Rainfall partitioning by vegetation in China: A quantitative synthesis

Rainfall partitioning into stemflow, throughfall, and interception loss by vegetation alters hydrological and biogeochemical fluxes between vegetation and soil. Here, we compiled a comprehensive dataset of rainfall partitioning by forests, shrublands, croplands, and grasslands in China from 287 peer reviewed papers (71 in English and 216 in Chinese). Based on this dataset, we summarized the best-fit functions reported for rainfall partitioning (in both mm and %) as a function of rainfall amount, as well as the rainfall thresholds for throughfall and stemflow initiation. We explored the pattern of the proportions of stemflow, throughfall, and interception loss of vegetation in China, and performed boosted regression trees (BRT) analysis to model the relative effects of cross-site biotic and abiotic predictors on each of the rainfall partitioning fluxes (%). Our results identified the scarcity of rainfall partitioning data, particularly for grasslands. A substantial variability of each rainfall parti-tioning flux (mm) could be explained solely by rainfall amount, with median R2 values of 0.91, 0.99, and 0.82 for stemflow, throughfall, and interception loss, respectively, and with linear functions most often reported as the best-fit functions. Significant differences (p < 0.0001) were detected in rainfall thresholds for initiating stemflow (median: 3.3 mm; interquartile range, IQR: 1.8-5.4 mm) and throughfall (median: 1.2 mm; IQR: 0.8-2.2 mm). Stemflow (%) had a median (IQR) of 2.7 % (1.2-6.0 %), and that value was 74.3 % (66.7-80.3 %) for throughfall (%) and 21.6 % (16.3-28.5 %) for interception loss (%), respectively. Significant differences were detected in the proportion of stemflow (p < 0.001) and throughfall (p < 0.01) between forests and shrublands, respectively; whereas no significant differences in the proportion of interception loss were found among vegetation types. BRT analysis indicated that of the eleven biotic and abiotic predictors examined, six were classified as significant predictors in determining stemflow (%) and interception loss (%), respectively, whereas throughfall (%) had four significant predictors. Non-linear partial effects of predictors on rainfall partitioning fluxes were prevalent. This study avails a global readership to the findings of a large cache of Chinese studies that have been inaccessible hitherto, providing a mechanistic understanding of the effects of cross-site biotic and abiotic predictors on rainfall partitioning fluxes.