*3.1. Drought Indices*

These six widely used drought indices (PCI, VCI, TCI, VHI, SDCI, and TVDI) can be categorized into three types. One type comprises a single-factor index, including PCI, TCI, and VCI, calculated only from one of the three—precipitation, LST, and NDVI (Table 2). The second type is an index comprising a combination of factors; examples of this type are the VHI and SDCI, which are calculated from weighted combinations of multiple single-factor indices (Table 3). The other one is TVDI, which employs the spatial relationship between LST and NDVI to reflect drought information (Equation (1)).

$$\text{TVDI} = \frac{LST - LST\_{\text{min}}}{LST\_{\text{max}} - LST\_{\text{min}}} \tag{1}$$

$$LST\_{min} = a + b \times NDVI \tag{2}$$

$$LST\_{\text{max}} = c + d \times \text{NDVI} \tag{3}$$



**\*** The minimum and maximum values in these formulas were selected for the whole 2001–2018 period over the whole study area

**Table 3.** Drought indices based on a combination of factors.


In Equation (1), *LST* is the observed surface temperature at a given pixel; *LSTmin*, the minimum surface temperature in the triangle for a given *NDVI* defining the wet edge (Equation (2)); and *LSTmax*, the maximum surface temperature in the triangle for a given *NDVI* defining the dry edge (Equation (3)) [27]. Their coefficients (a, b, c, and d) can be estimated by fitting the dry and wet edges of the triangle (Figure 2).

**Figure 2.** Simplified land surface temperature (LST)- normalized difference vegetation index (NDVI) triangle (Adopted from [27]).

MODIS and TRMM datasets were acquired and processed as the inputs to generate six drought indices. The values of each of these indices range from 0 to 1, with a higher value indicating less drought for most indices. However, for TVDI, a higher value indicates a more severe drought. Table 4 lists the classification scheme of these indices for drought levels according to [34].


**Table 4.** Classification scheme of the drought indices employed in this study [34].
