**5. Discussion**

Spatiotemporal analyses of *DI* values showed that approximately 80% of the GCM experienced stable drought conditions without significant changes over the last 18 years. In general, precipitation-based indices (PCI and SDCI) showed similar spatiotemporal patterns, while two vegetation-based indices (VCI and VHI) exhibited similar patterns. The other two indices based on LST revealed vague changing patterns of drought, with pixels indicating significant changes distributed sparsely across the whole study area. This may be because the LST is extremely sensitive to location and land cover type and varies with time. By comparing the annual pattern of change of drought indices with the meteorological factors such as precipitation and temperature, we found that the LST-based indices (TCI and TVDI) were not closely related to either annual average precipitation or temperature. Analysis of the relationship of the land cover type with drought indices as well as their slope showed that grasslands are more easily affected by droughts than are deciduous broadleaf forests. The changing rates of the drought situation of three forests (i.e., evergreen needleleaf forests, deciduous broadleaf forests, and deciduous needleleaf forests) are clearly greater than those of the other land cover types.

As the GCM is a mountainous area, the terrain too may affect the drought situation. We extracted landforms of the GCM based on the terrain information extracted from Shuttle Radar Topography Mission digital elevation model according to [41]. Eight geomorphic units with different combinations of altitude and relief were determined. Relief is the difference between the maximum and minimum altitude in a landform unit. It is usually calculated in square or circular sampling units. These units include the low-altitude plain (LAP), low-altitude hill (LAH), low-relief and low-altitude mountain (LR-LAM), middle-relief and low-altitude mountain (MR-LAM), middle-altitude plain, middle-altitude hill (MAH), low-relief and middle-altitude mountain (LR-MAM), and middle-relief and middle-altitude mountain (MR-MAM) (Figure 8a). Zonal statistics of the six *DI* maps of 2018, shown in Figure 8, were obtained based on the geomorphic unit map in order to calculate the mean *DI* values in different units. The relative drought situation was consistent regardless of the index used. Overall, the MR-MAM and MAH have smaller TVDI values and higher values for the other five indices, indicating a mild drought condition; this may be a result of the higher altitude of the MR-MAM and MAH along with the lower temperature and the generally greater precipitation and better forest coverage in these regions. Drier areas generally correspond to the geomorphic units of LAP and LAH, which have low altitude and poorer vegetation coverage compared to high-altitude areas.

**Figure 8.** (**a**) Landforms of the GCM; (**b**–**g**) Mean values of drought indices in different geomorphic units (**b**: PCI; **c**: TCI; **d**: VCI; **e**: VHI; **f**: SDCI; and **g**: TVDI).
