Time-Varying Meteorological Drought Index for a Changing Climate

Drought is considered as one of the most complicated natural disasters, whose adverse effects span over different domains such as agriculture, ecosystem, and economy. One of the widely used meteorological drought index is Standardized Precipitation Index (SPI), which is based on the stationary assumption (i.e., the statistical parameters does not change over time). Nevertheless, numerous studies have reported that the precipitation series has undergone remarkable changes, which emphasizes the need for developing a drought index incorporating the dynamic behavior of the precipitation. Hence, in this study, a non-stationary SPI (NSPI) is developed to capture the temporal dynamics of the precipitation and to identify the meteorological drought-prone areas over India. Before modelling, the non-stationarity in the distribution parameters of precipitation series are detected. If non-stationary is observed in any of the parameters, then that particular parameter is modelled as non-stationary, otherwise it is modelled as stationary. The proposed index provides a probability-based description of drought status and its uncertainty bounds, which are computed using Bayesian Inference. Results reveals that the traditional SPI is biased by the lowest magnitude of precipitation leading to overestimation of drought where frequent severe dry events are clustered, which is overcome by NSPI. Additionally, NSPI captured the historical drought, capturing the temporal dynamics of precipitation series in India and is more reliable than SPI. The proposed NSPI is found to be a potential index for drought monitoring in a nonstationary climate.