19th International Congress on Modelling and Simulation, Perth, Australia, 12–16 December 2011

http://mssanz.org.au/modsim2011

APSIM and DSSAT models as decision support tools

Mukhtar Ahmeda and Fayyaz-ul-Hassana
a
Department of Agronomy, PMAS-Arid Agriculture University, Rawalpindi, Pakistan

Corresponding author email address: ahmadmukhtar@uaar.edu.pk

ABSTRACT

Climatic variations, continuously increasing population pressure and market infrastructures are driven
forces to reduce agricultural productivity. New management options and appropriate genotypes are need of the
day to be considered for sustainable production. Crop simulation models are complementary tools in field
experiments to develop innovative crop management systems. In this perspective, two crop growth models viz.,
APSIM (Agricultural Production System Simulator) and DSSAT (Decision Support System for Agro Technology
Transfer) were calibrated and validated to predict growth and yield of wheat under rainfed conditions for Pothwar
region of Pakistan. Five wheat cultivars; Tatara, NARC-2009, Sehar-2006, SKD-1 and F-Sarhad were planted in
plot size of 5 m x 3 m with row spacing of 25 cm arranged in Randomized Complete Block Design (RCBD)
replicated thrice during 2008-09 and 2009-10 at Islamabad. Both the models were parameterized using different
agronomic parameters (phenological development, drymatter accumulation, leaf area index, physiological indices
and grain yield) and climatic data. Efficiency of both models was tested using model validation skill scores
including d-stat, RMSE and R2. RMSE between observed and predicted value by APSIM for drymatter for Tatara
genotype was 0.34 t ha-1 as compared to DSSAT with RMSE 1.40 t ha-1. Similarly, results observed for grain
yield and other agronomic parameters generated by APSIM were more precise while less accurate results were
depicted by DSSAT. Poor simulations were observed for Sehar-2006 by both models. While, predicted results for
NARC-2009 were close to Tatara, whereas, SKD-1 and F-Sarhad showed similar behaviour. Findings of the
study highlighted that crop models as a decision support tool to design agricultural production and crop
management system, select suitable crop and genotypes under changing climatic scenario, fertilizer management,
weeds management and appropriate sowing time. Similarly, these models are helpful in the managements of
biotic and abiotic stresses. Meanwhile, this study will introduce both models under rainfed conditions of Pakistan
which will further help to assess the constraints that influenced the model parameterization and validation. A set
of validation skill scores were used to evaluate the models to examine their usefulness, representativeness and
robustness. Similarly, the focus of the models comparison was based upon crop, soil and climatic data, technique
and model outcomes. The results generated by the models will help in evaluating strengths and weakness of each
of the modeling approaches, and based on these outcomes recommendations about managements could be made.

The results led to the conclusion that APSIM model was more efficient and suitable to simulate wheat
growth and development for arid area of Pakistan. Therefore, it can further be used for decision making to
mitigate climatic risk, selection of suitable genotypes and management options for agricultural sustainability.
Similarly, climatic scenarios analysis might be performed in the APSIM model to mitigate future problems
related to climatic extremes, therefore, it’s important to design future strategies based upon models outputs.

