Using 30m GAP image as an Indicator for Neo-tropical Bird Diversity in Poison Canyon Quadrangle

INTRODUCTION

An important objective of wildlife managers is to identify and manage specific geographic areas as habitat, or potential habitat for wildlife. In the past, problems have arisen in this assessment due to the inaccessability of many areas in which wildlife reside. Rugged terrain and the sheer size of many of the areas of interest have made investigations of habitat criteria and species correlations extremely labor intensive or next to impossible. For this and many other reasons, Geographic Information Systems (GIS) have come to play an increasingly important role in natural resource management. More specifically they have become an important tool in the area of wildlife management because of their ability to model species distribution and habitat (Davis et al., 1990, Clark et al., 1993). In addition, Remote sensing techniques are offering new and more efficient methods of assessing wildlife habitat relations at large spatial scales. This remote sensing data allows natural resource managers to easily identify vegetation classes, obtain qualitative and quantitative spatial assessment, as well as monitor habitat changes over time (Homer et al., 1993).

The presence of Neo-tropical birds (those birds which migrate to the tropics during winter) is considered by many experts to be an indicator of ecosystem health and biodiversity in general (Gill, 1990). As a result, the species richness (S), defined as the number of different species in a given area, becomes a key factor in management decisions and proposed designs for the area. However, for wildlife managers, determination of where these birds are or might be can be a daunting prospect. Collection of this data from the ground would be extremely time consuming and expensive. Therefore, if a relationship can be drawn between habitat and neotropical bird diversity, then a GIS database of bird habitat for a given area may be used to predict bird diversity for that area, greatly facilitating this process (Shaw et al., 1990).

The objective of this project is to determine whether there is a direct correlation between Neo-tropical bird (S) and habitat diversity. A number of different methods will be implemented to come up with a diversity value for the bird census points within the study area, all of which will be compared with our bird diversity data. We hypothesize that as habitat diversity, juxtaposition, and patch richness increase so to will (S) (Olsen et al., 1993) Should a relationship be discovered between diversity of habitat and (S), this study could have management implications for areas in which biodiversity is of concern.

STUDY AREA

The study area consisted of a single quadrangle located in Southeastern Utah (figure 1).

Figure 1.

Go to the Poison Canyon Quadrangle!

The "Poison Canyon" quadrangle is located on USFS land on the Manti LaSal National Forest, Monticello Ranger District, on a mountain range known as Elk Ridge. Elk Ridge is a flat plateau type of a mountain with a few isolated buttes. Elk Ridge runs north and south and is located some 30km west of Blanding, Utah. Elevation ranges from 1650 meters in the canyon bottoms to as high as 3000 meters at the Big Notch and the Bear's ears. The Poison Canyon quadrangle lays in the heart of Elk Ridge. Plant communitites consist mainly of Ponderosa pine Pinus, ponderosa and Gambel's oak Qurceus, gambelii intermixed with various other types of mountain shrubs, and grassy meadows on top. As the ridge slopes away from the top, to the east and the west Pinyon pinus, edulis and Pinyon-Juniper Juniperous, osteosperma become the more dominate habitat types.

Figures 2 and 3. Habitat types shown along the top (2) and off in the canyons (3) oak, ponderosa, and pinyon juniper.

The table below gives a descripton of the habitat types found in the Poison Canyon Quadrangle according to the GAP 30m data. Descriptions include: classification, histogram of the number of pixels in each class, color of each habitat type, area (acres and hectares), and percentages of each type within our study area.

Table 1.

METHODS

Map Generation

Before we really dove into our project we spent the first week or so learning how to compose and generate a map using the 30m GAP image of the quadrangle (figure 4). This was done mainly for Dr. David Winn who needed a map as part of his preliminary report for the project. This was a real interesting and hair pulling experience, but we now know how to generate a map using imagine.

Figure 4.

Make it Bigger if You Want!

Bird data

Bird census were to be taken following the Breeding Bird Surveys (BBS) protocal as described by Ralph (1980). The surveys were taken along several secondary and terrtiary roads along the top, mainly towards the southeastern corner of the quad. It must be noted here that there are some questions about the reliability of the bird data collected. This subject will be covered in more detail in the conclusion section of this report.

Data Manipulation and Analysis

A preliminary step to our analysis was to examine the distribution of our bird data collection points within a dataset of predetermined habitat distribution for our area of interest. The GAP analysis 30m dataset was chosen as an appropriate representation of habitat because of its previous application in biodiversity analysis. GAP was concieved by Dr, J. Michael Scott of the National Biological Survey in 1987. "GAP Analysis works by overlaying maps of land cover and species occurrence onto maps of protected areas, using Geographic Information System (GIS) technology. The resulting maps show the relationship between areas of biological significance and the level of protection afforded these areas"(Biggs, 1995).

