Post on 24-Jul-2018
THE AGRICULTURAL MODEL INTER-COMPARISON AND IMPROVEMENT PROJECT (AGMIP)
END OF PROJECT REPORT
TENYWA MOSES
MAKERERE UNIVERSITY
1. Introduction
Uganda, located in eastern Africa has a population of 32.3 million people and one of the highest
population growth rates in the world at 3.56% per annum (UNDP, 2010). With 199,807.4 Sq.km
land area (accounting for 82.7% of all surface area) Uganda relies on its agricultural sector
contributing 23.2% of the GDP, and supporting more than 70% of the population for their
livelihood. Uganda’s industrial sector is composed mainly of food processing engaging up to
65% of total employment. The agricultural sector is supported by vast fertile soils suitable for
production of a range of food crops including banana, maize, beans, cassava, millet, rice, wheat,
fresh vegetables and pulses. Coffee, tea and tobacco are her main cash crops. Ugandans keep
animals (especially cows, goats and sheep) and crops (both cash and food crops) on vast soils
largely regarded as ‘fertile’. Farmers are grouped into three major categories: Subsistence
farmers (70%); Semi commercial farmers (25%) and Commercial farmers (5%) (AGRA, 2010).
The agricultural sector is also sustained by the diverse climate patterns and the influence of
several large rivers, bodies of water, and mountain ranges to the east and west.
Uganda, like many other countries in SSA is continuing to record very low levels of farm
productivity. Available evidence indicates that farm level yields are several times lower than at
agricultural research stations for similar crops. Farmers achieve between thirteen and thirty %
attainable yield at research stations (MAAIF, 1996). This gap is likely to be widened by climate
change and variability, if proper climate change adaptation practices are not put in place.
Furthermore, the use of some agricultural land for bio-fuel production, which is projected to
double in the next decade, is likely to reduce further the agricultural production and therefore
contribute to food insecurity. Biofuels are estimated to reduce greenhouse gas emissions by 10–
90 % relative to fossil fuels, depending on the type of feedstock and production technology.
Biofuels currently account for 0.2 % of total global energy consumption, 1.5 % of total road
transport fuels, 2 percent of global cropland, 7 percent of global coarse grain use and 9 percent of
global vegetable oil use. The increase in bio-fuel use has been driven largely by policy support
measures in the developed countries, seeking to mitigate climate change, enhance energy
security, and support the agricultural sector.
2. Methodology
2.1 Study area description
In Uganda, the study was conducted in two districts located in the Albertine region, namely
Hoima and Masindi, that straddle the Lake Albert bordering Uganda and DRC (Fig.1). The two
districts are located in the western Mid-Altitude Farmlands and Semliki Flats (MAFSF) Agro-
ecological zone (Wortmann and Eledu, 1992). Hoima district is located between 1° 00'-2° 00' N
and 30° 30'-31°45' E. It is bordered by Lake Albert in the west, Bundibugyo and Kibaale in the
south, Masindi in the northeast and Kiboga in the east. The district lies within an altitude range
of 621 m and 1,158 m above sea level, making it one of the lowest and hottest areas in the
country. It has an area of 5,932 sq km (Table 1), of which 2,268.6 sq km is occupied by water
bodies (mostly Lake Albert) and 712.3 sq km is forest. Masindi District is located next Hoima in
the Western Region of Uganda between 1o
22'-2o
20' N and 31o
22'-32o
23' E. It borders Gulu in
the north, Apac in the east, Nakasongola in the southeast, Kiboga in the south, Hoima in the
southwest and the Democratic Republic of Congo in the west. Both district lies at an altitude
range of 621m to 1,158m above sea level (Table 2). It comprises a total area of 9,326 sq km, of
which 8,087 sq km is land, 2,843 sq km wildlife-protected area, 1,031 sq km forest reserves, and
799.6 sq km water (Table 1).
Fig. 1. Map showing target areas AGMIP project in Hoima and Masindi in Uganda.
Table 1: Land use and area coverage in the districts of Hoima and Masindi
Particulars Land size (sq km)
Hoima Masindi
Total area 5,932.80 9,326.90
Broad leaved plantation 0.5 2.8
Conifers 4.3 1.1
Fully stocked forests 484.4 509.7
Degraded forests 267 20
Woodland 848.9 3,934.20
Bush land 86 267.30
Grassland 715.60 2,014.60
Swamp 58.20 130.40
Subsistence farmland 1,183.20 1,645.10
Large-scale farmland 12.90 108.90
Built up area 3.40 9.4
Open water 2,268.70 799.6
Open water (% of total district area) 38.24 8.46
Impediments (rocks) 0 0.1
Source: SCRIP (IFPRI, Kampala) for PRIME-WEST
The average annual rainfall in Hoima and Masindi ranges between 700-1,000 and 800-1630 mm,
respectively with a bi-modal distribution and peaks in March-May and August-November.
Table 2: Altitude and climate in the studied Agro-ecological zone of Uganda
Agro-ecology District Altitude (m) Mean
Temperature
(0C)
Annual Rainfall (mm)
Mid-Altitude Farmlands
and Semliki Flats
(MAFSF)
Masindi 621-1158 22.7-24.2 800-1630
Hoima 621-1158 23.4-25.6 700 -1000
Majority of the people in the region are smallholder farmers. The dominant farming systems in
the region are the western banana coffee cattle system and banana millet cotton system (Osiru,
2006). Major crops in the region under large-scale farming are maize, tea, sugarcane, while
small-scale farming includes beans, ground-nuts, rice, sweet potatoes, cassava, millet, pigeon
peas, banana, and simsim (Mubiru et al., 2007). However, the productivity of most smallholder
farms vary from high to low potential and depends on soil type and management. Most of the
farmers rarely use the inorganic fertilizer and a few apply organic manure.
Table 3: Characteristics of smallholder farms in the studied agro-ecological zone of Uganda
(AGMIP survey, 2012)
District Mean
Household
size
Mean
Fertilizer use
(kg N/ha)
Dominant maize
variety
Average
maize yields
(kg/ha)
Hoima 7 0 kg ha-1
Local (traditional) 1521
Longe 5 1725
Longe 9 1917
Masindi 6 0 kg ha-1
Local (traditional) 1478
Longe 5 1769
Longe 9 2128
2.2. Soils of the study area
The dominant soils in the region include Hoima catena (Petric Plinthosols), Naitondo series
(Dystric Regosols), and Kigumba series (Acric Ferralsol). The distribution of the three soils in
the districts of Hoima and Masindi is given in Figure 2. The productivity of these soils are
believed to be lowest in with Hoima catena, medium for Kigumba and highest for Naitondo.
