Temporal Rainfall Variations Induce Forecast Errors in Rainfed Agriculture in the Brazilian State of Ceara, Brazil

The research aims to: a) assess the instabilities associated with rainfall and the variables that define the production of rice, beans, cassava and corn in the state of Cear¨¢ between 1945 and 2020; b) estimate models that can be used to make projections of harvested areas, yields and prices for these crops between 1945 and 2020; c) assess the impact of rainfall on the estimated forecasting models; d) assess how rainfall affects the likelihood of farmers making forecasts of the variables that define agricultural production. Rainfall data was obtained from the National Centers for Environmental Information (NOAA). Crop yield data came from the Brazilian Institute of Geography and Statistics (IBGE). Instabilities were measured by the coefficients of variation. ARIMA models (autoregressive, integrated and moving average model) were used to make the forecasts. The hypothesis that the residuals generated by the models are influenced by annual rainfall was tested. The results showed high instabilities in annual rainfall, which spread to the variables that define crop yields. Parsimonious and robust adjustments were obtained from a statistical point of view and it was shown that the errors generated, including their magnitudes, in the models used to forecast all the variables that define bean and corn yields, harvested areas and rice yields, as well as cassava yields, are influenced by annual rainfall in Cear¨¢ between 1945 and 2020.