North Carolina
Cooperative Extension Service
NORTH CAROLINA STATE UNIVERSITY
COLLEGE OF AGRICULTURAL & LIFE SCIENCES

NWQEP NOTES
The NCSU Water Quality Group Newsletter

Number  66            July 1994         ISSN 1062-9149

NONPOINT SOURCE PROGRAM NEWS

Massachusetts Implements Clean Water Policy

Arleen O'Donnell, Acting Deputy Commissioner
Massachusetts Department of Environmental Protection

In 1993, the Massachusetts Department of Environmental Protection (DEP) completed a Clean Water Strategy. The overall goal of the Strategy is to protect the environmental integrity of the state's water resources by putting the necessary tools in place to set resource-based priorities, integrate programs geographically, and improve the effectiveness and efficiency of programs that cross division (organizational) lines.

The Strategy establishes as its centerpiece a river basin approach to resource management. The basin has been identified as the integrating theme for resource protection programs including monitoring, assessment, and regulatory activities. The DEP has begun to synchronize activities that previously occurred in isolation: water quality monitoring; water withdrawal permitting; and National Pollutant Discharge Elimination System (NPDES) permitting. The agency is currently developing a system for assessing nonpoint source contribution to total pollution loading within basins. By coordinating these activities and focusing them in a particular basin, the relationship between water quality and water quantity can be better understood and cumulative impacts of multiple withdrawals and discharges can be better assessed in relation to critical resources that need protection.

Pilot basin permitting initiatives were undertaken during 1992 in the Housatonic and Stony Brook River Basins. In the Housatonic, water quality sampling locations were chosen in areas where water withdrawals and wastewater discharges were a potential concern. The sampling provided a snapshot of the health of the ecosystem and created a baseline for evaluating withdrawal and discharge permits on a basin basis to facilitate a coordinated look at resource impacts. In the Stony Brook pilot, a geographic information system (GIS) was used to promote wellhead protection efforts at the local level. The Stony Brook pilot gave DEP information on how much effort is required for, and the relative value of, generating a number of GIS data layers for the basin.

The Clean Water Strategy effort is being rapidly expanded. The DEP is working toward integration of two other major initiatives within the department. The FIRST program (Facility Inspection to Reduce the Source of Toxics) targets multimedia inspections to those facilities defined as contributing to the non-attainment of Water Quality Standards. In the DEP's "site discovery" program, staff actively seek to uncover unknown hazardous waste sites in areas identified as high risk critical resources.

Plans are underway for involving other environmental agencies and facilitating citizen participation to help set goals for each basin, gather information, identify sources of pollution, provide input on permit decisions, and, hopefully, take action locally to control nonpoint source pollution. In this way, DEP's permitting decisions will be more strongly informed by a concerned and educated constituency, and partnerships can be created for a shared vision of the watershed and broad participation in formulating solutions to reach the vision.

In order to ensure the coordination and accountability necessary to implement the Strategy, the Office of Watershed Management (OWM) was created within DEP. The OWM is responsible for implementing basin permitting and the water quality assessment work needed to support it. GIS is being used to establish water resource protection priorities by identifying the most sensitive watershed in the state and overlaying water resource attributes of statewide significance. The co-occurrence of significant characteristics will identify critical areas for extra protection efforts, such as stricter regulatory controls through designation of Outstanding Resource Waters or Areas of Critical Environmental Concern (ACECs), or prime sites for state land acquisition. This effort will augment DEP's current policies, which recognize public water supply wells and Outstanding Resource Waters (such as designated ACECs, certified vernal pools and surface drinking water supplies) as the state's highest priority water resources.

By evaluating the relationship between critical areas information in the GIS and DEP's regulated facility data base, DEP staff can further target permitting, compliance, enforcement, and technical assistance efforts to those facilities and activities that threaten critical resource areas. The GIS/facility data base interface, which will integrate these two systems, is currently under development. In addition, resource threats will be assessed using water quality data bases and the 305(b) report of statewide water quality, nonpoint source assessments (and information in the state's Nonpoint Source Management Plan), hazardous waste site locations, and other information systems. Thus, for each river basin and statewide, DEP can set priorities with regard to critical resource areas and those activities that pose the greatest threats, and can target its programs accordingly.