Figure 2: Distribution of soils in Hoima and Masindi districts
The Hoima catena (a low productive soil unit) covers a large part of south-west Bunyoro
overlying the sedimentary beds predominated by tillites and phyllites with subsidiary counts of
sandstones and conglomerates as basal members (Table 4a). The soils are usually very acid, poor
in bases and available phosphorus and with lower amounts of organic material than their dark
colour would suggest.
Table 4a: Hoima profile of Petric Plinthisols (Low productivity)
Depth (cm) Silt
(%)
Clay
(%)
Ca
(m.e/100g
soil)
Mg
(m.e/100g
soil)
K
(m.e/100g
soil)
Na
(m.e/100g
soil)
Mn
(m.e/100g
soil)
Total
bases
(m.e/100g
soil)
Exch
H
(m.e
%)
CEC
(m.e)
% BS pH OC (%) Truog
P205
(ppm)
20 8 35 1.4 1 0.16 0 0.11 2.67 8.2 10.87 24.6 4.9 1.46 2
50.8 5 5 0.8 0.6 Tr. 0 0 1.4 9.7 11.1 14.4 5.1 0.59 2
Critical
values
5.2 0.5 0.22 <1.0 5.5 1.75
0.14%N in top 0-20 cm, C/N = 10.8 with low P205 and bases
The Dystric Regosols are believed to be a variant of the Hoima catena overlying the valley
slopes with deeper accumulation of soil material giving rise to reddish-brown clay loams. It
occurs at lower altitude. Magnesium and potassium contents are rather variable and calcium
seldom shows a deficiency. Soil reaction throughout is very acid which appears to be a fairly
constant feature of these phyllite soils even when bases are present in adequate amounts (Table
4b). Crop growth is poor on the shallow acid members of this catena and intensive cultivation is
confined mainly to the deeper soils on the valley slopes (Soil Memoirs, 1960).
Table4b: Naitondo Profile of Dystric Regosols of High Productivity
Depth (cm) Silt
(%)
Clay
(%)
Ca (m.e/100g
soil)
Mg
(m.e/100g
soil)
K (m.e/100g
soil)
Na (m.e/100g
soil)
Mn
(m.e/100g
soil)
Total bases
(m.e/100g
soil)
Exch
H (m.e
%)
CEC
(m.e)
% BS pH OC
(%)
Truog
p205
(ppm)
20 10 23 1.8 2.4 0.14 0 0.19 4.53 10.6 15.13 15.1 5.4 2.69 21
43 20 33 0.8 0.3 0.16 0 0.86 10 10.86 7.9 4.8 0.98 12
81 14 39 1 0.3 0.14 0 1.14 7.9 9.04 12.6 4.8 0.91 7
117 14 43 1.6 0.3 0.15 0 0.13 1.88 9 10.88 17.3 4.7 0.41 8
173 18 35 2 0 0 0 0 2 7.6 na na 4.8 0.31 6
203 20 23 0.4 0.3 0.13 0 0.02 0.15 5.5 5.65 2.7 4.9 0.05 7
Critical
values
5.2 0.5 <1.0 5.5 1.75
0.26%N for top 0-20 cm C/N 10.5 with high Mn & acidity as well
The Acric Ferralsols (Kigumba series) are reddish-brown loams to clay loams. This soil lies with
topography that appears as broad flat-topped or very slightly convex ridges separated by wide
swamp tracts. The pH values are fairly steady around 5.0, bases are low, deficient in magnesium
and the amounts reduce with depth (Soil Memoirs, 1960).
Table 4c: Kigumba profile under Acric Ferralsols of medium productivity
Depth (cm) Silt
(%)
Clay
(%)
Ca
(m.e/100g
soil)
Mg
(m.e/100g
soil)
K
(m.e/100g
soil)
Na
(m.e/100g
soil)
Mn
(m.e/100g
soil)
Total
bases
(m.e/100g
soil)
Exch
H
(m.e
%)
CEC
(m.e)
% BS pH OC
(%)
Truog
P205
(ppm)
13 6 26 2.7 0.8 1.27 0 0.03 4.8 7.4 12.2 39.4 5.1 2.78 29
46 4 40 0.6 0.3 0.63 0 0.1 1.33 8.3 9.63 13.8 4.8 0.92 2
84 8 42 1.3 0.3 0 0.3 0.12 2.02 5.6 7.62 25.5 5 0.51 5
122 8 42 1.1 0.4 0 0.5 0.08 2.08 4.3 6.38 32.6 5.3 0.35 7
152 4 38 1.1 0.3 0 0.5 0.03 1.63 4.2 5.83 28 5 0.22 7
183 4 42 1.1 0.7 0.08 1.1 0.02 3 4.5 7.5 40 4.9 0.2 12
213 4 32 0.4 0.3 0 +1 0.01 1 5.3 6.8 15.9 4.6 0.32 8
Critical
values
5.2 0.5 0.22 <1.0 5.5 1.75
0.14%N in top 0-13 cm, C/N 19.8 low P2O5 & bases
2.3 Data and methods of assessment
The assessment used soil conditions that are homogeneous with regard to their capacity to support
production of a wide range of food and cash crops as the unit for evaluating the impacts of climate
variability and change. Relevant data required to calibrate, validate and apply climate, crop and
economic models was collected publications (Kaizzi et al., 2004), East African Seed, Soroti Catholic
Diocese Integrated Development Organisation (SOCADIDO), and Sango Bay maize farm. In addition, a
survey was carried out in all the target regions to characterize the smallholder farming systems (farms
enterprises, management, productivity, sources of non-farm income). Table 5 shows the number of
households surveyed per district.
Table 5: Sampled households in each AEZ
Country AEZ District Number of HHs
Uganda MAFSF Hoima 76
Masindi 231
2.3.1 Climate Data and Scenarios in Uganda
Long-term daily observed historical climate data (1980-2010) was obtained from the Department of
Meteorology in Uganda for Masindi district. For Hoima district (Kyangwali), the NASA’s Modern Era-
Retrospective Analysis for Research and Applications (MERRA) (Rienecker et al., 2011) was
applied. MERRA is the NASA’s climate model data repository. MERRA data are produce using
their 4-Dimensional Variational (4D-Var) assimilation system and were provided in AgMIP
extension (Ruane and Goldberg, in preparation). MERRA data were used to fill temperature gaps
for Masindi and its rainfall data from Kyangwali was adjusted using ten year data obtained from
Bulindi weather station. The period 1980-2009 was considered as reference period because of the
availability of complete MERRA climate data sets. The data collected include rainfall, minimum
and maximum temperature, wind, relative humidity and solar radiation. This information was
used to characterize the variability in the observed climate, develop future scenarios and also for
use in DSSAT and APSIM crop simulation models. The minimum data set required for this
analysis included daily records of rainfall, minimum and maximum air temperatures and solar
radiation.
In order to assess the climate change scenarios, we used a significance test (Δs), that is the delta
method. The delta method was used to downscale the different projected climate scenarios for
both Masindi and Hoima districts.
