Correlation vs. Kappa statistic: a new perspective on comparing meteorological drought indices

Comparing meteorological drought indices is essential for selecting the most appropriate index for specific regions and purposes, as well as for validating new or improved drought indices. This study evaluates four meteorological drought indices—Standardized Precipitation Index (SPI), Rainfall Anomaly Index (RAI), Standardized Precipitation Evapotranspiration Index (SPEI), and Reconnaissance Drought Index (RDI)—at 1, 3, and 6-month time scales using data from 36 stations over 40 years in the Saurashtra region of Gujarat, India. In addition to the commonly used correlation coefficient method, this study explores relatively new alternative approach of Kappa statistic for comparing these indices. The two methods yielded distinct results; while RAI showed strong correlations (ranging from 0.90 to 0.95) with other indices, its Kappa Coefficient values were significantly lower, ranging from 0.16 to 0.42. The correlation coefficient indicated strong relationships (r > 0.92) between SPEI and SPI/RDI across all time scales, whereas the Kappa coefficient ranged from 0.67 to 0.81. The results were validated using (i) drought severities and frequencies (ii) mapping spatial patterns of major historic droughts and (iii) crop yield loss for major crops (cotton and groundnut) and corresponding drought severity. While RAI indicated very high frequency and severity of drought, SPI and RDI showed nearly identical drought detection patterns. SPEI proved better at capturing mild and moderate droughts compared to SPI and RDI but underestimated extreme drought conditions. Despite high correlations among the indices, each one exhibited unique characteristics, highlighting different aspects of drought conditions. The Kappa Coefficient was found to be more effective for comparing indices as it aligned more accurately with observed drought conditions. This improved understanding will aid in more efficient drought declaration, monitoring, and the development of mitigation strategies.