Test and Application of the DSSAT Model to Assess Rice Productivity in the Accra Plains of Ghana

Abstract

Field studies on the growth and development of the rice variety TOX 3107 were conducted at the Agricultural Research Station (A.R.S,) Kpong with the aim of calibrating and validating the CERES-RICE model of DSSAT (version 3.5). Rice was grown for two seasons namely 1999 minor cropping season (August to December) and 2000 major cropping season (April to August). In the 1999 minor season experiment, rice was grown under rainfall with supplementary irrigation. There were two simultaneous experiments in the 2000 major cropping season. In one experiment, rice was grown under rainfall with supplementary irrigation and in the second, rice was grown under solely rainfed conditions. Three levels of nitrogen fertilization namely 0-, 45- and 90 kg N ha-1 were used in all experiments. There was a basal application of 45 kg P ha-1 and 35 kg K ha-1 in each case. Randomised complete block design was used with four replications. The growth of the irrigated rice crop in the 2000 major season was better than the irrigated crop in the 1999 minor season. The irrigated rice crop in both seasons performed better than the rice grown under rainfed conditions in the 2000 major season. The durations of the phenological growth stages of rice grown in the minor season were shorter than those in the major season under both irrigated and rainfed conditions. Crop, soil, weather and field management data from the 1999 minor season experiment were used to calibrate the CERES-RICE model. The calibrated model failed to predict crop growth and development well for the 2000 major season experiment and for an independent data. The goodness of fit for simulations of the calibrated model did not give an indication of a high predictive capability. This constrained its application to assess the output of different nutrient levels, management or weather scenarios.