Keyword: Crop Simulation Modeling, APSIM, DSSAT, Climate, Arid Region

1174
Ahmed, M. and Fayyaz-ul-Hassan, APSIM and DSSAT models as decision support tools

1. INTRODUCTION
Climatic variability, moisture stress, continuously increasing population and market infrastructures are
driving forces to alter agricultural productivity. Changes in climatic scenarios are of vital importance for rainfed
agriculture as a change in one climatic variable alters other variables which include temperature, precipitation and
solar radiation. Associated impact of increasing temperature, changing rainfall pattern and intensity has led to
reduced agricultural productivity and yield over the world. Pakistan is developing country. Water availability
through rainfall has become a limiting factor for crop productivity due to lack of decision support and
management strategies like suitable sowing time, appropriate genotypes and cropping systems under changing
climate. Global earth temperature may increase up to 0.6-2.5 oC in the coming fifty years and by 1.1 – 6.4 oC
during the next century as temperature has been boosted up by 0.6 to 0.9 oC over the last 20th century on global
scale (IPCC, 2007). The climate is being changed abruptly due to variation in rainfall pattern, dry spells,
intermittent droughts and floods. Reduction and failure in crop production was projected due to these seasonal
variations (Mishra et al., 2008). Peterson et al. (1993) suggested that a systematic approach to the study of soil
and crop management problems is useful for testing present research knowledge to answer practical agricultural
problems and simultaneously identify gaps in basic research knowledge. Finding the best management options
and mimic climatic degradation are two key factors under contemplation in agronomic research to enhance crop
productivity. Crop simulation models proved to be efficient substitute for agricultural systems under diverse
climatic conditions. These models aid in decision making tools for better and sustainable agriculture (Amanullah
et al., 2007).
Crop simulation models consider the complex interactions between weather, soil properties and
management factors which influence crop performance. Crop simulation modeling has developed over many
years in recital with advances in crop physiology, crop ecology and computing technology. Modeling
performance had been envisaged with several opportunities in future including scientific investigation, decision
making by crop managers and key contributor in understanding and advancing the genetic regulation of plant
performance and plant improvement. An important task in experimenting with models is the testing their
performance in a wide range of circumstances to identify their scope of validity and limitations. As crop
simulation models are site and crop specific in nature and cannot be used in other areas until and unless validated
under local conditions. For simulating yield and biomass of maize, Clemente et al., (2005) compared CERES-
Maize and CropSyst modeling approach. Similarly, CERES-Wheat and CropSyst models were parameterized to
simulate water-nitrogen interaction in wheat (Singh et al., 2008). Comparison of different modeling approaches is
beneficial to select suitable crop model for a specific locality and climate so that it could be further used to
predict and simulate agricultural productivity under changing climatic scenarios.
Various models are being used around the world as a tool for yield forecasting. The same has been
initiated in Pakistan too, but requires testing of various models to identify their scope and limitations. It has been
observed that comparative evaluation of DSSAT and APSIM models has not been undertaken for wheat growth
and development in Pakistan yet. Based on this, the present study was carried out with the objectives to compare
the performance of APSIM and DSSAT models to simulate wheat growth and yield for forecasting of wheat crop
in high rainfall zone of Pakistan under different climatic scenarios.
2. APSIM (Agricultural Production System Simulator) overview
Agricultural Production Systems Simulator (APSIM) is a software tool that enables sub-models (or
modules) to be linked to simulate agricultural systems (McCown et al., 1996). APSIM has various modules
grouped and categorised as Plant, Environment and Management. It simulates the mechanistic growth of crops,
soil processes, and range of management options considering cropping systems perspective. APSIM-Wheat

1175
Ahmed, M. and Fayyaz-ul-Hassan, APSIM and DSSAT models as decision support tools

module simulates the growth and development of a wheat crop in a daily time-step on area basis (per square
meter, not single plant). APSIM module required inputs including weather, soil and crop data along with
management options (Fig. 1). Wheat growth and development in this module responds to climate, soil water
supply and soil nitrogen. The module returns information on its soil water and nitrogen uptake to the Soilwat and
SoilN modules on a daily basis for reset of these systems. Information on crop cover also be provided to the
Soilwat module for calculation of evaporation rates and runoff. Wheat stover and root residues were ‘passed'
from wheat to the surface Residue and SoilN module respectively at harvest of the plant crop. The SoilWater
module is a cascading water balance model that owes much to its precursors in CERES (Jones and Kiniry, 1986)
and PERFECT (Littleboy et al., 1992). SoilWater in APSIM is on daily basis, and typical of such models the
various processes are calculated consecutively. The SoilN module describes the dynamics of both carbon and
nitrogen in soil. The APSIM Met module provided daily meteorological information to all modules within an
APSIM simulation.
2.1 Model Parameterization and Evaluation
The data collected from the field experiments was used for model evaluation. Model calibration and
validation were described as different ways of model evaluation by Otter-Nacke et al. (1987). Specific cultivar
coefficients for the genotypes used in this experiment was not in the list of genotypes available with the model,
therefore, evaluation was done using basic information for the cultivar coefficients provided with the model. The
cultivar coefficients were adjusted, until main growth and development stages were simulated within 10% of the
measured values. Simulated observed comparisons were made for growth and development parameters, the
purpose being sensitivity analyses of the model and improvement of the coefficients. Coefficients were increased
or decreased using a small step if needed.
2.2 APSIM Model Utilization
APSIM was used to study the rainfall variability in rainfed areas of Islamabad in relation to global
mechanisms of El Niño Southern Oscillation / Southern Oscillation Index (ENSO/SOI) and its impact on weather
changes in Pakistan. To study the impact of ENSO/SOI phases on rainfall variability in this area, attempt was
made to establish a relationship between rainfall variability during October-November (wheat sowing time) and
monthly SOI phase in July (selected on the basis of rainfall data analysis). A probabilistic approach was used to
describe the chances of exceeding median rainfall. The median rainfall calculated from the long term (1961-
2010) actual rainfall data. Rainfall data (1961 - 2010) was analyzed using STATISTICA version 8 by plotting the
total monthly actual rainfall received over that period of time to get the pattern in Islamabad during summer
(locally called as kharif season starting from May extended until September) and winter (locally known as rabi
season, the duration of which is from October to April) seasons (1961-2010). Actual rainfall data was aggregated
over a span of 3 years. Range of deviation from 3 years’ mean was calculated against each aggregated value.
APSIM-wheat module used to simulate wheat crop using long term (1961-2010) climatic data. The averaged
yield data partitioned on the basis of July SOI phase. The purpose of this partitioning was to explore the use of
seasonal climate forecasting based on SOI phase for the selection of suitable cultivar. The APSIM, when
parameterized in this pattern, was further used to explore farmer cropping decision options given the variable
climate to enhance the resilience in wheat based cropping system of the area. A scenario analysis was done using
a long term historical weather data to explore the feasibility of wheat based cropping system in the area as
proposed by Asim et al. (2006). Similarly, use of APSIM model to study long term data of Pothwar from 1961-
2010 revealed an obvious trend of increasing summer rainfall and decreased winter rainfall. The observed trend
also indicated a significant tendency toward increased rainfall in July-September and decreased rains in
September. Based on information generated the analysis revealed an opportunity of early sowing and hold true
for selection of optimum sowing time in the mid of October as the selection of sowing time in this rainfed area
depend on moisture availability.

