PCA–based composite drought index for drought assessment in Marathwada region of Maharashtra state, India.

This paper presents a composite approach for drought characterization and monitoring using in situ and remote sensing-based drought indicators. The study was carried out on one of the most drought-prone areas of India, i.e., the Marathwada region of Maharashtra. Meteorological, hydrological, and agricultural drought indices, namely standardized precipitation index (SPI), streamflow drought index (SDI), and vegetation condition index (VCI), respectively, were integrated to develop the composite drought index (CDI) using principal component analysis (PCA). SPI and SDI were computed using in situ precipitation and streamflow data, respectively, for 35 years (1980–2014), whereas VCI was computed using MODIS satellite data (500-m resolution) for 15 years (2000–2014) at 1-, 3-, and 5-month time scales. The time scales of drought indices were evaluated using historical drought years and foodgrain production of the region. The drought areas observed by SPI, SDI, and VCI at different time scales were correlated with foodgrain production during the kharif crop growing season for 15 years (2000–2014). The maximum correlation of foodgrain production was observed with 3-month SPI (r = − 0.72), 5-month SDI (r = − 0.40), and 5-month VCI (r = − 0.81) for meteorological, hydrological, and agricultural drought, respectively. Three-month SPI, 5-month SDI, and 5-month VCI were selected from each drought category to develop CDI. These drought indices were combined using weights derived through the PCA technique. The maximum weight was obtained for 3-month SPI (45.4%) followed by 5-month VCI (42.8%) and 5-month SDI (11.8%). The developed CDI products showed a strong relationship (r = − 0.85) with foodgrain production. The drought years observed by CDI were also closest to drought years declared by the State Government. The time series trend of the drought-affected area observed by CDI, 3-month SPI, and 5-month VCI resembled the drought patterns very closely, especially during the drought years. The spatio-temporal analysis of individual drought indices and CDI with foodgrain production deviation showed that CDI was better for capturing drought conditions than individual indicators. The study suggested that an individual or single indicator is not sufficient to capture the actual drought severity and its magnitudes; therefore, using a composite approach could be a good choice for effective drought assessment and monitoring in the region. [ABSTRACT FROM AUTHOR]