Application of Automated Geospatial Watershed Assessment Tool to Watersheds in Arid Region

Abstract

This study applied the AGWA ArcGIS tool to the 3.7 hectare Lucky Hills watershed, and three large watersheds (Yuma Desert, Coyote Wash, and Mohawk Wash) in the Barry Marsh Goldwater Range (BMGR) in Arizona. The Kineros model was applied to Lucky Hills watershed to calibrate modeling parameters. Events were grouped by similar precipitation patterns, runoff hydrographs, and precipitation volumes to optimize four key parameters for infiltration and surface runoff. These parameters include the soil suction parameter (G), saturated hydraulic conductivity (Ks), and Manning’s roughness for both the channels and planes. Model calibration was carried on event by event, with 10,000 simulations for each event. The set of parameters that yielded the minimum Root Mean Square Error was chosen as the best-fit parameters. Then, the SWAT and Kineros models were applied to three large watersheds in BMGR to determine possible causes of erosion in the range. Model calibration was attempted based on the observed bed elevation changes using the Digital Elevation Model (DEM) over a period of 13 years. Results for the Lucky Hills showed that grouping events based on precipitation volume yielded varying ranges of best-fit parameters. The six smallest volume precipitation events yielded parameter ranges for G, Ks, cn, and pn as 143.09-147.34 mm, 5.98-7.79 mm/hr, .03-.06, and .03-.04, for each parameter respectively. The six medium sized events yielded parameter ranges for G, Ks, cn, and pn as 145.61-147.34 mm, 5.98-11.32 mm/hr, .03, and .03-.06, for each parameter respectively. The six largest volume precipitation events yielded parameter ranges for G, Ks, cn, and pn as 143.09-236.77 mm, 5.98-14.65 mm/hr, .01-.03, and .03-.07 for each parameter set respectively. For SWAT error is significant in each watershed. For the Yuma Desert basin the minimum error found over 44 sub basins is 24.11%, with only this watershed having an error lower than 30 percent. Of 44 sub basins modelled, 34 under-predicted erosion. Based on the DEM subtraction 6 of the sub basins deposited sediment. For Mohawk Wash, the minimum error found over 33 sub basins was 79.72%. Nine of the sub basins deposited sediment, and 22 underpredicted erosion. For watershed Coyote Wash, of the 35 sub basins modelled the minimum error found was 26.84%. Thirteen sub basins deposited sediment, and 18 underpredicted erosion. Three of the sub basins had errors under 30%. In general, it is apparent that the BMGR range is an erosional environment. SWAT could not accurately predict results for any watershed. The lowest errors were slightly under 30 percent, and most sub basins significantly underpredicted erosion. Kineros results also were poor. To compare Kineros with SWAT results, the results of Kineros planes are added together to form the same grid code as SWAT. Results are shown in figure 5.4.4. Erosion is significantly underpredicted for the Kineros model. For the Yuma Desert basin, the minimum error found for all sub basins was 84.06%. For the Mohawk Wash basin, the minimum error was 90.68%. The minimum error found for Coyote Wash was 25.48%. This indicated that the observed erosion rate was poorly correlated with these selected characteristics of sub-basins. Excluding these parameters, other properties or unobserved hydrologic data, such as precipitation distributions, may contribute to the observed erosion rates.