentration, equal to 0.210 mol mol−1 ; CO2i is the CO2 concentration in the intercellular spaces of the leaves (mol mol−1 ); α3 is the intrinsic quantum efficiency for C3 plants (mol CO2 mol−1 photons); τ is the CO2 /O2 ratio of kinetic parameters; and APAR is the density flux of the absorbed photosynthetically (12) where Kc and Ko are Michaelis constants, the first constant is used to fix CO2 and the second to inhibit oxygenation (mol mol−1 ), respectively. The leaf respiration (mol CO2 m−2 s−1 ) is calculated as: Ru = γVm Tm (13) where γ is the leaf respiration coefficient of the Rubisco enzyme (Collatz et al., 1991). The maximum Rubisco enzyme capacity (Vm , mol CO2 m−2 s−1 ) for the carboxylase function is calculated through the Vmax parameter and attenuated by soil moisture stress (St) and temperature stress (Tvm ). Vm = Vmax Tvm St St = 1 − exp(Stfac awc) 1 − exp(Stfac ) (14) (15) where Stfac is the coefficient for soil moisture stress and awc is the available water content in the soil. The physiological parameters used by the model are shown in Table 2. Table 2 Physiologic parameters used in the SITE model, where the values of τ m , Kc , Ko , and Tvm are obtained from the respective equation, and the other values (Vmax , b, m, α3 , γ) are obtained from the literature and optimised during the calibration process Variable Symbol CO2 /O2 ratio of kinetic parameters τ Michaelis constants (fix CO2 ) (mol mol−1 ) Kc Michaelis constants (inhibit O2 ) (mol mol−1 ) Ko Factor of temperature stress (K) Tvm Maximum Rubisco enzyme capacity (mol CO2 m−2 s−1 ) Coefficient for minimum stomatal conductance when net photosynthesis is zero broadleaf trees (mol CO2 m−2 s−1 ) Coefficient for stomatal angular conductance for broadleaf trees Intrinsic quantum efficiency for C3 plant (mol CO2 mol−1 photons) Leaf respiration coefficient of the broadleaf trees Vmax b 75 × 10−6 0.010 m α3 γ 10 0.060 0.0150 Tf = Tu − 273.16. Source: Collatz et al. (1991, 1992); Leuning (1995). Used value 1 1 4500exp −5000 − 288.16 Tf 1 1 1.5 × 10−4 exp 6000 − 288.16 Tf 1 1 0.25 exp 1400 − 288.16 Tf exp[3500((1/288.16) − (1/Tf ))] [1 + exp(2 − 0.40Tf )][1 + exp(0.40Tf − 20)] S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 297 2.5. Balance of water intercepted by the canopy T q The balance of water stored in the leaves (Wu ) and stems (Ws ) (kg H2 O m−2 ) surfaces is described through linear differential equations (Pollard and Thompson, 1995): dWu Wu = Pu − Eiu − dt τdrip (16) dWs Ws = Ps − Es − dt τdrip (17) E H Eu Hu Tu Wu Ts Ws T12 q12 ss Eg Hg where τ drip is a time constant for leaf and stem liquid drip (12 h); Wu /τ drip (or dripu ) represents the dripping of water stored on the leaves; and Ws /τ drip (or drips ) represents the dripping of water stored on the stems. The precipitation rate intercepted by leaves (Pu ) and stems (Ps ) is given by: Pu = P(1 − e−0.5LAI ) (18) Ps = P(1 − e−0.5SAI ) (19) where P is the pluviometric precipitation rate measured above the canopy (kg H2 O m−2 s−1 ).The evaporation of the water intercepted by the leaves, Eiu , and stems, Es (kg H2 O m−2 s−1 ) are given by: Eiu = fwetu su (qu − q12 )LAI (20) Es = fwets ss (qs − q12 )SAI (21) where fwetu = Wu /Wumax is the wet fraction of upper canopy leaf area and fwets = Ws /Wsmax is the wet fraction of upper canopy stem area. When qu < q12 , dew is allowed to condense. In this case, the water flux is not proportional to the wet fraction of leaves/stems, and the condensation is not allowed to exceed the storage capacity of leaves and stems. In Eqs. (20) and (21), it is assumed that only the upper part of leaves and stems are wet. The leaf evapotranspiration flux (Eu ) is equal to the sum of evaporation of the water stored on the leaf surface and transpiration. Eu = Etu + Eiu su Es Hs σa (22) Maximum interception of water by the canopy leaf area, Wumax , and by the stems, Wsmax (kg H2 O m−2 ), are calculated assuming that the maximum height of water stored in the leaves or stems is 0.1 mm, or 0.1 kg per m2 of leaf. Gg σg Tg qg Fig. 6. Schematic representation of the mass and energy transport in the atmosphere. 2.6. Water and energy transport in the atmosphere The parameterisations of mass and energy transport in the biosphere–atmosphere are based on the mass and energy conservation principles. Prognostic temperature of leaves (Tu ), stems (Ts ), soil surface (Tg ), and air in the middle of the leaves (T12 ) are calculated through the balance of energy, according to Fig. 6 and Eqs. (23)–(25). (Ci + cw Wi ) dTi = Si + I i − H i − L v Ei dt dTg = Sg + Ig − Hg − Lv Eg − Gg dt dT12 C12 = Hg + H u + H s − H a dt Cg (23) (24) (25) where the subscript i denote leaves (i = u) or stems (i = s); cw is the specific heat of water; Lv is the latent heat of vaporisation of water; Lv Ei is the latent heat flux (W m−2 ); Ci is the heat capacity of leaves or stems (J m−2 K−1 ); Cg is the heat capacity of the top soil layer (J m−2 K−1 ) and Gg is the heat flux into the soil, calculated in Section 2.8. 298 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 The calculation of sensible heat flux requires the computation of the components of the canopy Hu and Hs , of the soil, Hg and of the atmosphere, H Hu = ρcp su (Tu − T12 )2LAI (26) Hs = ρcp ss (Ts − T12 )2SAI (27) H = ρcp σa (T12 − T) (28) Hg = ρcp σg (Tg − T12 ) (29) Pg SR zg +ss (Ts − T12 )2SAI = 0 (30) Air specific humidity inside of the canopy, q12 (kg kg−1 ), is calculated using the principle of water mass conservation: dq12 (31) = Eg + Eu + Es − E dt where Eg = σg ρ(qg − q12 ) is the evaporative flux of water from the soil, and E = σg ρ(q12 − q) is the total evaporation flux from canopy into the atmosphere. Neglecting the variation of specific humidity of the air inside of the canopy as a function of time, the following solution is given: eg Fd ed zd Neglecting the heat capacity of the air inside of the canopy (C12 ), substituting Eqs. (26)–(29) in (25) and dividing by ρcp , Eq. (25) can be rewritten as: σa (T − T12 ) + σg (T12 − Tg ) + su (Tu − T12 )2LAI Fg D Fig. 7. Schematic representation of the soil water flux. to satisfy the conditions above. Sensitivity tests were conducted to verify the numeric stability of the sensible heat flux of the soil (Hg ), and the thickness eg = 7.5 cm was chosen. Obviously, the use of more soil layers would improve the representation of the physical processes in the