TREX-SMA: A multi-event hybrid hydrologic model applied at California Gulch, Colorado.

摘要

This dissertation describes a hydrologic model, Two-Dimensional Runoff Erosion and Export (TREX) Soil Moisture Accounting (SMA), created from adding the Sacramento Soil Moisture Accounting model (SAC-SMA) to the TREX surface hydrology model. TREX-SMA combines the capabilities of TREX as a distributed physical surface hydrology model with a conceptual rendering of infiltration and return flow as found in SAC-SMA. In order to form the hybrid, infiltrated water (computed as a distributed function on the surface) is aggregated as an input to a system of soil moisture accounting zones, underlying the entire watershed.;In each model time step, TREX-SMA releases baseflow from the accumulated infiltrated water according to simple transfer functions. Evapotranspiration (ET) losses from the soil moisture zones are computed based on potential ET demand and available water. As baseflow and ET are released between precipitation events, TREX-SMA recovers capacity in the soil moisture zones. Based on the simulated recovery, the model then re-initializes the infiltration parameters of the surface model to prepare for the next event, allowing continuous simulation of multiple events.;The capabilities of the TREX-SMA model to continuously simulate soil moisture, infiltration, and rainfall-runoff are demonstrated with an application to multi-event modeling on the 30 km2 California Gulch watershed, near Leadville, Colorado, United States. Precipitation inputs are derived from measurements at a system of six precipitation and stream flow gauges providing ten-minute data for the summer of 2006. Eight major events were recorded during this time with runoff produced at all gauges. One additional event with partial watershed response was also evaluated for a total of 54 event hydrographs in the 50-day simulated series. Time steps in the simulation ranged between 2.0 and 4.0 seconds.;Parameters for the surface hydrology were obtained from a prior calibration of TREX and were distributed across 34,000 grid cells based on the 30-meter United States Geological Survey (USGS) Digital Elevation Model (DEM). Parameters for the soil moisture zones were obtained from a-priori estimates used by the Arkansas Basin River Forecast Center of the National Weather Service (NWS) of the National Oceanographic and Atmospheric Administration (NOAA) in their real-time operational flood forecasting model for the Arkansas River. Using conceptual soil moisture states to re-initialize distributed infiltration parameters, the simulation results with TREX-SMA improved relative to results from the unmodified TREX model with constant infiltration parameters.;Model results are processed using gnuplot to create real-time hydrograph plots as the simulation progresses. Gnu R scripts produce real-time plots of simulated minus observed residual and statistical analyses as the simulation progresses. Statistics generated for each gauge include Nash-Sutcliffe, percent bias, absolute percent bias, Pearson correlation and modified Pearson correlation, and mean-squared error. These statistics were generated both for the entire simulation series and for each individual storm event. The gnuplot and R plots are produced using web-based technology for instantaneous sharing via the Internet. Model results such as surface and channel water depth are processed with GRASS GIS and KML scripts to create 2.5 dimensional, browseable animations overlaid on a Google Earth terrain.;Statistical measures of the improvement of TREX-SMA over TREX are presented in this dissertation. The overall accuracy, measured by the Nash-Sutcliffe coefficient, improved in four out of six gauges. Peak over-estimation was corrected in a majority of the 54 peaks evaluated. Implementation of the TREX-SMA soil moisture accounting algorithm to re-initialize the infiltration parameters reduces the total absolute peak error from 180% to 135% of the observed peak flow rates. The Nash-Sutcliffe model efficiency improved over standard TREX simulations by 43%, 11%, 5%, and 10% at CG-1, CG-4, CG-6, and SHG09A.