About the EMDS System and Its Use in Salmonid Recovery Planning

Overview

The Ecosystem Management Decision Support (EMDS) system is a type of spatial decision support system (SDSS) designed for ecological, landscape assessments. It provides a framework for integrating geographic information system (GIS) with knowledge-based reasoning and decision modeling technologies to support resource managers, scientists, planners, and so forth in the adaptive management process of ecosystem management.

The GIS functionalities of the EMDS system are the product of a detailed knowledge base that describes the relationships between the factors that influence ecological function and quality. A knowledge base is not data, but a rule-based model or framework of criteria through which data are evaluated and interpreted. Because the EMDS system works within a GIS environment, it allows users to examine ecosystem processes and conditions at a landscape level. Defined geographic areas, such as watersheds or basins, can be systematically evaluated against the criteria specified within the knowledge base, and compared with one another. The EMDS system generates data tables and spatial data files that can be mapped or used in further analysis. It should be understood that the EMDS system is not a mathematical model; it is a type of rule-based modeling that uses “fuzzy logic” (described further below), and a type of SDSS that can serve as a valuable screening tool in ecosystem and resource management. There are several advantages, as well as considerations, in using the EMDS system, which are discussed further below.

The knowledge bases used by the EMDS system are highly flexible and can be customized to allow for a range of ecological applications. Therefore, the EMDS system can be a widely used tool, but is dependent on a well-constructed criteria framework to address a specific management issue of concern. The EMDS system is being used for various ecological applications, including evaluating forest ecosystem sustainability, analyzing site suitability for research reserves, and predicting moose distribution based on foraging habitat (visit the EMDS Web site for more information).

The EMDS system has also been used to assess limiting factors for endangered salmonid populations in watersheds of Northern California. Two knowledge bases were developed for application with the EMDS system to evaluate the suitability of habitat for sustaining healthy populations of salmonids. They evaluate potential sediment production and in-stream fish habitat quality across California’s planning watersheds to help guide agencies, landowners, watershed groups, and other stakeholders in their efforts to improve watershed and fisheries conditions. These knowledge bases, to be run with the EMDS system, were designed by a collaborative group of experts under the California Resources Agency for use in the North Coast Watershed Assessment Program (NCWAP). Additional knowledge bases to evaluate watershed suitability with respect to water quality and availability of fish food were also proposed by NCWAP but never developed. NCWAP’s knowledge bases are discussed below, but for more detailed information, visit the NCWAP Web site.

NCWAP’s Watershed Condition modeling approach proposes the use of four knowledge bases, that are based on critical limiting factors, for assessing habitat suitability for salmonids: potential sediment production, fish habitat quality, water quality, and fish food availability. Image courtesy of the North Coast Watershed Assessment Program.

Background on the EMDS System

The EMDS system was developed by the Forest Service of the United States Department of Agriculture (USDA). The latest release, version 3.0, is designed to work with Microsoft Windows® 2000 (and higher) and Environmental System Research Institute’s (ESRI) ArcGIS® software. It operates as an ArcGIS extension (so users must have some GIS experience), but it does not run alone and requires other files. The EMDS system can be obtained for free over the Internet or from the developers.

The knowledge base of the EMDS system provides the framework for conducting landscape evaluation and planning. This framework consists of a system of logical questions or criteria that defines the relationships between various landscape attributes and their joint contribution to overall ecosystem quality for a user-defined study area. Knowledge-based reasoning is a useful modeling methodology when current knowledge of the system or measurements of factors within the system is too incomplete or imprecise to support mathematical modeling (Reynolds 2002).

Knowledge bases used by the EMDS system are developed outside the ArcGIS environment using NetWeaver™software. NetWeaver works within the Microsoft Windows platform, and provides graphical tools for building and evaluating knowledge bases. The resulting knowledge base displays the underlying logic and assumptions of the analyses that will be performed (Reynolds 2002). NetWeaver software is required to build or modify a knowledge base; it is not required, however, to execute the final knowledge base within ArcGIS. A single license for Netweaver can be purchased from Rules of Thumb, Inc. for approximately $500 (visit the Rules of Thumb Web site for more information).