soil, but it leads to an undesirable additional complexity. Finally, we should point out that the 7.5 cm layer g includes, in addition to the soil, a thin layer of litter. The balance of water stored in the layers g and d of the soil is calculated applying the water mass conservation in each layer (32) dWg = Fg − Fd − Eg dt (33) where qg is the specific humidity of air in the soil pores (kg kg−1 ). dWd = Fd − D − Etu dt (34) 2.7. Soil water flux where Fg is the infiltration into the layer g; Fd is the infiltration into the layer d; and D = 1 × 10−8 mm s−1 is the deep drainage below the layer d (boundary condition). The precipitation rate intercepted by the soil (Pg , kg H2 O m−2 s−1 ) reaches the soil through direct precipitation through canopy openings (P − Pu − Ps ) and through dripping from leaves and stems (Bonan, 1996). σa (q − q12 ) + σg (qg − q12 ) + Eu + Es = 0 The treatment of soil physical processes in SITE requests the division of the soil in two layers. Considering the soil water flux, the ground layer g provides water to free evaporation, while the deep layer d determines the maximum depth where the root system extracts water for transpiration (Fig. 7). Considering the soil heat flux (described in Section 2.8), the temperature of the layer g depends mainly on its radiation balance, while the layer d, deeper, works as a large heat reservoir. These combined characteristics require a layer g as thin as possible, respected the numerical stability of the model, and a layer d thick enough Pg = P − Pu − Ps + dripu + drips (35) where dripu = Wu /τdrip represents the dripping of water from leaves and drips = Ws /τdrip represents the dripping of water from stems. S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 The fraction of soil moisture in the layers (θ i , m3 H2 O m−3 soil) is expressed as a function of the water mass stored in the respective soil layers (Wi , kg H2 O m−2 ): θi = Wi ei ρw (36) where the subscript i represents the ground layer (i = g) or the deep layer (i = d). The capacity of the roots to remove water from the soil depends on soil type and on a complex interaction of forces, known as water matric potential in the soil (ψ). The hydraulic properties (hydraulic conductivity and matric potential) are calculated from the average moisture content in each layer: ψi = ψe xi−b (37) ki = ks xi2b+3 (38) where ks is the saturation hydraulic conductivity (mm s−1 ); ψe is the air entry water potential (mm); xi = θi /θ s is the relative soil saturation in the layers g and d; and θ s is the saturation water content. Moisture content at field capacity (θ cc ) and wilting point of the soil (θ WILT ) are calculated inverting Eq. (37) and setting ψ to 3.30 m (1/3 atm) and ψ to 150 m (15 atm), respectively. The parameterisations of infiltration and runoff follow Bonan (1996), where the maximum infiltration velocity of the soil (Imax ) is expressed from Darcy’s law, at saturation conditions (Entenkhabi and Eagleson, 1989). Imax = ks (vxg − v + 1) (39) where v = −bψe /103 eg is an infiltration parameter. Fg (mm s−1 ), the effective infiltration of water into the soil, is the minimum between Imax and Pg and SR = Pg − Imax is the surface runoff. The soil water 299 flux between layers d and g is calculated using the expression obtained by Bonan (1996), pp.98). 2.8. Soil heat flux The temperature of the soil g layer is calculated as a function of the energy balance at the surface (Section 2.6). Soil temperature in the d layer is calculated as a prognostic variable (Bonan, 1996) using Cd Td − Tg dTd = = Gg dt (eg /2κg ) + (ed /2κd ) (40) where κ is the soil thermal conductivity (W m−1 K−1 ). The volumetric heat capacity of the soil (J m−2 K−1 ) depends on the volumetric fractions and the specific heat capacity of each soil component (Campbell and Norman, 1998): Cg = ρm eg φm cm + mo φo co + cw wg (41) Cd = ρm ed φm cm + cw wd (42) where m, o, w are indexes related to the minerals, organic materials and water present in the soil; mo = 2(Lu + Ls ) is the mass of organic matter of the soil (kg organic matter m−2 ), as seen in Section 2.9—it is assumed here that carbon accounts for half of the dry biomass; and φm = 1 − φ is the volumetric fraction of soil minerals. The thermal properties of the several soil components are in Table 3. Following Campbell and Norman (1998), soil thermal conductivity in the layer g, κg , and in the d layer, κd (W m−1 K−1 ), is computed by the normalised sum of the products of the volumetric fractions and the thermal conductivities of the several soil components: the mineral components (m), water (w) and gas phase component (air). κi = θi ξw κw + φair ξair κair + φm ξm κm ei (θi ξw + φarg ξi + φm ξm ) (43) Table 3 Soil thermal properties and other values used for calculating the thermal capacity of the soil Material Density, ρ (kg m−3 ) Specific heat capacity, c (J kg−1 K−1 ) Thermal conductivity, κ (W m−1 K−1 ) Correction factor, ξ Organic matter (o) Soil minerals (m) Water (w) Air 1300 2650 1000 – 1.92 × 103 0.87 × 103 4.48 × 103 – – 2.50 0.596 0.025 – 0.54 0.92 1.75 Values derived from Campbell and Norman (1998), pp. 118. 300 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 where φair = φ − φi is the air volumetric fraction in the g and d layers; ei is the thickness of the g and d layers. CO2a NEE 2.9. Carbon balance An The parameterisation of the carbon balance, adapted from the IBIS model (Foley et al., 1996), calculates the carbon flux through prognostic equations. The carbon fixed by photosynthesis can be allocated in four different reservoirs of the plants: leaves, stems, fine roots and coarse roots (Eq. (44)). The amount of carbon allocated to each canopy reservoir is calculated using a linear differential equation, assuming fixed fractions of allocation in each reservoir and fixed residence times. Fig. 8 presents a schematic representation of the distribution of carbon reservoirs and main fluxes. dCi Ci = ai NPP − (44) dt τi where the subscript i can be equal to u, s, f and r, that denote leaves, stems, fine roots and coarse roots, respectively. The sum of the au , as , af and ar is equal to 1. The initial values of the amount of carbon in each reservoir are shown in Table 4. The net primary production (NPP, kg C m−2 s−1 ) is expressed as a function of the gross photosynthesis (Ag , mol CO2 m−2 s−1 ) and the canopy autotrophic respiration (leaves, stems and roots). NPP = 0.012(1 − n) (Ag − Ru − Rs − Rf − Rr )dt (45) where n = 0.3 is the fraction of carbon lost due to growth (Amthor, 1984); 0.012 is the conversion Cu Cs Rsoil Cs Cu Cf Fig. 8. Schematic representation of the carbon reservoirs and fluxes. The direction of the arrows represents the positive direction of the flux. factor from mol CO2 to kg C; Ru is the respiration rate of the leaves (mol CO2 m−2 s−1 ); Rs = 3.17 × 10−10 Cs f(Ts )λ is the respiration rate of the stems (mol CO2 m−2 s−1 ); Rf = 3.17×10−8 Cf f(Tg +Td /2) is the respiration rate of the fine roots (mol CO2 m−2 s−1 ); Rr = 3.17 × 10−10 Cr f(Tg + Td /2)λ is the respiration rate of the thick roots (mol CO2 m−2 s−1 ); λ = 0.10 is the sapwood fraction of the stem and coarse root biomass, Cs , Cf and Cr are the carbon mass stored in stems, fine roots and coarse roots. The leaf area index (LAI) is calculated as a function of the carbon biomass in the leaves (Cu ) and the leaf specific area (Sl ) (Table 1). LAI = Cu Sl (46) Table 4 Initial values of the storage parameters of carbon in the canopy and of carbon mass in litter Reservoir (τ)a Residence time Gross photosynthesis allocation fraction (a)a Carbon mass (L, D)a Carbon biomass (C) a b c Kucharik et al. (2000). Calibrated to obtain the initial value of LAI. Measured on site. Leaves (u) Stems (s) Fine roots (f) Coarse roots (r) 1 year 0.45 0.50 0.360b 25 years 0.40 0.25 18.000a 1 year 0.10 0.44 0.100c 25 years 0.05 0.135 0.750c S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 The soil carbon module is based on the decomposition of leaves and stems by bacteria, and is a function of soil temperature, soil humidity and carbon mass in the litter, being similar to the IBIS model, with some modifications. The organic matter decomposition by the soil respiration is proportional to the litter mass. The processes of soil carbon decomposition by heterotrophic respiration are described by Eqs. (47) and (48). dLi Ci − f g g g hi Li = dt τi (47) (fg + fd ) (gg + gd ) dDi Ci = − h i Di dt τi 2 2 (48) where Li is the carbon mass of leaf (i = u) or stem (i = s) litter; Di is the carbon mass of dead fine root (i = f) and coarse root (i = r) (kg C m−2 ) (the initial values of carbon mass in the soil carbon reservoirs are shown in Table 5); hi is the respiration rate of leaves (i = u), stems (i = s), fine roots (i = f) and coarse roots (i = r) litter (kg C kg−1 C s−1 ); f(Ti ) = 2((Ti −283.16)/10) is the soil temperature function in the g and d layers (dimensionless); gg = 0.25 + 0.75xi is the soil moisture function in the g and d layers (dimensionless). The parameterisation of the soil heterotrophic respiration (RH ) is solved by the sum of the products of the temperature and moisture functions, and the decomposition rate for each reservoir of soil organic matter. RH = fg gg (hu Lu + hs Ls ) + (fg + fd ) (gg + gd ) (hf Df + hr Dr ) 2 2 (49) Table 5 Hydraulic properties of the soil at the micrometeorological site in Caxiuanã, Pará State, Brazil (Cosby et al., 1984) Variable Symbol Exponent of the moisture release equation Air entry water potential (mm) Porosity (m3 m−3 ) Humidity content at field capacity (m3 m−3 ) Wilting point of the soil (m3 m−3 ) Saturation water content (m3 m−3 ) Saturation hydraulic conductivity (mm s−1 ) b Used value 301 Finally, the net ecosystem production (NEP) is expressed by the difference between RH and the net primary production (NPP). A negative value of NEP indicates assimilation of carbon by the ecosystem. NEP = RH − NPP (50) 3. Description of the experimental area and site specific biophysical parameters The data used for the initial calibration and validation of SITE are part of the research conducted by the large-scale biosphere–atmosphere experiment in Amazonia (LBA), which were collected at 30-min intervals between April and September 1999, in the experimental area of the primary tropical evergreen forest in the Caxiuanã Forest Reserve (latitude 01◦ 42 30 S, longitude 51◦ 31 45 W, altitude 60 m). The reserve has an area of approximately 33,000 ha and is located about 400 km west of Belém, Pará, Brazil. A series of 28 days of data, collected during the dry season, from August 29 to September 25, 1999 (241–268 Julian day), was used to calibrate the model and data collected from April 16 to May 27, 1999 (106–147 Julian day), a total of 41 days during the wet season, were used to validate it. The measured data used for calibration and validation of the model were latent heat, sensible heat, and CO2 fluxes. Tables 5 and 6 present some site-specific biophysical parameters. The distribution of the textural composition of the soil (sand, clay and silt fractions are 20, 37 and 23%, respectively) was obtained by Ruivo et al. (2001). The hydraulic parameters of the soil (Table 5) were calculated using Cosby et al. (1984) and the biophysical parameters used for the studied area are shown in Table 6. 4. Results and discussion 9.066 4.1. Sensitivity test ψe φ θ cc θ WILT θs ks 223.9 0.477 0.36 0.23 0.477 0.0032 Sensitivity analyses were performed to determine the relative importance of variation in the main factors controlling carbon, water vapour and energy fluxes in a tropical rain forest. Below, we present results from the analyses that showed the highest sensitivity for CO2 and water vapour fluxes. 302 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 Table 6 Site-specific biophysical parameters used in the simulation Variable Symbol Used value Source Height of data measurement (m) Thickness of the superficial layer (m) Thickness of the deep layer (m) Depths of the center of the g layer (m) Depths of the center of the d layer (m) Typical dimension of leaves Typical dimension of stems Absorbed photosynthetically active radiation (W m−2 ) Leaves transmissivity Stems transmissivity za eg ed zg zd du ds APAR τu τs 56 0.075 3.425 0.0375 1.75 0.072 0.10 0.44325 S 0.16 0.19 Measured on site Section 2.7 Measured on site eg /2 (eg + ed )/2 Measured on site Measured on site Oliveira (2000) Moura et al. (2000) Moura et al. (2000) The CO2 flux simulated by the model is most sensitive to the parameter Vmax (see Eq. (14)). We run several simulations varying Vmax from 50 to 95 mol m−2 s−1 , keeping the other parameters at their baseline values. The CO2 flux calculated by the model for two values of Vmax are plotted in Fig. 9, along with the observed