L. Jean O'Neil, Environmental Laboratory, U.S. Army Engineer Research and Development Center, Vicksburg, MS.
INTRODUCTION
Users are faced with both a long- and short-term problem in the application of most Habitat Suitability Index (HIS) models. In the long-term and for all applications, users want a model that gives accurate, precise, and understandable results. In the short-term, one that works pretty well may have to do.
Development of high-quality models generally requires a long-term approach: thoroughly testing a model under a variety of conditions and locations requires a structured and potentially rigorous program. In the short-term, however, a species expert may serve as a surrogate for multiple years of population data and habitat variable sampling.
USE OF EXPERTS
Experts in the habitat requirements of a species or group of species can contribute to habitat evaluation in four ways. They can:
• review models;
• develop models, i.e. provide opinions or data for model development;
• rate habitat; and
• serve as a standard of comparison for a model test.
A key assumption in using an expert in any of the four ways is that he/she has an image or mental model of the species' habitat requirements, obtained through research or practical experience. The expert must be able to articulate that image, to consistently and rationally explain the components and functioning of the model. The expert should not be asked to operate outside of his/her expertise or experience, which would lower the reliability of the results. Although a working knowledge of HSI models is not a prerequisite for an expert to contribute useful information, some basic concepts may have to be explained.
Review Models
Reviewing the text of a model (e.g., description of food habits) is similar to reviewing any manuscript, but the best way to examine the variables, curves, and equations is not always evident. The quality of a review is usually improved if the requester guides the process by listing points for the reviewer to consider and questions to answer.
The user may want to "guide" the outcome of the review as well as the process by being specific in the questions asked. The question "Are any additional variables appropriate?" is non-leading and non-limiting. Asking "Do you think a variable for quantity of stream bank vegetation should be added?" would guide the reviewer and potentially limit his/her thinking. It may be a good idea to raise the issue both ways.
Depending on the capabilities of the reviewer, you may wish to ask for consideration of the entire model (life history information, variables, curves or graphs, and equations) or just portions of it. Specific questions will come to mind for a particular model, but the topics listed below should be applicable to most models.
Suggested Questions to Ask a Model Reviewer
Is the intended application (geographic location, life stage, ecosystem, type of impact, restoration action, etc.) clear and reasonable?
Are the stated assumptions valid? Are there additional assumptions to list?
Are any aspects of the life history information incorrect or incomplete? Do you know of any other pertinent data or literature?
Do the life requisite categories adequately reflect the needs of the species and the intended application of the model?
Are the variables accurately and clearly defined?
Could any variables be deleted - are they redundant, insignificant, or based on unreliable information?
Should any variables be added - are there additional factors to be considered, or are the current ones incomplete?
Are spatial relationships expressed adequately and accurately?
How well are the suitability index curves or graphs drawn - are maximum and minimum points appropriate, are they more appropriately drawn in discrete or continuous form?
Do the weights and equations reflect the biology of the animal and the system, e.g., is nesting habitat twice as important (limiting) as broodrearing habitat?
Develop Models
Several models have been built in workshops involving one or more experts and a facilitator. The expert supplies the life history information and knowledge of habitat needs. The facilitator guides the expert's output into model form. The objectives of the model and its intended application must be understood before beginning.
Researchers or practitioners may have data from an area or time period that are useful in model construction. The user must understand how the data were taken and any limits to their use and interpretation, e.g., their applicability to a location different from where the original work was done. If the expert does not actively participate in model development, his/her review of the finished product is recommended. If the expert does participate, one or more additional experts should be asked to provide a review.
The final products of model construction may be used as short-term solutions to a pressing need and may perform quite well. Additional testing and development will still be needed to assure that a high-quality model has been written.
Rate Habitat
If a completed model is not available or practical to obtain, areas can be rated by an expert and the ratings used as HSI values. The expert must use HSI model conventions, i.e., assign a score on the 0 to 1.0 scale; follow geographic and ecosystem constraints, which require that the expert be familiar with the location where the ratings are to be obtained; and clearly state the assumptions used in the ratings.
The advantage of using expert ratings as HSI scores is that it is fairly easy to obtain a quantitative evaluation of habitat quality for use in HEP or management planning. If no model exists or there is not time to build one, an expert may provide adequate surrogate scores. On the negative side, no written model exists for future use - it goes home with the expert. Other disadvantages are that the expert may provide biased ratings, be inconsistent in ratings over time or at different locations, or assign different scores from those of other experts.
Standard of Comparison
HSI models are tested by comparing their assumptions or output with a standard of comparison that represents what the model is trying to portray. Most model tests have used population data, habitat use, or other "hard" standards. Because it is often difficult to select the most appropriate standard and obtain such data, a good alternative is use of expert opinion.