For more information, or to receive a copy of the Clean Water Strategy, contact the Bureau of Resource Protection, Massachusetts Department of Environmental Protection, One Winter St., Boston, MA 02108, Tel: 617-556-1172.

TECHNICAL NOTES

A brief description of the Rural Clean Water Program (RCWP) is in order to set the stage for the following article. The RCWP was a 10-year federally sponsored nonpoint source (NPS) pollution control program initiated in 1980 as an experimental effort to address agricultural NPS pollution on a watershed scale. The RCWP was administered by the U.S. Department of Agriculture-Agricultural Stabilization and Conservation Service in consultation with the U.S. Environmental Protection Agency. Many other federal, state, and local agencies also participated. Programmatic and project-level decisions were made by national, state, and local inter-agency coordinating committees. Twenty-one experimental watershed projects, representing a wide range of pollution problems, were initiated across the country. Each project involved the implementation of best management practices (BMPs) to reduce NPS pollution and water quality monitoring to evaluate the effects of the BMPs. Landowner participation was voluntary, with cost sharing and technical assistance offered as incentives for implementing BMPs.(For further information on the RCWP, refer to NWQEP NOTES issue 58 or Gale et al., 1993.)

RCWP Expert: A Knowledge- and GIS-Based Software System for Site-Specific Recommendation of Water Quality Control Practices

Michael A. Foster, Runxuan Zhao, and Paul D. Robillard
Department of Agricultural & Biological Engineering and Laboratory for AI Applications, Penn State University

Introduction

The storehouse of knowledge gained from the Rural Clean Water Program (RCWP) can be most effectively utilized when it has been properly integrated and packaged in an easily accessible form (Robillard, 1992). Expert systems, computer programs that organize and integrate human problem-solving expertise, are ideal for this purpose. Knowledge-based systems are a broader category of problem-solving software which contain expertise for problem-solving derived from computer models, data bases, and documents, as well as human experts. Based on the wide availability of powerful, low-cost personal computers, improved methodology and tools for expert systems development, and the well-documented advantages of expert systems over non-computerized methods of technology transfer (Coulson and Saunders, 1987; Travis et al., 1992), the RCWP Expert project was initiated in 1991 to integrate and synthesize the lessons learned from the RCWP. The intended users of the RCWP Expert system are primarily watershed scientists and engineers who monitor, select, and implement BMPs to control nutrient, pesticide, and sediment loads in agricultural watersheds.

We describe here the recent evolution of RCWP Expert from a single-site expert system for selecting and evaluating BMPs to a knowledge-based system which recommends BMPs at multiple spatial scales (single or multiple fields). Much of the data for RCWP Expert is stored and displayed in a geographic information system (GIS), a software system for storing, analyzing, and displaying maps of spatially referenced information. Therefore, we refer to RCWP Expert as a knowledge- and GIS-based system.

Methods

The RCWP Expert system is targeted for a rapidly emerging computing standard: UNIX-based workstations emphasizing a GIS data base that can be shared easily among numerous user groups. Because the data bases used by RCWP Expert are based in GRASS and INFORMIX, the software environment is GRASS 4.1 on SUN OS 4.3.1 with an X-Windows (X11R5, Motif 1.2.2) interface and associated DOS models running under SoftPC for UNIX (Insignia Solutions, Inc., Mountain View, CA).

The knowledge base for RCWP Expert is drawn from principles, learned from RCWP projects, that relate site-specific conditions to recommended BMPs (Robillard et al., 1990). The principles were converted to IF-THEN rules written directly in C-language or in GRASS GIS scripts called by a standard GRASS GIS function, r.infer. The rules are of the form: IF (contaminant of interest is X) AND (soil hydrologic group is W) AND (season is S) AND (application class is C), THEN BMPs 1,2,3,etc. are recommended.

