Landform features and seasonal precipitation predict shallow groundwater influence on temperature in headwater streams

Abstract

Headwater stream responses to climate change will depend in part on groundwater-surface water exchanges. We used linear modeling techniques to partition likely effects of shallow groundwater seepage and air temperature on stream temperatures for 79 sites in nine focal watersheds using hourly air and water temperature measurements collected during summer months from 2012 to 2015 in Shenandoah National Park, Virginia, USA. Shallow groundwater effects exhibited more variation within watersheds than between them, indicating the importance of reach-scale assessments and the limited capacity to extrapolate upstream groundwater influences from downstream measurements. Boosted regression tree (BRT) models revealed intricate interactions among geomorphological landform features (stream slope, elevation, network length, contributing area, and channel confinement) and seasonal precipitation patterns (winter, spring, and summer months) that together were robust predictors of spatial and temporal variation in groundwater influence on stream temperatures. The final BRT model performed well for training data and cross-validated samples (correlation = 0.984 and 0.760, respectively). Geomorphological and precipitation predictors of groundwater influence varied in their importance between watersheds, suggesting differences in spatial and temporal controls of recharge dynamics and the depth of the groundwater source. We demonstrate an application of the final BRT model to predict groundwater effects from landform and precipitation covariates at 1075 new sites distributed at 100 m increments within focal watersheds. Our study provides a framework to estimate effects of groundwater seepage on stream temperature in unsampled locations. We discuss applications for climate change research to account for groundwater-surface water interactions when projecting future thermal thresholds for stream biota.