profile of CO2 . Using the lowest Vmax (50 mol m−2 s−1 ) underestimates CO2 flux, especially during night time, while the highest Vmax (95 mol m−2 s−1 ) overestimates CO2 flux (not shown). The optimum value found for Vmax was 75 mol m−2 s−1 . The water vapour flux simulated by the model is most sensitive to the coefficient for stomatal angular conductance, m (Table 2). We run several simulations varying m from 5 to 15, keeping the other parameters at their baseline values. The water vapour flux calculated by the model for two values of m are shown in Fig. 10, along with the observed fluxes. For m = 5, the model underestimates the water vapour flux, while m = 10 results in a better fit between simulated and observed data. 4.2. Carbon flux The simulated mean values of the carbon flux (NEP) from April 16 (106) to May 27 (147), used in the validation, and from August 29 (241) to September 25 (268) of 1999, used in the calibration of the model, were −0.69 and −0.79 kg C per m2 per year, respectively. These results present an error of −6.5% for the validation period and 6.8% for the calibration period, when compared to the observed mean values. (In this study, negative values of NEP refer to assimilation -2 -1 CO2 Flux (kgC m yr ) 15 10 5 0 -5 -10 -15 255 256 257 258 259 260 261 262 -6 Day of the year Simulated(Vmax=75.10-6 ) Simulated(Vmax=50.10 ) Observed Fig. 9. Sensitivity analyses for Vmax (maximum Rubisco enzyme capacity) parameter on CO2 flux. S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 303 -2 -1 Water vapour flux (kg H2 O m h ) 1,00 0,80 0,60 0,40 0,20 0,00 255 256 257 258 259 Day of the year 260 261 262 Simulated(m=10) Simulated(m=5) Observed Fig. 10. Sensitivity analyses for m (coefficient for stomatal angular conductance for broadleaf trees) parameter on water vapour flux. of carbon and positive indicates release to the atmosphere). Carswell et al. (2002) found an observed value of −0.54 kg C per m2 per year for the site in study, based on a longer data collection period; Malhi et al. (1998) found a NEP of −0.57 kg C per m2 per year for Cuieiras, Manaus, for the period from 13 to November 21, 1995; Rocha et al. (1996), doing simulations with the SiB2 model, found −0.63 kg C per m2 per year in the Ducke Reserve, Manaus, for the period from September of 1983 to August of 1985. According to Goulden et al. (1996), the eddy correlation system does not provide good values of CO2 during calm nights. In addition, Miller et al. (2004) found a strong dependence between nighttime NEP and turbulence above the canopy, showing a reduction in NEP for friction velocity (u∗ ) smaller than 0.2 m s−1 . Therefore, in a few comparisons in this study, the raw CO2 data collected during the night were eliminated when u∗ < 0.2 m s−1 . The root mean square error (RMSE) for raw and filtered CO2 flux data are 0.15 and 0.13 kg C per m2 per year, respectively. In the remaining discussion in this paper, the elimination of data is not considered except when noted. Figs. 11 and 12 show the profile of selected processes associated to the CO2 flux, and simulated and observed NEP. In Figs. 11a and 12a, Je has a null value during the night, and a maximum close to noon, local time. On the other hand, Jc presents a maximum value during the night, due to the dynamics of carbon through the stomata. In addition, the inverse dependence of Kc and Ko with air temperature also contributes, in a secondary way, for the profile of Jc . Figs. 11b and 12b show the net primary productivity (NPP) and the soil heterotrophic respiration (RH ). While NPP basically follows the Ag pattern, in the peaks during the afternoon it is possible to see the dependence of RH with the soil temperature. The difference between RH and NPP is the net ecosystem production (Figs. 11c and 12c), where negative values indicate assimilation of CO2 by the ecosystem. The model does not simulate the transport of CO2 in the atmosphere—only the production of CO2 due to the carbon balance is simulated. Thus, in spite of the reproduction of the general profile of the CO2 flux by the model, some variations at the hourly scale cannot be captured. Fig. 11 shows that the photosynthesis rate rises quickly after sunrise. In the second half of the morning, the activity of the Rubisco enzyme starts to limit the photosynthetic activity, which reaches its peak before noon. The peak of the simulated NEP oscillated between −7 and −8 kg C per m2 per year. Carswell et al. (2002) obtained an average peak of −7.2 kg C per m2 per year in the Caxiuanã forest for the period from 108 to 114 for the same year in study, with peak time at 11:00 a.m., local time. Malhi et al. (1998) observed an average peak of −6.8 kg C per m2 per year in Cuieiras, central Amazonia. Williams et al. (1998), using the SPA model (soil–plant–atmosphere canopy 304 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 Gross photosynthesis (kgC m -2 -1 yr ) 25 20 15 10 5 0 109 110 111 (a) 112 113 114 115 116 Ag Je Jc Day of the year 12 -1 CO2 flux (kgC m yr ) 10 -2 8 6 4 2 0 -2 109 110 111 (b) 112 113 114 115 116 RH NPP Day of the year -2 -1 CO2 flux (kgC m yr ) 15 10 5 0 -5 -10 -15 109 (c) 110 111 112 113 Day of the year 114 115 116 Simulated Observed Fig. 11. Profile of the processes associated to the observed and simulated CO2 fluxes: (a) gross photosynthesis (Ag) and its components Je and Jc ; (b) soil heterotrophic respiration (Rsoil ) and net primary production (NPP); and (c) net ecosystem exchanges (NEP) from April 19 to 26 (109–116) of 1999. The negative sign corresponds to the accumulation of CO2 in the ecosystem. 305 25 Gross photosynthesis (kgC m -2 -1 yr ) S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 20 15 10 5 0 124 125 126 127 128 129 Ag Je Jc Day of the year (a) 12 -1 CO2 flux (kgC m yr ) 10 -2 8 6 4 2 0 -2 124 125 126 127 128 RH NPP Day of the year (b) 129 10 -2 -1 CO2 flux (kgC m yr ) 15 5 0 -5 -10 -15 124 (c) 125 126 127 Day of the year 128 129 Simulated Observed Fig. 12. Profile of the processes associated to the observed and simulated CO2 flux: (a) gross photosynthesis (Ag) and its components Je and Jc ; (b) soil heterotrophic respiration (Rsoil ) and net primary production (NPP); and (c) net ecosystem exchanges (NEP) from May 4 to 9 (124–129) of 1999. The negative sign corresponds to the accumulation of CO2 in the ecosystem. 