DHSVM – Annotated Bibliography

By Jeff Phillippe (Geo 565)

 

The bibliography below is a comprehensive listing of recent articles associated with the DHSVM (Distributed Hydrology Soils Vegetation Model) – a GIS based hydrologic model.  A review of each article summarizes the main points and contains a brief discussion of the appropriateness of using DHSVM in the Hood River Basin, OR.

                                                                                                    

                                                                                                                      

 

 

 

Wigmosta, M.S., Vail, L. W., and Lettenmaier, D.P.  A Distributed Hydrology- Vegetation Model for Complex Terrain.  Water Resources Res., 30, 1665-1679,

1994.

 

Waichler et al. present the findings of their own DHSVM, established in the early 1990’s.   The model is unique in that they developed it for mountainous areas and on a catchement scale.  Furthermore, unlike all other models of the time, it incorporates and outputs data on a pixel by pixel basis, with a user-defined resolution and time scale.  The DHSVM structure integrates a 2-layer canopy model for evapotranspiration, a doubled rooting zone model, a saturated groundwater flow model, and an energy balance model for snow accumulation and ablation.  Meteorologic values of precipitation, air temperature, solar radiation, wind speed, and vapor pressure are assigned to each pixel at a specified height above the overstory. 

 

They ran the model in the Middle Fork Flathead River Basin of NW Montana, calibrating it with 2 years of recorded precipitation and discharge, and verified it using snow coverage and discharge data the following two years.  Calibration was met through adjusting the lapse rate, soil thickness, rooting zone, thickness, and saturated hydraulic conductivity.  The simulated flows had an r-squared value of 0.95 with measured values.  The simulated spatial extent of snow was accurate but had a time lag.

 

The confirmation and applicability of DHSVM in this study area has good indications for similar use in the Hood River Basin for both basins are heavily dependent on snowmelt-derived runoff.  However there needs to be some modification of the model to serve the runoff characteristics of glaciers on Hood.

 

 

 

Bhanumurthy, V., Simhadrira, B., Srinivasarao, G., Rao, V. V., Raju, P. V., Siva

Sankar, E.,  Application of Remote Sensing and GIS in Water Resources Development, Water and Energy International,60, 38-50, 2003.

 

Bhanumurthy et al. describe the various applications of GIS and remote sensing (other than DHSVM) to water resources development – in other words, how to use these tools to get the “optimal management” of water supply.  Remote sensing/GIS are now being applied to water supply assessment, hydrologic/hydraulic modeling, flood management, irrigation management, sedimentation assessment of reservoirs, and seasonal snowmelt forecasting.

 

Although not the same as DHSVM, the team describes a similar GIS-based model SLURP, which was developed at NHRI in Canada.  It runs vertical water balances, incorporating meteorological, topographic, and satellite image data to calculate runoff. 

 

One observable flaw in the article is that the authors make no distinction between glacier and snowmelt periods.  It should be noted that glacier meltwater runoff has a significant lag time after melting whereas snowmelt runoff does not.  This fact has major consequences for and needs to be considered in the hydrologic modeling of the Hood River Basin.

 

 

Chennault, J. W., Modeling the contributions of glacial meltwater to streamflow in Thunder Creek, North Cascades National Park, Washington. MS Thesis,                      78p., 2004.

 

Chennault presents an excellent DHSVM analysis of glaciermelt in the North Cascades, WA.  All other runs of the model address meteorological and land type issues but this is the first one to incorporate the effects of glaciation on basin runoff.  He did so by inputting glacier cells into the vegetation grid.  By adjusting the number of glacier cells, Chennault modeled the effect of glacier retreat on total runoff.

 

Chennault found that in the Thunder Creek Watershed, glaciermelt contributions varied from 0.6% to 56.6% and the onset of glacier melt ranged from June 13 to July 26.  Greatest contributions were provided during warm, dry years, when water stresses were most likely the highest.  This is likely the case for the Hood River Basin, as it experiences a similar climate and glacier-fed streams.

 

 

Kenward, T., Lettenmaier, D., Wood, E., Fielding, E.  Effects of Digital Elevation Model     Accuracy on Hydrologoic Predictions, Remote Sensing and the                                                                                                          Environment, 74, 432-444, 2000.

In this study, Kenward et al. quantifies the deviation of both spatial accuracy and hydrological modeling output of two DEMs to a reference DEM.  They compared a standard USGS 7.5’ DEM and the remotely sensed SIR-C 30 meter DEM to a 5 m-resolution product of low-altitude aerial photography, all from Mahantango Creek.  The USGS DEM-generated water basin area matched the reference area and reference elevation points deviated by a mean of only 4.03 m, whereas the SIR-C DEM was 3.6% larger and reference points deviated by 23.9 m.

 

The DHSVM was applied using all three DEMs.  The USGS- and SIR-C DEM- simulated annual runoffs were 0.3% and 7% larger than that of the reference DEM respectively.  Additionally, peak runoff was generally lower for the two test DEMs.  In conclusion, Kenwood et al. show that the accuracy of the input DEM is essential to hydrologic modeling, and for the DHSVM in particular.  Most likely, it would be more appropriate to use the USGS DEMs over SIR-C for the use of DHSVM in the Hood River Basin.  This type of research could further benefit from exploring the impact of resolution size on hydrologic modeling.

