The interception of rainfall by vegetation canopies plays an important role in the hydrologic process of ecosystems. Most estimates of canopy rainfall interception in present studies are mainly through field observations at the plot region. However, it is difficult, yet important, to map the regional rainfall interception by vegetation canopy at a larger scale, especially in the southern rainy areas of China. To obtain a better understanding of the spatiotemporal variation of vegetation canopy rainfall interception with regard to the basin scale in this region, we extended a rainfall interception model by combining the observed rainfall data and moderate resolution imaging spectroradiometer leaf area index (MODIS_LAI) data to quantitatively estimate the vegetation canopy rainfall interception rate (CRIR) at small/medium basin scales in Guangdong Province, which is undergoing large changes in vegetation cover due to rapid urban expansion in the area. The results showed that the CRIR in Guangdong declined continuously during 2004-2012, but increased slightly in 2016, and the spatial variability of CRIR showed a diminishing yearly trend. The CRIR also exhibited a distinctive spatial pattern, with a higher rate to the east and west of the mountainous areas and a lower rate in the central mountainous and coastal areas. This pattern was more closely related to the spatial variation of the LAI than that of rainfall due to frequent extreme rainfall events saturating vegetation leaves. Further analysis demonstrated that forest coverage, instead of background climate, has a certain impact on the canopy rainfall interception, especially the proportion of broad-leaved forests in the basin, but more in-depth study is warranted in the future. In conclusion, the results of this study provide insights into the spatiotemporal variation of canopy rainfall interception at the basin scale of the Guangdong Province, and suggest that forest cover should be increased by adjusting the species composition to increase the proportion of native broad-leaved species based on the local condition within the basin. In addition, these results would be helpful in accurately assessing the impacts of forest ecosystems on regional water cycling, and provide scientific and practical implications for water resources management.

Droughts cause extreme anomalies in tropical forest growth, but the direction and magnitude of tropical forests in response to droughts are still widely debated. Here, we used four satellite-based canopy growth proxies (CGPs), including three optical and one passive microwave, and in situ fluxes observations from eddy covariance (EC) measurements for a retrospective investigation of the impacts of historical droughts on tropical forest growth from a statistical point of view. Results indicate two opposite directions in drought-related canopy dynamics across pantropical forests. The canopy of tropical forests with higher CGPs is more vulnerable to drought stress and recovers faster in the post-drought recovery period. In contrast, the canopy of tropical forests with lower CGPs increases during the drought period and declines in the subsequent recovery period, which is beyond general expectation. In situ measurements from eddy-covariance flux towers showed that forests with higher gross primary production and latent heat flux decreased photosynthesis and evapotranspiration during the drought period but increased photosynthesis and evapotranspiration faster during the post-drought recovery period, supporting the findings from satellite observations. Our statistical analysis against climatic factors predicts that higher-CGPs tress with probably taller and bigger canopies are more responsive to shortage of water availability caused by drought; while lower-CGPs tress with shorter and smaller canopies are more responsive to sunlight availability and tend to increase their canopy leaves and enhance photosynthesis in sunnier days during the drought period. Our results highlight the differences in tropical forests in responding to drought stress, which are worth incorporated in Earth system models for time-series evaluations.