Some shrub-steppe habitat in the study area. Photo by Kristin Dust.
Multi Criteria Evaluation (MCE) was used to assess shrub steppe bird habitat suitability for a section of the South Okanagan Valley. The resulting distribution of best habitat showed that these birds are limited in their range. This was compared to a second MCE that described cattle grazing habitat. Multi Objective Land Allocation (Mola) was used to resolve conflict between best bird and best cattle habitat. The resulting image suggests areas that may be put aside as habitat reserves. Applications of these GIS modules for habitat management are discussed.
Brewer's sparrow photo by Nancy Mahony
The shrub-steppe of the Okanagan Valley, British Columbia,
is a
unique ecosystem that is threatened by fragmentation due to
development. Much of the remaining habitat is subjected to heavy
grazing pressure. This area provides habitat for a large number
of blue-listed (threatened) and red-listed (endangered) species.
Included among these are two red-listed, sagebrush nesting birds:
the Brewer's sparrow (Spizella breweri breweri) and the
sage thrasher (Oreoscoptes montanus). Both occur in the
Okanagan at the very northern edge of their breeding ranges in
North America. This habitat also
provides breeding grounds for a number of other songbirds
included vesper sparrows, lark sparrows and western meadowlarks.
To manage the habitat of any species, it is
important to know where that species occurs.
Since it is difficult to know bird distributions absolutely for
large areas, Geographic Information Systems may be employed to
model the distribution for a species, given assumptions based on
knowledge of the species' ecology and habitat requirements. The
resulting map of habitat suitability for the species can give
land managers a better idea as to where management efforts should
be focussed.
In this project Multi Criteria Evaluation will be used in Idrisi to rate a section of the South Okanagan for bird habitat. MCE allows the evaluation of any combination of factor criteria. Constraints may also be specified. A factor criterion is measured on a continuous scale and therefore supplies a range of habitat suitability values. Factors should first be standardized so that they are all measured on the same scale. The FUZZY module is used to standardize factors to a scale from 0 (worst suitability) to 255 (best suitability). Constraints are limiting factors that are in the form of boolean images. Each pixel has a value of zero (unsuitable) or one (suitable).
photo by Kristin Dust
Most of the shrub steppe this also supports
cattle grazing. The amount sagebrush increases with grazing
pressure, and the presence of sagebrush in the Okanagan may be
due to its spread by cattle. While some level of disturbance is
probably necessary to maintain a healthy shrub steppe,
overgrazing can be a problem. Although the effects of overgrazing
on bird numbers are not known for this area, there are
suggestions that it is detrimental (Harvey, 1992). Overgrazing
can alter plant communities, increase the amount of bare ground, and
cattle can damage sagebrush. Increases in exotic annuals such as
cheat grass are associated with overgrazing, and there is
evidence that Brewer's sparrows are negatively associated with
areas dominated by cheatgrass (Dobler et al. 1996). If crown land
is going to be set aside for habitat preservation, then cattle
grazing is usually the main conflicting land use objective. The
Multi Objective Land Allocation (Mola) module in Idrisi can be
used with the ranked results of MCE maps for conflicting
objectives. The amount of land to be allocated to each objective
is specified, along with the relative importance of each
objective. Mola provides a compromise between objectives that
trys to maximize the suitability of the land assigned to each
objective. Objectives may be weighted to reflect unequal
importance, and this is also considered. Where conflicts occur,
this module assigned the pixel to the objective with the highest
weighted suitability. In this study, Mola is used to find
potential areas to conserve for bird habitat that are a
compromise between best bird and best cattle habitat.
A raster based GIS, Idrisi, was used for all analyses. The study area encompassed two adjacent TRIM map sheets. The Canada/USA border defined the southern edge of the area. The area also includes part of Osoyoos Lake to the east, and part of the Similkameen River to the west. Habitat data was obtained from a BC Ministry of Enviroment biophysical mapping project (Lea et al. 1991). 1:20 000 airphotos were used to derive a map of habitat classes for the South Okanagan Conservation Strategy. This data set is still in the process of being corrected for errors, but I was allowed to use it for this project.
