Abstract
Using the hydrological modeling capabilities of ArcView Spatial Analyst, we are able to model and estimate water quality parameters for streams throughout a watershed. Various stream pollutant concentrations or loadings are modeled using expected mean pollutant concentrations for particular land cover classes. We also use a weighting factor for unfavorable land on steep slopes and a path distance model to define riparian areas where stream conditions are most sensitive to runoff. The ability to predict stream water quality effects of future clearcuts, strip mines, or agricultural development will prove to be a valuable analysis tool for planners and resource managers.
Introduction
We have developed a watershed modeling system that analyzes landscape pattern change and its affects on water quality. The modeling process combines weighted mass balance and rational methods with literature-based expected mean concentrations for land cover and land use classes to model in-stream nitrogen, phosphorous and total suspended solids. The true power in this watershed tool is the ability to perform “what if” analysis for proposed changes in the landscape to predict stream concentrations and loadings from the change. For example, the ability to predict the increased percentage of total suspended solid loadings from a proposed clear cut in a watershed will be demonstrated in this paper. The following section of this paper will focus on the methodology for creation of the grids and the interface that enables the scenario building.
Methods
The methodology will be discussed in four sections. The first will cover the creation of the hydrologic grids needed in the analysis, the second will cover the expected mean concentration (EMC) grids, and the third will cover how the hydro grids and EMC grids are used together. The fourth section will discuss the interface for changing the land use and cover classes and rerunning the model to evaluate the effects of land cover change.
Section 1. Creation of the hydrology grids
The input data required to develop the hydrology grids include a USGS 30 meter digital elevation model grid, an annual precipitation grid at 30 meters, 1:24,000 stream hydrology. The first step is to create a stream coverage that is fully dendritic, i.e. a layer that represents the drainage network with a fully connected set of single lines (Saunders, 1999). This includes using a single center arc segment instead of left and right bank lines for major rivers and streams. It also includes removing any stray un-networked streams and lakes or ponds. On ponds or lakes fed from drainage, a center stream arc through the water body should be used. For more guidance on vector hydrology pre-processing refer to Saunders (1999).
The processing method of reconditioning the DEM for hydrological modeling used four main steps; (1) rasterize the 1:24,000 scale fully dendritic hydrology to 30 meters and thin the raster grid cells, (2) assign 30 meter sink filled DEM elevation values to the grid cells of the raster stream network, (3) alter the stream network raster cells to ensure that elevations progress in a descending order toward the outlet(s), and (4) use a fixed elevation difference of 20 meters between the stream network raster cells and the surface DEM cell values. After step 4, this reconditioned DEM was then used with the Esri GRID flowdirection and flowaccumulation functions.
With the flow direction grid, watersheds were delineated at each USGS gauging station. For the contributing area above the gauge, the 30-year average annual precipitation was found from the precipitation grid. A linear relationship was created regressing the 30-year annual average flow rates at the USGS gauges to the 30-year annual precipitation value. This linear relationship was then used to create grids of runoff and a cumulative annual runoff grid. A grid of stream flow can also be generated from the runoff grids.
The methods used to create these runoff grids followed the work of Saunders and Maidment (1996). The grids described in this section combined with the expected mean concentration grids allow for instream water quality modeling.
Section 2. Expected mean concentration grid
The land cover and land use grid used for expected mean concentration modeling was the EPA MRLC 30-meter grid. The thirteen classes for West Virginia from this data set were aggregated to six general classes because loading values for nitrogen, phosphorous and total suspended solids were only available for those six classes. The aggregated classes and the corresponding EPA MRLC values included:
| Total Nitrogen | Total Phosphorous | Total Suspended Solids | |
| Urban |
1.89
|
0.009
|
166
|
| Open/Brush |
2.19
|
0.13
|
70
|
| Agriculture |
3.41
|
0.24
|
201
|
| Woodland |
0.79
|
0.006
|
39
|
| Barren |
3.90
|
0.10
|
2200
|
| Wetland |
0.79
|
0.006
|
39
|
The values above were field calibrated by NRCS in West Virginia and are part of a forthcoming report.
Our interface for using the values above allows the user to change the loading coefficients to test different weights for the respective land use/cover classes or update them, as new information becomes available.
