According to available land cover data, forests, grasslands, irrigated and rain-fed crops are the dominant vegetation cover types in the drylands (6.14%, 40.73%, 4.81%, and 9.19% of the total area, respectively) (Fig.S1). So, we believe that they can be used as typical vegetation for subsequent studies.
After calculating the water resources and fitting the relationship between ET and GPP, we tried to calculate the vegetation restoration potential with AWR constraints. Positive potential (values above 0) indicates that water resources within the area can still support an increase in that vegetation cover. Negative potential (values below 0) indicates that vegetation growth exceeded the area's water vegetation carrying capacity. As shown in Fig.\ref{416579}, additional GPP that could be restored under current water constraints ranges from 8% to 12% depending on vegetation types (10.23% for forests, 11.65% for grasslands, 7.67% for irrigated crops, and 9.49% for rain-fed crops) from 2003 to 2018. Overloaded vegetation occurs more frequently in this region's western and north-central parts, mainly due to the water deficit. Among them, overloading was most severe in northwestern grasslands, presumably because these overloaded areas are in the hyperarid zones. The higher revegetation potential is mainly found in grasslands and forests, mostly in the central and southern places (Fig.\ref{416579}a,\ref{416579}b), which is more related to water surpluses. 
The plots in Fig.\ref{416579} panels (a)-(d) showed the variation of vegetation restoration potential with drought gradients. It can be found that all four vegetation types were overloaded in hyperarid and arid regions, and the restoration potential increased with increasing moisture for all of them. Although the four vegetation types followed the same trend with the drought gradient, they were not overloaded by the same magnitude. Forests and rain-fed crops were overloaded to a greater extent in hyperarid zones, suggesting that these types should be avoided in subsequent local vegetation restoration decisions.

3.2 Conversion potential among different vegetation