In this study, Standardized Precipitation Index (SPI) derived from parametric and nonparametric methods using 0.25^∘ gridded rainfall data from 1901 to 2013 (113 years) generated by India Meteorological Department (IMD) was compared for understanding drought conditions over the Indian region. The parametric SPI was computed using a three-parameter Gamma distribution function, whereas nonparametric SPI was computed using Gringorten, Weibull, and Hazen plotting positions, on a 4-month cumulative rainfall data of June–September (SPI-4) representing the southwest monsoon season. Nonnormality is a major concern if equal-sized intervals are drawn for interpretation, and SPI being a normalized index wherein classes are standard deviations from normal, its impact on drought assessment needs to be understood. Accordingly, in our study, normality tests were performed using the Shapiro-Wilk method on SPI derived from both parametric and nonparametric methods. The SPI showed 100% of grid cells conforming to normality in the case of nonparametric methods, whereas in the case of parametric approach it was only 80%. The remaining 20% of nonnormality in parametric SPI is spread over montane, tropical wet, and semi-arid regions of India. Furthermore, differences in the estimation of dryness are observed in the range of 1.0 to 2.5% between nonparametric and parametric SPI for the drought years considered this study. The quantile analysis on all grid cells for the drought year 2002 showed an important fact that at 0.025 quantile only 2.6% of grid cells are in the extremely dry condition as per parametric SPI, whereas in the case of nonparametric SPI it is 6.9%. For the drought year 1939 in grid cells where normality is not followed in parametric SPI, Cohen’s kappa (κ = 0.15) under extreme dryness category indicates large disagreements between parametric and nonparametric SPI. The temporal analysis of Cohen’s kappa computed for each grid cell over drought years shows that in 22.5% of cases the drought category between nonparametric and parametric SPI is not in perfect agreement. Hence, the nonparametric SPI can better categorize the drought classes, representing well the extent of dryness and normality conditions, it is highly recommended for drought assessment over India.

在这项研究中，标准化降水指数（SPI）从参数和非参数的方法，使用0.25衍生^∘比较了印度气象局（IMD）生成的1901年至2013年（113年）的栅格化降雨数据，以了解印度地区的干旱状况。参数SPI是使用三参数Gamma分布函数计算的，而非参数SPI是使用Gringorten，Weibull和Hazen绘图位置计算的，该数据代表了西南季风的6月至9月（SPI-4）的4个月累积降雨数据季节。如果绘制大小相等的区间进行解释，并且SPI是归一化的指标（其中类别是与正常值的标准差），则非正态性是一个主要问题，需要了解其对干旱评估的影响。因此，在我们的研究中，对来自参数方法和非参数方法的SPI使用Shapiro-Wilk方法进行了正态性测试。在非参数方法的情况下，符合正常性的网格单元百分比为80 ％，而在参数方法的情况下，仅为80 ％。参数SPI中剩余的20 ％的非正态分布在印度的山地，热带湿润和半干旱地区。此外，在本研究的干旱年份，非参数SPI和参数SPI之间的干燥度估计差异在1.0％到2.5％之间。对2002年干旱年份所有网格单元的分位数分析表明，一个重要事实是，按照参数SPI，在0.025分位数下，只有2.6 ％的网格单元处于极端干燥条件下，而对于非参数SPI，则为6.9 ％。对于1939年的干旱年，在参数SPI中未遵循常态的网格单元中，极端干燥类别下的科恩卡帕（κ = 0.15）表明参数SPI与非参数SPI之间存在较大分歧。在干旱年中针对每个网格单元计算的科恩kappa的时间分析表明，在22.5 ％的情况下，非参数SPI和参数SPI之间的干旱类别不一致。因此，非参数SPI可以更好地对干旱类别进行分类，很好地表示干旱程度和正常状况的程度，强烈建议在印度进行干旱评估。