2.3.2 Household survey -sampling scheme
The number of households surveyed in three different soil units were 87 for Hoima catena, 98 for
Kigumba series, and 122 for Naitondo. The summary statistics from the household survey were
compiled. The variables generated included: yield, planting dates, planting density, harvesting
dates, soil type, moisture levels, soil fertility, fertilizer application rates, manure application
rates, maize variety, plot area, days to harvesting, seeding rate and yield. These variables were
provided for each of the 308 farm households interviewed for the survey for five crops of
interest: Maize, Beans, Cassava, Groundnuts and Banana. Two observations were excluded due
to incomplete information on all variables. Labour costs were computed using labour
requirement for men, women and children involved in various activities. Family labor was
excluded from the computation of variable labor costs. All monetary values were converted to
US dollars using the prevailing exchange rate. Below are summary statistics (means, standard
deviations and coefficients of variation) of prices, yield, acreage, total variable costs and net
returns from the 5 enterprises (Table 6).
Table 6: Summary statistics: Prices, output, area and yield of crops – Survey Data Measure Maize Beans Groundnuts Cassava Banana
Prices Shs/Kg 492.8 1152.978 1404.2 607.17 306.94
$/kg (1$=2500) 0.197 0.46 0.56 0.24 0.12
Standard Dev 172.34 423.54 603.73 321.86 149.71
Coefficient of Variation 34.97 36.73 42.99 53.01 48.78
Area Mean (ha) 0.63 0.34 0.35 0.41 0.07
Standard dev 0.37 0.11 0.09 0.13 0.075
Coefficient of Variation 57.83 32.79 23.95 30.57 107.11
Yield Mean (Kg/Ha) 1,892.64 1,266.74 755.39 1,516.90 5,030. 6
Standard dev 3,268.59 6,231.54 785.43 2,349.39 7,989.9
Coefficient of Variation 172.70 491.94 103.98 154.88 158.83
TVC Mean ($/farm) 94.67 82.21 75.25 402.13 78.59
Standard dev 140.86 111.79 110.22 1377.40 322.50
Coefficient of Variation 148.79 135.98 146.47 342.52 410.38
Net returns Mean 429.98 316.11 259.35 50.43 100.64
Standard dev 867.25 1329.20 269.66 480.86 264.49
Coefficient of Variation 201.70 420.48 103.98 953.47 262.80
As shown by the high values of coefficient of variation in Table 1 above, variability in these
variables was high due to a number of reasons: seasonality, subsistence production and poor
information flow. First, agricultural prices in the study area and in Uganda in general tend to
depress during the peak season (bumper harvest) and rise during the off peak season. Farmers’
level of knowledge regarding prices and whether the farmer participates in the market either as a
buyer or as a seller affects prices. In most rural areas of the country, farmers hardly participate in
the market. In other words, most farmers carry out subsistence production hence because no
participation in the market they may not have a true picture of what is prevailing in the market.
Hence the high variability may be due the few farmers who know exactly what is prevailing in
the market and the majority of the farmers who do not what may be prevailing in the market. The
situation is further compounded by market imperfections in the rural areas; because the
equilibrium price between the farmer and the trader is largely dependent on each other’s level of
knowledge regarding the market situation. Variability in output per farm is brought about by
some farmers adopting/using improved technologies which are yield enhancing resulting into
high yields compared to other farmers who were using traditional technologies which are less
rewarding hence low yields. The high variability in yield is also attributed to seasonality in
agricultural production and weather conditions. Very good climatic conditions lead to good
production whereas poor climatic conditions lead to poor yields.
Data handling involved annualizing seasonal yields, costs, revenues and returns. The raw data
was plotted into graphs to check for outliers and extreme values were replaced with the stratum
means.
2.3.3 Crop and soil Data calibration
Two crops models that is DSSAT and APSIM, were used to simulate the likely impact of climate
change on maize yield as depicted by the twenty GCMs downscaled in the study area for the
different projected climate scenarios. The two crop models were calibrated using data from
previous experimental research conducted in Uganda (Kaizzi et al., 2012). During the survey,
information on soil and crop management including varieties, inorganic fertilizer inputs, planting
windows/date, number of plants per hectare, days to maturity, amount of manure were collected.
In addition, each household farm was geo-referenced and the coordinates over-laid on soil maps
of the study area to extract information on type of soil. The number of households surveyed in
the three soil units were 87 for Hoima catena (Petric Plinthic), 98 for Kigumba series (Acric
Ferralsol), and 122 for Naitondo (dystric Regosols). In each soil unit, household farms were
categorised into good, medium, poor fertility based on farmers’ perceived historical land use
productivity. All soil profile data were obtained from the Uganda soil memoirs (Table 4a, 4b,
4c). For the model runs, each soil type in the agro-ecological zone was categorised into three
levels, that is, low, medium and high productive soil types based on total soil organic carbon
content (Table 7).
