Advancing the capability of hydrologic models in poorly monitored domains and disturbed catchments

Abstract

In recent decades, the western U.S. has experienced increasing magnitudes and frequencies of natural land cover disturbances that impact water budget partitioning. The natural land cover disturbances considered in this work include wildfire, insect induced forest mortality and drought. Watersheds in cold regions, such as those in the high elevation Rocky Mountains or high latitudes of Alaska are also vulnerable to warming temperatures because small changes in temperature can shift the phase of water from frozen to liquid, which has implications for water storage, as well as the timing and magnitude of annual hydrologic fluxes. In cold regions, water supply and infrastructure planning efforts are developed around seasonal freeze/melt cycles. These efforts depend on reliable hydrologic predictions that are becoming progressively difficult to achieve considering recent land cover and climate disturbances. Two major issues that water managers and hydrologists face today are 1) constraining water budgets in ungauged regions, and 2) understanding post-disturbance hydrologic response, especially when multiple disturbances overlap in space and/or time. The purpose of this dissertation is to explores solutions for quantifying cold region water budgets in both ungauged areas and disturbed catchments. These challenges are addressed first developing a method to constrain a parsimonious water balance model in a severely ungauged cold region (Alaska), then by evaluating the hydrologic response to overlapping wildfire and insect induced forest mortality in several catchments the Rio Grande Headwaters (RGH). This work also addresses these problems by adding quantitative vegetation representation to a water budget model, to better capture forest disturbances. This modified model is used to produce synthetic future streamflow scenarios under single and overlapping disturbances in the RGH.

The modeling framework developed for the ungauged Alaskan domain relies on gridded evapotranspiration and fractional snow-covered area products to constrain the model. Using this framework, the model performs best (NSE typically > 0.75) in the moderate-to-dry Alaskan interior – in the absence of glaciation and where permafrost is generally not continuous. Modeled runoff for 1992-2017 decades shows significantly (p < 0.05) increased runoff trends at lower latitudes (often + 0-10%) and slightly decreased (often -1-10%) runoff trends at higher latitudes. These findings point to a major need for better hydrologic characterization in Alaska, and reveal that reliance on non-runoff fluxes for model calibration is viable in this cold region. Catchments in the RGH that were evaluated for post-disturbance hydrologic response do not show significant changes in post-fire surface water budget partitioning, which is likely the result of decreased evapotranspiration due to insect-induced vegetation change in all study catchments prior to the fire. These findings highlight the importance of considering overlapping disturbances, especially in paired catchment studies, and in regions like the Rio Grande headwaters that are increasingly prone to drought, insect mortality and wildfire. Relative to a baseline scenario, synthetic future streamflow scenarios for the RGH predict 1) decreases in annual water yield under a hot-and-dry climate scenario (-14.4% by 2041-2050), except during the rising limb of annual snowmelt, when warm temperatures drive more rapid runoff, 2) increases in annual water yield and peak runoff under a fire-simulation (+32.0% by 2041-2050), 3) earlier and higher annual peak runoff under combined (fire + hot-and-dry) conditions. These findings demonstrate the strengths of hydrologic models in separating overlapping disturbance signals at the stream outlet, as well as the need for better quantitative vegetation representation within models to adequately represent dynamic disturbance conditions. Especially as climate continues to rise, and natural land cover disturbances (e.g. wildfire, insect mortality, forest disease, etc.) show no signs of slowing, the ability of hydrologic models to separate and reproduce disturbance signals is more critical than ever for water supply planning.