We chose the Sycamore Creek watershed in Ingham County, Michigan, as the initial site for RCWP Expert prototype development because it is one of 8 beta-test sites for the Hydrologic Unit Water Quality (HUWQ) software system under development by the U.S. Department of Agriculture - Soil Conservation Service (USDA-SCS, 1993). HUWQ is a UNIX-based tool with an X Window interface, based in part in work begun at Purdue University (Mitchell et al., 1993) and Michigan State University (He et al., 1993), which generates input files for several water quality models (AGNPS, SWRBB, EPIC, GLEAMS) from GRASS data layers and associated INFORMIX (RDBMS) attribute tables.

Since 1989, the Sycamore Creek watershed has been a USDA- and U.S. Environmental Protection Agency-designated demonstration project for nonpoint source controls in agricultural watersheds (USDA-SCS et al., 1990).

System Design and Applications

Rules for Control Systems Selection

The rules for RCWP Expert recommend one or more sets of control practices to the user, based on the following site-specific characteristics: the contaminant of interest; potential level of loading (low, medium, high); potential level of leaching (low, medium, high); soil hydrologic group (A, B, C, D); whether the time of year is within the growing season or not; and the type of land use (cropland, animal waste, or critical area). A separate set of rules has been developed for each of 16 general categories of BMPs used in the RCWP projects. For example, some form of conservation tillage is recommended to reduce runoff from cropland under conditions otherwise favoring loss through sediment transport, such as a contaminant strongly adsorbed to the soil (such as total phosphorus), the non-growing season, and soils with a relatively high runoff potential (soil group C or D). The user can consult RCWP Expert for control systems recommendations for either a single site or the entire watershed. RCWP Expert first recommends all the individual BMPs appropriate to each set of site-specific conditions based on the rules for siting individual control practices. The system then recommends alternative control systems (complementary sets of source, transfer, field, and delivery-type control practices) composed of sensible combinations of the recommended individual control practices. The selection is made from a lookup table of all plausible control systems for the given application class (such as cropland).

Generation of Expert System Input from the Soils and Fields Data Bases

Some inputs to the RCWP Expert system are entered through user dialogues. These include: the contaminant of interest, the choice of application class for a particular site, and whether it is growing season or not (Figure 1). Input for soil hydrologic group is taken directly from a digitized (electronic) version of the USDA's SSURGO soil series (SOILS5) data base. Other inputs, such as expected leaching potential and expected contaminant loading, are generated by custom functions (described below) from data in the underlying GRASS soils and fields data bases and their associated INFORMIX relational data tables. Essential data types for generating inputs are several GRASS data layers (such as topography, watershed boundaries, field boundaries, the stream network) and several field-specific INFORMIX data tables linked to each spatial field unit by ID number: a Field Identification Table with additional ID numbers pointing to tables for cropland operations (rotations and tillage), fertilization schedule, pesticide schedule, and irrigation schedule.

Linkage to AGNPS

AGNPS is a distributed-parameter, storm-event-based model which estimates runoff, sedimentation, and nutrient yields in surface runoff within agricultural watersheds (Young et al., 1989). Outputs can be examined either at the watershed outlet or at the individual cell level to identify critical areas and to site and evaluate NPS control systems effectiveness. Recent efforts at AGNPS-GIS linkage in GRASS (He et al., 1993; Mitchell et al., 1993) have been incorporated into HUWQ (USDA-SCS, 1993). The prototype version of HUWQ generates an AGNPS input file for all cells in a watershed from the spatial and relational soils and fields data bases. This input file can then be used by UNIX and DOS versions of AGNPS. RCWP Expert can call either version of AGNPS directly from its X Window interface and display standard AGNPS model outputs for all the cells in the watershed. Within the RCWP Expert context, the purpose of AGNPS simulations is to estimate the potential effectiveness of control systems recommended by the system in improving surface water quality. As other water quality models are incorporated into the HUWQ framework (USDA-SCS, 1993), especially models with a ground water component such as SWRBB (Arnold et al., 1993), we will add software links to the models from the RCWP Expert interface. We are also adding intelligence to model-GIS links for output visualization and explanation.

