Jeremy S. Fried, Mark O. Zweifler, Michael A. Gold and Dan G. Brown
Vegetated riparian buffer strips (RBS) represent a promising approach to impeding the delivery of non-point source (NPS) pollutants into streams, thereby protecting or enhancing water quality. Watershed managers and planners are interested in deploying RBS in the field, but rarely have the resources to establish them along every meter of stream, creek and drain. At MSU, a statewide issues research project is addressing the issue of targeted siting of RBS via GIS modeling, a survey of riparian landowners, and interviews with key institutional players.
The ideal GIS model would prioritize locations for RBS installation by both type and magnitude of potential non-point source pollution problem and likelihood of RBS effectiveness, operate at a high enough resolution to guide decisions about efficient action on individual parcels, and be parsimonious in its requirements for data and skilled analysts. Although others (e.g., Inamdar et al., 1993) have sought to prioritize sites for edge-of-field buffers using GIS representations of slope and land cover, GIS modeling has not yet been integrated into RBS siting analysis. This poster summarizes several GIS approaches to riparian buffer strip siting (dubbed Models I-V) ordered along a continuum of complexity and information richness (of both inputs and outputs). The first three have been constructed for a sub-watershed of Sycamore Creek, near Lansing Michigan; the others are described as work in progress.
Common to all models is an investigative riparian buffer. This conceptualization greatly reduces the areal extent of the NPS management problem by restricting attention to that portion of the watershed through which NPS pollution gains entry to streams and in which most NPS pollution problems are generated.
Model I, an overlay of a fixed width buffer on aerial photo-interpreted land use, should be popular with those who consider land use to be the primary consideration in identifying water pollution sources. Fixed-width delineation of the riparian zone (e.g. 250 meters on each side of the creek) has little basis in hydrology, but is easy to construct. Land uses within the riparian buffer can then be ranked by probable contribution to water pollution and/or degree of inherent pollution mitigation capacity. For example, row crops might be rated high based on probable contribution; forest might be rated low by virtue of filtration capacity.
Model II, a multiplicative, map algebra model (Sivertun et al., 1988), uses ratings based on distance to stream, slope, soil K factor, and land use to calculate a critical value index for each cell. While its suitability index approach may be familiar to potential users, the model focuses on source areas. The model's tendency to rate isolated locations which are quite far from the stream as critical demonstrates that even with the inclusion of distance to stream, NPS pollution delivery is not well represented. Sophistication and accuracy may be somewhat greater than for Model I, as are its data requirements.
Based on the hypothesis that both detachment and transport are required for pollution to enter a waterway, Model III relies on the DYNWETG component of the Terrain Analysis for the Environmental Sciences (TAPES) software (Gallant and Wilson, 1996) to calculate dynamic wetness index (DWT=ln(Ae/S)), with or without distributed soils parameters, and stream power (PWR=AeS), (Ae = effective upslope contributing area, S=slope), as indices of detachment and transport. Fuzzy membership functions were defined for each index, such that the function is 1 for the 0-50th percentile, decreases linearly to 0.01 at the 95th percentile, and is 0.01 above that point. A unit cost surface was generated by calculating the product of these fuzzy sets (i.e., the fuzzy intersection). The cost surface values range from 0.0001 to 1 and are least where there is both high dynamic wetness and high stream power. Distance from the stream can then be accumulated over this cost surface to generate a cost distance map. The lowest n percent of values in the cost distance map can then be reclassified to generate a binary investigative riparian buffer, with the selection of n dependent on situation specific factors. Land use coincident with this buffer can serve as the basis for further prioritization of areas within this buffer based on NPS pollution potential, e.g., the binary investigative buffer can be used to clip a land use layer to create a hydrologically defensible, variable width, land use cognizant, "critical" buffer spanning the stream (Figure 1).
Model IV, not yet constructed, would essentially be a refinement of model III which would integrate DYNWETG's dynamic wetness algorithm into the TAPESG software to take advantage of TAPESG's Digital Elevation Model's (DEMON) flow routing option. Comparisons of output from models with D8 and DEMON flow routing algorithms conducted by the authors and others (e.g., Costa-Cabral and Burges, 1994) confirm that indices based on DEMON produce far more realistic looking maps of hydrologic flow, and eliminate the artifacts of "flow shadows" produced by D8.
With some additional programming, Model IV could be extended to incorporate a distributed coverage of weights representing pollution potential derived from a land use coverage into the flow accumulation calculations. Dubbed Model V, this representation would assign greater weight to surface flow with high stream power originating from catchments with higher concentrations of NPS generating land use.
Two types of validation were attempted: 1) a GPS based stream bank survey for concentrated runoff and ponding and 2) manual interpretation of USGS topographic quadrangle maps (Figure 1). Ultimately, the conveyance of pollutants across the riparian corridor and into the stream must be monitored during storm run-off events to assess the reliability of any of these models.
While technically deficient, Model I may still be a good choice as a first step towards an investigative buffer, and can be generated with minimal GIS expertise; however, up-to-date land use may be difficult to obtain in many areas. Model II is more sensitive to soils and terrain, but requires additional GIS coverages and somewhat greater analytic acumen. Either could be constructed by a typical GIS consulting firm. The far greater complexity of Models III-V, and the greater difficulty inherent in mastering the TAPES software make these more appropriate choices for managers with access to sophisticated GIS analysis support. Additional monitoring is needed to compare the efficacy of these models.
Implementation of any of these models poses real challenges for watershed managers, a clientele likely to have little or no experience using, building or fitting GIS models. After validation and testing of these models, our ultimate objective is to compare the predictive power of the high cost/accuracy/resolution models such as III-V with that of more affordable and parsimonious methods that would be accessible to watershed managers. Managers could then select the approach best suited to their information needs, analytic capabilities, and budget.
Figure 1. Left: Investigative buffer generated using Model III and a 15th percentile cost distance threshold for Barnard Drain, superimposed on land use. Right: Potential pollution entry sites located by the stream bank survey are represented as points; problem (concentrated flow) areas as determined from interpretation of topographic maps are represented by flow lines running the length of the presumed contributing areas.
Costa-Cabral, M.C. and S.J. Burges. 1994. Digital elevation model networks (DEMON): A model of flow over hillslopes for computation of contributing and dispersal areas. Water Resources Research 30:1681-1692.
Gallant, J.C. and J.P. Wilson. 1996. TAPES-G: A grid-based terrain analysis program for the environmental sciences. Computers and Geosciences 17(3):413-422.
Inamdar, S.P., S. Zacharias, C.D. Heatwole, and T.A. Dillaha. 1993. Spatial placement of filter strips using a GIS. Presented at the December, 1993 meeting of the American Society of Agricultural Engineers, Paper No. 93-3560.
Sivertun, Reinelt, and Castensson. 1988. A GIS method to aid in non-point source critical area analysis. Int. J. of Geographical Information Systems 2:365-378.
Jeremy S. Fried
Michigan State University
Department of Forestry
126 Natural Resources
East Lansing, MI 48824
517-432-3352
jeremy@msu.edu
Mark O. Zweifler
Michigan State University
Department of Forestry
126 Natural Resources
East Lansing, MI 48824
517-432-3537
mark@okemos.for.msu.edu
Michael A. Gold
Michigan State University
Department of Forestry
126 Natural Resources
East Lansing, MI 48824
517-353-4751
mgold@msu.edu
Daniel G. Brown
Michigan State University
Department of Geography
412 Natural Science
East Lansing, MI 48824
517-353-9811
brownda@pilot.msu.edu