Table 7. Soil profile characteristics used in DSSAT and APSIM crop models in Uganda
Soil
layer
base
depth
Lower limit Drained
upper limit
Saturation Bulk
density
Organic C Clay Silt Coarse
fraction
(stones)
pH Cation
exchange
capacity
Sat. hydraulic
conductivity,
macropore,
cm mm3/mm3 mm3/mm3 mm3/mm3 g/cm3 g[C]/100g % % % cmol/kg cm h-1
Petro Plinthic Good 15 0.155 0.366 0.512 1.290 2.770 23.00 8.00 0.00 5.50 3.50 0.10
35 0.172 0.370 0.496 1.330 1.530 29.00 10.00 0.00 5.60 3.50 0.18
70 0.205 0.400 0.503 1.320 1.100 37.00 8.00 0.00 4.40 3.50 0.13
101 0.232 0.423 0.585 1.360 0.460 43.00 4.00 0.00 4.60 3.70 0.10
Medium 15 0.153 0.359 0.502 1.320 2.355 23.00 8.00 0.00 5.50 3.50 0.10
35 0.171 0.368 0.491 1.350 1.301 29.00 10.00 0.00 5.60 3.50 0.18
70 0.204 0.399 0.598 1.330 0.935 37.00 8.00 0.00 4.40 3.50 0.13
101 0.232 0.423 0.596 1.340 0.391 43.00 4.00 0.00 4.60 3.70 0.10
Poor 15 0.150 0.351 0.490 1.350 1.939 23.00 8.00 0.00 5.50 3.50 0.10
35 0.170 0.365 0.486 1.360 1.071 29.00 10.00 0.00 5.60 3.50 0.18
70 0.204 0.398 0.593 1.340 0.770 37.00 8.00 0.00 4.40 3.50 0.13
101 0.232 0.423 0.593 1.340 0.322 43.00 4.00 0.00 4.60 3.70 0.10
Dystric regosols Good 15 0.160 0.406 0.561 1.160 3.228 23.00 10.00 0.00 5.40 15.13 0.10
43 0.189 0.408 0.524 1.260 1.176 33.00 20.00 0.00 4.80 10.86 0.18
81 0.216 0.425 0.530 1.250 1.092 39.00 14.00 0.00 4.80 9.04 0.13
117 0.235 0.444 0.613 1.290 0.492 43.00 14.00 0.00 4.70 10.88 0.10
173 0.195 0.403 0.593 1.340 0.372 35.00 18.00 0.00 4.80 5.65 0.10
203 0.138 0.335 0.448 1.460 0.060 23.00 20.00 0.00 4.90 5.65 0.18
Medium 15 0.164 0.399 0.553 1.190 2.690 23.00 10.00 0.00 5.40 15.13 0.13
43 0.188 0.405 0.515 1.290 0.980 33.00 20.00 0.00 4.80 10.86 0.10
81 0.216 0.424 0.525 1.260 0.910 39.00 14.00 0.00 4.80 9.04 0.10
117 0.235 0.444 0.610 1.200 0.410 43.00 14.00 0.00 4.70 10.88 0.18
173 0.194 0.403 0.591 1.350 0.310 35.00 18.00 0.00 4.80 5.65 0.13
203 0.138 0.335 0.448 1.450 0.050 23.00 20.00 0.00 4.90 5.65 0.10
Poor 15 0.159 0.384 0.534 1.230 2.152 23.00 10.00 0.00 5.40 15.13 0.10
43 0.187 0.403 0.510 1.300 0.784 33.00 20.00 0.00 4.80 10.86 0.18
81 0.215 0.422 0.616 1.280 0.728 39.00 14.00 0.00 4.80 9.04 0.13
117 0.235 0.444 0.605 1.310 0.328 43.00 14.00 0.00 4.70 10.88 0.10
173 0.194 0.402 0.588 1.360 0.248 35.00 18.00 0.00 4.80 5.65 0.10
203 0.138 0.335 0.448 1.460 0.040 23.00 20.00 0.00 4.90 5.65 0.18
Acric Ferralsol Good 13 0.175 0.405 0.564 1.150 3.336 26.00 6.00 0.00 5.10 12.20 0.13
46 0.219 0.411 0.525 1.260 1.104 40.00 4.00 0.00 4.80 9.63 0.10
84 0.229 0.427 0.613 1.290 0.612 42.00 8.00 0.00 5.00 7.62 0.10
122 0.228 0.426 0.604 1.310 0.420 42.00 8.00 0.00 5.30 6.38 0.18
152 0.208 0.396 0.587 1.360 0.264 38.00 4.00 0.00 5.00 5.83 0.13
183 0.181 0.364 0.471 1.400 0.240 32.00 4.00 0.00 4.80 7.50 0.10
213 0.182 0.367 0.479 1.380 0.384 32.00 4.00 0.00 4.60 6.30 0.10
Medium 13 0.175 0.402 0.561 1.160 2.780 26.00 6.00 0.00 5.10 12.20 0.18
46 0.219 0.410 0.518 1.280 0.920 40.00 4.00 0.00 4.80 9.63 0.13
84 0.228 0.426 0.609 1.300 0.510 42.00 8.00 0.00 5.00 7.62 0.10
122 0.228 0.426 0.601 1.320 0.350 42.00 8.00 0.00 5.30 6.38 0.10
152 0.208 0.395 0.585 1.370 0.220 38.00 4.00 0.00 5.00 5.83 0.18
183 0.181 0.363 0.468 1.410 0.200 32.00 4.00 0.00 4.80 7.50 0.13
213 0.182 0.366 0.476 1.390 0.320 32.00 4.00 0.00 4.60 6.30 0.10
Poor 13 0.171 0.388 0.543 1.210 2.240 26.00 6.00 0.00 5.10 12.20 0.10
46 0.218 0.409 0.511 1.290 0.736 40.00 4.00 0.00 4.80 9.63 0.18
84 0.228 0.426 0.604 1.310 0.408 42.00 8.00 0.00 5.00 7.62 0.13
122 0.228 0.425 0.599 1.330 0.280 42.00 8.00 0.00 5.30 6.38 0.10
152 0.208 0.395 0.582 1.370 0.176 38.00 4.00 0.00 5.00 5.83 0.10
183 0.181 0.363 0.468 1.410 0.160 32.00 4.00 0.00 4.80 7.50 0.15
213 0.181 0.363 0.470 1.410 0.256 32.00 4.00 0.00 4.60 6.30 0.16
Soil parameters such as hydraulic conductivity, Drainage Upper Limit (DUL) and Lower Limit
(DLL), air dry, saturation, bulk density were estimated using soil water characteristics Program
(Rawls et al., 1982) based on soil texture, soil organic carbon content, gravel content, salinity
and compaction. The soil profile depth in the models was arranged to a fixed depth per layer
(Table 7). In APSIM, values of soil organic carbon pools (Fbiom, Finert, HumC, BiomC,
InertC) as described in the model were dependant on SOC. In DSSAT, initial soil conditions of
stable SOC were set at 44%. The C/N ratio in both models was derived by calculation.
In DSSAT, site descriptions such as slope per cent, run-off curve numbers, drainage rate, run-off
potential, and soil fertility factor were estimated based on available literature (Soil memoirs). All
root growth factors were set at 1.0 as non-limiting in all soil profiles.
2.3.4 Crop management based on on-farm surveys
Crop management parameters used in setting up simulations for individual farms were derived
from the results of the survey conducted in Hoima and Masindi districts. Parameters used in
setting up on-farm simulations were derived from data collected in the survey. Varieties that are
most dominant in the region are local traditional, longe 5, and longe 9. Detailed Information on
Longe 5 and Longe 9 varieties was obtained from NASECO seeds. For the local traditional
variety, Katumani information was used because of similarities in physiology, phenology and
yield potential.
Table 8: Maize varieties used by farmers and the identified equivalent in the model
Variety used by farmer Duration (days) Grain Yields
(t ha-1
)
Variety in the Model
Nafa, Ndele, Longe 1,
Longe 4
100-105 1.5-1.8 Local traditional
(Katumani)
Longe 5, dk 115 4-5 Longe 5
Longe 2H, Longe 6H,
Longe 10H
120 7-8 Longe 9
Others Considered as local Local- traditional
Katumani
Source: NASECO seeds Ltd, Kampala, Uganda
Plant population that were estimated for each individual farms were set based on the amount of
seed used by farmers. The plant population normally ranged between 40,000 to 60,000 plants per
hectare. Three major categories were used ad per Table 7. In the survey, majority of the farmers
did not use inorganic fertilizers. While setting up the models for individual farmers, 5 kg N
fertilizer amounts was applied to allow model sensitivity. All model runs for the survey data was
conducted using AGMIP tools QuadUIi and ACMO UI that can be availed at
www.tools.agmip.org.
