Supra-permafrost aquifers within the active layer are present in the Arctic during summer. Permafrost thawing due to Arctic warming can liberate previously frozen particulate organic matter (POM) in soils to leach into groundwater as dissolved organic carbon (DOC). DOC transport from groundwater to surface water is poorly understood because of the unquantified variability in subsurface properties and hydrological environments. These dynamics must be better characterized because DOC transport to surface waters is critical to predict the long-term fate of recently thawed carbon in permafrost environments. Here, we used distributed Darcy’s Law calculations to quantify groundwater and DOC fluxes into Imnavait Creek, Alaska, a representative headwater stream in a continuous permafrost watershed. We developed a statistical ensemble approach to model the parameter variability and range of potential contributions of steady-state groundwater flow to the creek. We quantified the model prediction uncertainty using statistical sampling of in-situ, active-layer soil hydro-stratigraphy (water table, ice table, and soil stratigraphy), high-resolution topography data, and DOC data. Moreover, the predicted groundwater discharge values representing all possible hydrologic conditions towards the end of the thawing season were also considered given the potential variability in saturation. The model predictions were similar to and span most of the observed range of Imnavait Creek streamflow, especially during recession periods, and also during saturation excess overland flow. As the Arctic warms and supra-permafrost aquifers deepen, groundwater flow is expected to increase. This increase is expected to impact stream, river, and lake biogeochemical processes by dissolving and mobilizing more soil constituents in continuous permafrost regions. This study highlights how quantifying the uncertainty of hydro-stratigraphical input parameters helps understand and predict supra-permafrost aquifer dynamics and connectivity to aquatic systems using a simple, but scalable, modeling approach.
Read full abstract