Custom GIS Functions in RCWP Expert

We have completed a set of functions to generate contaminant loading potential (total nitrogen (N), total P, and sediment) and leaching potential from the soils and fields data bases described above. The numerical values obtained by these functions are then classified into high, medium, or low based on user preferences or default settings (break points of 30 and 80% of the calculated maximum for the watershed) to provide input to the expert system rules. A set of three GRASS functions is used to estimate relative contaminant loading (high, medium, low) from three components: baseload already in the soil, the contribution expected from manure application, and the contribution expected from fertilizer application. Baseload contributions of P and N are determined from organic content in the soils data base by the function r.baseload. Individual contributions from manure application and inorganic fertilizer application are calculated by r.manure and r.fert, respectively. The r.manure function calculates the total N and P application rate in lbs/acre per year on a farm based on the animal number and type in the confinement area and the user's manure distribution strategy (nearly uniform to concentrated on fields near the confinement area). The r.fert function calculates the nutrient application rate (N and P) expected from inorganic fertilizer (lbs/acre) based on field fertilizer schedules located in the INFORMIX data base tables. Finally, r.np.loading classifies the loading potential of N or P into three categories (low, medium and high) based on actual loading from fertilizer, manure, and baseload concentration. Sediment loading is calculated in relative terms by the custom function r.erosion as expected soil loss (tons/acre/yr, generated by the AGNPS model) divided by the erosion tolerance factor. The r.leaching.p function calculates leaching potential based on a complex function of percolation curve number (itself a function of soil hydrologic group), and upon annual and seasonal (fall and winter) precipitation for the watershed (Williams and Kissell, 1991). Default 30-year normals are supplied for precipitation, or the user may substitute values.

Discussion

The literature on software systems for managing nonpoint source pollution in agricultural watersheds is diverse and rapidly growing (Barnwell et al,. 1989; Engel et al., 1993; Ford et al. 1993, Hamlett et al. 1992, He et al. 1993, Heidtke and Auer, 1993; Kiker et al., 1992; Tim et al., 1992; Yakowitz et al. 1993). With few exceptions (Barnwell et al., 1989; Ford et al., 1993; Yakowitz et al., 1993), these decision support systems are purely model-based, GIS-based (Hamlett et al., 1992), or hybrid systems with models running within a GIS framework (Engel et al., 1993; He et al., 1993; Kiker et al., 1992; Tim et al., 1992). The addition of expert system components can overcome some of the difficulties in primarily model-based systems: overly intensive input data requirements, inability to handle missing or incomplete data, requirements that all inputs be numerically expressed, and the high degree of expertise needed to structure model input and explain model output relative to the user's problem context. Primarily GIS-based software systems are often similarly unable to handle incomplete or missing data and require a high degree of expertise to structure the problem and interpret output. The expert system component in RCWP Expert also reduces the number of model runs needed for decision support through preliminary, rule-based screening of sensible BMP combinations at each site of interest in the watershed.

There is a growing awareness of the need to develop sophisticated `dynamic data bases' that allow users to view changes in data over time and space (Keller, 1991). We are currently developing visualization routines, using IBM's Data Explorer, on both the SUN and RS6000 platforms, which can ultimately be linked as C language programs with the RCWP Expert system. When complete, these functions will enable users to more intuitively visualize complex sets of expert system and model outputs, such as: 1) multiple types of model output (e.g., N, P, sediment loading) in an animated, 3-D landscape with both height and color for numerical value, and 2) time series or scenario series (e.g., simulation of progressively `better' management practices) as an animated landscape of outputs. We are also developing explanation functions for both expert system recommendations and AGNPS model output through a combined parsing of the output text followed by rule-based reasoning. Finally, an Apple Macintosh-based ORACLE (Oracle Corp., Redwood Shores, CA) data base of references on control practice effectiveness will be ported to UNIX to provide users with a mode, besides models, for evaluating effectiveness of control systems recommended by RCWP Expert. The final form of the software system will provide users with highly intuitive and flexible guidance based on a wide range of expert knowledge derived from expert rules, water quality models, and scientific literature.