Table 9: Number of farmers using different plant populations in the three soil types in Uganda
Plant population
(plants/ha)
Acric Ferralsols Petric
Plinthisols
Dystric
Regosols
Total
40,000 60 69 53 162
50,000 18 16 32 66
60,000 20 18 21 59
Source: Uganda AGMIP survey data, 2012
2.3.5 Crop Model calibration and validation
APSIM and DSSAT Models were calibrated for Longe 5 and Longe 9 using experimental data
from Bulindi Zonal Agricultural Research Institute in Hoima (Kaizzi et al., 2012). Validation of
model outputs was done using information from National Agricultural Research Laboratories,
Kawanda (NARL). Both experiments were conducted for the two rain seasons of 2010. Crop
management and growth measurements over the seasons and growing period were conducted as
described by Kaizzi et al. (2012). All model parameter coefficients for DSSAT and APSIM
were derived after adjusting existing varieties typical of tropical conditions (Table 10 and 11)
and also by manipulating various growth parameters until the simulated phenology and yields
matched the observed. Models were run and relationship between the observed and simulated
was demonstrated using observed and simulated days to flowering, days to maturity, stover and
grain yield (Figure 3). All APSIM and DSSAT model runs were compared to observed values
using the highest fertilizer rates of 100 and 150 kg N ha-1
since model runs were under non-
limiting conditions of phosphorus, acidity, micronutrients and pH, thus providing highest yield
potential.
Table 10a: Data for calibration and validation of the crop models (Longe 5) in Uganda
Treatmen
t (N
kg/ha)
Stover
weight
Grain weight TOP
weight
Planting
date
(2009)
Anthesis
date
Physiological
maturity Date
Harvest
Date
2009B Julian Days
Long rain
(longe 5)
1 0 3.83 1.32 5.15 252 306 2(2010) 40 (2010)
2 50 4.69 2.96 7.65 252 306 2(2010) 40 (2010)
3 100 4.86 3.43 8.29 252 306 2(2010) 40 (2010)
4 150 5.23 3.54 8.77 252 306 2 (2010) 40 (2010)
2010A
Short rain
(Longe 5)
5 0 4.55 1.47 6.02 78 136 203 234 (2010)
6 50 7.25 3.66 10.91 78 136 203 234 (2010)
7 100 7.35 3.5 10.85 78 136 203 234 (2010)
Sourced from Kaizzi et al., 2005
Table 10b: Data for calibration and validation of the crop models (Longe 9 and local traditional
varieties) in Uganda 2010A Treatmen
t (N
kg/ha)
Stover
weight
Grain
weight
TOP
weight
Planting
date
(2009)
Anthesis
date
Physiological
maturity Date
Harvest
Date
Julian Days
Long rain
(longe 9)
1 0 5.25 4.16 9.41 78 136 199 230 (2010)
Long rain
(longe 9)
2 100 5.15 5.20 11.35 78 136 199 230 (2010)
Sourced from NASECO seeds Ltd
Long rain
(Local
tradition)
1 0 3.25 1.16 4.31 78 137 208 236 (2010)
Long rain (local
tradition)
2 100 4.33 3.02 7.35 78 138 208 238 (2010)
Sourced from AGMIP on-farm survey
Table 11. DSSAT model calibration crop coefficients for the three varieties in Uganda.
Cultivar
ID
Variety name P1 P2 P5 G2 G3 PHINT
IC0005 Long 5 200.8 . 0.500 508.5 450.0 10.50 45.60
IC0006 Longe 9 208.6 0.500 554.0 460.0 10.50 45.00
IC0007 Local tradition
(Katumani)
180.0 0.600 600.0 680.0 8.70 40.0
Where P1 = Thermal time from seedling emergence to the end of the juvenile phase (expressed in degree days above a base temperature of 8 oC) during which the plant is not responsive to changes in photoperiod. P2=Is the extent to which development (expressed as days) is delayed for each hour increase in photoperiod above the longest photoperiod at which development proceeds at a maximum rate (which is considered to be 12.5 hours P5= is the Thermal time from silking to physiological maturity (expressed in degree days above a base temperature of 8 oC).\ G2 = Maximum possible number of kernels per plant G3 is the Kernel filling rate during the linear grain filling stage and under optimum conditions (mg/day). PHINT is the Phylochron interval;
the interval in thermal time (degree days) between successive leaf tip appearances.
Table 12. APSIM model calibration crop coefficients for the three varieties in Uganda
Katumani (Uganda
local)
Longe 5 Longe 9
hi_incr (1/days) 0.018 0.018 0.018
hi_max_pot (g/g) 0.55 0.55 0.55
head_grain_no_max () 450 610 750
grain_gth_rate (mg/grain/day) 10.5 9.17 8.0
tt_emerg_to_endjuv (o C day) 150 210 240
est_days_endjuv_to_init () 20 20 20
pp_endjuv_to_init 10 10 10
tt_endjuv_to_init ( oC day) 0.0 0.0 0.0
photoperiod_crit1 (hours) 12.5 12.5 12.5
photoperiod_crit2 (hours) 24.0 24.0 24.0
photoperiod_slope ( oC/hour) 10.0 10.0 0
tt_flower_to_maturity (oC day) 660 660 980
tt_flag_to_flower (oC day) 10 10 50
tt_flower_to_start_grain ( oC
day)
120 120 120
tt_maturity_to_ripe (o C) 1 1 1
Figure 3. Simulated and observed yield and top weight for Longe 5, Longe 9 and local traditional
variety
2.3.6 Model sensitivity for APSIM and DSSAT in Uganda
Table 13 compares the average maize yields over 30 years simulated by the two models under different
climatic conditions under good soil conditions, 50 kg N ha-1 fertilization and longe 9 variety. APSIM
simulated high grain yield as compared to DSSAT. Both models were sensitive to changes in temperature
and rainfall. Simulations with APSIM showed an increase in yield parameters with increase temperature
and rainfall compared to the baseline. However, DSSAT showed an opposite trend.
