Nutrient balances and options for their improvement under different levels of
intensification of dairy production in Uganda
Mubiru S, Romney D, Halberg N and Tenywa J S (ph. D progress presentation)
Presentation for the LSRP annual review workshop 18th - 21st March 2003
Abstract
Livestock are an integral part of most farming systems in Uganda whose structure has evolved
over time to suit the agro ecological zones and the socioeconomic setting. There has been
increasing intensification of dairy production systems in Uganda mainly resulting from
increasing population pressure, growing market for milk and governments, NGOs and donor
promotion. For instances where intensification of dairy production was promoted, there were
anticipations that it was economically and ecologically sustainable. There exists therefore a
variety of dairy production systems ranging from highly intensive to highly extensive. However it
is not clear as yet which systems are ecologically beneficial and what aspects render them so.
Research was therefore undertaken to study the different dairy production systems in Uganda.
The major objectives of the study is to establish the status of nutrient management in each of the
systems and identify methods for improvement. Nutrient management indicators that can be used
to assess dairy production systems will also be developed. A characterization survey has been
carried out in Jinja, Masaka and Mbarara districts to develop dairy farm categories based on
level of intensification. A longitudinal survey is to be carried out to collect detailed information
on farm activities across seasons. Data collected will enable estimation of "nutrient balances"
which are the differences between the nutrients into and out of production systems and emphasis
here will be on Nitrogen (N), Phosphorus (P) and Potassium (K). A "nutrient balance" is defined
as an indicative measure for ecological sustainability. This estimation can therefore be used to
give an indication on the nutrient status of a production system. A spreadsheet model with
various parameters required for estimation of "nutrient balances" will be developed. This model
could be put to future use for similar purposes for environments that match those of the research
area. This research is expected to yield knowledge of the "nutrient balances" at various levels of
intensification of dairy production and systems at highest risk of nutrient drain, potential
interventions for improved management of nutrients and indicators of nutrient management
useful for assessing ecological competitiveness of dairy systems.
1. Introduction
Livestock are an integral part of most farming systems in Uganda whose structure has evolved
over time to suit the agroecological zones and the socioeconomic setting. Major livestock
species are cattle, goats, pigs, sheep, poultry and rabbits. Livestock production contributes about
7.5% to the total GDP and 17% to the AGDP. In most instances livestock production systems
include crop production for both subsistence and commercial purposes. It is estimated that these
mixed farming smallholder systems and the pastoralists own over 90% of the national cattle herd
as well as all the small ruminant and non-ruminant stock. These systems also produce the bulk of
the domestic milk and slaughter animals. Eighty percent of the national cattle herd is in the south
and western parts of the country where average number of cattle per household is 2.11 compared
to the northern part where it is 0.67 and the national average is 1.37 (FAO., 2000). Dairy
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 1
production is increasingly gaining importance with current national annual milk production at
511,000 tonnes from 457,000 tonnes in 1995 (FAO., 2001).
Major dairy production systems in Uganda are zero grazing, semi-intensive, fenced dairy farms,
tethering and communal, agro-pastoral and pastoral grazing. Zero grazing systems are prevalent
within vicinity of urban areas where land is scarce and the milk market is good (UN/WFP.,1993).
Semi-intensive systems are commonly found in peri-urban areas mainly in South Western,
Central and South Eastern parts of Uganda. The fenced dairy farms are also common in the same
areas, however, they are not necessarily close to urban centres. Communal and pastoral systems
are found in the South West, Central, North and North Eastern parts of the country. Tethering
system in not confined to any particular area but is common where crop production is a prime
activity and cattle numbers are very few (FAO., 2000). There has been increasing intensification
of dairy production systems in Uganda. Intensification here is defined as increased productivity
per unit of land. This has resulted from increasing population pressure, growing market for milk
and governments, NGOs and donor promotion. Improved cattle breeds that were primarily
introduced for dairy production, are currently largely kept under intensive management systems
on small and medium sized farms (FAO., 2000). The enterprise has been favoured by the ready
market for milk and available crop residues and agro industrial by-products for additional feed
resources.
