Capturing the variability of rainfall intensity and its impacts on mountain hydrology

Abstract

Storms in mountainous regions can develop quickly and cause significant flooding. Understanding how brief high intensity precipitation is influenced by terrain is important for evaluating the risk of flash floods and other natural hazards. The lack of precipitation gages in mountainous remote areas inhibits detailed monitoring of these hazardous events. As wildfires become more common in areas of complex terrain, increasing the likelihood of flash floods and debris flows, it is more important than ever to understand the potential for extreme precipitation. Precipitation estimates from remotely sensed data, such as radar and satellite, have improved in recent decades. Convection-permitting models also produce accurate precipitation estimates, that have outperformed interpolated gage datasets. In this dissertation, I examine the performance of these datasets at scales relevant to hazardous events in Colorado. Additionally, I investigate how considering the variability of natural rainfall observed in the mountains of Colorado influences runoff and erosion processes. Radar estimates of precipitation can fill the gaps in areas where gages are sparse, but the signal can be blocked by mountains, depending on where the storm is relative to the radar site. Because the error of radar estimates of precipitation can change based on where the storm is located in relation to the surrounding terrain and location of the radar, the reliability of these precipitation estimates is variable, adding to the difficulty of monitoring storms in mountains. To address this uncertainty, I developed a novel method of identifying where and when the radar estimates of precipitation are reliable, based on attributes of the region, rainfall, and storm events. The model can assist in deciding when to trust radar estimates of precipitation and in determining where more gages or radar sites are necessary. The error model uses the Multi-Radar Multi-Sensor (MRMS) product which incorporates radar, quantitative precipitation forecasts, and gage data at a high spatiotemporal resolution for the United States and southern Canada. For several time series samples of MRMS 15-minute intensity, various features related to the physical characteristics influencing MRMS performance are calculated from the topography, surrounding storms, and rainfall observed at the gage location. A gradient boosting regressor is trained and was used to predict a range of error throughout the mountains of Colorado during warm months. Mapping of this dataset by aggregating normalized RMSE over time reveals that areas further from radar sites in higher elevation terrain show consistently greater error. However, the model predicts larger performance variability in these regions compared to alternative error assessments. Precipitation gages provide high temporal resolution data; however, terrain induces significant variability in precipitation across Colorado. As a result, interpolated datasets or frequency analyses based on simple linear regression of gage data may fail to capture critical extremes. Anomalous precipitation events and variation at subdaily time scales are likely omitted from gage-based datasets due to low station density. To explore this uncertainty, I use several remotely-sensed and model-derived hourly datasets and re-evaluate the influence of terrain on the magnitude of subdaily precipitation intensity throughout Colorado. Precipitation–elevation relationships differ among basins: the Missouri and Arkansas show decreasing precipitation with elevation—an effect stronger for hourly than daily totals—whereas the Colorado and Rio Grande exhibit increasing precipitation with elevation, with daily totals rising more steeply and significantly than hourly ones. Gage‑based frequency studies, limited by sparse networks, miss the frequency of high intensity clusters along the Front Range and Pikes Peak shown by model data. Gage interpolation schemes might also fail to capture how particular terrain features, rather than elevation alone, affect precipitation development. After evaluating the capability of various datasets to represent precipitation variability in Colorado's mountains, I investigate how this variability impacts sediment mobility and runoff. Non-uniform rainfall profiles can result in significantly higher runoff and erosion rates compared to constant rainfall. In this study, I use a rainfall simulator capable of generating time varying intensity profiles similar to natural rainfall observed by gages in the mountains of Colorado. I examine the effects of time-varying rainfall intensity on both mulched and bare soil on a steeply sloped flume. The effectiveness of mulch varies markedly between fluctuating rainfall intensities and a steady intensity. The wood mulch treatment failed to significantly reduce total runoff under time-varying rainfall, though it did under constant rainfall. Erosion rates were reduced with mulch for all rainfall events, despite the increasing intensity rain event causing significantly more erosion on bare soil. Accounting for the variability of rainfall directly impacts management solutions for post-wildfire recovery. Rainfall variability in mountainous regions can be highly dynamic across both space and time, affecting surface processes in small plots and at larger scales. Accurately representing these fine-scale rainfall patterns remains challenging. Remotely sensed data and model outputs can complement or substitute gage data, especially when gages are sparse, enhancing the accuracy of rainfall estimates.