Performance of quantitative drought indices in the upland rice yield estimates

The objective of this work was to characterize the drought intensity and occurrence quantitative indices, and to evaluate relationship between these index and the upland rice yield historical data of Goiania micro region, GO, Brazil. The historical data set was adjusted in order to minimize the effects of region's climatic variability and technological advances on crop yield. Four indices were evaluated: Palmer's drought severity index (PDSI), Palmer's Z index (Z-index), rainfall anomaly index (RAI), and the standardized precipitation index (SPI). Drought indices were analyzed trough Pearson's correlation, number and frequency of drought occurrence, and the index accuracy related to the upland rice adjusted yield data. RAI quantified the highest number of drought extreme occurences, while PDSI did not estimate any case. Z-index showed the highest accuracy in relation to the variations in the adjusted yield. In periods with rice adjusted yield variations higher than 300 kg ha(-1), the Z-index, RAI and SPI indices showed the highest accuracies with, 78, 78 and 67%, respectively. The Z-index had the best performance in the estimation of upland rice adjusted yield.