*2.2. Data*

Satellite remote sensing data from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Tropical Rainfall Measurement Missions (TRMM) multi-satellite precipitation analyses (TMPA) were used as the inputs to generate di fferent drought index images. Four products including the NDVI, LST, Land cover and precipitation were employed as listed in Table 1. All data have a monthly or yearly temporal resolution. The monthly data were then aggregated to yearly data by averaging. For Land cover product (MCD12Q1), Land Cover Type 3 layer with annual Leaf Area Index classification scheme, which has 10 land cover classes, was used here. All these products have enough time scale to cover the whole of 2001–2018.


**Table 1.** Satellite remote sensing products used in this study

The annual average temperature and total precipitation data for 14 prefecture-level cities in the GCM area from 2001 to 2017 (data of 2018 were missing) were collected from local statistical yearbooks. They were used to explore the correlation between remotely sensed drought indices and meteorological observations.
