A dynamic hierarchical Bayesian approach for forecasting vegetation condition

Abstract. Agricultural drought, which occurs due to a significant reduction in the moisture required for vegetation growth, is most complex amongst all drought categories. The onset of agriculture slow and can occur over vast areas with varying spatial effects, differing particular land cover or specific agro-ecological sub-regions. These variations imply that monitoring forecasting agricultural require models consider given region interest. Hierarchical Bayesian are suited modelling such systems. Using partially pooled data sub-groups characterise differences, these capture sub-group variation while allowing flexibility information sharing between sub-groups. This paper's objective improve accuracy precision spatially diverse regions hierarchical model. Results showed model was better at capturing variability different zones covers compared regular auto-regression distributed lags forecasted condition associated probabilities were more accurate precise 4- 10-week lead times. Forecasts from exhibited higher hit rates low probability false alarms events semi-arid arid zones. also good transferable forecast skills counties not included training data.