摘要：Forest disturbance induced changes in the coupling of forest carbon and water have important implications for ecosystem functioning and sustainable forest management. However, this is rarely investigated at the large watershed scale with cumulative forest disturbance. We used a combination of techniques including modeling, statistical analysis, and machine learning to investigate the effects of cumulative forest disturbance on water use efficiency (WUE, a proxy for carbon and water coupling) in the 19,200 km2 Chilcotin watershed situated in the central interior of British Columbia, Canada. Harvesting, wildfire, and a severe Mountain Pine Beetle (MPB) infestation have gradually cumulated over the 45-year study period, and the watershed reached a cumulative equivalent clear-cut area of 10% in 1999 and then 40% in 2016. Surprisingly, with the dramatic forest disturbance increase from 2000 to 2016 which was mainly due to MPB, watershed-level carbon stocks and sequestration showed an insignificant reduction. This resilience was mainly due to landscape-level carbon dynamics that saw a balance between a variety of disturbance rates and types, an accumulation of older stand types, and fast growing young regenerated forests. Watershed-level carbon sequestration capacity was sustained, measured by Net Primary Production (NPP). A concurrent significant decrease in annual evapotranspiration (ET), led to a 19% increase in WUE (defined as the ratio of NPP to ET), which is contrary to common findings after disturbance at the forest stand-level. During this period of high disturbance, ET was the dominant driver of the WUE increase. We conclude that disturbance-driven forest dynamics and the appropriate scale must be considered when investigating carbon and water relationship. In contrast to the stand-level trade-off relationship between carbon and water, forested watersheds may be managed to maintain timber, carbon and water resources across large landscapes.
关键词：Forest carbon ; Hydrology ; Cumulative disturbance ; Water use efficiency ; Forest carbon and water coupling ; Large watershed ; Evapotranspiration ; Machine learning
