Monitoring and forecasting drought impact on dryland farming areas

Frequent drought amplifies the need for a warning system and forecasting models for damage to crop yields. This study developed an operational model to assess agricultural drought impact. The dryland areas of Kermanshah Province (Iran) were selected to test the proposed modelling system. The model predicted the consequences of drought damage on wheat crop during critical stages of growth (emergence, vegetative growth, initiation of flowering, grain filling, and maturity) as a drought loss indicator. Two types of input were evaluated to correlate climate conditions versus drought losses. The first group comprises the Palmer Drought Severity Index, Z-index, Crop Moisture Index, Crop-Specific Drought Index (CSDI), Standardized Precipitation Index, and Effective Drought Index with one- to three-month timescales used as meteorological indices. The second group, which is consistent of the vegetation condition index and temperature condition index, is based on satellite data. Also a new satellite-based version of CSDI, so-called standardized CSDI (S-CDSI), where evapotranspiration was estimated using surface energy balance algorithm for land, is used. Adaptive Neuro-Fuzzy Inference Systems (ANFIS) technique was used for forecasting with genetic algorithms applied to select appropriate inputs from among the large number of indices. It was concluded that the combination of meteorological and satellite indices performed best in forecasting crop yield. As expected, accuracy improved over the growth stages as the crop developed. Enhancement of the model with a GIS platform made it possible to present the results more suitably, hence helping users to make more realistic decisions. Copyright (c) 2012 Royal Meteorological Society