Extrapolating effects of conservation tillage on dry spell mitigation, yield and productivity of water using
simulation modelling
Mkoga, Z.J.a Tumbo, S.D.,a Kihupi, N., a Semoka, J.b
a
Sokoine University of Agriculture Department of Agricultural Engineering and Land planning, P.O. Box
3003, Morogoro, Tanzania
b
Sokoine University of Agriculture Department of Soil Science, P.O. Box 3003, Mororgoro, Tanzania
Corresponding author: email: mkogazj@yahoo.co.uk
There is big effort to disseminate conservation tillage practices in Tanzania. Despite wide spread field
demonstrations there has been some field experiments meant to assess and verify suitability of the tillage
options in local areas. Much of the experiments are short lived and thus long term effects of the tillage
options are unknown. Nevertheless experiments to study long term effects of the tillage options are
lacking because they are expensive and cannot be easily managed. Crop simulation models have ability to
use long term weather data and the local soil parameters to assess long term effects of the tillage
practices. The Agricultural Production Systems Simulator (APSIM) crop simulation model; was used to
simulate long term production series of soil moisture and grain yield based on the soil and weather
conditions in Mkoji sub catchment of the great Ruaha river basin in Tanzania. A 24 year simulated maize
yield series based on ox ploughing tillage practice was compared with similar yield series based on
conservation tillage (ripping tillage practice with crop residue mulch). Results showed that yield trend
was significantly higher in conservation tillage than in conventional tillage (P<0.001). Long term
analysis, using APSIM simulation model, showed that average soil moisture in the conservation tillage
was significantly higher (P<0.05) (about 0.29 mm/mm) than in conventional tillage (0.22 mm/mm)
treatment during the seasons which received rainfall between 468 and 770 mm. Similarly the conservation
tillage treatments recorded significantly higher yields (4.4 t/ha) and productivity of water (1.3 kg/m3)
(P<0.01) than the conventional tillage (3.6 t/ha, 0.23 kg/m3) treatment in the same range of seasonal
rainfall. It is concluded that conservation tillage practice where ripping and surface crop residues is used
is much more effective in mitigating dry spells and increase productivity in a seasonal rainfall range of
between 460 and 770 mm. It is recommended that farmers in the area adopt that type of conservation
tillage because rainfall was in this range (460 – 770 mm) in 12 out of the past 24 years, indicating
possibility of yield losses once in every two years. It was also found that simulation modelling is a
suitable tool to analyse long term effects of tillage before recommending to farmers.
Key words: Crop simulation modelling, Conservation tillage, productivity of water, Dry spell mitigation
1.0 Introduction
There is big effort to disseminate conservation tillage practices among smallholder farmers in Tanzania
(Shetto et al., 2007). This effort is accompanied with activities to validate some of conservation tillage
technologies such as ripping, soil surface cover with organic crop residues, cover crops and crop rotations
(Mkomwa et al., 2007, Ringo et al., 2007, Enfors, 2009). There were some experiments also to validate
ability of ripping with crop residue mulch in enhance productivity of water and moderating irrigation
scheduling in Mkoji sub catchment (Mulengera, 2008). Most of these experiments were meant to assess
and verify suitability of the tillage options in local areas with very little flexibility and transferability of
results. Further more much of the experiments were short lived while long term effects of the tillage
options are not well understood. There is therefore an incomplete picture of the interactions among
biophysical factors related to conservation tillage practices in the short and long term perspective. A
combination of field experiments and computer simulation model can be an appropriate option to capture
important biophysical factors and their interactions so to have a comprehensive elucidation of the effects
of conservation tillage practices on yield and productivity.
Crop Simulation Models (CSMs) were primarily developed to aid in research, education and support of
learning and in decision support processes (van Ittersum et al., 2003). CSMs have been used successfully
as powerful tools for irrigation and agricultural decision making (Hoogenboom, 2000). There are practical
applications of decision support systems (based on crop simulation models) for pests and diseases (Kropff
et al., 1995), breeding (Yin et al., 2000) and in assessing and optimizing irrigation water management
scheduling (e.g. Li et al., 2005) to mention but a few examples. They are also being used in evaluating
crop management practices; yield gap analyses, strategic and tactical decision-making, as educational
tools and study development processes which are either too complex or time-demanding to study in the
field (e.g., Ruiz-Nogueira et al., 2001, Graves et al., 2002; Kato et al., 2004, Shang et al., 2004).
APSIM model is one of the efficient tools in agro-ecosystem analysis. The model has been widely tested
in a variety of conditions in Australia and Europe (Asseng et al., 2004). The use of APSIM in smallholder
systems in Africa is still young, where a few instances of use in Kenya (Rao and Okwach, 2005) and
Zimbabwe can be sighted. While APSIM has a lot of capabilities shown to assist in exploring various
management options under African conditions (Carberry et al.2002), there has been limited evaluation of
APSIM to simulate observed crop response to conservation tillage in the field. This study will therefore
use field experiments data parameterize and calibrate the model and use it to enhance rigorous analysis of
the potentials of conservation tillage in improving productivity of water through mitigating dry spells by
extrapolating the field experimental results over 24 seasons. Therefore, the main objective was to study
the long term trends of the effects of conservation tillage on soil moisture, yield and productivity of water.
The specific objectives were (1) to calibrate and validate the APSIM model for long term simulation; and
(2) simulate the effects of conservation tillage practice on soil moisture, yield and productivity of water
for the last 24 years in Mkoji sub-catchment.
2.0 Materials and Methods
2.1 Location of the study
The study was conducted in the Great Ruaha River Catchment within the Rufiji basin specifically in the
Mkoji sub-catchment. The sub catchment is located in the Southwest of Tanzania, between latitudes 7048’
and 9025’ South, and longitudes 33040’ and 34009’ East. The experiment fields from which soil moisture
and yield data was sourced, were located at Igurusi village, in the Igurusi ya Zamani irrigation scheme.
The Mkoji River, which has given name to the sub-catchment, is the main river draining through the
whole sub-catchment. The study area has a unimodal type of rainfall between November and April. The
mean annual rainfall of Mkoji sub catchment is about 770 mm.
APSIM Model parameterization
The TMV1-ST maize cultivar parameters for this study were adopted from Patched Thirst model and
from Igbadun (2006), who worked in the same location. The parameters were the soil dependent
transpiration factor, harvest index adjustment factors, leaf area index module coefficients, potential
harvest index for the maize crop and maximum rooting depth. Others included maximum leaf area index,
crop canopy radiation extinction factor, and radiation use efficiency. Solar radiation, minimum and
maximum temperature data required by APSIM were obtained from Igurusi agromet station some seven
kilometers from the experimental plots. Rainfall data used was collected using a standard rain gauge
placed at the experimental site. Soil data needed by APSIM model was obtained from a standard soil pit
dug within the study location and described according to FAO methodology detailed in (FAO, 1999). The
soil data used in this study is presented in appendix 1.