For Further Information Contact

Michael Foster, Dept. of Agricultural & Biological Eng. and Laboratory for AI Applications, 501 ASI Building, Penn State University, University Park, PA 61802, Tel: 814-865-3375, email: mfoster@psupen.psu.edu

References

Arnold, J. G., P.M. Allen, and G. Bernhardt. 1993. A comprehensive surface-groundwater flow model, J. Hydrol. 142: 47-69.

Barnwell, T. O., L.C. Brown, and W. Marek. 1989. Application of expert systems technology in water quality modeling, Wat. Sci. Tech. 21: 1045-1056.

Coulson, R. N. and M.C. Saunders. 1987. Computer-assisted decision-making as applied to entomology, Annu. Rev. Entomol. 32: 415-437.

Engel, B.A., R. Srinivisan, J. Arnold, C. Rewerts, and S.J. Brown. 1993. Nonpoint source (NPS) pollution modeling using models integrated with geographic information systems (GIS), Wat. Sci. Tech. 28 (3-5): 685-690.

Ford, D. A., A.P. Kruzic, and R. L. Doneker. 1993. Using GLEAMS to evaluate the agricultural waste application rule-based decision support (AWARDS) computer program, Wat. Sci. Tech. 28(3-5): 625-634.

Gale, J.A., D.E. Line, D.L. Osmond, S.W. Coffey, J. Spooner, J.A. Arnold, T.J. Hoban, and R.C. Wimberley. 1993. Evaluation of the Experimental Rural Clean Water Program. NCSU Water Quality Group, Bio. & Ag. Engineering Dept., North Carolina State University, Raleigh, NC. EPA-841- R-93-005. 559p Hamlett, J.M., D.A. Miller, R.L. Day, G.A. Peterson, G.M. Baumer, and J. Russo. 1992. Statewide GIS-based ranking of watersheds for agricultural pollution prevention, J. Soil Water Cons. 47(5): 399-404.

He, C., J.F. Riggs, and Y. Kang. 1993. Integration of geographic information systems and a computer model to evaluate impacts of agricultural runoff on water quality, Wat. Res. Bull. 29(6): 891-900.

Heidtke, T. M. and M. T. Auer. 1993. Application of a GIS-based nonpoint source nutrient loading model for assessment of land development scenarios and water quality in Owasco Lake, New York, Wat. Sci. Tech. 268(3-5): 595-604.

Keller, C. P. 1991. Time-space analysis and GIS. Pp. 141-143 In: M. Heit and A. Shortreid (eds.), GIS Applications in Natural Resources. Fort Collins, CO. GIS World, 1991.

Kiker, G. A., G.M. Campbell, and J. Zhang. 1992. CREAMS-WT Linked with GIS to Simulate Phosphorus Loading. ASAE Paper No. 92-9016, American Society of Agricultural Engineers, St. Joseph, MI.

Mitchell, J. K., B.A. Engel, R. Srinivasan, and S. S. Young. 1993. Validation of AGNPS for small watersheds using an integrated AGNPS/GIS system, Wat. Res. Bull. 29(5): 833-842.

Robillard, P.D., P.H. Heinemann, and M.A. Foster. 1990. Expert System for the Design of Water Quality Control Practices. ASAE Technical Paper 90-721. Int. Summer Mtg for Amer. Soc. Agric. Eng., Columbus, OH, June 24-27, 1990.

