Climate change and human activities are changing the structure and function of dryland ecosystems at unprecedented rate, thus threatening the stability of ecosystems. The stability of dryland ecosystems is vital for ecological security and local livelihoods. However, the mechanisms that underlie ecosystem stability in drylands remain uncertain due to limited field data from regional studies. Combined with transect survey in the drylands of China along the aridity gradient and remote sensing data, we characterized community temporal stability and identified its driving mechanisms along the aridity gradient. The results showed the community temporal stability in drylands of China revealed a U-shaped curve with increasing aridity and its major driving mechanisms shifted at an aridity level of ~0.88. In regions where aridity is below 0.88, increasing precipitation and species richness resulted in higher community productivity and community stability. In regions where aridity is above 0.88, however, higher soil organic carbon content and species richness may lead to higher variability of community productivity and lower ecosystem stability. Overall, our findings revealed that there existed an aridity threshold leading to abrupt changes on community stability in drylands of China. Our study also suggested divergent driving mechanisms of community stability above and below the threshold, which should be considered in policy making regarding the ecosystem management of drylands.

Forests play a pivotal role in regulating climate and sustaining the hydrological cycle. The biophysical impacts of forest on clouds, however, remain unclear due to the lack of direct observations. In this first global-scale observational study, we use long-term satellite-derived cloud cover data to show that forests can have opposite effects on summer cloud cover. We find enhanced cloud cover over most temperate and boreal forests, but inhibited cloud cover over Amazon, central Africa, and Southeast US. These cloud effects mainly arise from convection processes associated with forests. The spatial variation in the sign of cloud effects is driven by sensible heating where cloud enhancement (inhibition) is more likely to occur when sensible heat in forest is larger (smaller) than nearby nonforest. Ongoing forest cover loss has led to opposite cloud cover changes, with local cloud increase over forest loss hotspots in the Amazon (+0.78%), Indonesia (+1.19%), and Southeast US (+0.09%), but cloud reduction in East Siberia (-0.20%) from 2002-2018. Our data-driven assessment informs the climate effects of local-scale forest cover change and improves mechanistic understanding of forest-cloud interactions, the latter of which remains uncertain in Earth system models.