APSIM Model validation
Maize grain and biomass yield data collected from a field experiment conducted at Igurusi ya zamani
farm im Mkoji sub catchment in the 2006/07 and 2007/08 seasons by Mkoga (2009) was used to validate
the model. The quantitative data for the comparisons of measured and simulated grain and biomass yield
is presented in Appendix 2. Also soil moisture data collected from the same experiment was used. This
soil moisture content was monitored in the experiment using an Institute of Hydrology Neutron probe
(Didcot 3965 NE Instrument Company, Abingdon). When taking the measurements, the probe sensor was
lowered through the access tubes to measure soil moisture content at 0-150, 150-300, 300-450, 450-600,
600-750, 750-1000 mm depths. The Neutron probe was calibrated for the experimental site at the
beginning of the experiment. The Neutron probe was calibrated for the 0-150 mm depth and for the 1501000 mm depths as recommended by the manufacturers. Due to the presence of mudstones and gravels
below the one-metre profile depth in the fields, access tubes could not go below 1.00 m depth. Therefore
soil moisture was only monitored in the one-metre soil profile depth.
The Models’ output variables were validated by comparing model-simulated results with field measured
data from field experiments in the 2006/07 and 2007/08 seasons. The validation was done by using both
qualitative (graphical) and quantitative (statistical) methods. The statistical expressions used for
comparing the simulated and field measured data were Average Error of bias (AE), Coefficient of
Variation (CV), Root Mean Square Error (RMSE), Modeling Efficiency (ME) and Coefficient of
Residual Mass (CRM).
Average Error of Bias ( AE ) 
1
n
n
 P  O 
i 1
i
1

n
Coefficient of Variation (CV )  100 * 
1
Root Mean Square Error ( RMSE )  
 n
Modelling Efficiency ( EF ) 
(1)
i

Pi  Oi  

i 1

Om
n
0.5
2
2 0.5

n
 P  O  
i
i 1
i

2
n
n

2
 Oi  Om    Pi  Oi  
i 1
 i  1

n
 P  O 
(3)
(4)
2
i
i 1
i
n
Coefficient of Re sidual Mass (CRM ) 
(2)
n
 Oi   Pi
i 1
i 1
n
O
i 1
(5)
i
Where Pi is simulated values; Oi is measured values; Om is mean of measured values, and n is
number of the observations.
The average error of bias (AE) is a measure of bias between the simulated and measured data. The
coefficient of variation (CV) is a measure of variability while the root mean square error (RMSE) is a
measure of precision. The modelling efficiency (EF) also referred to as the coefficient of Nash-Sutcliffe
(Mahdian and Gallichand, 1995) is a measure of degree of fit between simulated and measured data. It is
similar to coefficient of determination (R2). EF varies from negative infinity (-∞) for total lack of fit to 1
for an exact fitting (Mahdian and Gallichand, 1995). The coefficient of residual mass (CRM) is an
indicator of the tendency of the model to either over- or under- predict measured values. A positive value
of CRM indicates a tendency of underestimation, while a negative value indicates a tendency of
overestimation (Antonopoulos, 1997).
Long term simulation of effects of tillage practices yield and soil moisture
The calibrated and validated APSIM model was used to simulate long term scenarios based on two of the
treatments tested in the field experiment by Mkoga (2009). The treatments were the Conventional Tillage
(CT) practice (using ox plough) with slash and burning of crop residues commonly practiced by farmers
and the Ripping tillage practice with crop residues left on soil surface (RR). Simulations were done for
soil and weather conditions of Igurusi (representing middle Mkoji sub catchment). Simulation outputs of
daily soil moisture content and annual grain yield were plotted to give trends of soil moisture, yield and
productivity of water for the two different tillage practices over the 23 years’ period (1985 to 2008). Soil
moisture trend was plotted with moisture content at the 0.55 depletion (0.55 D), Drainage Upper Limit
(DUL) and the Lower Limit (LL). In this way it enabled to identify days with excess or deficient soil
moisture, rather pinpointing frequency and severity of agricultural dry spells over the years. The dry spell
pattern assists to explain the effectiveness or inefficiency of a tillage practice in terms of soil moisture
conservation. It also explains relationship between soil moisture pattern and the effects they impact to
yield and productivity of water. Trends of yield and productivity presented a comparative evaluation of
suitability of a medium maturity (TMV1-ST) and long maturity (SC 407) maize cultivars along with the
tillage practices in the soil and weather conditions of Mkoji sub catchment.
3.0 Results and Discussion
3.1 Validating the APSIM Model to simulate yield and soil moisture
3.1.1 Simulation of grain and biomass yield
The relationship between simulated and measured grain yield from 2006/07 and 2007/08 seasons with the
standard plant density of 4.4 plants/m2 21 days after sowing are presented in Figures 1 and 2, respectively.
These figures showed poor correlation between simulated and measured grain yield. The trend was
persistent over the two seasons of experiment. A conclusion was about to be made that the model was not
simulating grain yield close to observed data.
6
R2 = 0.0007
Simulated yield (t/ha)
5
4
3
2
1
0
1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
Measured yield (t/ha)
Figure 20: Comparison
of and
simulated
and measured
grain
at harvest
Figure 1: Comparison
of simulated
measured
grain yield
at yield
harvest
2006/07 season
2006/07 season
6.0
R2 = 0.13
Simulated yield (t/ha)
5.0
4.0
3.0
2.0
1.0
0.0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
Measured yield (t/ha)
Figure 2: Comparison between simulated and measured grain yield at harvest 2007/08 season
(standard plant density)
According to APSIM training literature (www.apsim.info) it was found that the model subroutines for
germination and emergence are mainly based on thermo-time (degree days) between germination and
emergence. In particular emergence is governed by some four factors: (i) Degree days of time lag before
linear coleoptile growth starts, (ii) A shoot growth rate factor (Degree day/mm), (iii) Water stress factor
for crop emergence, and (iv) Fraction of available soil moisture. None of these factors explains a fraction
of emerging plants as a function of spatial variability in soil moisture but they only govern the rate of crop
emergence assuming homogeneous soil moisture conditions. In APSIM all plants are killed if the crop
hasn't germinated within 40 days of sowing, due to lack of moisture. In the model, plants are also killed if
the crop does not emerge after 150 degree days of sowing, because it was sown too deep. So the model
basically kills all plants after being subjected to stress of moisture beyond set limits. The only stress
condition that is set to kill a proportion of plants is atmospheric temperature whereby APSIM kills a
fraction of plants when high temperature is subjected to seedlings immediately following emergence.
