Non-stationary streamflow responses to forest harvesting vary along topographic gradients in a sub-alpine watershed of Southwest China

Abstract

The relationship between forest and streamflow is becoming non-stationary under climate change and growing disturbances especially in high mountainous region. However, the non-stationary forest-streamflow relationship and its spatial variation remains unclear particularly due to the lack of ability to capture both non-stationarity and spatial variations in forest-water interactions by available hydrological models. In this study, we proposed a new methodology combining the modified double mass curve (MDMC) and Soil Water Assessment Tool (SWAT) models to investigate the non-stationary streamflow responses to forest harvesting and their variations along topographic gradients in the Upper Zagunao River watershed in Southwest China, a typical sub-alpine watershed. We firstly determined the non-stationarity of the forest-streamflow relationship by the MDMC and classified the study period into the stationary (1961–1969) and non-stationary (1970–1991 and 1992–2006) periods. We further ran SWAT models with dynamic parameter sets in stationary and non-stationary periods, and found that simulations with dynamic parameter sets effectively captured ecohydrological processes during non-stationary periods and outperformed simulations with fixed parameter sets. Based on the simulations, we quantified the responses of annual and seasonal (dry and wet season) mean flows, and high and low flows to various forest change scenarios that combined forest harvesting levels and topographic gradients. We found that the annual and seasonal mean flows, low and high flows increased after forest harvesting. Lower forest harvesting thresholds for detectable changes in the magnitudes of dry season mean flows (10 % of forest harvesting) and low flows (6 % of forest harvesting) were determined in comparison to the magnitudes of wet season mean flows (25 % of forest harvesting) and high flows (12 % of forest harvesting). Spatially, streamflow was more sensitive to forest harvesting at lower elevations (2800–3100 m) or on steeper slopes (30–45°), and less sensitive on south-facing slopes. This study provides an innovative assessment of non-stationary streamflow responses to forest harvesting across topographic gradients, and offers valuable insights for forest restoration and water resource management in sub-alpine watersheds under a changing environment.