Intensive land restoration profoundly alters the spatial and seasonal patterns of deep soil water storage at watershed scales

Abstract

Soil-water storage (SWS) highly correlates with changes to land use and land restoration, and is an important variable for managing water resources and environmental restoration in water-limited terrestrial ecosystems. The Gully Land Consolidation project (GLC) was implemented in 2011 on the Chinese Loess Plateau and it may influence the spatial distribution and temporal variability of SWS at the watershed scale. However, there have been limited studies assessing the magnitude and pattern of such changes. We measured SWS to a depth of 5 m nine times over four seasons across two watersheds－one treated by the GLC project (GT-T, n = 89) and the other untreated (GT-U, n = 72). We then used correlation analysis, principle component analysis, multiple linear regression analysis, generalized linear model, and regression kriging to evaluate the seasonal variability, as well as spatial patterns, of SWS. Mean SWS in the spring, summer, autumn, and winter ranged from 830 to 934 mm in the GT-T watershed and from 681 to 771 mm in the GT-U watershed. Land use was the most important factor for predicting SWS in the two watersheds. After land management as a result of the GLC project, SWS exhibited different spatial patterns in the GT-T watershed compared with the GT-U watershed. Dynamics of SWS between adjacent seasons in the GT-U watershed displayed more variability than that in the GT-T watershed. In both watersheds, soil water volume (SWV, i.e., SWS per unit area × watershed area) declined from spring to summer (-3.9% in the GT-T and -13.4% in the GT-U), and increased from summer to autumn (11.7% in the GT-T and 20.7% in the GT-U). After the whole year, SWV displayed different degrees of increase in the two watersheds (10.6% in the GT-T, 2.7% in the GT-U) due to the effect of land restoration. The ratios of SWV in the gully land filled in the GT-T watershed to the SWV of the whole GT-T watershed for the four seasons ranged from 21.4–23.3%, with a mean value of 22%. Our results provide new insights regarding how human engineering of landscapes (as exemplified by GLC project) affect the spatial distribution of SWS and the dynamics of SWV storing capacity at the watershed scale. Understanding the dynamics of SWS/SWV between seasons is essential to evaluate, model and manage the soil water resources in human-affected watersheds.