However, soil is naturally not homogeneous with in field variations of soil physical characteristics,
elevation and surface characteristics. These minor differences can accentuate differences in soil drainage
characteristics (Cox et al., 2006) and thus soil moisture spatial variability especially during intra-seasonal
dry spells. These differences in soil texture and topography, and the resultant variations in hydrologic
properties, are primary determinants of crop yield due to crop stand (Iqbal et al., 2005). The inherent
within-field variability also creates management challenges to insure timely tillage, field preparation and
planting, and contributes to spatial variability of crop establishment, growth and yield. Over the two
seasons, rainfall and tillage treatments induced high variability in crop emergence. Figure 3 shows
variability of plant population per unit area in the 2006/07 and 2007/08 seasons due to in field soil
moisture variability 21 days after planting.
3.5
Plant density (Plants/m2)
2006/07
2007/08
2.9a
3
2.9a
2.5
2
1.5
1.7b
1.3b
1.8b
1.4b
1
0.5
0
CT
CTR
CTRCv
TR
Tillage treatments
RR
RRCv
Figure 3: Variation of plant population 21 days after sowing between the different tillage
treatments in the 2006/07and 2007/08 seasons (CT = Ox ploughing bare soil, CTR = ox ploughing
with surface crop residues, CTRCv = ox ploughing with surface crop residues plus lablab dolicos
cover crop, TR = Tie ridging, RR = Ripping with surface crop residues, RRCv = Ripping with
surface crop residues plus lablab dolicos cover crop)
Plant population was in many instances lower than the standard 4.4 plants per square meter expected for
the used crop spacing (30 x 75 cm). There was no significant difference in plant establishment in the
2006/07 season. In the 2007/08 season more plants per unit area were established in tillage treatments
which have ability to conserve soil moisture. In this season both ripping with residues and with or without
cover crop treatments recorded 2.1 and 2.9 plants/m2, respectively, being significantly higher than the rest
of the treatments (P<0.05). The number of established plants closely correlated with final recorded yield
for each plot. The pooled data (of the three experimental seasons) showed that seedling emergence was
highly correlated to final crop yield. Figure 4 shows the relationship between yield and plant density at 21
days after sowing. It showed high correlation (R2 = 0.92) between plant population and crop yield. Hence
actual plant densities measured from the fields were used to simulate yield and this resulted into a
completely different picture from that reported above.
7
6
2
R = 0.92
Yield (t/ha)
5
4
3
2
1
0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
Plant population (plants/m2)
Figure 4: Relationship between established plant population 21days after sowing and yield of maize
at harvest
The relationship between simulated and measured grain yield from 2006/07 and 2007/08 seasons with
consideration of measured plant density 21 days after sowing are presented in Figures 5 and 6,
respectively. The quantitative data showing simulated versus actual yields for the different conservation
tillage methods together with the respective plant densities are presented in Appendix 1. The simulated
grain yields correlated well with measured grain yields with coefficient of determination (R2) of 0.86 and
0.99 for the 2006/07 and 2007/08 seasons, respectively.
3200
R2 = 0.86
3100
Simulated yield (kg/ha)
3000
2900
2800
2700
2600
2500
2400
2300
2200
2200
2300
2400
2500
2600
2700
2800
2900
Observed yield (kg/ha)
Figure 5: Comparison between simulated and actual grain yield (kg/ha) in the 2006/07 season
4500
R2 = 0.99
Simulated yield (kg/ha)
4000
3500
3000
2500
2000
1500
1500
2000
2500
3000
3500
4000
Observed yield (kg/ha)
Figure 6: Comparison between simulated and actual grain yield (kg/ha) 2007/08 in the season
Table 1 shows the quantitative comparison between the measured and simulated grain yield for the
2006/07 and 2007/08 seasons based on measured plant density. The coefficient of residual mass (CRM)
indicated tendencies of the model to over predict of about 7% and 5% in 2006/07 and 2007/08 seasons,
respectively (Table 1). The close agreement between model simulated and measured grain yield implies
that APSIM adequately simulated grain yield of maize in the two seasons. Therefore the APSIM model
closely simulates grain yield provided proper parameterisation of factors including plant density is done.
Table 1: Comparison between simulated and measured grain yield for the 2005/06, 2006/07 and
2007/08 seasons
Model performance indicator
AE
RMSE
CV (%)
CRM
Unit of AE and RMSE for grain yield is kg/ha
2006/07
176.3
193.1
7.3
-0.067
2007/08
129.8
139.3
5.1
-0.048
Similarly, Figure 7 and 8 show the relationship between measured and simulated biomass yield at harvest
for the 2006/07 and 2007/08 seasons, respectively. The model closely simulates biomass yield with
coefficients of determination (R2) of 0.95 and 0.99 for the 2006/07 and 2007/08 seasons, respectively.
The quantitative data of the simulated and measured biomass yield at harvest based on measured plant
density 21 DAP for the 2006/07 and 2007/08 seasons are shown in Appendix 1. Table 2 shows statistical
data for the comparison between simulated and measured biomass yield at harvest for the three seasons.
The model tendency to either over- or under predict field results was less than 5% in the two seasons
indicating a model prediction efficiency of over 95%. In this case APSIM model showed tendency to over
predict biomass yield. These results were considered good and the model performance to predict biomass
yield was considered adequate and acceptable.
10500
R2 = 0.95
Simulated biomass yield (kg/ha)
10000
9500
9000
8500
8000
7500
7000
8000
8500
9000
9500
10000
10500
11000
11500
Observed biomass yield (kg/ha)
Figure 7: Comparison between simulated and actual biomass (kg/ha) 2006/07 season
11000
R2 = 0.98
Simulated biomass (kg/ha)
10000
9000
8000
7000
6000
5000
4000
4000
5000
6000
7000
8000
9000
10000
Observed biomass yield (kg/ha)
Figure 8: Comparison between simulated and observed biomass yield in 2007/08 season
Table 2: Comparison between simulated and measured biomass yield at harvest for the 2005/06,
2006/07 and 2007/08 seasons
Model performance indicator
AE
2006/07
-821.2
2007/08
1169.8
RMSE
CV (%)
EF
CRM
Unit of AE and RMSE for grain yield is kg/ha
868.1
9.2
0.86
-0.047
1204
17.1
0.80
-0.048
3.1.2 Simulation of soil moisture content of the soil profile layers
The comparison between simulated and measured soil moisture contents for the six profile layers for the
conventional tillage practice (ox-ploughing no crop residues) for the 2006/07 and 2007/08 seasons are
presented in Figs.9 (a-f) and 10 (a-f). There was generally a close agreement between simulated and
measured volumetric soil moisture content in the profile layers in the two experimental seasons. Also the
soil moisture predictions highly corresponded to the trend of rainfall events, especially for the top 500
mm of soil. This indicates that the model can simulate the effect of soil-water-plant-atmosphere
interactions reasonably well. However some over- and under-predictions of soil moisture by the APSIM
model need to be noted. In most occasions the model over-predicted soil moisture. Observing Figures 9
and 10 indicate that the measured and simulated soil moisture content was sometimes widely departing
from each other. Even the paired t-test showed that none of the simulated series of the daily soil moisture
were identical with the observed values.