One of the current challenges is to understand how the ecosystems are altered by intensive
agriculture and in turn develop strategies that take advantage of the ecological interactions within
the production systems (Matson et al., 1997). To attain this understanding, it is necessary to
categorise the existing systems and study the nutrient movements within each of them. A study
was therefore undertaken to categorise dairy production systems in Uganda, estimate nutrient
balances (Nutrient balances, which are the difference between the sums of nutrients inputs and
nutrient outputs (Geurts et al., 1999; Mohamed Saleem., 1998)) within each system, identify
means for improvement of nutrient management within each system and develop nutrient
management indicators. Nutrient balances are an effective means of assessing the vulnerability
of land use systems to nutrient degradation (Geurts et al., 1999; Mohamed Saleem., 1998). The
study is expected to yield information on sustainability of the different dairy production systems,
options for improvement of nutrient management in the less sustainable systems, indicators for
nutrient management that can be used by farmers, extension personnel and policy makers in
decision making and a spreadsheet model that can be used in the future for estimation of nutrient
balances.
2.
Methodology
2.1
Characterisation survey
2.1.1 Site familiarisation visit
Three districts Jinja, Masaka and Mbarara were selected for the study to capture as much as
possible existing variability within dairy production practices in Uganda. On basis of secondary
information there are contrasting cattle and human population densities (table 1).
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 2
Table 1: Human and cattle population densities in the three districts selected for the
characterization survey from Mugisha (1997/98)
Sub
Area km2 Sub county Cattle pop Cattle/ km2 Pop.
Pop/km2
counties
size km2
Mbarara 34
10154
299
607396
60
930772
92
Masaka 16
10611
663
528017
50
838736
79
Jinja
8
734
92
14031
19
289476
394
Mbarara is distinctive with more extensive cattle production systems and larger numbers of cattle
dominated by long horned Ankole cattle. Masaka and Jinja have similar cattle management
systems but with slightly better market opportunities in Jinja compared to Masaka.
2.1.2 Development and pre-testing of characterisation survey questionnaire
A first draft of the characterisation survey questionnaire was completed in the month of October
2001, it was then pre-tested in the three proposed districts for the survey. Three questionnaires
were administered in each of the 3 districts and thereafter appropriate adjustments were made to
respond to unforeseen features and issues encountered during the pre-testing exercise. During
this period information was also obtained from each district regarding the diversity of dairy
production systems and their distribution. This information together with that obtained from
ILRI on clusters was used in selection of sub-counties for the survey in each of the districts.
2.1.3 Selection of study areas
In order to assist selection of sub-counties and ensure that farmers were selected from a broad
range of market, climate and population density conditions, cluster analysis was carried out on
GIS spatial layers of human population density, climate (using ppe) and market access (road
density). The output from this analysis is shown in figures 1-3. Five clusters were identified
across the three districts with characteristics shown in table 2. Nine sub-counties, 3 from each
district, were selected to give a broad spread across clusters (Table 2).
Table 2: Characteristics of the 5 GIS clusters and Sub-counties (SCs) selected in each district
Cluster
Mean
population
density
483
Mean
annual ppe
1
No.