 

 

Stork, P., Bowling, L., Wetherbee, P., Lettenmaier, D.  Application of a GIS-based distributed hydrology model for prediction of forest harvest effects on peak stream flow in the Pacific Northwest,  Hydrol. Proc., 12, 889-904, 1998.

 

Stork et al. use the DHSVM in a practical application; they investigate the impact of clear-cutting and road installation on local peak discharges. They provide a thorough summary of the evolution of hydrologic modeling in the second half of the last century and present excellent graphics and explanations of how the DHSVM works.

 

They credit the availability of DEMs and land data and speedier processing (for the use of GIS) to the development of sophisticated models.  Furthermore, the paper suggests useful tips on the use of GIS in the modeling.  Basin boundaries are delineated with GIS utilities that track flow directions through a watershed (ie. ARC/INFO FLOW DIRECTION and WATERSHED routines).  DHSVM has the power to automatically adjust temperature and humidity values for different elevations.  DHSVM automatically assigns porosity, field capacity, wilting point, and vertical hydraulic conductivity to each pixel based on soil type input.  The same goes for vegetation type.

 

A new add-on is road networks, which will require attributes, and is applicable to the Hood River basin.

 

 

VanShaar, J.R., I. Haddeland, and D.P. Lettenmaier, Effects of land cover changes on the hydrologic response of interior Columbia River Basin forested catchments, Hydrol. Proc., 16, 2499-2520, 2002

VanShaar et al. use the DHSVM to determine the hydrological effects of different land coverage for several catchments within the Columbia River Basin.  The DHSVM indicated that a historical decrease in leaf area index has caused the increase in snow accumulation and runoff, and decreased evapotranspiration.  The effect of this changing landcover is most pronounced during the spring snowmelt season when soils are wetter.

 

This study emphasizes the point that setting up initial variables of interception storage, soil moisture, snow water equivalent content, and saturation extent requires one year of model run through.  This is a problem for places such as the Hood River Basin where long-term hydrological data is limited.

 

 

Waichler, S.R., Wemple, B.C., Wigmosta, M.S. Simulation of water balance and forest treatment effects at the H.J. Andrews Experimental Forest. Hydrol.

              Proces., 19, 3177-3199, 2005.

 

Waichler et al. assess the effectiveness of the DHSVM in reproducing observed water balances in several catchments of the H.J. Andrews Experimental Forest, OR.  The model accurately depicted streamflow from 1958 to 1998 on an hourly basis; however it underestimated low flows, probably due to inadequate storage and groundwater base flow.  The model slightly underpredicted high flows, probably because of undermeasuring snowmelt and downslope water movement.

 

In general though, the study confirms the effectiveness of DHSVM and bodes well for the modeling of the Hood River Basin, which like H.J. Andrews, is located in the snow-rain transition zone.

 

 

Waichler, S.R., Wigmosta, M.S.  Development of Hourly Meteorlogical Values from Daily Data and Significance to Hydrological Modelling at H.J. Andrews

              Experimental Forest, Journal of Hydrometeorology, 4, 251-263, 2003.

 

Since the DHSVM is heavily reliant on an extensive amount of meteorological and hydrologic data (both spatially and temporally), it is often necessary to interpolate missing figures.  For instance, one might be running the model on a 2-hour timescale but might only have temperature data for daily minimums and maximums.  In this paper, Waichler and Wigmosta outline tested methods for interpolating hourly data, based on model runs in the H.J. Andrews Experimental Forest, OR.  Given a data set of daily minimum and maximum air temperatures, precipitation, minimum and maximum relative humidity, and wind speed, one can generate hourly values of air temperature, precipitation, relative humidity, wind speed, and shortwave radiation.  Therefore to run the DHSVM one needs a minimum of the aforementioned data.  With the exception of wind speed, all sets were found to interpolate (with a selected equation) with a high degree of efficiency.

 

With the lack of extensive historical meteorological data in the Hood River Basin, it will be necessary to use these interpolation techniques to run the DHSVM calibration.

 

 

Wang, S., Huang, R., Ding, Y., Leung, L.R., Wigmosta, M. S., and Vail, L.W.   Improvements of a Distributed Hydrology Model DHSVM and its 

              Climatological-Hydrological Off-Line Simulational Experiments, Acta Meteorologica Sinica, 16, 374-387.

 

Written by Chinese climatologists and published in a European journal, this research represents the international importance of the DHSVM in the science community.  Recognizing the differences in climate, hydrology, soils, and vegetation characteristics between the Haile-Loache China River basins and the catchments of the Pacific Northwest, Wang et al. apply important changes to their implementation of the DHSVM. 

 

The paper reproduces most of the methods discussed in Wigmosta et al. (1994).  It uses the improved Penman-Monteith (IPM) which is more appropriate to the evaporation variability in North China.  They also refined the model by subdividing the basins into 4 sub-basins and using a unique meteorological station for each.  The changes in the model have increased efficiency values from 0.82 to 0.89 for two basins.