Simulated
Actual
At 0-150 mm soil profile depth 2006/07 season
0.36
0.39
0.34
Days after planting
Days after planting
a
b
Actual
At 370-500 mm depth of soil profile 2006/07 season
Actual
At 500-650 mm depth of soil profile 2006/07 season
Simulated
Simulat
ed
0.39
0.39
Soil moisture (mm/mm)
0.35
0.33
0.31
0.29
0.27
0.37
0.35
0.33
0.31
0.29
0.27
Days after planting
Days after planting
d
83
81
78
74
70
65
45
40
36
34
26
1
83
81
78
74
70
65
45
40
36
34
26
1
20
0.25
0.25
20
Soil moisture (mm/mm)
0.37
c
83
81
78
74
70
65
1
83
81
78
74
70
65
45
40
36
0.25
34
0.27
0.2
26
0.22
45
0.29
40
0.24
0.31
36
0.26
0.33
34
0.28
0.35
26
0.3
20
Soil moisture (mm/mm)
0.37
0.32
1
20
Soil moisture (mm/mm)
Actual
Simulated
At 150-370 mm depth of soil profile 2006/07
Actual
0.37
0.37
0.35
0.35
0.33
83
81
78
74
70
65
1
20
26
83
81
78
74
70
65
45
0.25
40
0.25
36
0.27
34
0.27
26
0.29
1
0.29
45
0.31
40
0.31
Simulated
36
Soil moisture
0.39
20
Soil moisture (mm/mm)
0.39
0.33
Actual
At 800-1000 mm depth of soil profile 2006/07 season
Simulated
34
At 650-800 mm depth of soil profile 2006/07 season
Days after planting
Days after planting
e
f
Figure 9: (a-f): Comparison between measured and simulated soil moisture content in the different
horizons of the soil profile in 2006/07 season
At a depth of 0-150mm of soil profile 2007/08 season
simulated
actual
at 150-370mm soil profile depth 2007/08 season
0.36
0.36
Soil moisture (mm/mm)
Soil moisture (mm/mm)
0.34
0.32
0.30
0.28
0.26
0.24
0.35
0.34
0.33
0.32
0.31
0.30
0.29
0.28
0.22
10 11 13 16 18 19 23 46 53 57 59 60 62 64 66 77 85 91
10 11 13 16 18 19 23 46 53 57 59 60 62 64 66 77 85 91
Days after Planting
Days after planting
a
at 370-500mm profile depth 2007/08 season
b
simulated
actual
at 500-650mm profile depth 2007/08 season
simulated
actual
0.39
0.39
0.37
Soil moisture (mm/mm)
Soil moisture (mm/mm)
simulated
actual
0.35
0.33
0.31
0.29
0.27
0.37
0.35
0.33
0.31
0.29
0.27
10 11 13 16 18 19 23 46 53 57 59 60 62 64 66 77 85 91
10 11 13 16 18 19 23 46 53 57 59 60 62 64 66 77 85 91
Days after planting
Days after planting
c
d
at 650-800mm profile depth 2007/08 season
simulated
800-100mm profile depth 2007/08 season
0.39
0.39
0.37
0.37
Soil moisture (mm/mm)
Soil moisture (mm/mm)
actual
0.35
0.33
0.31
0.29
simulated
actual
0.35
0.33
0.31
0.29
0.27
0.27
0.25
10 11 13 16 18 19 23 46 53 57 59 60 62 64 66 77 85 91
10 11 13 16 18 19 23 46 53 57 59 60 62 64 66 77 85 91
Days after planting
Days after planting
e
f
Figure 10: (a-f): Comparison between measured and simulated soil moisture content at the
different horizons of the soil profile in the conventional ox-plough tillage no residues treatment in
the 2007/08 season
However, The Modelling Efficiency (EF), RMSE, and average Error of Bias (AE) values gave a more
detailed picture of the model performance. Tables 3 and 4 show the statistics of the quantitative
comparison between simulated and measured volumetric soil moisture content of the different horizons of
soil profile for the 2006/07 and 2007/08 seasons, respectively. The average error of bias (AE) was
between -0.03 and 0.03 mm/mm in 2006/07 season and between -0.03 and 0.05 mm/mm in 2007/08
season. The RMSE values were between 0.03 and 0.05 mm/mm in 2006/07 and between 0.02 and 0.04 in
2007/08 season. In both seasons the modelling efficiencies were poor for the top soil layers, but became
fairly good as soil depth increased. They ranged between 0.46 and 0.57 in the top soil layer. On the
contrary modelling efficiency (EF) was all above 0.6 and ranged between 0.67 and 0.94 in the deeper soil
layers. This can be very logical because the neutron probe is not very efficient in measuring top soil
moisture content.
Table 3: Statistical results of the comparison between simulated and measured soil moisture in the
respective profile depths in the 2006/07 season
Model performance
Indicator
AE (m2/m2)
RMSE (m2/m2)
CV (%)
EF
CRM
0-150
-0.03
0.03
11.9
0.46
-.0.09
150-370
-0.02
0.03
10.1
0.67
0.05
Soil profile depth (mm)
370-500
500-650
650-800
-0.02
-0.01
0.02
0.05
0.05
0.05
15.6
14.4
13.3
0.95
0.84
0.92
0.07
-0.02
0.05
800-1000
0.03
0.05
14.7
0.94
0.07
Table 4: Statistical results of the comparison between simulated and measured soil moisture in the
respective profile depths in the 2007/08 season
Model performance
Indicator
AE (m2/m2)
RMSE (m2/m2)
CV (%)
EF
CRM
0-150
-0.03
0.04
13.4
0.52
-0.1
150-370
-0.01
0.02
5.5
0.60
-.003
Soil profile depth (mm)
370-500
500-650
650-800
-0.02
-0.02
0.05
0.02
0.02
0.04
7.9
7.7
11.7
0.82
0.82
0.8
-0.05
-0.06
0.049
800-1000
0.02
0.03
9.9
0.8
0.05
Figures 11 and 12 show the comparison between simulated and measured profile soil moisture for the
2006/07 and 2007/08 seasons. The trend of soil moisture is similar to that explained in the individual
profile layers. Table 5 shows the comparison between the measured and simulated volumetric soil
moisture content of the soil profile. The average error of bias (AE) between simulated and measured data
was between -0.03 mm/mm and 0.04 mm/mm in the three seasons. The RMSE was between 0.01 and
0.04 mm/mm. The coefficients of determination (R2) were poor (i.e. 0.40) in 2006/07 seasons and was
good (i.e. 0.97) in the 2007/08 season.