Table 13: Model sensitivity to changes in climatic conditions in Uganda
Treatment
APSIM DSSAT
Biomass Yield
(kg/ha)(CV)
Grain Yield
(kg/ha)(CV)
Biomass Yield
(kg/ha) (CV)
Grain Yield
(kg/ha)(CV)
Effect of temperature and rainfall
Base Climate 5748 (18.9) 2753.72 (15.13) 11087 (6.8) 3126 (12.2)
Base+10C
9007 (7.2) 4755.15 (12.4) 10716.8 (7.4) 2924 (12.1)
Base+30C
7901(8) 4136(15) 10579 (6.9) 2825 (15.3)
Base+50C 6926(9) 3431(19)
9139 (11.52) 1971 (20)
Base+10C+10%RF
9060 (5.5) 4843(11.5) 7774 (7) 2270.8 (9.3)
Base+30C+10%RF
7976 (6.4) 4227 (12.9) 7678 (7) 2088 (10.6)
Base+50C+10%RF
7020 (7.4) 3486 (17.2) 7334 (8) 1666 (12.2)
Base+10C-10%RF
8948 (10.9) 4659 (18) 8310 (41.7) 2061.3 (59.8)
Base+30C-10%RF
7793 (11) 4070 (17) 853 1 (38.5) 2174.9 (52.7)
Base+50C-10%RF
6833 (11) 3378 (20.4) 9342 (22) 2664.5 (22.8)
2.3.7 Comparison of DSSAT and APSIM model runs with on-farm observed for Uganda
After the calibration, APSIM and DSSAT model were used to simulate the yields of 307 households
farms that were conducted in the studied agro-ecological zone in Uganda. The on-farm survey was set up
by considering specific on-farm climate, soil, and crop and management parameters. In all, simulated
yields of DSSAT were much higher than that of APSIM. There was no relationship between the on-farm
yield and the two models simulated values with R2<0.2 (Figure 8).
Figure 4: Relationship between DSSAT and APSIM simulated yields and yields reported in on-farm
survey in Uganda (Red dotted line is 1)
2.4 Socio-economic impact assessment
TOA model was used to assess the socio-economic impacts of future climate to the households
in the studied area. For parameterization of system 1, the survey yield was normalized to the
current period by computing a unitless factor β obtained as a ratio of historical yield to the
average per farm survey yield. This factor was used to adjust the survey yield thus:
ij = βx per farm survey yield in kg per farm
The per farm output obtained asij x Area was used to obtain per farm net returns given prices
and costs of operation for each crop activity in the farmer’s enterprise mix.
To parameterize system 2, outputs, costs, revenues and net returns were computed using the
following procedures:
1. The per farm output for system 2 was obtained by factoring system 1 output with the per
farm relative yield.
2. The per farm output for system 1 was obtained by adjusting system 1 per farm output
using the relative yield obtained as a ratio between the simulated future yield and the
simulated baseline yield.
3. Crop returns were computed as the product of prices, area and per farm system 2 output.
4. The costs to returns ratio was used to adjust system 1 costs to obtain costs per returns for
system 2
5. Net revenue for system 2 was obtained as the difference between revenue and costs
To obtain results for the first objective, the model was set up with system 1 being the current
production system under current climate while system 2 was future climate under the unchanged
production system (i.e the case of no adaptation to climate change) thus:
Climate change without adaptation: System 1 = base climate, base technology (1980-2010)
System 2 = changed climate, base technology (2040-70)
The second objective was achieved using two procedures: The first procedure was to establish
the expected trend in producer prices and rain-fed production using global economic models.
The next stage was to conduct representative agricultural pathways (RAPs) to determine the
changes both in the direction and magnitude of key socio-economic variables including
enterprise composition, production, labor and other costs, area, changes in demographic
composition). This analysis scenario involved setting up TOA with these systems
Climate change with trend: System 1 = base climate, base technology (1980-2010)
System 2 = changed climate, changed technology (2040-70)
The above processes and procedures generated (5 climate models x 2 crop models x 2 economic
scenarios =) 20 sets of TOA-MD simulations.
To compute the growth rates of each variable X, the following three-part procedure was
performed:
1. Obtaining the change in growth of the variable
ΔXt Where X = Price, Productivity, Acreage,
Costs
2. Obtaining the mean of the growth rate
ɸ
3. Computing the mean over the period of analysis
ɣ=(1+ ɸ)T
This process resulted into the following growth rates which were then used to adjust system 1
variables, supplemented with information from the RAPs and from literature including published
reports and databases. Table 5 shows the assumptions, data sources and variables used for system
2.
Table 14: With Adjustment factor (for Trend)
Activity Productivity Prices Acreage Costs
Maize Factor +495% +220% +5.6% +213%
Source Global models Global models RAPs
+20% in farm size
per farm but 12%
decline in number of
farms
Global models and
RAPs
Beans Factor +50% 60% +5.6% +10%
Source RAPs RAPs RAPs RAPs Rise in
agricultural wages
Groundnuts Factor +190% +101% 5.6% +194%
Source Global models Global models RAPs Global models and
RAPs
Cassava Factor +7% 7% 5.6% +10%
Source Literature RAPs RAPs Rise in agricultural
wages
Banana Factor -10% 0% -21% +35%
Source RAPs
Diseases
RAPs
Most production is
not value added,
Not traded on
international market
Shift to less land-
demanding
enterprises
RAPs
Disease management
Rise in agricultural
wages
3. Integrated Assessment Results
3.1 Climate characteristics of the studied Agro-ecological zone
Masindi and Hoima district receive more than 1000 mm annually with Masindi slightly higher than
Hoima. In both locations, rainfall is bi-modal with short-rain season (August-November) receiving
slightly higher rainfall than the long-rain season (March-May) (Table 11). The difference in rainfall
amounts between the two seasons is more pronounced in Hoima (39.6% higher) than in Masindi (2.2%).
Generally, annual rainfall amounts at both sides show a low temporal variability with a CV of <16%
(Table 4). This was even lower for both seasons with a CV less than 10%. Temperature tends to be
relatively higher in Hoima than in Masindi, and during the long rain season than the short rain season.
Annual rainfall amounts were not showing any trends overtime. However, analysis of variability of long
rains in Hoima showed a rapid decline in the CV from 30% in 1980-1985 to less than 15 % in 1990 and
later remained stable up to 2000 before another increase to 30% in 2005. For the short rains, CV slightly
increased gradually from 13% to 19% in 1995, declined to about 10% in 2000 and increased to 20% in
2005. In Masindi, the long-rain CV fluctuated between 20 and 15% up to 1990. Since then, it stabilised
to about 20%. For the short-rain, CV declined gradually from 27% to 15% around 1990; increased
slightly, and then fluctuated slightly within 18- 21% (Table 15)
Table 15: Key climate characteristics at the studied agro-ecological zones in Uganda
Variable Hoima Masindi
Avg annual rainfall (mm) 1197.0 (15.5) 1292 (12.3)
Average LR season rainfall (mm) 346.9 (7.7) 412 (6.3)
Average SR season rainfall (mm) 483.8 (6.9) 421 (7.2)
Average annual Temperature (0C) 24.3 23.5
Average annual MaxT (0C) 30.0 29.2
Average annual MiniT (0C) 18.6 17.8
Average LR season Temperature (0C) 24.7 24.2
Average LR season MaxT (0C) 30.3 29.6
Average LR season MiniT (0C) 19.07 18.7
Average SR season Temperature (0C) 23.8 23.1
Average SR season MaxT (0C) 29.3 28.6
Average SR season MiniT (0C) 18.4 17.5
Note: Figures in parenthesis represent Coefficient of Variation (CV)
Generally, average temperature within the short and long rain seasons were increasing in both locations of
Masindi and Hoima (Figure 5). The increase in temperature became pronounced in Hoima sites in 1997
for both seasons. In Masindi, temperature increase during the short rain is observed since 1981 (Figure
6).
