*2.2. Data*

NDVI has been widely applied to detect and quantify the dynamic changes of vegetation in an extensive range [54]. Currently, various remote sensing satellite instruments can provide NDVI data, such as MODIS, SPOT/VEGETATION, and NOAA/AVHRR, etc. Compared with other vegetation index dataset, the GIMMS-NDVI3g, featured by its long time series and wide coverage, has proven to be one of the best datasets in describing vegetation growth dynamic changes [55]. Previous studies have shown that GIMMS NDVI dataset is significantly better than that of MODIS NDVI in reflecting dynamic changes over the Qinghai-Tibet Plateau [56]. In this study, the NOAA/AVHRR GIMMS production with a spatial resolution of 8 km × 8 km was used to calculate the NDVI. The data from January 1982 to December 2015 was derived from the third generation GIMMS NDV3g dataset, developed by the Goddard Aerospace Agency (http://ecocast.arc.nasa.gov/data/pub/gimms/3g/). Meanwhile, to further minimize the impact of clouds, atmosphere, and solar radiation angles on the NDVI values, the GIMMS NDVI3g data was preprocessed by employing S-G Filtering and Maximum Value Composite techniques to ensure the reliability of the research data and the accuracy of the results.

The Global Land Data Assimilation System (GLDAS), consisting of four different land surface models, i.e., CLM, NOAH, MOSAIC, and VIC [57], is a high-resolution land surface data assimilation system that is jointly managed by the American Goddard Space Flight Center and Environmental Forecast Center (http://ldas.gsfc.nasa.gov/gldas/GLDASvegetation.php), with two spatial resolutions

(0.25◦ × 0.25◦ and 0.5◦ × 0.5◦) and two temporal resolutions (3 hours and 1 month). The dataset with extensive sources is a combination of the surface observed data and the remote sensing satellite data. Compared to other remote sensing datasets, the GLDAS-NOAH data has a higher spatial and temporal resolution, a longer time span (1970 to present), and 28 variables (precipitation, air temperature, and soil moisture content, etc.). In this study, due to the limited number of meteorological gauging stations in the YZR basin, especially in the upper reaches, the monthly GLDAS-NOAH data at the 0.25◦ × 0.25◦ spatial resolution from 1982 to 2015 were used to analyze the dry-wet transitions of the YZR basin, and to calculate the SPEI based on the performance evaluation of the GLDAS-NOAH data.

In-situ observations of the precipitation and surface air temperature from twenty meteorological gauging stations (as shown in Figure 1) in the YZR basin were used to evaluate the performance of the GLDAS-NOAH data.