The statistical indices reported here were comparable with values obtained by Clemente et al. (1994)
when they compared performance of three soil water flow models: SWATRE (Belmans et al., 1983),
LEACHW (Wagenet and Hutson, 1989, as cited by Clemente et al., 1994) and SWASIM (Hayhoe and de
Jong, 1982, as cited by Clemente et al., 1994). They reported an average error of bias between simulated
and field-measured soil moisture content which ranged from -0.03 - 0.04 m3/m3 and the RMSE ranging
from 0.0083 - 0.0475 m3/m3 for these models. Modelling efficiency values are compatible with values
obtained by Eitzinger et al. (2004) when comparing CERES, WOFOST and SWAP models ability in
simulating soil moisture in Australia. Eitzinger et al. (2004) recorded EF ranging between 0.4 and 0.93,
and between 0.68 and 0.97 under spring barley and winter wheat, respectively.
0.39
Actual
Soil moisture (mm/mm)
0.37
Simulated
0.35
0.33
0.31
0.29
0.27
0.25
1 14 20 22 26 33 34 35 36 37 40 42 45 57 65 67 70 73 74 77 78 79 81 82 83
Days after planting
Figure 11: Comparison between simulated and measured soi moisture of the 100mm soil profile in
2006/07 season
0.37
actual
simulated
0.36
Soil moisture (mm/mm)
0.35
0.34
0.33
0.32
0.31
0.30
0.29
0.28
0.27
10
11
13
16
18
19
23
46
53
57
59
60
62
64
66
77
85
91
Days after planting
Figure 12: comparison between simulated and measured profile soil moisture 2007/08 season
Table 5: Statistical results of the comparison between simulated and measured profile soil moisture
in the conservation tillage treatment with no residue
Model performance
Indicator
AE (m2/m2)
RMSE (m2/m2)
CV (%)
EF
CRM
Soil profile depth (0-1000mm)
2005/06
2006/07
0.04
-0.01
0.04
0.01
17.6
4.4
0.97
0.4
0.17
-0.04
2007/08
-0.003
0.03
9.9
0.4
-0.04
3.2 Long term trend of soil moisture, dry spell, yield and productivity of water
3.2.1 Soil moisture and dry spell trend
In Figure 13 soil moisture trend under ox ploughing without surface crop residues (CT) was compared
with soil moisture under ripping with surface crop residues (RR) for the period between 1985/86 to
2007/08. The soil moisture trend responded well with the trend of seasonal rainfall for which, soil
moisture under RR was higher than that under CT throughout the 24 seasons (Fig. 13). The overall paired
t-test of mean soil moisture between RR and CT tillage practices showed that soil moisture in RR was
higher and differed significantly from CT at P<0.01 significant level. Mean soil moisture over the seasons
in RR was about 18% higher (0.29 (mm/mm)) than CT practice (0.24 (mm/mm)). This trend is much
more conspicuous during the seasons that received below the long term average rainfall for which there
were nine such seasons between 1985/86 and 2007/08.
400
Rain/ Soil moisture (mm)
350
300
250
200
150
100
50
18/10/2007
01/10/2006
14/09/2005
28/08/2004
12/08/2003
26/07/2002
09/07/2001
22/06/2000
06/06/1999
20/05/1998
03/05/1997
16/04/1996
01/04/1995
15/03/1994
26/02/1993
10/02/1992
24/01/1991
07/01/1990
21/12/1988
05/12/1987
18/11/1986
01/11/1985
0
Date
LL
DUL
CT
RR
Rain
0.55D
Figure 13: Comparison of simulated 24 year trend (1985 to 2008) of soil moisture between conventional (CT) and conservation (RR) tillage
practice at Igurusi, Mkoji sub-catchment. (LL = Drainage lower limit, DUL = Drainage upper limit, CT = Ox ploughing bare soil surface,
RR = Ripping with surface crop residues)
6000.0
5000.0
4000.0
3000.0
2000.0
1000.0
0.0
1985/86
1986/87
1987/88
1988/89
1989/90
1990/91
1991/92
1992/93
1993/94
1994/95
1995/96
1996/97
1997/98
1998/99
1999/00
2000/01
2001/02
2002/03
2003/04
2004/05
2005/06
2006/07
2007/08
Yield (kg/ha)/ rainfall (mm)
Mean soil moisture in the seasons with seasonal rainfall total below long term average in RR
treatment was about 24% higher (0.29 mm/mm) than CT practice (0.22 mm/mm) at P<0.01
significance level. It is in these seasons that soil moisture in the CT treatment drops below the
critical soil moisture (0.55 D) indicating periods of in intra-seasonal dry spells in which crops
suffer soil moisture stress. On the contrary in the RR treatments have indicated consistent ability
to maintain soil moisture above 0.55 D and always around the readily available soil moisture
range and sometimes far above the drainage upper limit (DUL) (Fig. 13). However, soil moisture
trend during the seasons that received rainfall above long term average showed that the moisture
in CT treatments was statistically the same as in RR treatments at P<0.01. This had a bearing in
the trends of yield and productivity of water. Figure 14 shows the trend of grain yield for the CT
and RR practices between the 1985/86 and 2007/08 seasons. The figure shows a very similar
trend as that of soil moisture and seasonal rainfall amount.
Seasons
Rain
RR
CT
Figure 14: Comparison of trend of yield between ox ploughing without surface crop
residues (CT) and Ripping with surface crop residues (RR) over the 1985/86 to 2007/8
seasons at Igurusi Mkoji sub-catchments. (CT = Ox ploughing bare soil surface, RR =
Ripping with surface crop residues)
In the long term, both RR and CT treatments recorded statistically the same magnitudes of yield
(about 4.5 t/ha) at P<0.05 level of significance. The RR treatment recorded significantly higher
yields (4.4 t/ha) than the CT treatments (3.6 t/ha) in seasons with below normal rainfall (P<0.05),
being an increase of about 25%. In such seasons it is presumed that high soil moisture was
maintained in the RR than in the CT treatments. It was found that five seasons of 1999/00,
2002/03, 2003/04, 2004/05 and 2005/06 recorded some intra-seasonal dry spells (Fig. 13) which
might have lowered grain yield in CT treatments. This might have been enabled by the ability of
RR practice to conserve soil moisture and mitigate effects of dry spells. Figure 15 shows this
effect in the dry spell stricken 1989/90, 1999/00 and 2003/04 seasons in which soil moisture in
the CT treatments went down close to and below the wilting point. Even in such occasions the RR
treatment maintained soil moisture well above the critical soil moisture.
