Sensitivity to Arroyo Development Scenarios: Insights from a Distributed Hydrologic Model

Rainfall-runoff relationships in semi-arid environments are highly non-linear due to the spatial and temporal variability of precipitation and watershed antecedent soil moisture conditions. Specifically within dryland regions, precipitation is often short in duration, limited in spatial extent, and characterized by high intensities. When combined with seasonal periods of elevated basin wetness, rainfall in semi-arid watersheds can lead to large flood and flash flood events within ephemeral channels. Because channel network geomorphology strongly influences flood wave generation and propagation in semi-arid watersheds, it is important to understand geomorphologic controls on streamflow behavior for both flood forecasting and water resources management. Specifically, semi-arid channel networks have experienced rapid geomorphological change that can significantly impact the timing, peak magnitude, and duration of flood events. As an example, during the initial incision stage of arroyo development, streams cut downward through the valley alluvium forming deep, narrow channel networks. Down-cutting through valley fill increases arroyo cross sectional area, thus allowing flood transmission through the channel network without incurring discharge losses to overbank flow. In addition, upslope extension of the channel network head results in the dissection of watershed terrain. The increase in arroyo drainage density reduces overland runoff path lengths and concentrates hillslope runoff in the channel network, which can lead to progressively larger flood events. Furthermore, during the latter stages of arroyo development, establishment of riparian vegetation on the inner floodplain promotes arroyo stability. Vegetative root systems anchor streambank sediments and also increase the resistance to streamflow along channel margins. As flood waves interact with vegetation, sediment deposition occurs and the arroyo network narrows and aggrades thereby increasing the propensity for overbank flow. Consequently, observed flood events may decrease in magnitude once arroyos shift from a generally down-cutting and widening phase to a period of channel aggradation. In this study, the TIN-based Real-time Integrated Basin Simulator (tRIBS) is used to simulate rainfall-runoff transformations and flood production within a large (> 1000 km2) semi-arid watershed under different scenarios of arroyo development. In the initial phase of the study, model parameters are calibrated using observed stream gauge data from the Upper Río Puerco in northwestern New Mexico. Model calibration is focused on simulating flood events that developed in early September 2003 in response to widespread convective activity associated with the North American Monsoon. Results for the calibration phase of the study demonstrate that the tRIBS model reproduces the major sequence of flood events observed in the Upper Río Puerco during September 2003 despite large uncertainties in calibrated soil parameter values and NEXRAD rainfall estimates (e.g, spatial and temporal resolution, absolute magnitude). However, tRIBS does not consistently replicate time to peak discharge, peak discharge magnitudes, or rapid recession limb characteristics observed in the stream gauge discharge time series. Following model calibration, parameters for the Upper Río Puerco tRIBS simulation are used in a series of modelling exercises designed to investigate streamflow response under different scenarios of arroyo development. Model simulations attempt to quantify changes in watershed flood production according to a conceptual hypothesis of arroyo development formulated from observations within Río Puerco watershed by Elliott et al (1999). Within a small test basin watershed (< 20 km2), combinations of channel length, width, and roughness are used to simulate streamflow response to different phases of the arroyo geomorphological cycle. Following simulations in the test basin, model runs are performed in the Upper Río Puerco for similar variations in tRIBS channel network representation. Results for model simulations that investigate different stages of arroyo incision reveal the greatest model sensitivity to changes in channel network length. Channel network extension associated with arroyo incision increases flood event magnitude and decreases time to peak. As arroyo headcuts migrate upslope, the distance to the channel decreases, which results in quick runoff arrival at the stream network. Once runoff is in the channel, longer stream networks efficiently route water to the basin outlet. tRIBS model simulations that considered different parameterizations of channel roughness demonstrated a slight increase in peak discharge and a decrease in time to peak for lower roughness values. However, for the basin scales and flood events explored in this study, model sensitivity to changes in channel roughness is limited. Likewise, the different channel width representations utilized in this study had minimal influence on simulated streamflow response. As a result, for this work, the effect of arroyo development is to magnify flood events primarily through the dissection of the watershed terrain with minor contributions from changes in channel roughness or width.