The Poison Canyon Quadrangle was clipped out from the State of Utah GAP database here at Utah State University. This yielded an attributed GIS layer of habitat types within the boundaries of the quad., upon which we could overlay our bird census points (figure 5.). With the spatial relationship between our points and the habitat data established, the problem was to come up with a habitat diversity value for the area immediately surrounding each of our census points. A radius of 60m around each point was believed to be a sufficient buffer to account for any error made in coordinate positioning by the technician (Winn, 1995 per comm.). Therefore, our diversity value needed to incorporate habitat within at least a 60m radius if not more.

Figure 5.

Zoom on Census Points!

We attempted a number of different strategies for coming up with our habitat diversity value. The first strategy incorporated the use of the inquire cursor in imagine, with which we detirmined the value of each pixel within a two pixel radius of our census point. With these values recorded in a consistent fashion for each point, we were able to determine the number of habitat types and their percentages. However, in order to calculate an accurate diversity value, the number of zones enclosing a unique habitat within the radius must be included as well (Winn,1995 per comm.). A weighted habitat diversity value was found for each point using the formula:

D=log(%1)+log(%2)+log(%3)+ . . . . . . +log(n%)+(zones-habitats)

Our second strategy was to run a fractal analysis for the pixel clusters surrounding our census points, to come up with a diversity value. The theory behind this analysis is that there should be a positive linear relationship between habitat diversity and fractal value because fractal value is a function of edge, juxtoposition, and diversity of reflectance (Olsen et al., 1993). Therefore, a higher diversity of habitat within a cluster should yeild a higher fractal value for that cluster. In order to perform the fractal analysis, the coordinate from the upper left corner of the area of interest must be determined. Then the cluster can be defined from there. Consequently, in order to define an area with at least a 60m radius around our bird data collection points, a 5x5 pixel window was created for each point, with the pixel within which the point fell acting as the center. The coordinate for the top left hand corner of each of these fifty nine windows was used as our data input.

The program for running the fractal value calculation is called "geco" and is located in ERDAS. The running of this program with our input data produced an unexpected 1200 different fractal values and their respective coordinates. We later learned that the program had been altered for a specific project, and that it would be months before the programmer would be able to change it back. As a result, we had to manually detirmine which of these coordinates were closest to one of our 59 bird points, 40 meters being the maximum acceptable distance from our census point (Ramsey, 1995 per comm.). This process, produced only 22 values which were of use to us, so there were some gaps in our data when we attempted to compare this output with our other diversity outputs.

Our last method of analysis was to take the GAP image and and reclassify it by running a 5x5 window accross the Poison Canyon quad. This gave each indvidual pixel a new value determined by the number of different classes within the 5x5 window (figure 6).

Figure 6 and 7

Therefore, our new image represented a pixel by pixel diversity reclassification. We then created a grid image which consisted of 60m buffers surrounding each of our census points. By overlaying these coverages, we were able to generate a summary describing the number of commonly recoded pixels for each pixel reclassification value within individual buffers. In effect, this exagerated the diversity of habitat for each of our buffer zones. A unique diversity value for each point was obtained by multiplying the individual pixel diversity values by the number of occurances of this value within the buffer:

D=(PDV1 x #)+(PDV2 x #)+. . . . +(PDVn x #)

This diversity value was different than our other two diversity values in that it incorported the habitat values of pixels outside of the 60m buffer on the original GAP coverage. However, this was viewed as appropriate because birds are known to forage outside of their nesting/breeding territory.

Table 2. shows the point number with corresponding bird diversity and habitat diversity

Results

To begin our analysis we plotted each of our diversity datasets against the species richness data for the individual points. This was done to determine whether or not there was a correlation between an increasing species richness and habitat diversity. Our initial hypothesis stated that, as (S) increased, so too would the habitat diversity. The results of our plot showed no clear relationship between these variables (table 3). However, by running a linear regression model, the reclassification diversity dataset showed a slight upward trend, while the fractal and manual interpretation values showed a slight downward trend, with increasing (S) (table 4). We also ran a correlation coefficient in MINITAB for all diversity datasets against (S), which gave us an r value of 0.11 for reclassification diversity, -0.05 for fractal diversity, and -0.09 for manual interpretation. We also ran a correlation between our manual interpretation diversity values and the reclassification diversity values. This gave us an r value of 0.817. This indicated that our manual interpretation formula was valid.

Table 3

Table 4

Discussion

The results of our analysis show that there is no direct correlation between (S) and habitat diversity. Therefore, according to this study, it would be difficult for wildlife managers to predict bird diversity based on habitat type. This is exemplified by the contradiction between diversity trends despite an r value near to 1.0. However, we still believe that our hypothesis holds true, and that there is a relationship between these two indices. Based on personal field observations and data collection, a homogeneous habitat yields fewer species of birds, as opposed to a fragmented or heterogeneous habitat (Black, 1994).