For example in a serious dry spell which occurred between 85th and 130th day after planting in
the 1989/90 season (in the CT treatment) was mitigated in the RR treatment by maintaining soil
moisture above the critical soil moisture and well above the wilting point (Fig 15). In the 1989/90
season the RR treatment recorded significantly higher (24%) average profile soil moisture (274
mm) than the CT (221 mm) at (P<0.01). The 221 mm profile soil moisture was below the critical
soil moisture of 244 mm, for which the crop was under moisture stress over the 45 days period
mitigated to zero day by the RR treatment. Similar effect was recorded in the 1999/00 and
2003/04 seasons with respectively 469 and 555 mm of rainfall. For example in a serious dry spell
which occurred in the 2003/04 season (in which soil moisture in the CT treatment was always
below the critical soil moisture) was mitigated by the RR treatment by maintaining soil moisture
above the critical soil moisture (Fig. 15). In the 2003/04 season, mean soil moisture of 300 mm
was recorded in the RR which was significantly higher (46%) than 205 mm recorded in the CT
treatment.(P<0.01).
However some yield reduction was recorded in the RR treatment for the 1986/87 and 1995/96
seasons, which had seasonal rainfall total higher than long term average (Fig 16). Significantly,
lower yields (33%) were recorded with the RR (3.1 t/ha) than in the CT (4.8 t/ha) treatments in
the 1986/87 and 1995/96 seasons at P<0.05 level of significance. These seasons received
respectively 1276 and 1268 mm of seasonal rainfall well above the long term average. This might
have induced water logging conditions with resulting effect in reducing grain yield. Maize crop is
highly sensitive to water logging conditions and may explain the reason for lower yields in the
1986/87 and 1995/96 seasons for the RR practice than that recorded in CT practice (Fig.14).
370
1989/90 season
Soil moisture (mm)
350
330
310
290
270
250
230
210
190
170
0
10
20
30
40
50
60
70
80
90
Days after planting
100 110 120 130 140
soil moisture (mm)
390
370
350
330
310
290
270
250
230
210
190
170
1999/2000 season
1
11
21
31
41
51
61
71
81
91
101
Days after planting
Soil moisture (mm)
400
2003/04 season
350
300
250
200
134
127
120
113
106
99
92
85
78
71
64
57
50
43
36
29
22
15
8
1
150
Days after planting
LL
DUL
CT
RR
0.55D
Figure 15: Comparison of simulated soil moisture between conventional (CT) and
conservation (RR) tillage practice in the drought stricken seasons at Igurusi, Mkoji subcatchment. (LL = Drainage lower limit, DUL = Drainage upper limit, CT = Ox ploughing
bare soil surface, RR = Ripping with surface crop residues)
1986/87 season: 1276 mm fainfall
Soil moisture (mm)
370
320
270
220
170
1
11
21
31
41
51
61
71
81
91 101 111 121 131 141
Days after planting
Soil moisture (mm)
420
1995/96 season: 1268 mm rainfall
370
320
270
220
170
1
11
21
31
41
51
61
71
81
91
101 111 121 131 141
Days after planting
LL
DUL
CT
RR
0.55D
Figure 16: Comparison of simulated soil moisture between conventional (CT) and
conservation (RR) tillage practice in the seasons which experienced soil water higher than
field capacity at Igurusi, Mkoji sub-catchment. (LL = Drainage lower limit, DUL =
Drainage upper limit, CT = Ox ploughing bare soil surface, RR = Ripping with surface crop
residues)
3.2.2 Long term trend of yield and productivity of water
Figure 14 presented in section 3.2.1 shows that despite the fluctuations in seasonal rainfall
amounts there was no record of complete grain yield loss even in the conventional tillage
practice, throughout the 24 year period. It suggests that the dry spells recorded in some seasons
had effect in reducing yield but were not severe enough to impact a complete yield loss. Figure 17
presents a comparison of yield from TMV1 and SC 407 cultivars under CT and RR practices at
Igurusi in Mkoji sub catchment. It shows a similar trend as that presented in Figure 14. A paired
t-test between grain yield from RR and CT treatments over the 24 year period showed no
difference at P<0.05 level of significance. This was from both SC 407 and TMV1 maize cultivar.
This suggests that it may not be necessary to adopt conservation tillage practice basing on risk of
grain loss in locations where rainfall distribution is relatively good. The SC 407 maize cultivar
produced higher yields than TMV1 at P<0.01 level of significance. An average of 5.8 t/ha of
grain was obtained from SC 407 cultivar compared to 4.5 t/ha in the same range of seasons.
8000
Grain yield (kg/ha)
7000
6000
5000
4000
3000
2000
2007/08
2005/06
2003/04
2001/02
1999/00
1997/98
1995/96
1993/94
1991/92
1989/90
1987/88
0
1985/86
1000
Seasons
SC407-CT
SC407-RR
TM V1-CT
TM V1-RR
Figure 17: Comparison of trend of simulated yield between TMV1 and SC407 maize
cultivars under conventional and conservation tillage treatments over the 1985/86 to 2007/8
seasons at Igurusi Mkoji sub-catchments. (SC407-CT = SC407 maize cultivar under ox
ploughing bare soil surface, SC407-RR = SC407 maize cultivar under ripping with surface
crop residues, TMV1-RR = TMV1 maize cultivar under ripping with surface crop residues,
TMV1 maize cultivar under ox ploughing bare soil surface)
Figure 18 presents the trend of productivity of water for the TMV1 and SC407 maize cultivars
under conventional and conservation tillage treatments over the 1985/86 to 2007/8 seasons at
Igurusi, Mkoji sub-catchments. This trend is similar to that of yield presented in Figure 17.
Generally productivity of water ranged between 0.24 kg/m3 to 1.3 kg/m3 during the f 24 years and
in both practices and cultivars. There was no significant difference in productivity of water
between the tillage practices.