1176
Ahmed, M. and Fayyaz-ul-Hassan, APSIM and DSSAT models as decision support tools

System Control
Manager
Cloc Report

Canopy Met

Wheat SoilWat
Crops E
N
SWI
Maiz
G Soil
I
Sorg N SoilPH
E
Legume SoilP
Other Crops
Soil
New Module

Irrigate Erosion

Fertilize Manure
Management Economics Residue

Fig.1. Working mechanism of APSIM model

3. DSSAT (Decision Support System for Agrotechnology Transfer) Overview

The DSSAT cropping model simulates growth, development and yield of crop growing under described
managements over the time. The DSSAT is structured using modular approach as reported by Jones et al., (2001).
It permits easy incorporation of diverse application packages because of well defined and documented interface
to modules. DSSAT is collection of independent programs which then operated together. Databases include
weather, soil, experimental conditions and genotypes to apply the models for different situations. These
application softwares aid to prepare these databases and to compare simulated results with observed values and to
improve model’s efficiency and accuracy. DSSAT crop model allows users to simulate options for crop
management to assess risks.
3.1 DSSAT Parameterization and calibration
A soil input file required with detailed hydraulic and physical properties of soil to simulate DSSAT. It
does not offer automated procedure for its calibration and validation. Changes in its input parameters must be
done to validate it under local conditions of a locality. In application of DSSAT for precision agriculture,
different files need to be generated for various management zones. Several types of modules are generated for
DSSAT to get simulated results which includes soil module, weather module, soil plant atmosphere module and
crop module. Several genetic coefficients generated for DSSAT parameterization. These coefficients includes

