3.1.2. MODIS

MODIS provides researchers with stable, long time series global remote sensing data. Some global land use/land cover (LULC) datasets based on MODIS data have been generated [41,42]. MOD09GA and MOD16A2 provide daily surface reflectance with a spatial resolution of 500 m and evapotranspiration of 500 m every 8 days [43,44]. MOD09GA and MOD16A2 were used in this study for irrigated area downscaling, while the latter is based on 8-day synthetic data and does not require further processing. MOD09GA is daily surface reflectance data, and cloud pollution has a large impact on the use of data. First, the NDVI is calculated based on the MOD09GA dataset.

$$NDVI = \frac{NIR - RED}{NIR + RED} \tag{3}$$

where *NIR* and *RED* are the surface reflectance factors for the presented wavelengths. Second, the 8-day maximum value composite (MVC) method was used for the daily NDVI time series dataset, and the method is to composite a new NDVI image by using the daily maximum value of NDVI within 8 days of each pixel in the image as a valid pixel value [45]. This processing method reduces the impact of clouds on the dataset and keeps the time resolution of the two MODIS products consistent. The batch redefinition projection and raster attribute conversion of MODIS data are based on MRT (MODIS Reprojection Tool supported by NASA, referenced by Dwyer et al. [46]), and the maximum synthesis of the NDVI is based on MATLAB 2018b.