The difference in productivity of water was however evident between the maize cultivars. The SC
407 cultivar recorded a productivity of water of 0.77 kg/m3 which differed from 0.59 kg/m3
obtained from TMV1 at a P<0.05 level of significance. The lowest productivity of water were
1.35
1.15
0.95
0.75
0.55
0.35
0.15
1985/86
1986/87
1987/88
1988/89
1989/90
1990/91
1991/92
1992/93
1993/94
1994/95
1995/96
1996/97
1997/98
1998/99
1999/00
2000/01
2001/02
2002/03
2003/04
2004/05
2005/06
2006/07
2007/08
Productivity of water (kg/m3)
recorded during the highest rainfall seasons of 1986/87 (0.24 kg/m3), 1995/96 (0.32 kg/m3),
1996/97 (0.28 kg/m3) and 1997/98 (0.23 kg/m3). These seasons recorded rainfall of 1275 mm,
1268 mm, 1583 mm and 1755 mm, respectively. During these seasons soil moisture was most of
the days higher that filed capacity indicating the possibility of loss of most rainwater through
runoff and or deep percolation. Based on these results the SC 407 cultivar and those of similar
characteristics can be best to be used in locations with rainfall pattern as that of Igurusi.
Season
SC407-CT
SC407-RR
TMV1-CT
TMV1-RR
Figure 18: Comparison of trend of productivity of water between TMV1 and SC407 maize
cultivars under conventional and conservation tillage treatments over the 1985/86 to 2007/8
seasons at Igurusi, Mkoji sub-catchments. (SC407-CT = SC407 maize cultivar under ox
ploughing bare soil surface, SC407-RR = SC407 maize cultivar under ripping with surface
crop residues, TMV1-RR = TMV1 maize cultivar under ripping with surface crop residues,
TMV1 maize cultivar under ox ploughing bare soil surface)
4.0 Conclusions
Crop simulation is a useful tool to extrapolate short term experimental results. Models cannot be
used until well parameterized and validated results shown here have exhibited ability of the
APSIM model to represent soil moisture dynamics and yield of both yield and biomass. The
model can therefore be used reasonably to study long term trends of moisture and yield with
respect to dynamics of dry spells. Long term analysis of soil moisture enabled a wider
examination of short term field experimental results on ability of conservation tillage to mitigate
intra-seasonal dry spells. It was a useful exercise because the two year experiment might not give
firm conclusive findings on some issues. For example it was concluded from the experiment that
conservation tillage results into higher productivity of water compared to conventional tillage
practice. However, it was found here that with the type of rains at Igurusi representing Mkoji sub
catchment conservation tillage involving ripping and surface cover crops may not always give
higher productivity of water. Higher yield and productivity of water in the RR treatment was
evident only in seasons which received rainfall lower than long term average and experienced
intra-seasonal dry spells. In such seasons there was much more gain in yield and smaller risk of
grain loss when adopting conservation tillage practice. Seasons that received rainfall higher than
long term average experienced reduction in yield with conservation tillage due to possibility
induced water logging conditions. The rainfall conditions of Mkoji sub-catchment necessitate the
use of medium duration, higher yielding cultivars to realize benefits of tillage at the potential
level.
Acknowledgements
I acknowledge the Comprehensive Assessment Programme of the International Water
Management Institute (IWMI) for the financial support through the Soil Water Management
Research Group (SWMRG) of Sokoine University of Agriculture. I also acknowledge my
employer, the Ministry of Agriculture, Food Security and Cooperatives, Tanzania for
complementary funding of this study.
Appendices
Appendix 1: Physical and chemical soil properties of soil profile pit in the Igurusi field
experiment
Table 1a:
Soil
profile
depth
(mm)
0-150
150-370
370-500
500-650
650-800
8001000
Soil physical properties
Moistur Moistur Soil
e
e
bulk
content content densit
at field
at
y
capacity wilting
(dry)
(m3/m3) point
(g/cm3
3
3
(m /m ) )
0.262
0.127
1.44
0.295
0.163
1.39
0.305
0.226
1.45
0.305
0.226
1.45
0.278
0.212
1.38
0.278
0.212
1.38
Table 1b: Soil chemical properties
Field
Soil pH
EC
Fe
referen (in
(in
mS/c Mg/k
ce
wate KC
m
g
r)
L)
0-150
5.95
5.20
0.41
150370
370-
6.27
5.08
0.07
6.49
5.41
0.13
42.8
6
35.9
8
25.4
Soil
bulk
density
at Field
capacit
y
(g/cm3)
1.86
1.89
1.84
1.84
1.76
1.76
Clay
%
Silt
%
Sand
%
19
31
33
33
36
36
18
17
22
22
19
19
64
52
45
45
45
45
Soil Textural
Class
Sand loam
Sand clay loam
Sand clay loam
Sand clay loam
Sandy clay
Sandy clay
TN
%
OC
%
Exch
P
(mg/k
g)
CEC
Cmol(+)/
kg
Exchangeable Bases
Ca2 Mg K+ Na
+
+
2+
0.1
1
0.0
6
0.0
1.6
7
1.0
7
0.4
31.14
16.2
2.21
15.8
0.58
16.0
4.5
0
4.3
5
4.2
1.65
1.79
2.26
0.7
5
0.6
2
0.5
0.2
3
0.2
6
0.3
500
500650
650800
8001000
6.49
5.41
0.13
6.49
5.41
0.13
6.49
5.41
0.13
4
25.4
4
25.4
4
25.4
4
8
0.0
8
0.0
8
0.0
8
6
0.4
6
0.4
6
0.4
6
0.58
16.0
0.58
16.0
0.58
16.0
8
4.2
8
4.2
8
4.2
8
2.26
2.26
2.26
5
0.5
5
0.5
5
0.5
5
3
0.3
3
0.3
3
0.3
3
Appendix 2: Quantitative data of simulated and actual grain and biomass yield for the
2006/07 and 2007/08
Table 7a: Comparison between simulated and observed grain yield (kg/ha) 2006/07 and 2007/08
based on the standard plant density of 4.4 plants/m2
Treatment
CT
CTR
CTRCv
TR
RR
RRCv
Actual
2.3
2.8
2.7
2.7
2.3
2.4
2006/07
Simulated
4.9
5.2
3.4
4.3
4.7
3.4
Actual
1.7
2.0
2.4
2.7
3.8
3.8
2007/08
Simulated
3.3
5.0
4.1
4.8
5.0
4.1
Table 7b: Comparison between simulated and observed grain yield (kg/ha) 2006/07 and 2007/08
based on measured plant density (plants/m2) at 21 DAP
Treatment
CT
CTR
CTRCv
TR
RR
RRCv
Actual
2843
2866
2742
2690
2299
2392
2006/07
Simulated
3097
3097
2787
2784
2537
2588
Actual
1703
1970
2370
2696
3822
3822
2007/08
Simulated
1901
2047
2486
2779
3927
4022
Table 7c: Plant density for the different conservation tillage practices at 21 DAP
Treatment
2006/07
2007/08
CT
2.3
1.3
CTR
2.3
1.4
CTRCv
2.5
1.7
TR
2
1.9
RR
2
2.9
RRCv
2.1
3
Table 7d: Comparison between simulated and observed biomass 2006/07 and 2007/08 wet
seasons based on measured plant density 21 DAP
Treatment
Actual
2006/07
Simulated
2007/08
Actual
Simulated
CT
CTR
CTRCv
TR
RR
RRCv
10730
11013
9308
9409
7884
8079
10032
10130
8080
8295
7407
7552
4875
5776
6442
7034
9118
9118
6065
6385
7547
8242
10513
10630
References
Antonopoulos, V.Z. (1997). Simulation of soil moisture dynamics in irrigated cotton in semiarid climates. Agricultural Water Management 34, 233-246.