Two constraint images were produced for the bird MCE. Each polygon in the SOCS database has a broad vegetation class assigned to it. This field was used to produce a boolean image of habitat types that were considered to be constraints for both the bird and cattle MCEs (Appendix A). A second constraint was produced for the bird MCE. While distance from roads was not considered to be a factor since birds are often found nesting along roadsides, birds can not nest in the roads. INITIAL and LINERAS was used to create a rasterized roads map. The roads database was filtered using SQL, and ASSIGN was used to create a boolean image of roads which was used as a constraint in the bird MCE.
Three factors were considered for bird habitat. SQL and ASSIGN were used again on the SOCS database to produce a boolean image of habitat comprised of forest of any type. It was assumed that habitat suitability increased with distance away from these edges since trees provide habitat and perches for corvids (ravens, jays, crows) which prey on nests. This factor also reduced the likelihood that small fragments of shrub-steppe within forested areas could be considered good habitat. The FUZZY module was used on a distance image of this module with a monotonically increasing sigmoidal function. The control points chosen were 50 and 750 meters, since benefit of being away from forest edge was considered minimal by this point.
SQL was further used in the SOCS database to give subjective ratings on a scale from 0 (worst) to 255 (best) for those broad vegetation classes that were not considered as constraints or forest. Ratings are presented in Appendix A. This resulted in a factor image named "habfacb".
SURFACE was used to produce an aspect image from the DEM. Since flat areas are given a value of -1, OVERLAY was used to add this image to another image that had an initial value of 1 for all cells. The FUZZY module was executed on this resulting image with a user defined membership function that gave the highest ratings to flat ground and east to southerly aspects. The assumption behind this factor is that since these birds are at the northern edge of their range, they will prefer warmer slopes. Morning is the most important time for songbird activity and it was noted last summer that nest densities seemed to be higher for east/southesterly aspects in this area.
The MCE module was used with these two constraints and three factors. Factors were considered to be equal in weight and therefore each had a weight of 0.333.
The cattle grazing MCE had one constraint. The same broad vegetation constraint image from the bird MCE was used.
Ratings for the broad vegetation class for cattle were invented (Appendix A). Forests are included in the cattle ratings since the forest of this area is open and often associated with a grass understory.
Habitat suitability was assumed to increase with distance to roads as this would facilitate the collection of cattle. DISTANCE was used on the reversed boolean image of roads created as the bird constraint. This map was then given a FUZZY defined membership function that was linearly decreasing with control points at 150m and the maximum distance from roads: 3414 meters.
SQL and ASSIGN was used on the river database to create a boolean image of lakes and permanent streams. It was assumed that cattle habitat suitability would increase with distance to drinking water. The DISTANCE module was applied to this image. The resulting map was given a FUZZY monotonically decreasing sigmoidal membership function with control points at 100m and the maximum value of 2341m.
SURFACE was used to produce a slope image from the DEM. It was assumed that steep slopes provided poor habitat for cattle. Slope was given a FUZZY monotonically decreasing sigmoidal membership function with control points at 15 and 35 degrees.
The MCE module was used with the one constraint and the four factors. Factors were considered to be equal in importance and therefore each had a weight of 0.25.
Since only crown land is available for management, the results of the bird and cattle MCEs were multiplied with a boolean image that had crown land = 1. Provinicial Forest was included with crown land in this analysis. An existing ecological reserve "E12" in the study area was also examined to see how much of its area could be considered good bird habitat. A boolean image with land in E12 = 1 was also multiplied with the bird MCE. EXTRACT was used to obtain information on the distribution of MCE values in crown and private land. These numbers were brought into EXCEL and used to obtain charts of amount of area per MCE score.
The bird and cattle MCE images for crown land were each reclassed (using RECLASS) so as to have two categories: land with 0 - 126 MCE score, and land with 127-255 MCE score. These were then crosstabbed (using CROSSTAB) so that conflicts between best bird and best cattle habitat could be observed.
