Evaluation of agricultural production systems simulator (apsim) as a yield predictor for maize cropping systems in sri lanka

Abstract

Contemporary agriculture undergoes enormous challenges in order to compensate the increasing food demand of the rapidly growing global population. Crop intensification is crucial to address that rising food requirement whilst focusing on the potential environmental impacts of intensive agriculture. Crop models are deployed in crop intensification process ensuring sustainability through virtualization of interactions between climate, soil, management practices and genotypes. Among the crop models, Agricultural Production Systems Simulator (APSIM) which is comprised of sub models capable of predicting crop yields and phonological parameters through integration of underlying factors. Present study evaluates the applicability of APSIM-maize sub model in yield prediction under Sri Lankan conditions. The study was undertaken in collaboration with CIC Grains (Pvt.) Ltd. for the existing maize fields in Siyambalanduwa area. The model was validated using the growth parameters viz., plant height and number of leaves, for 2016/17 maha cropping season. With the secondary soil, climate and management data, the model simulated growth parameters at weekly intervals and were compared statistically with the observed data. Plant height and number of leaves were predicted by the model with a strong correlation with observed data with calculated r2 values of 0.82 and 0.99. Student’s paired t-test showed no significant difference where P value was 0.73 and 0.67 at 95% confidence level. Positive modelling efficiency of 0.60 for plant height and 0.79 for number of leaves affirmed that the model predictions were acceptable. The model achieved RMSE of 33.61 for plant height with a standard deviation of 73.3and RMSE value of 1.15 with a standard deviation of 5.51. However, the model overestimated the yield for 2016 (observed yield=5.24 mt/ha, Predicted yield=6.27 mt/ha) and 2015 (observed yield=3.85mt/ha, predicted yield=5.22 mt/ha by 21%. This discrepancy could have been aroused due to the generalization of the soil data for the entire cultivated area under the study.