Scale Compatibility in Distributed Hydrological Modeling

Scale Compatibility in Distributed Hydrological Modeling: Optimizing Model Operation for High- and Low-Resolution Soil Data in the Regional Hydro-Ecological Simulation System (RHESSys)

CONTACT PERSON(S)


Trevor Quinn
Email: trevor_quinn@hotmail.com

OBJECTIVE


To find the optimal hydrologic model resolutions to minimize discrepancies in model output resulting from soil spatial data resolution incompatibility.

DESCRIPTION


Background


Increasingly GIS is used to parameterize the landscape for mesoscale hydrological models. One of the emergent stumbling blocks in distributed modeling is the problem of integrating GIS data sets of varying spatial resolution or scale. Over-generalized resolution for any one of the modeled hydrological sub-processes may propagate uncertainty to the model output. This is particularly a problem when researchers attempt to model a process at a more detailed level of resolution than the input data can match.

This research evaluates the effects of input layer scale incompatibility by comparing the response of the Regional Hydro-Ecological Simulation System (RHESSys) at multiple scales using soil information from two sources: low-resolution digitized soil survey maps and high-resolution data derived from the Soil-Land Inference Model (SoLIM). Appropriate scales at which to operate the model are determined by finding operation resolutions that minimize the difference between model output using the low-resolution soil data and model output using high-resolution soil data.


Study Area


The study area is a portion of the Lubrecht Experimental Forest in western Montana. The area is 36 km2 and characterized by moderate to strong relief. The climate is semi-arid to semi-humid, with mean annual precipitation between 50 and 76 cm (Ross and Hunter 1976). Moisture conditions vary by slope aspect and elevation: low-elevation, south-facing slopes are generally drier than higher-elevation, north-facing slopes (Zhu and Mackay 2001).

Mountain slopes in the study area are forested mainly with second growth Douglas-fir, along with smaller amounts of western larch and ponderosa pine. Ponderosa pine forests dominate at lower elevations, while Douglas-fir forest continue to elevations of roughly 1650 m, beyond which subalpine fir and Engelmann-spruce are predominant (Zhu and Mackay 2001).

There are three major geology types in the study area: Belt rocks, granite, and limestone, which have weathered to form twelve soil series (Zhu and Mackay 2001). Approximately 90 percent of the soils in the study area are Inceptisols - poorly developed soils with minimal organic content.


Proposed Tasks


1. Acquire source data (terrain, vegetation, metereological, and soils) for the entire study area watershed.

2. Generate a soil polygon area frequency distribution for the digitized soil survey and analyzed to determine the mean and modal polygon sizes.

3. Develop a set of UDA thresholds ranging from the minimum to the maximum soil polygon sizes of the conventional soil data. Using these hillslope partition scales (UDA threshold values), the GIS data layers will be converted into a RHESSys "worldfile" containing the landscape representation and parameters associated with each level of landscape partitioning.

4. Complete RHESSys model runs using both the lumped and distributed parameter approaches. Model output using both detailed and conventional soil information and both distributed and lumped parameters approaches will be evaluated at each of the hillslope scales.

5. Examine the total difference in the model output for both streamflow and ET over the 365-day study period between the detailed soil information run and the traditional soil survey run at the full range of partitioning scales. Determine the model operation resolution(s) (the appropriate UDA threshold or thresholds) at which the difference between model response using detailed soil information and model response using conventional soil information is minimized.