The two bird and cattle MCE images were then ranked in descending order so that the multi objective land allocation (Mola) module could be used. A hypothetical example was considered where Mola was used to find the best 1,000 ha of bird habitat and best 6,000 ha of cattle grazing land. The bird habitat objective was given a higher weight (0.6) than the cattle grazing objective (0.4).
The EXTRACT module was used to process the bird and cattle MCE images with the Mola image as the feature definition image. This was to see what the maximum, minimum, average and standard deviation of the MCE values were for the two categories (bird and cattle) in the final Mola image.
Broad vegetation classes for the study area were assigned numbers so that an image could be produced (Figure 1). Tenure classes for the study area are also displayed (Figure 2).
Execution of the Multi Criteria Evaluation module resulted in a final MCE image for shrub-steppe bird habitat (Figure 3), and a second image for best cattle grazing habitat (Figure 4). It is obvious from comparing these two images that much more of the land in the study area is considered to be highly suitable for grazing. According to the assumptions of these evaluations, shrub-steppe birds are far more limited in their available habitat.
The relative amounts of suitable habitat for private and crown land was compared. A boolean image for crown land was overlayed with the bird habitat MCE (Figure 5). Figure 6 gives the same information for the cattle grazing MCE. Figure 7 shows the MCE scores for ecological reserve E12.
Plots of the amount of land per class of MCE score are shown in Charts 1-5. Distribution of MCE scores was found to be similar for private and crown land, indicating that there is a substantial amount of land available to manage. Again it is clear that much more of the land in the cow MCE has a high MCE score. Total number of hectares for each MCE within private and crown land were summed for MCE scores between 200 and 255 (Table 1). The E12 ecological reserve has a relatively high proportion of good bird habitat (45%).
| Total ha (MCE = 200-255) | Total ha (MCE = 1-255) | |
| Bird/Crown land | 1504 |
7802 |
| Bird/Private land | 1764 |
9088 |
| Cattle/Crown | 4462 |
8302 |
| Cattle/Private | 7705 |
9894 |
| Eco. Reserve E12 | 127 |
306 |
The CROSSTAB of the images with bird and cattle MCE scores between 127 and 255 showed that most of the bird habitat fell into land that was also considered good grazing habitat (Figure 8). This is not surprising since most of the land was rated as good cattle habitat. For the CROSSTAB, 4214 ha were in conflict, 3625 ha were good cow habitat, and 103 ha were good bird habitat. The MOLA module produce an image that reconciled a hypothetical decision to allocate 6,000 ha to grazing and 1,000 ha to shrub-steppe bird preservation (Figure 9). A relatively large area of grazing habitat was specified since so much more of the land was good cattle habitat.
The performance of the Mola was evaluated by viewing the extracted statistics for each MCE image. The average bird MCE score for land allocated to bird habitat is higher (167) than the for land allocated to cattle (115) (Table 2). Likewise, the average cattle MCE score is higher (228) for land allocated to cattle grazing than for land allocated to bird habitat (210) (Table 3). This difference is not as great, presumably because so much of the land was rated highly for cattle grazing. Obviously some compromise of optimal bird habitat has occured. The bird MCE average score is 167 for the 1 000 ha specified for the Mola, but there were 1 504 ha with an MCE score greater than 200 (Table 1).
Bird Habitat |
Cattle Habitat |
|
| Minimum | 6 |
0 |
| Maximum | 236 |
236 |
| Average | 167 |
115 |
| Standard dev. | 36.11 |
43.48 |
Bird Habitat |
Cattle Habitat |
|
| Minimum | 80 |
206 |
| Maximum | 255 |
255 |
| Average | 210 |
228 |
| Standard dev. | 29.6 |
12.8 |
Another sagebrush scene, photo by Kristen Dust
The resulting habitat suitability model correlates well with
my personal observations of shrub-steppe nesting birds in the
study area last summer. However, although this analysis suceeded
in producing some nice maps, the results should be viewed with
caution. The SOCS data itself is still under development and its
level of accuracy is unknown. All of my assumptions were
completely subjective and based on one summer's worth of
experience in the Okanagan. It could be argued, for example, that
shrub-steppe birds may benefit from nesting close to riparian or
forested areas if these habitat are shown to provide good
foraging grounds. This MCE could be improved considerably with
more knowledge on what habitat variables are the most important
to nesting birds. This method should be applied with caution in
the case of managing wildlife, since bogus assumptions could lead
to poor management decisions. Models should be carefully
constructed and groundtruthed to give the best possible
representation of reality.