ARI UYOLE (2005) Annual Progress Report. Season 2002/03. Pp 33- 50.
Asseng, S.; Dunin, F. X.; Fillery, I. R. P.; Tennant, D. and Keating, B. A. (2004) Potential deep
drainage under wheat crops in a Mediterranean climate. II. Management opportunities to
control drainage. Australian Journal of Agricultural Research 52(1) 57 – 66.
Belmans, C., Wesseling, J.G. and Feddes, R.A. (1983). Simulation of the water balance of a
cropped soil: SWATRE. Journal of Hydrology, 63: 271-286.
Carberry, P.S., Probert, M.E., Dimes, J.P., Keating, B.A. and McCown, R.L. (2002). Role of
modelling in improving nutrient efficiency in cropping systems. Plant and Soil, 245, 193–
203.
Clemente, R. S., de Jong, R., Hayhoe, H. N., Reynolds, W. D. and Hare, M. (1994). Testing and
comparison of three unsaturated soil water flow models. Agricultural Water management
25: 135-152.
Cox, M.S., Gerard, P.D., Abshire, M.J. (2006).. Selected soil properties’ variability and their
relationships with yield in three Mississippi fields. Soil Science. 171:541-551.
Eitzinger, J., Trnka, M., Hösch, J. Žalud, Z. and Dubrovský, M. (2004). Comparison of CERES,
WOFOST and SWAP models in simulating soil water content during growing season under
different soil conditions. Ecological Modelling 171: 223–246.
Enfors, E. (2009) Traps and Transformations. Exploring the potential of Water Systems
Innovations in dryland sub-Saharan Africa. Doctoral Thesis. Stockholm University. Pp 164.
FAO (1999). New Concepts and approaches to land management in the tropics with emphasis on
steep lands. FAO Soils Bulletin 75. pp212.
Graves, A., Hess, T. and Matthews, R. (2002). Using models as tools in education and training.
In: Crop-Soil Simulation Models Applications in Developing Countries. (Edited by
Matthews, R. B. and Stephens, W.), CABI Publishing, New York, pp. 151-178.
Hoogenboom, G. (2000). Contribution of agro-meteorology to the simulation of crop production
and its applications. Agricultural Forest Meteorology. 103: 137-157.
Igbadun, H.I. (2006). Evaluation of irrigation scheduling strategies for improving water
productivity: Computer-based simulation approach. PhD. Thesis. Sokoine University of
Agriculture, Morogoro, Tanzania, Pp 245.
Iqbal, J., Read, J.J., Thomasson, A.J., Jenkins, J.N. 2005. Relationships between soil-landscape
and dryland cotton lint yield. Soil Sci. Soc. Am. J. 69:872-882.
Kato, T., Kimura, R. and Kamichika, M. (2004). Estimation of evapotranspiration, transpiration
ratio and water use efficiency from a spare canopy using a compartment model.
Agricultural Water Management 65:173-191.
Kropff, M.J., Teng, P.S. Rabbinge, R., (1995) The challenge of linking pest and crop models.
Agricultural Systems 49: 413 – 434.
Li, J., Inanaga, S., Li, Z. and Eneji, A. E. (2005). Optimizing irrigation scheduling for winter
wheat in the North China Plain. Agricultural Water Management 76: 8-23.
Mahdian, M. H. and Gallichand, J. (1995). Validation of the SUBSTOR model for simulating soil
water content. Transactions of the American Society of Agricultural Engineering 38: 513520.
Makurira, H., Savenije, H.H.G., Uhlenbrook, s., Rockstrom, J., Sanzanje, A. (2007). Towards a
better understanding of water partitioning processes for improved smallholder rainfed
agricultural systems: A case study of Makanya catchment, Tanzania. Physics and
Chemistry of the Earth. 32: 1082 – 1089.
Mkomwa, S., Mussei, A., Meakimbwala, R., Mulengera, N., Kiranga, E. (2007). Mbeya district
case study. In, Shetto, R. and Owenya, M., (2007). Editors, Conservation agriculture as
practiced in Tanzania: Three case studies. Nairobi. African Conservation Tillage Network,
Cente de Cooperation Internationale de Reserche Agronomique pour le Development, Food
and Agriculture Organization of The United Nations, pp 147.
Rao K.P.C. and Okwach G.E., (2005) Enhancing Productivity of Water Under Variable Climate.
Proceedings of The East Africa Integrated River Basin Management Conference held at
Sokoine University of Agriculture in Morogoro, Tanzania, between 7-9th, March 2005, pp.
Ringo, D.E., Maguzu, C., Mariki, W., Marietha, O. and Njumbo, F.S. (2007). Karatu district case
study. In, Shetto, R. and Owenya, M., (2007). Editors, Conservation agriculture as
practiced in Tanzania: Three case studies. Nairobi. African Conservation Tillage Network,
Cente de Cooperation Internationale de Reserche Agronomique pour le Development, Food
and Agriculture Organization of The United Nations, pp 147.
Ruiz-Nogueira, B., Boote, K. J. and Sau, F. (2001). Calibration and use of CROPGRO-Soybean
model for improving soybean management under rain fed conditions. Agricultural Systems
68: 151-173.
Shang, S., Li, X., Mao, X. and Lei, Z. (2004). Simulation of water dynamics and irrigation
scheduling for winter wheat and maize in seasonal frost areas. Agricultural Water
Management 68: 117-133.
Shetto, R. and Owenya, M. (2007). Conservation agriculture as practiced in Tanzania: Three case
studies. Nairobi. African Conservation Tillage Network, Cente de Cooperation
Internationale de Reserche Agronomique pour le Development, Food and Agriculture
Organization of The United Nations, pp 147.
Van Ittersum, M.K., Leffelaar, P.A., van Keulen, H., Kropff, M.J., Bastiaans, L., and Gourdriaan,
J. (2003) On approaches and applications of the Wagenigen crop models. European
Journal of Agronomy 18 (2003) 201 – 234.
Yin, X., Kropff, MJ., Goudriaan, J., Stam, P., (2000) A model analysis of yield differences among
recombinant inbreed lines in barley. Agronomy Journal 92: 114 – 120.