1177
Ahmed, M. and Fayyaz-ul-Hassan, APSIM and DSSAT models as decision support tools

P1D (Photoperiod sensitivity coefficient), P1V (Vernalization sensitivity coefficient), P5 (Thermal time from the
onset of linear fill to maturity), G1 (Kernel number per unit stem/spike weight at anthesis), G2 (Potential kernel
growth rate), G3 (Tiller death coefficient. Standard stem/spike weight when elongation ceases) and PHINT
(Thermal time between the appearance of leaf tips). Afterward, model evaluation being done by comparing
observations and simulated results and measuring its efficacy and suitability for specific predictions and locality
(Jones et al., 2003).
4. Implementation era
Simulation model testing consists of two main activities (i) establishing the source codes representing
the models performance as intended, and (ii) confirming that simulation models accurately reproduce empirical
data (Meinke, 1996).These two activities were referred as model verification and validation (V & V) (Kleijnen,
1995). In the present study, APSIM and DSSAT module were tested for wheat phenology, days after sowing, leaf
area index, drymatter accumulation (t ha-1) and grain yield (t ha-1) as these are major constituents to determine
optimal productivity of crop. Both the models were calibrated under local weather conditions to simulate wheat
genotypes of diverse origins under rainfed conditions. The data collected proved to be beneficial and useful to
evaluate both the tactical and modeling approaches. Models were tested by validation skill scores including root
mean square error (RMSE), normalized mean square error (NMSE), d-stat and R2 which allow comparative
assessment of model performance at particular location whereas, linear regression line expressed model stability
across variable climatic conditions. Both the models performed well to simulate wheat growth. However,
predicted results derived from APSIM model were much better to observed ones for all parameters as compared
to DSSAT. APSIM model predicted all the parameters with significant R2 than DSSAT for all wheat genotypes.
However, both simulation approaches showed better and precise results for zadok phenology and days after
sowing as predicted and observed values were close. Simulated results by APSIM for leaf area index, drymatter
accumulation and grain yield were close to observed ones with more R2 and RMSE than observed for DSSAT
which highlighted the performance of APSIM module. From the results, it was observed that APSIM module
gave unbiased and precise predictions of zadok phenology, leaf area index, drymatter accumulation and grain
yield for five genotypes as close association was demonstrated between observed and predicted values. In the
same way, Zhang et al., (2008) reported that yield simulation might be improved if model could simulate more
accurate days after sowing under variable climatic scenarios. Simulated results for leaf area index depicted that
leaf area at different zadok growth stages was close to predicted values by APSIM which was near to 1:1 line as
compared to DSSAT (Table 1 (a)).
Both simulation approaches were parameterized to simulate wheat drymatter accumulation from three
leaf upto maturity and results highlighted that APSIM model simulated better results for drymatter accumulation
with higher R2 and results closely associated to 1:1 line (Table 1 (b)). The use of models to simulate above
ground biomass was reported with good accuracy by Chen et al., (2010) who concluded that APSIM-wheat
module can simulate biomass and model was able to explain more than 90% variation in crop biomass. Grain
yield is of more and vital importance as all experimental studies focused upon yield. Likewise, other attributes,
APSIM module performed efficiently for grain yield with R2 more than 0.90 % while, observed and predicted
values by DSSAT were not closely related and far away from 1:1 line (Table 1 (c)). APSIM model proved to be
more efficient and reliable to simulate wheat crop under rainfed conditions. The overall prediction of grain yield
by model was reported satisfactory by Singh et al., (2008) with R2 values (0.88). The results depicted that APSIM
model performed well and its efficacy can be determined by validation skill scores.

5. Conclusion

Crop modeling is becoming a valuable tool to understand and mimic climatic constraints and yield gaps. The
outcomes of the study clearly depicted that APSIM module predicted wheat crop growth and yield with more
accuracy than DSSAT. So, it is more appropriate and can be parameterized to simulate crop growth under
changing climatic scenarios to select suitable genotypes, sowing time, cropping pattern, fertilizer and weed

1178
Ahmed, M. and Fayyaz-ul-Hassan, APSIM and DSSAT models as decision support tools

management strategies enabling crop to cope with environmental hazards. Therefore, studies have clearly
depicted crop simulation model as potential agronomic and decision making tool to understand wheat crop
biodynamism under variable climatic conditions of dryland agriculture. Furthermore, the evaluation of models all
over Pakistan will enhance our knowledge to pick variability and build resilience in various temperature and
rainfall regimes and prove a better tool for crop yield and rainfall forecast of the region.

Table 1. Model Validation Skill Scores for (a) Leaf Area Index, (b) Drymatter accumulation (t ha-1) and (c) Grain
Yield (t ha-1)

(a)

APSIM DSSAT
Genotypes
R2 d-stat RMSE NMSE R2 d-stat RMSE NMSE
Tatara 0.99 0.98 0.34 0.02 0.56 0.70 1.40 0.39
NARC-2009 0.97 0.97 0.38 0.02 0.51 0.81 1.11 0.29
Sehar-2006 0.94 0.94 0.48 0.05 0.48 0.62 1.30 0.59
SKD-1 0.93 0.94 0.51 0.06 0.59 0.61 1.30 0.55
F-Sarhad 0.91 0.92 0.65 0.08 0.51 0.52 1.63 0.91

(b)
APSIM DSSAT
Genotypes
R2 d-stat RMSE NMSE R2 d-stat RMSE NMSE
Tatara 0.99 0.99 1.13 0.05 0.81 0.76 4.58 0.91
NARC-2009 0.97 0.96 1.70 0.09 0.83 0.80 3.80 0.62
Sehar-2006 0.89 0.96 1.86 0.12 0.83 0.79 3.72 0.67
SKD-1 0.87 0.99 0.96 0.03 0.78 0.85 2.87 0.46
F-Sarhad 0.90 0.98 1.12 0.04 0.81 0.83 3.22 0.54