The cattle grazing MCE indicates that most of the land in the study area will support cattle. This is not surprising since there are cattle or signs of cattle all over this area. The cattle MCE failed to give as great a distinction between suitability levels of the area as the bird MCE did. This is probably because cattle are more flexible in the habitat they can use. This MCE could have been improved with more knowledge about actual forage quality for the different broad vegetation types. Forest classes should probably have been rated as lower than they were. Incorporating cost surfaces (using slope, for example) into the distance to road and water factors may have given more distinction between good and poor habitat.
The distributions of bird and cattle habitat MCE scores were similar for private and crown land. It is useful to know how much crown land with good bird habitat is available, since crown land can be managed. A further value of using MCE is that it allows managers to assess available bird habitat on private land. Although this land can not be managed, it is interesting to see how much habitat is available. In this area, there is a lot of good bird habitat on private land, most of which is owned by cattle ranchers. These ranches also depend on their tenures of adjacent crown grazing land to sustain their herds. While overgrazing might possibly be shown to have negative effects on bird populations, bird populations would be more seriously impacted if cattle ranchers ceased to be able to make a profit and sold all their private land for condos and golf courses. It is vital to the sustained protection of this area that land managers and ranchers cooperate. Multi Objective Land Allocation in GIS provides a means for groups with conflicting interests to find a compromise for land use. In this analysis there was definately a few scraps of crown land that would be suitable areas to preserve to sustain bird habitat. However, cattle ranching and bird habitat are not mutually exclusive. In retrospect a better study might have been to find a compromise between land allocation for bird habitat, and a building development which would definately wipe out habitat.
The use of Mola in this project may not have been especially informative since birds were so much more limited in their habitat, and plenty of good cattle grazing habitat was available. It did, however, serve to indicate where exactly the best 1 000 ha of bird habitat was, taking the need for grazing land into consideration. The average bird habitat MCE score was higher the areas identified as bird habitat by the Mola. It would have been interesting to compare this Mola with a RECLASS of the best 1,000 ha of bird habitat, unfortunately whenever I tried to RECLASS this file Idrisi always crashed except when I tried the equal interval RECLASS. It is evident that the 1,000 ha in the Mola does not represent the very best land, since there were 1 504 ha with an MCE score greater than 200.
GIS has many many potential applications in assessing habitat suitability of species for management purposes. It is important to remember that the Multi Criteria Evaluations are just models with a lot of unseen error associated with them, and should not be taken as fact. Vastly different images could have been produced with a different reasonable set of assumptions. If care is taken to produce MCEs for different objectives that are as accurate as possible, Multi Objective Land Allocation can be applied to resolve conflicts between differing parties. Compromises can be acheived where limited areas of land may be set aside for habitat preservation. Knowledge of this technology could help guide management decisions.
Photos for this page were taken by Kristin Dust and Nancy Mahony, and used with permission. Thank you!
Lea, E. C, Maxwell, R. E. and W. L. Harper. 1991. Biophysical Habitat Units of the South Okanagan Study Area. Wildlife Working Report WR-XX. Wildlife Branch, Ministry of Environment.
Dobler, F. C., J. Eby, C. Perry, S. Richardson & M. Vander Hagen. 1996. Status of Washington's shrub-steppe ecosystem: extent, ownership, and wildlife/vegetation relationships. Research Report. Wash. Dept. Fish and Wildlife, Olympia. 39pp.
Harvey, D. H. 1992. The distribution, density and habitat of Brewer's sparrows Spizella Breweri. Unpublished report, Simon Fraser University.