Analyzing mountain hydrology in post-wildfire landscapes

Abstract

Estimating water balance components in mountainous areas is challenging due to high spatial and temporal variability. Wildfires exacerbate this challenge and are increasing in size and severity while burning at higher elevations in the western U.S. The region is also expected to experience more extreme rainfall events due to climate change. This dissertation aims to examine rainfall spatial variability, suitability of gridded climate products in hydrologic applications, and catchment characteristics that could be influencing post-fire streamflow responses. Chapter 2 investigates the influence of elevation on precipitation patterns across the Front Range in northern Colorado and evaluates a gridded radar-based rainfall product compared to gauge data. Flash floods in the Colorado Rocky Mountains are triggered by intense rainfall during convective storms, and the likelihood and severity of flash flooding is increased further following wildfires. The most extreme floods in this region have occurred at lower to middle elevations (~1500 to 2500 m), within a landscape that extends above 4000 m, indicating that rainfall intensity may vary with elevation; however, the variability has not been quantified relative to orographic position (i.e., east or west of mountain ranges). This chapter combines Multi-Radar/Multi-Sensor System (MRMS), Parameter-elevation Regressions on Independent Slopes Model (PRISM), Gridded Surface Meteorological (gridMET), and tipping bucket precipitation datasets to evaluate how rainfall intensities vary with elevation along Colorado’s northern Front Range mountains. The MRMS, PRISM, and gridMET datasets showed greater total warm season (i.e., June-September) precipitation east of the Colorado’s northern Front Range mountains than west, but the magnitudes of rainfall differed between datasets. The MRMS dataset had greater warm season rainfall than gridMET and PRISM east of the mountains but lower amounts to the west. The relationship between elevation and rainfall intensity varied between the three warm seasons evaluated, but the highest rainfall intensities were consistently observed east of the mountains. A network of stream stage sensors installed within the Cameron Peak and East Troublesome fires was used to evaluate the ability of MRMS and tipping bucket rainfall intensities to predict stream occurrence in burned areas using a logistic regression. The MRMS rainfall intensities predicted stream occurrence better east of the mountains, while tipping buckets performed better to the west; this difference likely results from poor radar coverage observed west of the mountains. The MRMS dataset may outperform tipping buckets for streamflow predictions in areas with high radar coverage, likely due to its broader spatial coverage than most rain gauge networks. Chapter 3 evaluates how streamflow responses to rainfall vary between fires on the east and west side of the Front Range and between snow zones and year post-fire. Stream gauges installed at 26 burned and unburned catchments in the Cameron Peak and East Troublesome fires were used to quantify streamflow response as stage rises and lags to peak, and the MRMS data were used to quantify the 60-minute maximum rainfall intensities for each storm. Between the two fires, stage rises in the East Troublesome Fire streams were higher than those in Cameron Peak Fire. The differences in stage rise between fires were associated with higher clay content in soils and more valleys and hollows in the East Troublesome Fire compared to Cameron Peak. Across both fires, catchments in the intermittent snow zone (i.e., areas that do not hold a consistent winter snowpack) experienced higher rainfall intensities and have a greater proportion of their area at elevations above the 75th percentile of the catchment’s elevation range, leading to greater stage rises and shorter lags to peak than catchments in the seasonal snow zone. The timing of intermittent snow zone streamflow responses was consistent with infiltration excess overland flow, indicating that these areas are at high risk for flash flooding. The results highlight that post-fire streamflow responses have high variability even within a given region and that catchment morphology is also important to consider in post-fire studies. Chapter 4 examined how gridded and station climate data sources affect hydrologic model calibration, performance, and behavior in a Colorado Front Range catchment. A small portion of this watershed was burned a year prior to the start of the study and was calibrated independently from the unburned portion. A streamflow model was run with a combination of the PRISM and Daymet products as the gridded climate inputs and a network of sensors as the station climate inputs. The two climate inputs were evaluated prior to modeling, and differences were likely due to elevation biases from the gridded climate data algorithms. The differences in climate inputs led to distinct parameter calibrations between the two datasets, and of the calibrations, the station climate inputs yielded better hydrologic model performance for streamflow at the catchment outlet. The gridded inputs resulted in greater snow accumulation and earlier, faster snowmelt due to lower snow interception and higher snowmelt parameter values compared to the station climate inputs. The burned area had different parameter constraints, including reduced interception, reduced maximum summer infiltration, and increased maximum depression storage, to reflect changes post-fire. These different parameter constraints for the burned area led to greater snow accumulation, higher soil saturation, and increased interflow and shallow groundwater. Overall, these results demonstrate that hydrologic model parameter optimizations can change depending on the climate inputs. This highlights the importance of accurate input data and illustrates the importance of not only calibrating to streamflow but also internal state variables.