Robillard, P. D. 1992. Extending the RCWP Knowledge Base to Future Nonpoint Source Control Projects. In: Proceedings of The National RCWP Symposium. EPA/625/R-92/006, US-EPA, Washington, D. C., pp. 385-392.

Tim, U. S., S. Mostaghimi, and V. O. Shanholtz. 1992. Identification of critical nonpoint pollution source areas using geographic information systems and water quality modeling, Wat. Res. Bull. 28(5): 877-887.

Travis, J.W. and R.X. Latin. 1992. Development, implementation, and adoption of expert systems in plant pathology, Annu. Rev. Phytopathol. 29: 343-360.

USDA-SCS, -CES, and -ASCS. 1990. Sycamore Creek Watershed Water Quality Plan. January 1990. Prepared by USDA-SCS, CES. 9 pp.

USDA-SCS. 1993. Water Quality Model/Grass Interface. Discovery Prototype Version. USDA Soil Conservation Service, Technology Information Systems Division, Fort Collins, CO. July 1993.

Williams, J.R. and D.E. Kissell. 1991. Water percolation: an indicator of nitrogen-leaching potential. Pp. 59-84 In: Follett, R. F. et al.(eds), Managing Nitrogen for Groundwater Quality and Farm Profitability. Soil Science Society of America, Inc., Madison, Wisconsin, USA, 1991.

Yakowitz, D.S., J.J. Stone, L.J. Lane, P. Heilman, J. Masterson, J. Abolt, and B. Imam. 1993. A decision support system for evaluating the effects of alternative farm management systems on water quality and economics, Wat. Sci. Tech. 28(3-5): 47-54.

Young, R.A., C.A. Onstad, D.D. Bosch, and W.P. Anderson. 1989. AGNPS: a nonpoint-source pollution model for evaluating agricultural watersheds, J. Soil Wat. Cons. 44(2): 168-173.

INFORMATION

Rhode Island Stormwater Design and Installation Standards Manual

Rhode Island Department of Environmental Management and Rhode Island Coastal Resources Management Council. 1993. State of Rhode Island Stormwater Design and Installation Standards Manual. Rhode Island Department of Environmental Management and Rhode Island Coastal Resources Management Council, Wakefield, RI. 92p.

The Stormwater Design and Installation Standards Manual is designed to assist property owners, developers, engineers, consultants, contractors, municipal planners, and others in planning and designing effective stormwater best management practices. Although specific to Rhode Island, the manual should be of interest to folks in other states who are developing educational materials and programs aimed at effective management of stormwater runoff. The manual addresses: 1) stormwater management performance standards; 2) calculating sediment accumulation volumes; 3) wet ponds; 4) extended detention ponds; 5) general requirements for infiltration practices; 6) pretreatment devices; 7) infiltration basins; 8) infiltration trenches and dry wells; 9) vegetated filter strips; 10) grassed swales; 11) general maintenance requirements for BMPs; 12) site plan requirements; and 13) calculating pollutant loadings.

Copies may be ordered ($13 per copy including postage/handling) from the Rhode Island Coastal Resources Management Council, 4808 Tower Hill Road, Wakefield, RI 02879-1900, Tel: 401-277-2476.

EDITOR'S NOTE

NWQEP NOTES is issued bimonthly. Subscriptions are free within the United States (contact: Publications Coordinator at the address below or via email at wq_puborder@ncsu.edu). A list of publications on nonpoint source pollution distributed by the NCSU Water Quality Group is included with each hardcopy issue of the newsletter.

I welcome your views, findings, information, and suggestions for articles. Please feel free to contact me.

Judith A. Gale, Editor
Water Quality Extension Specialist
North Carolina State University Water Quality Group
Campus Box 7637
North Carolina State University
Raleigh, NC 27695
Tel: 919-515-3723
Fax: 919-515-7448
Internet: notes_editor@ncsu.edu

Production of NWQEP NOTES is funded through U.S. Environmental Protection Agency Grant No. X818397.