Karst regi
ons in southwest China are characterized by vulnerable ecological envir
onment. Knowledge
on the driving factors of vegetati
on cover change could provide valuable informati
on for ecological restorati
on management. However, quantitative identifificati
on of the key drivers for the vegetati
on restorati
on remains challenging in karst trough valleys. In this study, we used normalized difffference vegetati
on index (NDVI) time series (2000–2016), Theil-Sen median analysis, Mann-Kendall trend test, and Hurst exp
onent to analyze the vegetati
on cover trends in a karst trough valley. The performance of multiple
linear regressi
on (MLR), generalized additive models (GAM), support vector machine (SVM), and random forest (RF) in accounting for vegetati
on cover change were compared. The results showed that vegetati
on cover trends for increasing, stable and decreasing accounted for 71.44%, 28.16% and 0.40% of the study area, respectively. Lithology had a signifificant effffect
on spatial patterns of temporal change and future sustainability in NDVI (p < .01). RF performed much better than MLR, GAM and SVM in accounting for vegetati
on cover change. The RF model had much lower fifitting error indices (MAE = 1.46*10−3 , RMSE = 1.92*10−3 ) and higher R2 (0.65) than MLR, GAM and SVM models. Thus, RF model was applied to identify impacts of driving factors
on vegetati
on cover change quantitatively. Precipitati
on change, lithology and elevati
on were key factors for vegetati
on cover change. The vegetati
on restorati
on and rec
onstructi
on projects should pay more attenti
on to the regi
on where limest
one and above- 900 m elevati
on dominate, due to relatively slow vegetati
on growth in these regi
ons. The new understandings obtained in this study enrich our knowledge of the effffects of lithology and topography
on the vegetati
on cover
change and are necessary to guide sustainable projects of ecological recovery in karst trough valleys.