Experimental analysis and modelling of the rainfed rice cropping systems in West Africa

Abstract

There is a need to improve rice productivity to meet the increasing demand for rice in West Africa since it is acknowledged that existing rainfed rice cultivation practices deal with irregularities of climate like drought or submergence with iron toxicity risk on the one hand and on the other hand land management associated with soil fertility and the topography. This study addressed the above issues by investigating experimental results in both rainfed lowland and upland system.<br /> In a rainfed lowland system, the study examined first the constraints to rice production in inland valleys in West Africa which depend on the rainfall distribution and the heterogeneity of the topography that leads frequently to mobilization of Fe2+ and runoff causing erosion and loss of N. During 4 years (2007 to 2010), a three factorial trial showed that the grain yield across the seasons had quite diverse response with respect to slope position (up and down) and management practices (bunds and fertilizer). The impact of fertilizer has been significant in the year 2009 leading to the increase of grain yield by 0.45 Mgha-1 with fertilizer compared to the control. Negative correlation with Fe concentration in rice was only found at the upper slope position. Our findings showed that, at the upslope, Fe concentration in rice is higher with bunding. At downslope position, rice yield was significantly correlated with ponding water level in the first month, cation exchange capacity and organic C of the soil and N concentration in the rice tissue.<br /> As the exploitation of lowland inland valleys for rice production requires improved understanding of the effect of management practices on soil water, nutrient dynamics and rice yield, the crop model EPIC (Environmental Policy Integrated Climate) was further applied to the upper slope position in order to capture processes involved in crop development and yield in temporarily inundated rice fields and to assess the suitability of the model for this specific agroecosystem. The model was parameterized using observed soil water characteristics and crop parameters and run against observation data. The simulated LAI development, aboveground biomass and grain yield compared well with field observations. MRE (mean relative error) of simulated yield was 6 to 18 % except for with bund plots in 2009 and 2010, where grain yield was overestimated by the model when no fertilizer was applied (MRE=45%). This was due to the negative effect of elevated iron concentration in the rice plant, which the model was not able to consider in the simulations.<br /> In upland rice experiments, our study was motivated by the challenge for increasing productivity to grow rice on low-input farmland. Therefore, we assessed improved upland varieties in 6 sites of Benin Republic. Although uniform fertilizer input was applied across the experiments, the effect of site and interaction between site and years appeared as factors that strongly influence rice production. Environments with higher organic carbon coupled with sufficient rainfall water during the cropping period led to higher grain yield. These conclusions therefore confirmed that the test of the performance of field scale crop models under different agro-ecological conditions is a prerequisite for the evaluation of the impact of management strategies for larger scales. Therefore, the EPIC model was again tested for upland land rice production by taking into account seasonal variability in Guinean and Guinean-Sudanian zones in Benin and Nigeria (West Africa). The results showed the accuracy of the model to simulate LAI, total above ground biomass and grain yield. The model exhibited more variability in yield for increasing N fertilizer application than P. In addition, general precision in model output is reduced when considering farmer’s field condition. Large root mean square RMSE in calibration (<35) and the validation (>100) suggested that robustness of the model became restrictive under severe drought condition while the rice response to N fertilizer became reduced. <br /> The general use of the model for rainfed rice production at a large scale requires identification of areas with iron toxicity, drought and flooding risk and improvement of the model with respect to the impacts of iron toxicity and drought on rainfed rice.