306 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 model), simulated peaks between −7.6 and −9.5 kg C per m2 per year. Correlation coefficient between simulated and observed hourly CO2 fluxes for Caxiuanã Reserve in the period from April 16 to May 27 (106–147) of 1999 is 0.88. Despite the good correlation, the model is not able to simulate high values of CO2 flux during night time (Fig. 12). Williams et al. (1998), using the SPA model, obtained a correlation coefficient of 0.73 between simulated and observed hourly fluxes, while Zhan et al. (2003), using two versions of the SSiB model, obtained hourly fluxes correlation coefficients of 0.73 and 0.79. The average conditions around noon indicate a slight closing of the stomata and the decrease of the gross photosynthesis, causing a reduction in the loss of water by evaporation and in the CO2 assimilation rates. It is observed that between peak time and late afternoon (period called the dark phase) the gross photosynthesis fluxes are no longer determined by the solar energy, but are limited by the activity of the Rubisco enzyme. In late afternoon, the CO2 assimilation becomes limited by solar energy again (Figs. 11a and 12a). NPP has a mean value of approximately 2 kg C per m2 per year (Figs. 11b and 12b). The NPP simulated by the model is close to the simulated by Kucharik et al. (2000) who, using the IBIS model, obtained a mean value of 1.9 kg C per m2 per year for tropical rainforests. In Figs. 12a and 12b, some oscillations in the peaks of Ag and NPP close to noon can be verified in days 125 and 129, consequently altering the NEP profile (Fig. 12c). These oscillations are related to the presence of cloudiness (not shown), demonstrating consistency in the simulated values of Je . Clark et al. (2001) reviewed the estimates of NPP from 39 tropical forests experimental sites around the world. In these sites, the lower limit of NPP ranged from 0.17 to 1.18 kg C per m2 per year, while the upper limit ranged from 0.31 to 2.17 kg C per m2 per year. In Amazonia, NPP estimates ranged from 0.67 to 0.92 kg C per m2 per year (lower limit) and 1.22 to 1.68 kg C per m2 per year (upper limit). Our simulated values of soil heterotrophic respiration (averaging 1.3 kg C per m2 per year) are also close to the observed values. Trumbore et al. (1995) and Davidson et al. (2000) obtained mean values of 1.5 and 2 kg C per m2 per year in the Victoria farm, Paragominas, Eastern Amazonia. Most of the references available in the literature report the total soil respiration (autotrophic and heterotrophic). Carswell et al. (2002) verified a seasonal variation for the night respiration from 2.7 to 3.5 kg C per m2 per year, for the Caxiuanã Reserve Forest. Rocha et al. (1996), using the SiB2 model, simulated a mean value of approximately 2.4 kg C per m2 per year in the Ducke Reserve, corresponding to the period from September 1983 to August 1985, having also verified that the soil contributes with approximately 70–80% of the total carbon emitted to the atmosphere. Malhi et al. (1998) obtained a mean value around 2.5 kg C m−2 h−1 for the soil respiration in Cuieiras, Manaus. Grace et al. (1996) reported values of night respiration ranging from 2.3 to 2.6 kg C per m2 per year in the Jaru Reserve, Rondônia. Fig. 13 shows the variation of the simulated canopy conductance through the period 109–116, with mean peak around 0.8 mol H2 O m−2 s−1 , which is similar to the value observed in the same site by Carswell et al. (2002), who found a maximum peak ranging seasonally from 0.7 to 0.8 mol H2 O m−2 s−1 . Grace et al. (1996) reported gs ranging from 0.4 to 1.0 mol H2 O m−2 s−1 in the morning, and a decrease along the day in the Jaru Reserve, Rondônia, in 1993. However, Roberts et al. (1996), obtained peak values ranging from 0.35 to 0.4 mol H2 O m−2 s−1 , for the experimental areas of forests of the ABRACOS project. The difference among gs values is possibly due to the strong influence of the specific humidity deficit that exists in different experimental areas. 4.3. Water vapour flux The average simulated water vapour flux (E) ranged from 0.14 kg H2 O m−2 h−1 (for the period from 106 to 147) to 0.19 kg H2 O m−2 h−1 (for the period from 241 to 268). The RMSE for raw and filtered water vapour flux were 0.006 and 0.005 kg H2 O m−2 h−1 , respectively. Although the simulated values of E overestimate the observed values by approximately 17%, the simulated mean values are in the range estimated for other areas of the Amazon forest, such as at the Ducke Reserve, in Manaus, where Shuttleworth (1988), Rocha et al. (1996) and Hodnett et al. (1996) obtained mean values Stomatal conductance (mol H2O m-2 s-1) S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 307 1.20 1.00 0.80 0.60 0.40 0.20 0.00 109 110 111 112 113 114 Day of the year 115 116 Fig. 13. Temporal variation of the stomatal conductance. of 0.15, 0.15, and 0.16 kg H2 O m−2 h−1 , respectively. Nepstad et al. (1994) found a mean rate of 0.15 kg H2 O m−2 h−1 in the dry season in Pará, and Costa and Foley (1997), using a modified version of the LSX model, estimated a mean value of 0.18 kg H2 O m−2 h−1 for the tropical Amazonian rainforest. Fig. 14 illustrates simulated and observed E profiles for the periods from April 18 to 25 (108–115), from April 30 to May 8 (120–128) and from May 13 to 20 (133–140), periods used to validate the model. The mean simulated peaks of E were 0.50, 0.50 and 0.35 kg H2 O m−2 h−1 for the three periods, respectively. The peaks of the simulated E in the night of the days 108, 111, 123, 136, and 138 were overestimated by the model due to the effect of wind bursts that happened during these periods. We believe that this discrepancy found in the simulation is due to the following factors: (a) integration interval (dt) relatively high (1 h) – models like LSX, LSM, and SiB2 generally use integration intervals of the order of 20 min or less; (b) high sensitivity of E to the strong night wind bursts, when the air inside of the canopy is saturated or near saturation; (c) numerical instability in conditions of neutral atmosphere, similar to the observed in other models with similar structure. Although the referred problems could be solved through the use of new numeric methods for the mass and energy transport, the reduction of the integration interval (dt) or the increase in the complexity of the parameterisation of the atmospheric conductance, we believe that the results are satisfactory (ρ = 0.64 for the entire validation period), considering the desired intermediate complexity of the code. More sophisticated models (SSiB, SPA) typically obtain values in the range of 0.80–0.90 (Zhan et al., 2003; Williams et al., 1998). 