A multi-spatial scale assessment of land-use stress on water quality in headwater streams in the Platinum Belt, South Africa

River water quality is affected by various stressors (land-uses) operating at different hydrological spatial scales. Few studies have employed a multi-scaled analyses to differentiate effects of natural grasslands and woodlands, agriculture, impoundments, urban and mining stressors on headwater streams. Using a multi-scaled modeling approach, this study disentangled the distinct spatial signatures and mechanistic effects of specific stressors and topographic drivers on individual water quality parameters in tributaries of the Gwathle River Catchment in the Platinum Belt of South Africa. Water samples were collected on six occasions from 15 sites on three rivers over 12-months. Physio-chemical parameters as well as major anions, cations and metals were measured. Five key water quality parameters were identified using principal components analysis: sulfate, ammonium, copper, turbidity, and pH to characterise catchment water quality conditions. Using class-level composition (PLAND) and connectedness (COHESION) metrics together with topographic data, generalized linear mixed models were developed at multiple scales (sub-basin, cumulative catchment, riparian buffers) to identify the most parsimonious model with the dominant drivers of each water quality parameter. Ammonium concentrations were best explained by urban stress, Cu increased with mining and agriculture, turbidity increased with elevation heterogeneity, agriculture, urbanisation and fallow lands all at the sub-basin scale. River pH was positively predicted by slope heterogeneity, mining cover and impoundment connectivity at the catchment scale. Sulfate increased with mining and agriculture composition in the 100 m riparian buffer. Hierarchical cluster analysis of water quality and scale-dependent parsimonious drivers separated the river sites into three distinct groups distinguishing pristine, moderately impacted, and heavily mined sites. By demonstrating stressor- and scale-dependent water quality responses, this multi-scale nested modeling approach reveals the importance of developing adaptive, targeted management plans at hydrologically meaningful scales to sustain water quality amid intensifying land use.