Figure 5: Ten year moving coefficient of variation (CV) of rainfall starting from 1980 during short rain
season at the two sites in Hoima and Masindi districts Uganda
Figure 6: Average air temperature during short and long rain season during the period 1980-2010 (bars
represent observed average values and line represents five year moving average) in Hoima and
Masindi district, Uganda
3.2 Future climate projections
The projected climate change and variability in both Masindi and Kyangwali are presented in Figure 5
and Figure 6. Most of the models projected an increment in rainfall, Tmax and Tmin for both sites,
except CSIRO-Mk3-6-0, MIROC5 and MIROC-ESM which predicted negative change in rainfall. However,
the magnitude of the change varied from one model to another. High value rainfall increments were
predicted by BNU-ESM, GFDL-ESM2M, IPSL-CM5A-LR, IPSL-CM5A-MR, lowest rainfall increments were
predicted by MRI-CGCM3, Inmcm4, CSIRO-Mk3-6-0, bcc-CSM1 and Access1-0 for both sites. High values
of change in temperature were predicted by ACCESS1-0, CanESM2, bcc-CSM1-1, CSIRO-Mk3-6-0,
HadGEM2-ES, HadGEM2-CC, IPSL-CM5A-LR, IPSL-CM5A-MR and MRI-CGCM3. HadGEM2-CC predicted
small negative change in rainfall for 4.5 Mid Century and 4.5 End Century, while HadGEM2-ES, predicted
same trend for 4.5 Mid Century.
Figure 7: Projected change in Rainfall, Minimum and Maximum Temperature of Masindi district,
Uganda
Figure 8: Projected change in Rainfall, Minimum and Maximum Temperature of Hoima district,
Uganda
3.2 Sensitivity of current agricultural production systems to climate change
Simulations were carried out with both DSSAT and APSIM for baseline and climate change scenarios for
all combinations of RCPs 4.5 and 8.5 and time periods mid and end centuries for all the 20 AOGCMs.
Generally, both APSIM and DDSAT predicted that maize yields will reduce for all soils and seasons under
future climate scenarios, except for the petric plinthisols were there will be a relative increment for RCP
4.5. Magnitude of the reduction varied from one crop model to another. DSSAT seems to show higher
relative changes than APSIM for RCP 8.5.
Fig 9: APSIM predicted future maize yield
-35
-30
-25
-20
-15
-10
-5
0
Both LR SR Both LR SR Both LR SR
Acric Ferralsol Dystric regosol Petric Plinthisols
% D
evia
tio
n
DSSAT
4.5MID
4.5END
8.5MID
8.5END
Fig 10: DSSAT predicted future maize yield
3.3 Socio-economic implications
Table 15 below shows the means, standard deviations of the variables disaggregated by strata.
Three strata based on soil type were identified.
Table 15: Yields, Total variable costs and Net returns disaggregated by soil type
Soil Type Maize Beans Groundnuts Cassava Banana
Total Variable Costs ($/ha) Acric 95.99 98 95 240 50
Dystric 68.91 72 59 624 81
Petric 129.41 78 76 273 107
Yields (Kg/ha) Acric 1685.00 761.54 688.02 1040.65 3952.57
Dystric 2043.13 1846.13 818.42 1011.51 6385.39
Petric 1917.22 1029.99 743.61 932.51 4360.55
Mean net returns Acric 407.39 198.67 231.05 24.62 115.48
Dystric 446.15 449.70 286.39 40.73 92.81
Petric 432.94 262.61 253.62 93.00 94.83
Standard Deviation of net
returns Acric 750.16 162.48 177.95 495.51 326.57
Dystric 1126.65 2063.96 315.61 584.56 179.44
Petric 510.45 498.30 284.98 252.16 287.23
Fig. 11: Total variable costs of Maize
Fig. 12: Survey data – Maize yields
Fig. 13: Comparison between the average of DSSAT yields and baseline yield
Table 15 above shows that by comparison, APSIM simulated yields were lower than DSSAT
simulated yields. Figure 14 below shows that with the exception of the model CCSM4, all the
other models show no significant divergence between average DSSAT yields of the three
different soil types. It is evident that although the average simulated yields of CCSM4 are higher
than both the baseline yields and observed yields, they are lower than simulated yield averages of
the other 4 models – GFDL, HadGEM-2, MIROC-5 and MPI-ESM. In fact, a further assessment
shows that simulated average yields of GFDL and MIROC-5 are not different from each other
implying that these two models predict similar yields. Figure 6 below also shows a near perfect
match for HadGEM-2 and MPI-ESM. These comparisons indicate that simulation results are
likely to be similar for models that predict similar patterns and for the three different soil types.
In addition, these yields may be an indication that the study population was homogenous in terms
of soil type – that yields across soil types do not vary significantly. It is important to note that
within each soil type, there were variations in quality ranging from poor to good in which case
comparable yields were obtained from different soil types of relative equality.
Fig 14: Comparison of DSSAT simulated Base scenario yields by soil type/stratum
Fig 15: Comparison of APSIM simulated Base scenario yields by soil type/stratum
Fig 16: DSSAT Simulated yields in comparison with observed by stratum
Fig 17:.APSIM Simulated Base scenario yields in comparison with observed by stratum
The figures below show results of TOA model simulations using DSSAT yields. With the
exception of CCSM4, all other models predict a higher percentage of gainers within the stratum
Petric followed by Dystric and lastly Acric Feralsols. The four models predict that the gainers (as
a percentage of farms) are fewer in acric soil type compared to the other two soil types.
The table below shows results of DSSAT model runs for the Base Scenario. The per capita
income ranges from 461.56$ per person per year to 477.63$ per person per year depending on
the GCM. GFDL model predicts the lowest poverty rate of 51.08% compared with 52.09%
predicted by the CCSM4 model.
Table 17: Gainers, net returns and poverty rate
GCM Model
Gainers
(%)
Net Returns
(%/person/year)
Population
Poverty Rate (%)
Percapita income
($/person/year)
CCSM4 53.21 1005.22 52.09 461.56
GFDL 64.77 1082.45 51.08 477.63
HadGEM2 62.27 1064.34 51.38 473.49
MIROC-5 64.49 1079.64 51.12 477.07
MPI-ESM 61.83 1059.97 51.44 472.53
Fig 18.Adoption rates under different models for the base scenario.
