Alterations of throughfall and stemflow chemistry by vegetation in China: Insights from a machine learning based meta-analysis

Abstract

In vegetated ecosystems, the chemical composition of rainwater is strongly altered after gross rainfall (GR) partitioning into throughfall (TF) and stemflow (SF) before entering the under-canopy soil. However, quantitative syntheses of the chemical alteration are lacking, rendering an insufficient comprehension of its overarching patterns and magnitudes. Here, we furnish a comprehensive dataset related to the chemistry of rainfall partitioning by vegetation in China compiled from 131 peer-reviewed papers published between 1988 and 2023. A meta-analysis was conducted to assess the chemical alterations as GR partitions into TF and SF. We discerned a marked acidification effect in SF but not in TF, with the pH being significantly lower in SF than in GR. Additionally, we observed a persistent concentration-based chemical enrichment of total nitrogen (TN) and total phosphorous (TP) as well as 12 commonly reported nutrient-ions (except for Cu2+) in both TF and SF, with SF generally exhibiting a greater degree of chemical enrichment than TF. Using a machine learning method (boosted regression trees), we identified the key drivers from twelve biotic and abiotic predictor variables on the effect sizes of pH, TN, TP, and each nutrient-ion, and further elucidated the prevalent non-linear partial effects of these key drivers on the magnitude of chemical alterations. For instance, three predictor variables were identified as key drivers for the effect sizes of TN for SF, with a descending relative influence of diameter at breast height (56.9 %; positive effect), Human Footprint (16.5 %; positive), and bark texture (14.0 %; stands with mixed bark textures exerting a greater impact). Our study has significant implications for an accurate estimation of nutrient budgets in vegetated ecosystems throughout China, contributes to a deeper understanding of the intricate ways in which biotic and abiotic predictors influence chemical alterations in rainfall partitioning, and aids in evaluating forest health and developing sensible forest management strategies.