An evaluation of agricultural drought indices for the Canadian prairies

A comparative performance analysis was carried out on four agricultural drought indices to determine the most appropriate index for monitoring agricultural drought and predicting Canada Western Red Spring wheat (Triticum aestivum L.) yield on the Canadian prairies. A series of curvilinear regression-based crop yield models were generated for each of the 43 crop districts (20 in Saskatchewan, 12 in Manitoba, and 11 in Alberta) in the study region based on four commonly used measures of agricultural drought (Palmer Drought Severity Index, Palmer's Z-index, Standardized Precipitation Index, and NOAA Drought Index). The yield models were evaluated by comparing the model predicted yields to the observed yields (1961-1999) using four goodness-of-fit measures: the coefficient of determination (R-2), the index of agreement (d), the root mean square error (RMSE), and the mean absolute error (MAE). The analysis indicated that Palmer's Z-index is the most appropriate index for measuring agricultural drought in the Canadian prairies. The model evaluation indicated that the Z-index is best suited for predicting yield when there is significant moisture stress. There is a statistically significant relationship between the Z-index and Red Spring wheat yield in all crop districts, but the strength of the relationship varies significantly by crop district due to the influence of factors other than moisture availability (e.g. disease, pests, storm damage, and soil characteristics). The significant variations in model performance between the four agricultural drought indices underscores the necessity of carrying out a performance evaluation prior to selecting the most appropriate agricultural drought index for a particular application. (C) 2003 Elsevier Science B.V. All rights reserved.