Section 3. Combining hydrology grids and expected mean concentration grids
With the values as attribute fields for the land cover grid, we created a cell based loading grid by multiplying the EMC value for each grid by the annual runoff grid. A conversion based on cell size loading was also used to keep units consistent. With the cell based loading grid as the weight grid in the flow accumulation GRID command we then calculated a cumulative loading grid. The cumulative loading grid is the annual loading from the landscape cover classes to the stream. By dividing this grid by the annual stream flow grid we could calculate the concentration of pollutants for each stream cell. The advantage of this raster modeling allows a potentially unique value to be generated for each 30 meters stream cell. These cells can be queried for their associated loading and concentration values within the ArcView interface.
Section 4. Changing land cover and evaluation of percentage change
The procedure for evaluation of a land cover or land cover change consists of first setting a study area extent by zooming to a location in the display. Next, the pull down menu choice for evaluation of the EMC pollutant modeling is chosen as shown in Figure 1.
Figure 1.
The choice of pollution parameter is the next message box for the user to select as shown in Figure 2.
Figure 2.
After a pollution parameter is chosen the model will run as creating the grids as specified in Section 3 described earlier. The results are two new themes added to the table of contents for Total suspended solids in mg/L and Kg/Year as shown in Figure 3 A and B.
Figures 3 A and B.
The southern barren area as shown in the two figures is obviously contributing much to the total suspended solid loading (345 mg/L and 1,369,888 Kg/Year) at the highest points. If the southern strip mine were converted to grassland by re-vegetating the landscape, what would be the decrease in total suspended solids? To evaluate this question, we would simply change the land cover grid by digitizing in the polygon of change and attributing it with the changed land use and merging it into the MRLC land cover grid as shown in Figure 4.
Figure 4.
After rerunning the model with the changed land cover location from coal mine and quarry (barren) to forested revegetated grassland, we observe that total suspended solid concentrations decreased to 18 mg/L and 22,956 Kg/Year at the highest points in the watershed. The updated results are shown in Figure 5 A and B.
Figures 5A and 5B.
Discussion
The major limitations for this type of modeling exist in the land cover scale of data and the assumptions made for the water quality results. The scale of the MRLC land cover grid (30 meter resolution) did not allow for very accurate small-scale interpretations of features. It would have been preferable to assign loading rates to a land cover layer that had a better scale and smaller minimum mapping unit. However, as better land cover becomes available the methodology in this study can be incorporated.
The water quality modeled results are subject to the following limitations. The transport of pollutants is considered to be conservative (values are averaged over changing flow conditions only) no loss or decay of pollutants is considered. It is essentially a landscape-based fate transport advection dispersion model that uses a runoff grid and flow grid for downstream concentrations and loads. It assumes that the streams have the same hydrogeometric properties (stream slope, roughness, width, and depth) and also assumed are that the streams have the same ecological rate constants (reareation rates, pollution decay rates and sediment oxygen demand rate). Other limitations are that the model does not consider infiltration, or ground water flow additions or does it include atmospheric conditions such as evapotranspiration.
However, despite these assumptions and limitations, the techniques do have the following advantage; the output is in an easy to analyze visual format and query in the raster framework and you can include point sources of discharges by treating them as a non-point source in the loading grid calculation.
The techniques are primarily a landscape watershed model as compared to a receiving water quality model. Future work includes the ability to incorporate additional EMC loading values and to model the effects of additional classes besides the general six used in this study (urban, agriculture, open/brush, woodland, wetland, and barren)
References
Saunders, W. K. and D. R. Maidment (1996), A GIS Assessment of Nonpoint Pollution in the San Antonio-Nueces Coastal Basin, Center for Research in Water Resources Online Report 96-1, University of Texas, Austin, TX.
Saunders, W. K. (1999) Preparation of DEMS for use in Environmental Modeling Analysis, In: Conference Proceedings: 1999 Esri User Conference, Environmental Systems Research Institute, Redlands, CA.
Michael P. Strager
Research Coordinator
Natural Resource Analysis Center
West Virginia University
Morgantown, WV 26508-6108
Phone: (304) 293-6253
Fax: (304) 293-3752
Email: mstrager@wvu.edu
Jerald J. Fletcher
Professor
Natural Resource Analysis Center
West Virginia University
Morgantown, WV 26508-6108
Phone: (304) 293-6253
Fax: (304) 293-3752
Email: jfletch@wvu.edu
Charles B. Yuill
Professor
Natural Resource Analysis Center
West Virginia University
Morgantown, WV 26508-6108
Phone: (304) 293-6253
Fax: (304) 293-3752
Email: cyuill@wvu.edu
Jacquelyn M. Strager
Research Coordinator
West Virginia University
Morgantown, WV 26508-6108
Phone: (304) 293-6253
Fax: (304) 293-3752
Email: jrowe@wvu.edu