Experts provide a major advantage in being relatively easy to locate, and experience shows that they are enthusiastic about participating. Another advantage is that the interpretations and explanations given by the expert can be very helpful in modifying the model. A disadvantage is that the expert, who should not be familiar with the model, may be thinking or operating in a different framework so that comparisons are not meaningful. A second disadvantage is that the end result will not be, in traditional modeling language, a validated model.
Illustration
In a test of an HSI model for the hairy woodpecker, experts evaluated 40 sites in Tennessee. They each recorded an independent score for each site on a scale of 0 to 1.0 and explained their thinking. The model was changed following analysis of the scores and the original model output. Two variables were redrawn, one variable was added, and the equation was altered. The process is more fully described in O'Neil et al. (1988).
PROCESS FOR RATING HABITAT AND TESTING MODELS
The result of these two activities differs but the process is similar. The quality of the results depends on the caliber and experience of the expert, how well the expert and the sponsor understand the model framework and what is desired at the end of the exercise, and how the rating or testing exercise is conducted.
An adequate number of evaluation sites is needed to insure that the entire range of habitat quality existing in the area is represented. The sites should be clearly described and marked by flagging, topographic features, or whatever it takes to be certain that the expert evaluates the correct area. The size of each site depends on the species and the size of its range, tempered with practical concerns on logistics and access.
If one expert is used, you get one answer. With two experts, you may get two answers and not know which one to accept. With three, you should get a majority opinion. If more than one expert is retained, their scores should be evaluated for consistency and perhaps averaged to give one number per site. If they are not consistent, the scores should be discussed to determine why they differ and an attempt should be made to obtain a consensus. If no consensus is possible, the ratings can be reduced to rankings. Ranked data, however, are of less value for purposes of the evaluation.
The exercise should be conducted with as little bias as possible. For example, if the subject of the model is seen or heard, there is a good chance that the site will receive a higher rating than a similar site with no animal observed. If the animal is observed during evaluation, that fact should be noted by the sponsor and the expert asked to not factor it into the rating or score.
A second source of bias may come from the sponsor or person accompanying the expert. Comments may tip the expert one way of another, e.g., "This is one of my favorite sites on the whole area" or "We'll only need to spend a little time here". Site visits should occur in random order if possible, to avoid any personal trends such as saving the best for last.
A template you may find helpful in discussions with an expert is provided below.
Example Briefing for Expert Evaluation of Habitat
Background
You are being asked to evaluate habitat quality for the Loch Ness Monster at several locations in Lake George. We are using your expertise as a mental model against which we will compare our written Habitat Suitability Index (HSI) model. HSI models provide numerical indices for a given habitat's ability to support a species, and are assumed to be positively related to the carrying capacity of the environment. The scale of an HSI ranges from 0.0 to 1.0. Unsuitable habitat is 0.0 and optimum is 1.0, with a score of 0.3 indicating conditions half as good as habitat rated 0.6. The variables in HSI models are usually measurable features of topography, water, and vegetation. Other determinants of carrying capacity, such as interspecific relationships, can only rarely be included in a model. An HSI is for a defined geographic area; in this case, our reference is limited to one lake in Scotland. The HSI score encompasses all life requisites for the Monster. However, if you think certain life requisites (e.g., food, cover) should be considered separately, give a rating for each one.
Expected Products
You will visit 12 separate sites. At each site, please assign a rating on the scale of 0 to 1.0. You may give two ratings or a range if conditions warrant, but this must be accompanied by an explanation and one overall final score. I also require written comments on why a site gets the rating it does. If it is high, why? What would make it even higher? If low, why? What is the optimum condition against which you are comparing these sites? Comments are critical in interpreting results when we compare your model with the HSI model. For each evaluation, provide the assigned site number and a phrase for the location, e.g., 100 m west of Lookout Rock.
Constraints and Conditions
Several conditions must exist to ensure that we conduct a valid test.
*You may not see the HSI model before assigning your scores.
*Another individual has evaluated the sites; we will not give you his scores, nor should you talk to him until after the evaluation.
*The site boundaries have been delineated and will be explained to you. Please paddle, swim, dive, and look around, taking as much time as you want but staying within the bounds of the site. If a feature outside the site catches your attention, you may consider it, but please explain how it might affect your rating; give a rating with and without its being considered.
*Some or all sites may be too small to support the monster. For purposes of this evaluation, assume that the same conditions extend to a sufficiently large area.
*If time permits, you may revisit any site and change your scores as you see new sites. Please try to make your final ratings in a consistent and evenly-scale fashion.