Figure 18 above shows that with the exception of CCSM4, the other 4 models predict above 60%
adoption rate. This implies that the percent gainers are higher than 50% of the population.
Table 18: Correlations between net returns
Rho Crops in
system 1
Crops in
system 2
Acric 0.866567 0.0119 0.0119
Dystric 0.957566 0.01163527 0.01163527
Petric 0.973811 0.103 0.103
Fig 19: Opportunity cost and adoption rate
Overall, 58% of farms would adopt system 2. These are the farms whose opportunity cost of
adoption is less than zero. However this general picture masks the differences within the
different soil types. For instance, within farms under acric soil type, the threshold adoption rate is
53.5% while for Dystric and Petric farms, the adoption rate is higher than the overall at 59.4%
(Figure 18). However, the area under the graphs is higher for acric soil types implying that there
are higher gains under acric soil type compared to farms under Dystric or Petric soils. These
figures indicate that climate change is likely to have different effects across farms under different
soil types.
Fig 20: Adoption rate disaggregated by strata
The gains from climate change are shown below. The results show that mean net returns are
higher for Petric and Dystric soil types before and after climate change.
In general results showed that there are benefits to climate change, though the gains are low and
vary by soil type.
Fig 21: Cumulative gains with adoption disaggregated by strata: DSSAT results
From DSSAT model runs, gains from adoption are rising for farms under all soil types.
However, farms under petric soil types do not have any benefits to adoption before the 35% level
of adoption while at that level of adoption, farms under acric or dystric soil types have 15% and
10% gains from adoption. Additionally, there are no further benefits to adoption beyond 42%
level of adoption for farms under dystric soil types (Figure 21).
The poverty rate of the adopters and the poverty rate of the overall population is equal at zero
percent adoption and declines upto a point of optimum adoption rate (of 58%). On the other
hand, the poverty rate of non-adopters increases with adoption and is equal to the poverty rate of
the overall population at maximum adoption (of 100%). Overall in the population, as adoption
increases (upto the threshold of 58%) poverty rate decreases. Switching from system 1 to system
2 improves welfare before the threshold point. However, there are variations by strata. The figure
below shows that up to the optimum adoption rate, switching from system 1 to system 2 reduces
the poverty rate. However, climate change does not have a significant effect on poverty rate on
farms under petric much soils as it has on the other two strata.
Fig 21: Poverty rate and CC
Fig 22: Net returns from adoption among adopters, non adopters and the entire population
Fig 23: Per capita income across the entire population, among the adopters and non
adopters
As adoption increases, the per capita income of adopters reduces. In the same vein, as more
farms adopt, the per capita income of non-adopters increases.
Figure 24: Poverty rate across the entire population, among the adopters and non adopters
4. Stakeholders involvement and feedback
Results from this study were shared with stakeholders (see appendix) and the AGMIP protocols
have been used in a number studies in Uganda and D.R. Congo. A stakeholder meeting was
organized at Makerere University on 13th
June 2014 with the objective of sharing the results of
the project and get feedback from them.
Four presentations were made by the AGMIP-Uganda group members summarizing the key
findings of the projects. Questions and comments were collected and discussion and working
groups formed.
Fig 25: Participants to the stakeholders meeting organized on 13th June 2014, Makerere University,
Kampala-Uganda
Fig 26: Group discussion during the stakeholders meeting organized on 13th June 2014, Makerere
University, Kampala-Uganda
All participants agreed on the climate trends that:
The short rains have changed too (Sept-early Nov) but extended to December in Western
Uganda
The long rain season is much longer (March-May) but now to June, depending on
location
There is a very short dry season (in JULY)
There is change in rainfall distribution pattern but not the total amount
Temperature increase evident like in Kabale and Kasese districts
Economic patterns in results expected
and confirmed that projected values are within the range projected by different authors in the
region.
The following are the key recommendations of the stakeholders’ working groups:
a) Awareness
Create awareness on climate variability and training of the farmers on soil and land
management practices
b) Policy and byelaw
Need of bylaws to regulate the use of agricultural inputs at the farm and the households
Need to strengthen land management practices and use
c) Crop husbandry and agricultural inputs
Matching crop to season length eg. Late maturing for long crops and Early maturing for
short seasons
Develop of short and long maturing varieties
Increasing rain water harvesting techniques for LR to short rains seasons
Following timely agronomic practices like timely planting, fertilizer use, weeding etc
Adoption of tolerant cultivars to very dry and wet conditions
Application of adequate fertilizers to make use of long-rains and maximize yields
d) Early warning
Timely access and utilization of weather forecasts and data
In addition the AGMIP protocol was used to downscale climate projections for 32 stations across
Uganda for the Climate Change Task Force (Uganda Climate Change Unit), for 3 stations for
Environmental Alert (NGO in Uganda), IUCN-Uganda and WWF-DR. Congo. Three students
have been trained to use the script for downscaling climate projection, and ten of them trained in
the use of the crop models.
References
Wilby, R.L., Charles, S., Zorita, E., Timbal, B., Whetton, P., Mearns, L.2004. Guidelines for use of
climate scenarios developed from statistical downscaling methods. In: IPCC Supporting Material,
available from the DDC of IPPC TGCIA.
Moss, R.H., Edmonds, J.A., Hibbard, K.A., Manning, M.R., Rose, S.K., van Vuuren, D.P., Carter,
T.R., Emori, S., Kainuma, M., Kram, T., Meehl, G.A., Mitchell, J.F.B., Nakicenovic, N., Riahi, K.,
Smith, S.J., Stouffer, R.J., Thomson, A.M., Weyant, J.P., Wilbanks, T., 2010. The next generation
of scenarios for climate change research and assessment. Nature 463, 747–756.
Taylor K.E., R.J. Stouffer, and G.A. Meehl. 2012: An overview of CMIP5 and the experiment design.
Bulletin of the American Meteorological Society 93(4):485-498, doi: http://dx.doi.org/
10.1175/BAMS-D-11-00094.1
Rosenzweig, C., Jones, J., Hatfield, J., Antle, J., Ruane, A., Boote, K., Thorburn, P., Valdivia, R.,
Porter, C., Janssen, S., Mutter, C. 2013. AgMIP Guide for Regional Integrated Assessments:
Handbook of Methods and Procedures Version 5.0.
Rienecker, M. R. and co-authors, 2011: MERRA - NASA's Modern-Era Retrospective Analysis for
Research and Applications. J. Climate 24, 3624–3648.
Ruane, A.C., and R. Goldberg: "AgMIP Hybrid Baseline Climate Datasets: Shifted Reanalyses for Gap-
filling and Historical Climate Series Estimation" (In preparation).
MAAF.1996. Modernisation of Agriculture in Uganda. The Way forward 1996-2001. Medium-term
Agricultural sector modernisation plan for Ministry of Agriculture Animal Industry and Fisheries, Uganda.