Meteorological influences on process-based spatial-temporal pattern of throughfall of a xerophytic shrub in arid lands of northern China

Highlights: • Throughfall were monitored at 10-min intervals at different radial directions beneath shrub canopies. • Temporal heterogeneity of rainfall clearly affected the timing of throughfall inputs. • Throughfall differed markedly among different radial directions beneath shrub canopies. • Principal Components Analysis was performed on meteorological variables. • Three principal components were introduced into a multiple regression model to predict throughfall. Numerous field experiments had demonstrated great spatial variability and temporal stability of throughfall under tree canopies within forested ecosystems. Nonetheless, no known studies have investigated the intrastorm spatial-temporal variability of throughfall beneath xerophytic shrub canopies within arid desert ecosystems where water is typically the principal limiting factor determining the structure and dynamics of ecosystems. Here we investigated the spatial-temporal pattern of throughfall at intrastorm scale, and systematically examined the effects of meteorological variables on throughfall based on the principal components analysis (PCA) and a multiple regression model. Throughfall was monitored at 10-min intervals by placing tipping-bucket rain gauges at different radial directions beneath 3 shrubs of Caragana korshinskii during the growing season of 2016 within a water-limited arid desert ecosystem of northern China. We found the temporal heterogeneity of rainfall clearly affected the timing of throughfall beneath shrub canopies within discrete rainfall events. Throughfall also differed markedly among different radial directions beneath shrub canopies, which was found to be well associated with wind directions during rainfall events. PCA on meteorological variables indicated that three principal components accounted for 84.2% of the total variance, and we found that the second principal component (loaded strongly on rainfall amount and maximum 10-min rainfall intensity) was the dominant component controlling throughfall and its spatial variability after introducing three principal components into a multiple linear regression model. Our findings highlight the spatial-temporal variability of throughfall at the intrastorm scale, and are expected to be helpful for an improved process-based characterization and modelling of throughfall in vast arid desert ecosystems.