Year of Award

2007

Document Type

Thesis

Degree Type

Master of Science (MS)

Degree Name

Resource Conservation

Department or School/College

College of Forestry and Conservation

Committee Chair

Donald Potts

Commitee Members

Joel Henry, LLoyd Queen

Keywords

cumulative watershed effects, DHSVM, distributed hydrologic models, fire effects on hydrology, model calibration, model validation, post-fire hydrologic assessment

Publisher

University of Montana

Abstract

The Distributed Hydrology-Soil-Vegetation Model (DHSVM) was applied to the Eightmile Creek watershed in western Montana. The purpose of this research was primarily to assess the applicability of the model as a cumulative effects assessment tool in the post-fire landscape of a forested watershed in this region. The model was first calibrated to the pre-fire watershed conditions using six years of historic streamflow data. DHSVM was able to accurately simulate the general shape of the measured hydrograph for each of the six simulated water years, and the normalized median absolute error statistics were below the target threshold of 50% for each year simulated. This relative success of the calibration efforts is particularly surprising when one considers the significant limitations presented by the lack of any sub-daily or high-elevation meteorological data for use in driving the calibration simulations. Because the accuracy of DHSVM simulations were greatly improved through rigorous calibration, this research demonstrates the need for model calibration to a watershed of interest, prior to hydrologic simulations of different landscape scenarios. Next, two different calibrated versions of DHSVM, including DHSVM version 2.0.1 and the DHSVM fire model, were each used to simulate runoff in the Eightmile Creek watershed following a near catchment-wide stand-replacing forest fire. Due to weather anomalies and limited, discontinuous streamflow data, no decisive conclusions could be made regarding the performance of either version of the model in the validation efforts. Results do suggest, though, that the DHSVM fire model has the potential to outperform the standard model version in fire-affected landscapes. Further research utilizing the DHSVM fire model with more substantial post-fire streamflow records for model validation is warranted.

Share

COinS
 

© Copyright 2007 Crystal S. Stonesifer