Isotopic evaluation of the National Water Model reveals human water use influences on streamflow across the western United States
On this page:
The National Water Model (NWM) provides critical analyses and projections of streamflow that support water management decisions. However, the NWM performs poorly in higher-order rivers of the Western US. The accuracy of NWM depends on 1) the fidelity of the model inputs and 2) representation of modeled processes, including evapotranspiration, partitioning of overland and subsurface flow, and human water use and management practices. To diagnose water cycle process errors in climate and hydrologic models, prior studies have leveraged stable isotopes of water (δ18O and δ2H). We performed an isotope mass balance using mean summer (JJA) NWM hydrologic fluxes between 2000 and 2019, with gridded precipitation and groundwater isotope ratio inputs. We compared the NWM-flux-estimated isotope ratios to 4503 in-stream observations from EPA, USGS, and project-specific collections in 877 catchments across 5 basins in the Western US. A simple regression between observed and model-predicted isotope ratios explained 57.9% (δ18O) and 67.1% (δ2H) of variance, though observations were 0.5‰ (δ18O) and 4.8‰ (δ2H) higher, on average, than model predictions. The observation-model bias and the unexplained variance suggest that not all water sources are correctly characterized and/or the model is missing processes that cause isotope fractionation (e.g., evaporation). To understand the possible sources of model error, we evaluated patterns in observation-model differences (δ18Odiff) including their d-excess (ddiff = (δ2Hobs-δ2Hmod) – 8*( δ18Oobs-δ18Omod)). We detected evapoconcentration of observations relative to model estimates (negative ddiff and positive δ18Odiff) at lower elevation sites with greater agricultural water use and reservoir influences. Our results suggests that the isotope mass balance approach can identify missing processes affecting streamflow in higher-order streams in the Western US and may support quantitative estimates of agricultural return flows to these waterways.