of
SCs
in
cluster
5
District where
cluster selected
SCs selected
0.80
Mean
market
distance
22
Jinja
Budondo; Mafubira
2
2
304
0.81
17
Jinja
Butagaya
3
6
91
0.64
34
Mbarara
Buremba
4
9
154
0.76
23
Masaka
5
21
102
0.68
17
Mbarara
Kalungu;
Kisekka;
Kyanamukaka
Bisheshe; Rwanyamahembe
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 3
Figure 1: Clusters for Jinja district
Figure 2: Clusters for Masaka district
Figure 3: Clusters for Mbarara district
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 4
Two villages were then selected randomly from each of the sub-counties using lists of parishes
and villages, in some cases referred to as cells, obtained from sub-county headquarters. Within
the selected sub-counties, cells that were small in size and population (particularly considering
number of cattle owning households) were merged to create what were termed as 'villages' for
the purpose of the survey, however, where cells were big enough they were considered villages
in themselves. Using statistics obtained from the village Local Council leaders for human
populations and number of households or tax payers in these villages, numbers of cattle and noncattle households to be interviewed in each of the villages were apportioned. It was planned to
have 34 respondent-households per sub-county of which 70% would be cattle households and
30% non-cattle households.
In each village 2 transects were used for sampling. Village maps were developed together with
resource people from the sub-counties, key sites were marked and 2 pairs of key sites were
selected at random. All households along the most direct route between the paired sites were
marked. Each of these households was visited and a record of their family name and whether
they had cattle or not was taken. This information was later used in selecting the households for
interview ensuring that the sample sizes required were obtained with the required numbers of
cattle and non-cattle households. Selection of farms along either side of the route was alternated
at specific intervals to ensure that much of the 'route' is covered and with more less equal number
of farms on either side of the route. The sampling interval along the routes depended on
household density along the route. Two and four households were being skipped on either side of
the transect in low household and high household concentration areas respectively. Each of the
selected households was visited to make an appointment for when the interview would be held.
In some sub-counties the required sample size, particularly the total number of cattle households
could not be realised from the village selected. In such cases, two or three neighbouring villages
were combined before transects were redrawn.
This was particularly common in
Rwanyamahembe and Bisheshe in Mbarara and Kyanamukaaka in Masaka.
2.1.4
Sample selection
i. In order to obtain 100 households from the district, each sub-county would have a sample size
of 34 households (100/3 = 34 approx.)
ii. The proportion of sample allocation of the sample to per village (Pv) was obtained using the
formula:
Pv
=
Household population OR Village population
of one of the villages from the sub-county (ie a or b)
____________________________________
Pop. Village (a) + Pop. Village (b)
X 34
iii. Sample selection per village was based the total human or household population per village.
In each village the ratio of cattle to non-cattle households was 7:3 the following calculation was
used
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 5
Chh = 0.7 x PV
NChh = 0.3 x PV
Where: Chh = Number of cattle households to be sampled from the village
NChh = Number of non-cattle households to be sampled from the village
The sub-counties, villages and enumerators were coded for systematic recording of data.
2.1.5 Enumerator training
Enumerator training was carried out separately in each of the 3 districts and included: an
overview of the project; survey techniques; detailed discussion of the questionnaire; and
discussion of the sampling techniques to be used. A practical session was also carried out for the
enumerators to practice carrying out the interview using the questionnaire. Areas for revision
were identified during these sessions and appropriate changes in the questionnaire were made.
2.1.6 Monitoring of the survey process
The questionnaires were collected regularly from the enumerators, reviewed for any missing
data, incorrect or inconsistent records and returned to the enumerator for correction if necessary.
Some of the irregularities could only be corrected by visiting the interviewed household again.
The questionnaires were reviewed for a second time after the corrections to ensure that the areas
with irregularities were clarified. In some instances questionnaires had to be returned as many as
four times. A total of 303 Questionnaires were administered.
2.1.7 GPS data collection
Each of the households that were included in the characterisation survey were visited to take a
GPS reading so that the households could be mapped and so that at a later stage GIS layers can
be used as parameters in the analysis. This was done on separate visits to the households after
completing the questionnaires. The waypoints recorded during this process were plotted in "Map
Source".
2.1.8 Data management
The data was entered into an access database extracted into excel and analysed in SPSS. Also
respective annual precipitation and population density data generated from the GIS readings have
been entered in the database. Analysis done include descriptive analysis, principle component
analysis and cluster analysis.