When studying wildlife and habitat interactions it is important to consider the data and data collection including: its relative quality, integrity, and general informational content as well as the limitations imposed by different methods of analysis (Stoms, et al., 1993). The following is a list of data limitations that may have influenced the results of our study.

1. Uncertainty of census point locations: A GPS unit was not used to determine UTM coordinates, therefore certainty of the precise location is questionable. Point location determinations were limited to technician map interpretation.
2. Not following BBS protocal: It was later found that BBS protocal may not have been followed by the techicians. If data were collected in the afternoon instead of between the hours of 0530 and 1030, birds are more likely to have been out foraging instead of displaying and defending within their territories.
3. Non-random point distribution: The census points were limited to the southeastern 1/4 section of the quadrangle. This was due to accessability of this region by motorized vehicle. A more random sample may have yeilded a more representive sample of bird diversity for the area.
4. Proximity to roads: The BBS protocal calls for census points to be placed 50m off secondary or tertiary roads (those roads which are not graveled or well maintained by area managers). Data points were placed right on the actual roads by the technician. The 60m buffer that we placed around each point may have been insufficient in limiting the effect of the road on bird diversity for our study area.
5. Observer errors in bird ID and habitat type: It is uncertain how well trained the observers were and how well they were able to ID the birds. For example in our data set for the Census, the code"juch" was used as a code for a bird over and over again. However, in the North America Bird Banding Manual (USFWS et al., 1977), no such bird even exists. Another error that may have occured is that the observer was not in homogeneous habitat types. This was evident to us when we overlaid the census points on top of the GAP image. Many of the points fell right on the edge of two different habitat types or in a collage of different habitat types.
6. Data analysises and Methods: It should also be noted that our methods and techniques to analize the data may be in question. It may have been more appropriate to use the raw 30m TM data instead of the GAP image. This would have allowed for different types of anaylsis and interpretation of the data. It would have allowed us to analize the data based on continious dataset as apposed to a more generalized coverage.

CONCLUSIONS

Wildlife managers are facing many important resource related problems with reguards to human impacts on wildlife and their habitat, changing land use and land management policies. With the use of a GIS and remote sensing data sound decisions about these problems can be made only when adequate and correct data are provided (Mayer, 1994).

LITTERATURE CITED

Biggs, B., 1995. Encyclopedia of GAP Analysis, Introduction to GAP Analysis. Utah State University U.S.B.S..

Black, T. A.. 1994. 1994 Species list of the LaSal Mountains, Moab, Utah. Presented to The Manti LaSal National Forest, Moab Ranger District.

Clark, Joseph D., J.E. Dunn, K.G. Smith. 1993. A Multivariate Model of Female Black Bear Habitat Use for a Geographical Information System. J. Wildl. Manage. 57(3):519-526.

Davis, F.W., D.M. Stoms, J.E. Etes, J. Scepan, and J.M. Scott. 1990. An Information Systems Approach to the Preservation of Biological Diversity.International Journal of Geographic Information Systems. 4:55-78.

Gill, F.B.. 1990. Ornothology. Witt, Freeman, and Co.. New York, New York.

Herr, A.M., L.P. Queen. 1993. Crane Habitat Evaluation Using GIS and Remote Sensing. PE & RS. 59:10 1531-38.

Homer, C.G., T.C. Edwards, R.D. Ramsey, K.P. Price. 1993. Use of Remote Sensing Methods in Modelling Sage Grouise Winter Habitat. J. Wild. Manage. 57:1 78-84.

Mayer, K.E.. 1984. A Review of Selected Remote Sensing and Computer Technologies Applied to Wildlife Habitat Inventories. Calif. Fish and Game. 70:2 102-112.

Olsen, E.R., R.D. Ramsey, and D.S. Winn. 1993. A Modified Fractal Dimension as a Measure of Landscape Diversity. PE & RS. 59:10 1517-1520.

Ralph, C.J., 1980. Estimating numbers of terrestrial birds: Proceedings of an international symposium. Cooper Ornithological Society, Lawrence Kanas.

Ramsey, R.D., 1995. Professor of GIS and Remote Sensing, Department of Geography and Earth Resources, Utah State University. Personal Converation.

Stoms, David M.. 1992. Effects of Habitat Map Generalization in Biodiversity Assessment. PE & RS. 58:11 1587-1591.

Stoms, David M., Frank W. Davis, and Christopher B. Cogan. 1992. Sensitivity of Wildlife Habitat Models to Uncertainties in GIS Data. PE & RS. 58:6 843-850.

USFWS and CWS. 1977. North America Bird Banding Manual. vol. 2.

Winn, D.S.. 1995. Assistant Professor, Utah State University. Department of Geography and Earth Resources, GIS and Remote Sensing. Personal Conservation.