4.4. Energy balance and sensible heat flux In general, the simulated sensible heat flux was smaller than the observed in the two studied periods. The underestimation of the sensible heat flux, is related to the overestimation of the latent heat flux. Although the mean simulated sensible heat flux is smaller than the measured at the site, it was similar to the simulated values found by Rocha et al. (1996) of 13.9 W m−2 for Ducke Reserve in the period from September of 1983 to August of 1985. Table 7 illustrates the comparison of the partition of the energy balance, showing approximately 10 days in the periods used for validation and calibration. The model overestimates the latent heat flux and underestimates the sensible heat flux, and the RMSE for the sensible heat flux (non-filtered) is 3.8 W m−2 . The soil heat flux and the variation in the energy stored in the ecosystem (G + :S) are obtained by the difference between the other energy fluxes (Rn − Lv E − H). The partitions of energy between the sensible and latent heat flux were different between the periods. In the period from 106 to 116, the solar radiation was attenuated by the high presence of cloudiness, presenting a mean value of approximately 145.3 W m−2 , where 75% of this value was used in the evapotranspiration, and 7% was used to heat up the atmosphere. 308 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 -2 -1 Water vapour flux (kg H2 O m h ) 1,00 0,80 0,60 0,40 0,20 0,00 108 109 110 111 (a) 112 113 114 115 Simulated Observed Day of the Year -2 -1 Water vapour flux (kg H2 O m h ) 1,00 0,80 0,60 0,40 0,20 0,00 120 121 122 (b) 123 124 125 126 127 Day of the Year 128 Simulated Observed -2 -1 Water vapour flux (kg H2 O m h ) 1,00 0,80 0,60 0,40 0,20 0,00 133 (c) 134 135 136 137 Day of the Year 138 139 140 Simulated Observed Fig. 14. Simulated (3-h running mean) and observed profile of water vapour flux (E) from April 18 to 25 (108–115), from April 30 to May 8 (120–128) and from May 13 to 20 (133–140) of 1999. S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 309 Table 7 Mean values of the partition of the simulated and observed energy balance: net radiation (Rn), latent heat flux (Lv E), sensible heat flux (H) and residue (G + :S), in periods from 16 to 26 of April (106–116) and from 11 to 20 of September (254–263) of 1999 Variable (W m−2 ) Rn Lv E H G + :S Period 106–116 Period 254–263 Simulated (%) Observed (%) RMSE (non-filtered) 145.3 109.5 9.1 26.6 145.3 82.6 29.7 33.0 – 4.4 3.8 – (100) (75) (7) (18) (100) (57) (20) (23) Simulated (%) 147.1 146.5 11.7 −11.1 The heat flux into the soil and the variation of the energy stored in the ecosystem removes approximately 18% of the radiation balance: In the period from 254 to 263, the net radiation was more intense and the CO2 flux and the canopy conductance were higher. Consequently, the available energy was almost totally used for the evapotranspiration. The simulated Lv E was approximately equal to the net radiation, although 20% greater than the Lv E observed. In general, it can be observed that the values simulated by the model agree well to the values observed in the experimental area in study, being in the range of uncertainty of the method of flux mensuration (∼20%). In this period, RMSE for latent and sensible heat fluxes are 10.9 and 5.2 W m−2 , respectively. Delire and Foley (1999), using the IBIS model at Reserva Jaru, Rondônia, obtained a RMSE of 43 and 31 W m−2 for the latent and sensible heat fluxes, respectively. (100) (100) (8) (−8) Observed (%) 147.1 122.3 31.9 −7.1 (100) (83) (22) (−5) RMSE (non-filtered) – 10.9 5.2 – Nobre et al. (1996) found mean values of 138.9, 112.3 and 26.1 W m−2 , for Rn, Lv E and H, respectively, for the Ji-Paraná forest, Rondônia, in the period of July of 1993. Reviewing the measurements of energy balance in the Amazon tropical forest, Pereira (1997) reported an average radiation balance of 123.8 W m−2 , being 64% of Rn transformed in latent heat, and 29% transformed in sensible heat, for the rainy season. Galvão and Fisch (2000) evaluated the energy balance in the area of Ji-Paraná, Rondônia, finding a ratio of Lv E/Rn and H/Rn of 79 and 17%, respectively. During the dry season, Lv E and H corresponds to 62 and 18% of Rn, respectively. Using the SiB2 model, Rocha et al. (1996) found the approximate simulated ratios of Lv E/Rn and H/Rn of 82 and 18% for the Reserva Ducke, 79 and 21% for the Reserva Jaru, and 88 and 12% for the Reserva Vale do Rio Doce, respectively. -2 -1 Water vapour flux (kg H2 O m h ) 1,00 0,80 0,60 0,40 0,20 0,00 133 134 135 Day of the Year 136 137 138 139 140 Simulated(with moving average) Simulated(without moving average) Observed Fig. 15. Simulated (with and without 3-h moving average) and observed water vapour fluxes profiles for 8-day period from May 13 to 20 (133–140) of 1999. 310 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 4.5. Numerical stability of the model The results discussed in the previous section indicates that the SITE CO2 simulation is much more accurate than the water vapour and sensible heat flux simulations. The major reason for the lower quality of the water vapour and energy simulations is a weak numerical instability of the model. Eqs. (23), (24) and (30), when solved through finite differences, are highly unstable numerically. To avoid this problem, an implicit numerical method was used through the solution of a system of four linear equations. We should note here, however, that even the solutions presented are somewhat unstable numerically, especially for the time step of one hour. To avoid the numerical instability, we should substitute Eqs. (9), (20), (21) and (22) in Eqs. (23) and (32), introducing undesired mathematical complexities, caused by the if-clauses associated to Eqs. (9), (20) and (21), and to the iterative (two-step) calculations of Wu and Ws . Were SITE coupled to an atmospheric model, this instability would be unacceptable, as it would propagate into the atmospheric circulation. However, we figured out that the effects of the instability in the output variables virtually disappear if a moving average of the output variables is plotted. The numerical instability can be seen in Fig. 15, where simulated raw and 3-hourly moving averages water vapour fluxes are plotted. 