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 6
2.1.9 Participatory rural appraisal
Short PRA sessions were held with a selection of the farmers who were involved in the
characterization survey. Issues discussed during these sessions included the following, farmers
perceptions of farming systems in their areas and their attributes, resource flow maps, systems
diagrams, seasonal calendars, labour profiles and resource use and control.
3
Results
3.1
Farm tasks and household members
Among the tasks carried our by household members, the household head was found to contribute
labour more than other member except in the case of subsistence crop field activities.
Subsistence crop field management was mainly done by adult females within the household.
Results for this are shown in table 3.
Table 3: Roles of different household members (figure shown is the total frequency for which a
record indicates that that household member participates in the named activity)
Household
members
Grazing
animals
Cut & carry
of feeds
Household head
Adult
males
other
than
household head
Adult females
other
than
household head
Any adult in the
household
Any household
member
Male children
Female children
Male long-term
labourers
Female
long
term labourers
Male
casual
labourers
Female casual
labourers
80
14
3.2
Milking
Marketing
of milk
Other
livestock
activities
Subsistence
crops field
activities
Commercial
crop field
activities
50
7
Planting
and
weeding
forages
41
6
55
9
30
6
65
10
400
56
265
33
35
31
30
8
5
52
418
197
18
19
18
8
10
21
201
113
21
13
10
4
7
28
173
86
27
3
57
20
0
14
10
2
10
31
5
41
19
2
22
23
8
11
51
48
58
32
22
37
0
0
0
0
0
0
8
7
8
4
10
10
2
8
134
104
0
0
2
0
0
1
112
63
Use of organic and inorganic inputs
Both organic and inorganic inputs are minimally used among the farms. Organic inputs were
used by 21.1% of farm households and inorganic inputs by 1.41% of the farm households.Table
4 shows the detailed results.
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 7
Table 4:
Frequency of use of organic and inorganic inputs
Input
Manure
Slurry
Mulch
Plant only compost
Inorganic fertilizer
No use of inorganic fertilizers
No use of organic inputs
Frequency of use by %
14.54
1.95
3.10
1.33
1.41
98.59
78.90
The mean proportion of land manured by districts was also obtained and the largest proportion
was for Masaka district and the least for Mbarara district (Table 5).
Table 5:
Mean proportion of land manured by districts
District
Mean proportion of Standard
land manured
deviation
Jinja
0.20a
0.338
Masaka
0.57b
0.451
Mbarara
0.18a
0.424
NB: Values with the same letter indicate no significant difference at 5% level of significance
3.3
Land use
Mean acreages under bananas, coffee and natural pastures were highest in Masaka, Masaka and
Mbarara respectively. See figures 4-6.
3.4
Dairy production categories
Six categories were obtained from cluster analysis using major variables with inference on the
level of intensification. Clusters 1-4 were for farms with cattle while clusters 5-6 were for noncattle farms. Table 6 shows the major attributes of each cluster and table 7 shows the average
values for some of the variable used by clusters. Output from the PRA sessions held in the 3
districts showed that farmers tended to create a category of farmers (crop-livestock) with
indigenous cattle with no proportion of breed improvement in the herd. However all the clusters
generated from SPSS had a proportion of the herd improved. The farmers also associated high
dairy intensification with wealth however, there was a cluster (cluster 2) with very high dairy
intensification and low household income. These differences result from the number of factors
considered during the classification. SPSS package will obviously use as many variables as
possible to cluster and the farmers only use major factors. Farms have been selected from each
cluster for the longitudinal survey. Selection of farm households for the longitudinal survey has
been made and it includes households from each of the 3 districts.
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 8
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 9
Mean acreage of bananas in
pure stand (acres)
Figure 4:
Mean acreage under bananas in Jinja, Masaka and
Mbarara (means are not significantly different)
2.0
1.5
1.923
1.0
1.073
0.681
Masaka
Mbarara
Jinja
District
Mean acreage under coffee in pure
stand(acres)
Figure 5:
Mean acreage under coffee in Jinja, Masaka and Mbarara (There are significant differences.