Each execution of the EMDS system generates new spatial data files that were scored from 1 (true) to -1 (false) (based on existing data) for each question or criterion in the knowledge base. Reference curves with “break points” are defined in the knowledge base and used to evaluate each proposition of the assessment to determine its degree of truth/falsehood (for example, values approaching +1 are closer to true and those approaching -1 are closer to false). These values serve as metrics for expressing the degree of truth (or falsehood) of assertions about ecological states and processes (Reynolds 2002). Such “fuzzy logic” used by the EMDS system is rather unique among decision support systems.

This example reference curve addresses the proposition “stream temperature is suitable for salmon.” The horizontal axis shows temperature in degrees Fahrenheit, while the vertical axis is labeled “Truth Value” and ranges from -1 to +1. The line shows what are fully unsuitable temperatures (-1), fully suitable temperatures (+1), and those that are on the continuum in between (>-1 and <+1). Break points (where the slope of the reference curve changes) occur at 45, 50, 60, and 68 degrees Fahrenheit. Image courtesy of the North Coast Watershed Assessment Program.

There are several advantages to using the EMDS system as a decision support tool. Because it works within a GIS, it can be applied to ecological assessments at any geographic scale. Analyses can also be documented and repeated. The “fuzzy logic” used allowed with the EMDS system, generates scores that can be ranked, instead of just true (1) or false (-1). Outputs from any stage of an assessment, not just the final rankings, can be displayed in ArcMap's data views, graphs, and tables (Reynolds 2002). Outputs can also be used to create maps or for spatial overlay, additional spatial analysis, or modeling within the ArcGIS environment.

Another valuable aspect of the EMDS system that makes it ideal for use in environmental assessments is that it can reason with incomplete information. It allows for partial evaluations of system states and processes based on the available information, and ranks any missing data in order of relative importance to the analysis (Reynolds 2002). Understanding which data are required for an analysis and the influence of missing data helps users focus limited resources on areas that would improve the assessment.

The Netweaver component of the EMDS system also allows for a single knowledge base to incorporate a wide variety of topics. This is can be valuable for ecological assessments where different topics and subtopics are related, such as terrestrial vegetation, wildlife, fish populations, recreation, water and air quality, aesthetic concerns, commercial concerns, and so forth. The number of topics and the inter-relations that can be represented in a knowledge base for the EMDS system are only limited by a computer’s dynamic memory (Reynolds 2002).

There are also several considerations to using the EMDS system. It is not an empirical, mathematical, biological, or a Monte Carlo simulation model. The EMDS system, used in conjunction with a knowledge base, is a static, rule-based modeling tool. The quality of results from the EMDS system are dependent upon the design of the knowledge base it uses, and the availability and quality of input data. The knowledge bases may be made up of dependency rules that oversimplify ecosystem processes and functions. In addition, literature or reference values are also needed to implement the “fuzzy logic” of this rule-based tool, but such studies and data are often lacking. Although decision support systems like the EMDS system usually allow for repeatability of analyses, they can also be difficult to validate or confirm. Certain skills and expertise, including an understanding of spatial analyses and GIS skills, are needed to program the knowledge bases and operate the EMDS system. Even though the EMDS system provides data results in GIS format, they are "aspatial" in that they do not reflect analysis of the effects of neighboring areas of interest (such as upstream or downstream watersheds from a selected watershed). Finally, the EMDS system can be considered a coarse modeling tool, where small to moderate changes in certain environmental factors may have little or no effect on the overall analysis results (Rich Walker, CDF, personal communication, 2004). Therefore, the EMDS system should be used to complement, not replace, other types of models used in ecological assessments.

For more information on the EMDS system developed by the USDA Forest Service, visit their EMDS Web site.

About NCWAP’s Models

In 2001, the North Coast Watershed Assessment Program (NCWAP) began exploring the EMDS system as an SDSS for evaluating suitable watershed and stream conditions for salmonids in Northern California (NCWAP 2002). Since the EMDS system supplies only the framework for conducting any type of ecological assessment, NCWAP needed to develop customized knowledge bases for the EMDS system to use in analyzing common limiting factors for Pacific salmonids, not just in Northern California, such as sedimentation, water quality, large woody debris, and so forth. (See the Factors Limiting Salmonid Production section of this Web site for more information.) The result was the development of two knowledge bases that focused on different geographic and management scales: the Stream Reach Condition Model (developed primarily by the California Department of Fish and Game [CDFG]) and the Potential Sediment Production Model (developed primarily by the California Department of Forestry [CDF]).

