Simulation of potential vegetation distribution in arid areas at regional scale

Awareness of potential natural vegetation (PNV) is imperative in ecological restoration. Extensive global and national PNV simulation with coarse spatial resolution is increasingly seen to mitigate land degradation under climate change. However, precise PNV modeling in drylands at a small regional scale with high enough spatial resolution usually receives little attention. This study developed a precise regional multiple PNV framework in drylands based on climatic variables, soil properties, groundwater depth, and topographical variables through a maximum entropy algorithm (MaxEnt). The effects of three different input variable combinations and five background point numbers on PNV prediction were compared. Applying the model to the lower Tarim River Basin, Northwest China, the results showed that the model developed incorporating groundwater depth exhibited better performance in regional PNV simulation, and the AUC of the tree, grass, and shrub was improved by 0.21, 0.18 and 0.09, respectively. Furthermore, the results showed that the probability of growing natural vegetation decreased with the increase of distance from the river within 100 m; and the probability of occurring tree, grass, and shrub was higher when the groundwater depth was close to -3 m, -2 m, and -7 m, respectively. Moreover, the suitable habitat of the tree, grass, and shrub was mainly concentrated in the riparian zone, accounting for 34.5%, 21.9%, and 72.4% of the study area, respectively. Comparisons of reconstructed PNV and existing land use patterns revealed that the lower Tarim River still has enormous land potential for ecological restoration. The model developed here can assist in precise ecosystem restoration in drylands.

Automated summary

Awareness of potential natural vegetation (PNV) is crucial in ecological restoration. This study developed a precise regional multiple PNV framework in drylands based on climatic variables, soil properties, groundwater depth, and topographical variables through a maximum entropy algorithm (MaxEnt). The results showed that incorporating groundwater depth improved the performance of regional PNV simulation, and the lower Tarim River Basin has significant potential for ecological restoration.