5. Summary and conclusions We have developed an intermediate complexity model to simulate the carbon, energy and water fluxes between a tropical ecosystem and the atmosphere. The Simple Tropical Ecosystem Model (SITE) also simulates the carbon dynamics in the ecosystem. In SITE, we combined a realistic representation of the processes involved in the ecosystem functioning with a simple mathematical framework, keeping the required mathematics at the linear differential equations and linear algebra level. In addition, the simple structure of SITE makes it easy for users to implement parameterisations of processes that are not yet represented. Although the model was tested against data collected at an 150-year-old Amazon tropical forest site, where species from families Sapotaceae, Chrisobalanaceae and Lauraceae dominated, it can be used in several other situations, including tropical forests under different management strategies, young forests and even other tropical ecosystems, provided that adequate input data and parameters are available. Even though SITE is considerably less complex than other models of similar goals, it reproduces well the hourly variability of the fluxes of CO2 and water vapour, and it simulates the balance of those elements in seasonal scale properly. In particular, SITE CO2 flux simulations have a very good correlation with the hourly-observed values. In most of the times, the model reproduces well the hourly variability in the fluxes, including some details usually difficult to reproduce, like the NEP peak that happens before noon, and the transient water vapour flux during the night. Nevertheless, the model is excessively sensitive to wind bursts that happen during the evening, when the canopy air is near saturation, overestimating the water vapour flux in these situations. SITE is available as a 1200-line FORTRAN code, and as a computer spreadsheet. We believe the model will be useful to help train the next generation of tropical ecologists in the use of ecosystem models. Acknowledgements The Brazilian agencies CAPES and CNPq, as well as WWF–Brazil, provided funding for this research. Data used for validation and calibration of the model was collected by the project ECOBIOMA, also funded by CNPq. We would like to thank the anonymous reviewers and the editor for their comments and recommendations. References Amthor, J.S., 1984. The role of maintenance respiration in plant growth. Plant Cell Environ. 7, 561–569. Bonan, G.B., 1996. A Land surface model (LSM version 1.0) for ecological, hydrological, and atmospheric studies: technical description and user’s guide. NCAR Technical Note TN417+STR, 150 pp. Campbell, G.S., Norman, J.M., 1998. An Introduction to Environmental Biophysics. 2nd ed. Springer-Verlag, New York, 286 pp. S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 Carswell, F.E., Costa, A.C.L., Palheta, M., Malhi, Y., Meir, P., Costa, J.P.R., Ruivo, M.L., Leal, L.S.M., Costa, J.M.N., Clemente, R.J., Grace, J., 2002. Seasonality in CO2 and H2 O flux at an eastern Amazonian Rain Forest. J. Geophys. Res. v.107, 8076, doi: 10.1029/20001D000284. Clark, D.A., Brown, S., Kicklighter, D.W., Chambers, J.Q., Thomlinson, J.R., Holland, J.eE.A., 2001. Net primary production in tropical forests: an evaluation and synthesis of existing field data. Ecol. Appl. 11, 371–384. Collatz, G.J., Ball, J.T., Grivet, C., Berry, E.J.A., 1991. Physiological and environmental regulation of stomatal conductance, photosynthesis and transpiration: a model that includes a laminar boundary layer. Agric. For. Meteorol. 54, 107–136. Collatz, G.J., Carbo, M.R., Berry, J.A., 1992. Coupled photosynthesis-stomatal conductance model for leaves of C4 plants. Aust. J. Plant Physiol. 19, 519–538. Cosby, B.J., Hornberger, R.B., Clapp, R.B., Ginn, T.R., 1984. A statistical exploration of the relationships of soil moisture characteristics to the physical properties of soils. Water Resour. Res. 20, 682–690. Costa, M.H., Foley, J.A., 1997. The water balance of the Amazon basin: dependence on vegetation cover and canopy conductance. J. Geophys. Res. 102, 23973–23990. Davidson, E.A., Verchot, L., Cattânio, J.H., Ackerman, I.L., Carvalho, J.E.M., 2000. Effects of soil water content on soil respiration in forests and cattle pastures of eastern Amazonia. Biogeochemistry 48, 53–69. Delire, C., Foley, J.A., 1999. Evaluating the performance of a land surface/ecosystem model with biophysical measurements from contrasting environments. J. Geophys. Res. 104, 16895–16909. Dickinson, R. E., Sellers, A.-H., Kennedy, P. J., Wilson, M. F., 1984. Biosphere-Atmosphere transfer scheme (BATS) for the NCAR community climate model, NCAR Technical Note TN275+STR, 69 pp. Entenkhabi, D., Eagleson, P.S., 1989. Land surface hydrology parameterization for atmospheric general circulation models including sub grid scale spatial variability. J. Climate 2, 816– 831. Farquhar, G.D., Caemmerer, S.V., Berry, J.A., 1980. A biochemical model of photosynthetic CO2 assimilation in leaves of C3 species. Planta 149, 78–90. Farquhar, G.D., Sharkey, T.D., 1982. Stomatal conductance and photosynthesis. Ann. Rev. Plant Physiol. 33, 317–345. Foley, J.A., Prentice, I.C., Ramankutty, N., Levis, S., Pollard, D., Sitch, S., Haxeltine, A., 1996. An integrated biosphere model of land surface processes. Global Biogeochem. Cycles 10, 603– 628. Galvão, J.A.C., Fisch, G., 2000. Energy budget in forest and pasture sites in Amazonia. Revista Brasileira de Meteorologia 15, 25–37. Goulden, M.L., Munger, J.W., Fan, S.