The purpose of the Stream Reach Condition Model (view model) is to evaluate if and where current stream reach conditions are suitable for sustaining healthy populations of native anadromous fish within a specific watershed. This knowledge base relies largely on data collected under CDFG’s stream survey protocols (NCWAP 2002). The data sets needed (in spatial format) to run this analysis with the EMDS system are summarized in the table below. The Stream Reach Condition Model has undergone limited validation and testing. Visit the NCWAP Web site for more information on this knowledge base.

Information for this table was obtained from NCWAP 2002.

To evaluate whether watershed conditions are suitable to sustain healthy populations of native anadromous salmonids among planning watersheds within a user-selected basin, four knowledge bases were proposed for the Planning Watershed Condition Model approach: the Potential Sediment Production Model (view model) and the Water Quality, Fish Habitat Quality, and Fish Food Availability Models (view models). All of the knowledge bases were conceptually designed except for the Fish Food Availability Model; yet of these four knowledge bases, only the Potential Sediment Production Model was actually developed and used by NCWAP (NCWAP 2002). Data requirements (in spatial format) for the Potential Sediment Production Model, as well as the Water Quality and Fish Habitat Quality Models, are summarized in the table below. Visit the NCWAP Web site for more information on the knowledge bases that make up the Planning Watershed Condition Model approach.

Outputs and Applications

NCWAP’s Potential Sediment Production Model (Version 1.1) has been used with the EMDS system for select basins with sufficient data for the planning watersheds, such as the Red Wood Creek basin shown in the figure below. Spatial data outputs from the analyses allowed for planning watersheds within a basin to be ranked from highest potential suitability to sustain healthy salmonid populations with respect to sediment input to streams (so less sediment loading) to lowest potential suitability (so greater sediment loading). Such information can be useful to natural resource managers and land use planners in various watershed management efforts, including setting restoration priorities, selecting road decommissioning or repair projects, or planning for future development. Results from the analyses can also help landowners, watershed groups, and restoration professionals in selecting the appropriate types and locations of restoration projects that best contribute to salmonid recovery (NCWAP 2002). Since sedimentation is not the only limiting factor for salmonids, management decisions should be made in conjunction with the results from the other knowledge bases, when they are complete, for a more comprehensive assessment of watershed conditions. Since information and data are lacking for some of the criteria of the knowledge bases, the knowledge bases and results are likely to change and improve as validation, testing, and scientific understanding increase (Rich Walker, CDF, personal communication, 2003).

Example enhanced map output from NCWAP’s Potential Sediment Production Model showing potential suitability for salmonids among planning watersheds within the Redwood Creek basin, based upon all sediment sources (natural and management-related). Planning watersheds with lower potential sedimentation are generally considered better for salmonids (shown in green), while those with higher potential sediment production are generally considered worse for salmonids (shown in brown). Image courtesy of the North Coast Watershed Assessment Program.

The experiences of NCWAP point to the need for better calibration, validation, and reporting of uncertainty of their models in the future. Review of the knowledge bases by an external science panel has played a critical role in improving the logic and validation of NCWAP’s models thus far (Walker 2003). Prior to being cancelled due to state budgetary constraints, NCWAP was working towards completing the Water Quality, Habitat Quality, and Fish Food Availability Models for the Planning Watershed Condition modeling approach, as well as incorporating the results of other external models such as a road-related sedimentation model (SEDMODL_V2) and a large woody debris model, into the analyses. However, the EMDS system is just one of several approaches that can be used to assess salmonid limiting factors in watersheds of California. It can be considered a framework, or a starting point, for additional analyses and better understanding in the management and decision-making processes of salmonid recovery and watershed management (NCWAP 2002). For more information on the uses of NCWAP’s knowledge bases, please visit NCWAP's Web site.

References

NCWAP. 2002. "Appendix: Ecological Management Decision Support (EMDS) Model." North Coast Watershed Assessment Program. Unpublished draft 7/17/02, 53 pp. View on-line document.

Reynolds, K. 2002. "Ecosystem Management Decision Support Users Guide (Version 3.0)." US Department of Agriculture, Forest Service, Pacific Northwest Research Station. Report C11651, 112 pp.

Back to top