(c)

APSIM DSSAT
Genotypes
R2 d-stat RMSE NMSE R2 d-stat RMSE NMSE
Tatara 0.95 0.97 0.26 0.00 0.62 0.65 0.73 0.03
NARC-2009 0.92 0.82 0.48 0.02 0.56 0.69 0.70 0.03
Sehar-2006 0.83 0.88 0.48 0.02 0.51 0.66 0.70 0.03
SKD-1 0.87 0.87 0.49 0.02 0.54 0.68 0.68 0.03
F-Sarhad 0.84 0.85 0.49 0.02 0.58 0.69 0.44 0.03
6. References

Amanullah, M., C. K. Mohamed, A. Safiullah, S. Selvam, and K. Siva- kumar ( 2007). Crop simulation growth
model in Cassava. Research Journal of Agriculture and Biological Science, 3 (4), 255–259.

Asim, M., M. Aslam, N. I. Hashmi, and N. S. Kisana (2006). Mungbean (Vigna radiata) in wheat based cropping
system: An option for resource conservation under rainfed ecosystem. Pakistan Journal of Botany, 37(4), 1197-
1204.

Chen, C., E. Wang, and Q. Yu (2010). Modelling the effects of climate variability and water management on crop
water productivity and water balance in the North China Plain. Agricultural Water Management, 97, 1175-1184.

1179
Ahmed, M. and Fayyaz-ul-Hassan, APSIM and DSSAT models as decision support tools

Clemente, R.S., M.E. Asadi, and P.N. Dixit (2005). Assessment and comparison of three crop growth models
under tropical climate conditions. Journal of Food, Agriculture Environment, 3 (2), 254–261.

IPCC. (2007). Summary for policymakers of climate change: the physical science basis. Contribution of Working
Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge
University Press, Cambridge.

JJones, J.W., G. Hoogenboom, C.H. Porter, K.J. Boote, W.D. Batchelor, L.A. Hunt, P.W. Wilkens, U. Singh,
A.J. Gijsman, and J.T. Ritchie ( 2003). The DSSAT cropping system model. European Journal of Agronomy, 18,
235-265.

Jones, C. A. and J. R. Kiniry (1986). CERES-Maize: A simulation model of maize growth and development.
Texas A & M University Press, College Station, Texas.

Jones, J.W., B.A. Keating, and C.H. Porter (2001). Approaches to modular model development. Agricultural
Systems. 70, 421-443.

Kleijnen, J.P.C (1995). Verification and validation of simulation models. European Journal of Operational
Research, 82 (1), 145-16.

Littleboy, M., D. M. Silburn, D. M. Freebairn, D. R. Woodruff, G. L. Hammer, and J. K. Leslie (1992). Impact of
soil erosion on production in cropping systems. I. Development and validation of a simulation model. Australian
Journal of Soil Research, 30, 757 – 774.

McCown, R. L., G. L. Hammer, J. N. G. Hargreaves, D. P. Holzworth, and D. M. Freebairn (1996). APSIM: a

novel software system for model development, model testing and simulation in agricultural systems research.
Agricultural Systems, 50, 255 – 271.

Meinke, K. (1996). Topological methods for algebraic specification. Theoretical Computer Science, 166, 263-
290.

Mishra, A., J. W. Hansen, M. Dingkuhn, C. Baron, S. B. Traore, O. Ndiaye, and M. N. Ward. (2008). Sorghum
yield prediction from seasonal rainfall forecasts in Burkina Faso. Agriculture Forest and Meteorology, 148,
1798–1814.

Otter-Nacke, S., D. C. Godwin, and J. T. Ritchie (1987). Testing and validating the CERES-Wheat model in
diverse environments. Michigan State University. Pp 1-147.

Peterson, G. A., D. G. Westfall, and C. V. Cole (1993). Agroecosystem approach to soil and crop management
research, Soil Science Society of America Journal, 57, 1354-1360.

Singh, A. K., R. Tripathy, and U. K. Chopra (2008). Evaluation of CERES-Wheat and CropSyst models for
water-nitrogen interactions in wheat crop. Agricultural Water Management, 95, 776-786.

Zhang, X., S. Chen, H. Sun, D. Pei, and Y. Wang (2008). Dry matter, harvest index, grain yield and water use
efficiency as affected by water supply in winter wheat. Irrigation Science, 27, 1-10.

1180