Drought is one of the most complicated but least understood natural hazards (Portela et al., 2015; Wilhite, 1996) with widespread impacts on water resources (Zhang et al., 2015), agricultural production (AlKaisi et al., 2013), ecosystem function (Bond et al., 2008), environment (Dijk et al., 2013), local and global economies (Ding et al., 2011). Generally, there are four types of drought: meteorological, agricultural, hydrological and socio-economic drought (Heim, 2002). Despite its widespread damages, drought identification and characterization are challenging due to its varying definition, great spatiotemporal variability, complicated physical process and non-structural impacts (Nam et al., 2012; Wu and Wilhite, 2004). Over the past decades, lots of efforts have been devoted to drought monitoring and characterization and a series of drought indices has been developed, such as the Standard Precipitation Index (SPI) (McKee et al., 1993), Palmer drought severity index (PDSI) (Palmer, 1965), Rainfall Deciles (RD) (Gibbs, 1967), China-Z Index (CZI) (Wu et al., 2001) and Crop Moisture Index (CMI) (Palmer, 1968). Among them, SPI and PDSI are more popular and frequently used. As the first water-budget-based index, PDSI is considered as a landmark in the efforts dedicated to developing drought indices (Heim, 2002; Vicente-Serrano et al., 2010). It is calculated from precipitation, temperature and soil water content by considering both water supply and demand. However, it also faces many deficiencies like the fixed timescale, strong dependence on the data calibration, limitations in spatial comparability and the arbitrary interpretation of drought conditions to the index values (Andreadis et al., 2005; Sheffield et al., 2009; Wells et al., 2004). In order to overcome these shortcomings, the selfcalibrating PDSI (scPDSI) was developed by Well et al. (Wells et al., 2004) and recently improved by Liu et al. (2017). However, the problem of fixed timescale has not been completely solved in scPDSI. Considering the important drought characteristics at multiple timescales, the SPI was developed with a lot of advantages, such as simple calculation, multiple timescales, good adaptability to different climates and an easy comparison between different regions. Owing to these advantages, the SPI has been widely applied to drought monitoring studies (Guo et al., 2016; Guo et al., 2017a; Vu et al., 2015; Zhu et al., 2016). However, SPI is only based on precipitation without considering temperature and evaporation which play an increasingly important role in drought occurrence as a result of global warming (Yoon et al., 2012). The SPEI (Vicente-Serrano et al., 2010) was developed by not only considering the sensibility of droughts to temperature but also remaining the multiple timescales and allowing for flexible comparability. The calculation of SPEI is based on the water balance equation represented by the difference between precipitation (P) and potential evapotranspiration (PET).