Standard deviations for Mbarara, Masaka & Jinja are 0.509, 0.453 & 0.326 respectively)
1.2
1.1
1.0
0.9
0.8
0.7
1.143
0.6
0.528
0.547
0.5
Mbarara
Masaka
Jinja
District
Mean acreage of fallow and
natural pasture (acres)
Figure 6:
Mean acreage under natural pasture in Jinja,
Masaka and Mbarara (differences are not significant)
10
5
10.57
1.54
0.75
0
Mbarara
Masaka
Jinja
District
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 10
Table 6:
Cluster
1
2
3
4
5
6
Table 7:
Major attributes of each cluster (intensification is in terms of dairy production)
Intensification
High
Very high
Low
High
NA (non-cattle cluster)
NA (non-cattle cluster)
Household income
Medium
Low
Low
High
High
Low
Acreage
Medium
Small
Large
Very large
Small
Small
Major variables used for clustering and average values per farm for each
cluster
Clusters
Variables
1
2
3
4
5
6
Total no. of cattle
43
30
95
15
15
63
Proportion of cattle which is
improved
0.2
0.5
0.2
0.4
Household income class
3.9
2.8
2.5
5.4
4.4
2.5
Monthly off-farm income
62791
7167
5941
200667
103867
11695
Total cattle TLU
4.4
1.2
5.3
10.7
No. of small ruminants
2.7
1.7
3.1
6.1
0.9
1.1
No. of poultry (excluding
chicks)
4.3
2.6
3.5
3.7
4.5
2.4
No. of rabbits
0.8
0.5
0.0
1.0
2.7
0.2
No. of Pigs
0.7
0.4
0.5
0.0
2.1
0.3
Total land
7.5
4.5
12.4
28.0
4.3
3.9
Cropland
5.1
3.5
7.5
5.3
3.5
3.4
Cattle TLU/acre
1.0
0.4
0.8
0.6
Milk/day
4.8
3.2
5.3
9.7
Milk/cow
1.7
2.1
1.4
1.5
Milk/acre
0.8
0.9
0.5
0.8
Milk/cattle TLU
1.3
2.1
1.1
1.8
Money spent on AI+vet/TLU
17488
61222
12378
16086
AI-cost/TLU
4480
15976
2250
4506
Vet drug cost/TLU
13007
45245
10128
11581
Amount spent on fodder/TLU
189
16370
826
91
Amount spent on concentrates
& minerals/TLU
7574
7038
4002
9108
NB:
1. Household income classes: 1 = <Ushs 10,000, 2 = Ushs 10,000 - 30,000, 3 = Ushs 30,000 - 60,000, 4 =
Ushs 60,000 - 100,000, 5 = Ushs 100,000 - 200,000, 6 = >Ushs200,000
2. TLU scale used:
Age group
Cattle
Sheep & Goats
Adult male
1.0
0.1
Adult female
0.7
0.1
Weaners
0.5
0.07
Pre-weaners
0.2
0.03
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 11
4.
Conclusions and way forward
The analyses done reveal a number of issues regarding farming systems. First it is clear that there
are some tasks dominated by certain household groups for instance marketing of milk is
dominated by men and women take the lead in management of subsistence crops. Secondly use
of organic and inorganic fertilizers is generally very low. The general proportion of land
manured is also low particularly in Jinja and Mbarara. Thirdly dairy production systems can be
categorized on basis of level of intensification. In this study 3 levels of intensification have been
developed. It is now necessary to carry out data collection on a cross section of farms
representing the categories developed on a long-term basis. This will provide data required for
studying nutrient management within each category, identifying methods for improvement of
nutrient management and development of nutrient management indicators.
____________________________________________________
Presentation at LSRP annual review workshop: 18th - 21st March 2003
Page 12