-M., Daube, B.C., Wofsy, S.C., 1996. Measurements of carbon sequestration by long-term eddy covariance: methods and a critical evaluation of accuracy. Global Change Biol. 2, 169–182. Grace, J., Lloyd, J., Mcintyre, J., Miranda, A.C., Meir, P., 1996. Carbon dioxide flux over Amazon rainforest in Rondônia. In: Gash, J.H.C., Nobre, C.A., Roberts, J.M., Victoria, R.L. (Eds.), 311 Amazonian Deforestation and Climate, 1st ed. John Wiley and Sons, Chichester, pp. 307–318. Haxeltine, A., Prentice, I.C., 1996. Biome 3: an equilibrium terrestrial biosphere model based on ecophysiological constraints. Global Biogeochem. Cycles 10, 693–709. Hodnett, M.G., Tomasella, J., Marques Filho, A., De, O., Oyama, M.D., 1996. Deep soil water uptake by forest and pasture in central Amazonian: predictions from long-term daily rainfall data using a simple water balance model. In: Gash, J.H.C., Nobre, C.A., Roberts, J.M., Victoria, R.L. (Eds.), Amazonian Deforestation and Climate, 1st ed. John Wiley and Sons, Chichester, pp. 81–99. Hurtt, G.C., Moorcroft, P.R., Pacala, S.W., Levin, S.A., 1998. Terrestrial models and global change: challenges for the future. Global Change Biol. 4, 581–590. Kucharik, C.J., Foley, J.A., Delire, C., Fisher, V.A., Coe, M.T., Gower, S.T., Lenters, J.D., Young-Molling, C., Norman, J.M., Ramankutty, N., 2000. Testing the performance of a dynamic global ecosystem model: water balance. Global Biogeochem. Cycles 14, 795–825. Leuning, R., 1995. A critical appraisal of a combined stomatalphotosynthesis model for C3 plants. Plant Cell Environ. 18, 339–355. Malhi, Y., Nobre, A.D., Grace, J., Kruijt, B., Pereira, A.C., Scott, S., 1998. Carbon dioxide transfer over a central Amazonian rain forest. J. Geophys. Res. 31, 31593–31612. Medina, E., Cuevas, E., 1996. Biomass production in nutrientlimited rainforest: implications for responses to global change. In: Gash, J.H.C., Nobre, C.A., Roberts, J.M., Victoria, R.L. (Eds.), Amazonian Deforestation and Climate, 1st ed. John Wiley and Sons, Chichester, pp. 221–239. Melillo, J.M., Mcguire, A.D., Kicklighter, D.W., Moore III, B., Vörösmarty, C.J., Schloss, A.L., 1993. Global climate change and terrestrial net primary production. Nature 363, 234–240. Miller, S.D., Goulden, M.L., Menton, M.C., Rocha, H.R., Freitas, H.C., 2004. Annual CO2 exchange by a tropical Forest. Ecological Applications. In press. Moura, R.G., Manzi, A.O., Nogueira, V.S., Mendes, D., 2000. Comparação entre os perfis de radiação solar dentro de ambiente de floresta para dias com diferentes coberturas de nebulosidade. XI Congresso Brasileiro de Meteorologia, Rio de Janeiro-RJ. Anais. Neilson, R.P., 1995. A model for predicting continental scale vegetation distribution and water balance. Ecol. Appl. 5, 362– 385. Nepstad, D.C., Carvalho, C.R., Davidson, E.A., Jipp, P.H., Lefebvre, P.A., Negreiros, G.H., Silva, E.D., Stone, T.A., Trumbore, S.E., Vieira, S., 1994. The role of deep roots in the hydrological and carbon cycles of Amazonian forests and pastures. Nature 372, 666–669. Nobre, C.A., Fisch, G., Rocha, H.R., Lyra, R.F., Rocha, E.P., Costa, A.C.L., Ubarana, V.N., 1996. Observations of the atmospheric boundary layer in Rondônia. In: Gash, J.H.C., Nobre, C.A., Roberts, J.M., Victoria, R.L. (Eds.), Amazonian deforestation and climate, 1st ed. John Wiley and Sons, Chichester, pp. 413–423. 312 S.N. Monteiro Santos, M.H. Costa / Ecological Modelling 176 (2004) 291–312 Oliveira, J.B., 2000. Análise do Balanço de Radiação na Região Amazônica. Viçosa, MG: UFV, M.S. thesis dissertation, Federal University of Viçosa, Brazil, 79 pp. Parton, W., Stewart, J., Cole, C., 1988. Dynamics of C, N, P and S in grassland soils: a model. Biogeochemistry 5, 109–131. Prentice, I.C., Cramer, W., Harrison, S.P., Leemans, R., Monserud, R.A., Solomon, A.M., 1992. A global biome model based on plant physiology and dominance. J. Biogeography 16, 117– 134. Pereira, A.R., 1997. Radiation regime of tropical rain forest. Revista Brasileira de Agrometeorologia, vol. 5. pp. i–viii. Pollard, D., Thompson, S.L., 1995. The effect of doubling stomatal resistance in a global climate model. Global Planet. Change 10, 129–161. Raich, J.W., Rastetter, E.B., Melillo, J.M., Kicklighter, D.W., Steudier, P.A., Peterson, B.J., Grace, A.L., Moore, B., Vörösmarty, C.J., 1991. Potential net primary productivity in South America: application of a global model. Ecol. Appl. 1, 399–429. Roberts, J., Cabral, O.M.R., Costa, J.P., Mcwilliam, A.L.C., 1996. An overview of the leaf area index and physiological measurements during ABRACOS. In: Gash, J.H.C., Nobre, C.A., Roberts, J.M., Victoria, R.L. (Eds.), Amazonian Deforestation and Climate, 1st ed. John Wiley and Sons, Chichester, pp. 287–306. Rocha, H.R., Sellers, P.J., Collatz, G.J., Wright, I.R., Grace, J., 1996. Calibration and use of the SiB2 model to estimate water vapour and carbon exchange at the ABRACOS forest sites. In: Gash, J.H.C., Nobre, C.A., Roberts, J. M., Victoria, R.L. (Eds.), Amazonian Deforestation and Climate, 1st ed. John Wiley and Sons, Chichester, pp. 460–471. Ruivo, M.L.P., Quanz, B., Sales, M.E.C., Meir, P., 2001. Solos dos Sı́tios do Experimento esecaflor-Caxiuanã, Pa. In: Lisboa, P.L.B. (Ed.), Caxiuanã. Populações Tradicionais, Meio Fı́sico e Biodiversidade, Belém, Museu Paraense Emilio Goeldi, Belém, pp. 206–213. Running, S.W., Gower, S.T., 1991. Forest-BGC, a general model of forest ecosystem processes for regional applications, II, dynamic allocation and nitrogen budgets. Tree Physiol. 9, 147–160. Sellers, P.J., Mintz, Y., Sud, Y., Dalcher, A., 1986. A simple biosphere model (SiB) for use within general circulation models. J. Atm. Sci. 43, 505–531. Sellers, P.J., Bounoua, L., Collatz, G.J., Randall, D.A., Dazlich, D.A., Los, S.O., Berry, J.A., Fung, I., Tucker, C.J., Field, C.B., Jensen, T.G., 1996. Comparison of radiative and physiological effects of atmospheric CO2 on climate. Science 271, 1402– 1406. Shuttleworth, W.J., Gash, J.H.C., Lloyd, C.R., Moore, C.J., Roberts, J., Marques, A.D., Fisch, G., Silva, V.D., Ribeiro, M.D.G., Molion, L.C.B., Sá, L.D.D., Nobre, C.A., Cabral, O.M.R., Patel, S.R., Moraes, J.C., 1984. Eddy-correlation measurements of energy partition for Amazonian forest. Quart. J. R. Met. Soc. 110, 1143–1162. Shuttleworth, W.J., 1988. Evaporation from Amazonian rainforest. Proc. R. Soc. Lond. 33, BS321–346. Trumbore, S.E., Davidson, E.A., Camargo, P.B., Nepstad, D.C., Martinelli, L.A., 1995. Below ground cycling of carbon in forests and pastures of eastern Amazonia. Global Biogeochem. Cycles 9, 515–528. Williams, M., Malhi, Y., Nobre, A.D., Rastetter, E.B., Grace, J., Pereira, M.G.P., 1998. Seasonal variation in net carbon exchange and evapotranspiration in a Brazilian rain forest: a modelling analysis. Plant Cell Environ. 21, 953–968. Woodward, F.I., 1987. Climate and Plant Distribution. Cambridge University Press, Cambridge. Zhan, X., Xue, Y., Collatz, G.J., 2003. An analytical approach for estimating CO2 and heat fluxes over the Amazonian region. Ecol. Model. 162, 97–117.
© Copyright 2026 Paperzz