Project Team

John Gallo
office: (805) 893-8652
home: (805) 971-6052
email

Angela Barale
home: (805) 968-3934
email

Meghan McCord
home: (805) 685-1227
email

Jeff Scott
home: (805) 571-7422
email

Part 1:

Summary

Expanse of South Coast Region
Location Map
Species Background

Ecological Principals
Envisioning a Protected and Connected Future

GIS Application

Objectives

Reference to Related Works

Part 2:
Data Collection, Database Design/Management, Methodology
Flowchart

Suitability of habitat type for reproduction (WHR Model Processs)
GAP data (habitat suitability)
Table 1: Wildlife Habitat Values of GAP Polygons
Habitat Quality Layer

Digital Line Graph: Roads
The Road Impact Layer
Table 2: Roads Weighting
Identifying Source Areas

National Land Cover Dataset (NLCD, EPA MRLC 2000)

Implementation, Hardware/Software

Analysis/Discussion

Outlook/Conclusion

Bibliography

Figures 1-5

Appendix

Related Links

Part 1:
Summary
    This project will help identify regional land conservation priorities by examining the local mountain lion population.  The predicted range of the mountain lions will be determined by a Geographic Information System (GIS) analysis using habitat data and roads data. The landscape linkages between core mountain lion populations will be predicted based on habitat quality and barriers to movement, namely roads and cities.  Weights based on research and collected data will be set accordingly.  A habitat layer based on the Wildlife Habitat Relationships (WHR) model will be assigned a value from 1 to 10, with 1 representing a low cost for movement.  Similarly, a roads impact layer based on the quality and type of roads in an area will be weighted using the same assignment.  Based on quantity of forest cover in an area, the dataset will be re-sampled to 200m, reclassified as a binary grid and smoothed using a focal mean operation to account for edge effects and miscalculation.  Here, a range from 1 to 6 will typify the habitat, with 1 representing the most suitable or densely covered area. Ultimately, a grid where the value of connected cells is the “cost” for a mountain lion will be generated by an overlay operation. At this point, the “least cost path” between two points can be determined. The end result will be a component of Conception Coast’s Project’s “Conception Coast Wildlands Network” that identifies a proactive conservation plan for the region based on ecological requirements.

Expanse of South Coast Region
    The study area is centered about the mountain lion metapopulation that resides in the ecologically rich transition zone between the south and central coast ecoregions and connects the lion metapopulations of these regions to each other and those of the Sierra Nevada Mountains.  Thus, the study area is crucial to California’s mountain lion population and will include Santa Barbara and Ventura Counties, the Sierra Foothills of Kern County and the coastal areas of San Luis Obispo.

Species Background
    Occupying habitats from the Yukon to the Strait of Magellan and from the Atlantic to the Pacific, mountain lions were at one time known to range throughout the largest geographic area of any land mammal in the Western hemisphere.  Varying inversely with human population, this unique species has been imperiled by drastic changes in land use.
    Human population impacts such as roads, destruction of forest cover, and poaching threaten the livelihood of human interaction threaten the livelihood of the California Mountain Lion.  Felis concolor californica has evolved as a ‘solitary-living predator (an animal dependent upon killing and eating other animals, prey, for its own survival). Backed by a history of changing ideals and legal status, monetary incentives were offered for slaughter of this “bountied predator” between 1907 and 1963.  In 1969 popularity in gaming and a lack of understanding and appreciation generated the term “game mammal” for the carnivore population.  As public awareness and environmental research began to facilitate an ‘eco-centric’ understanding, the mountain lion became known as a “special protected mammal,” generating appreciation and public awareness due to the increasing amount of human encroachment.

Ecological Principals
Envisioning a protected and connected future
    As human habitat forcefully dominates control, increasingly diminishing the habitat available for mountain lion livelihood, a need for “minor changes in management of multi-use and commodity production (of) lands can help conserve biodiversity” (Hunter 2000).  Related work in progress stresses a need for rewilding by examining wilderness, connectivity and viability in order to create an ‘umbrella of protection.’ “Rewilding arises from the need to look at biodiversity conservation from a broader perspective to incorporate the basic processes that promote ecological integrity” (Hunter 2000). It is therefore important for “scientists who study top predators like wolves, mountain lion and bobcat” to understand the critical roles these large carnivores play in the ecological systems.
    The mountain lion is a keystone species that maintains an ecological balance among lower tropic levels, and is an umbrella species such that attaining it’s landscape requirements also attains the requirements of myriad other species vital to ecological integrity.  “The scientists have shown that these large carnivore species are needed to provide balance for interactions between different levels of the food web.  These species require very large territories that are connected to provide for unimpeded movement between them” (Hunter 2000).  Related works include analysis such as those done by The South Coast Wildlands Project.  Using the ‘Three C’s- cores, connectivity and carnivores to connect and protect the future, various approaches to the issue of fragmentation have been provided. By recognizing that “habitat loss and fragmentation are the root cause of increased levels of inbreeding in small, isolated populations” (Pray 1999), the mountain lion is an ideal surrogate for determining viable landscape connectivity for a region. “Moreover, maintaining corridors that would allow for both patterns of movement may be critical for the conservation of these large felids. Finally, extensive overlap in the distribution of mountain lions, especially the association of one group of individuals on winter range and another on summer range for mountain lions with disjunctive distributions, indicates a more flexible social system than previously described” (Pray 1999).

GIS Application
    Proper assessment and implementation of plans is a fundamental necessity, which concerns scientists.  Analysis was performed with a GIS because it is useful to overlay several different categories of land use and land cover.  Analysis conducted by modeling mountain lion habitat affinity as well as dispersed routes were facilitated and data greatly enhanced by GIS application.

Objectives
Goal: Identify a landscape that, if preserved, will maintain connectivity between mountain lion populations.
    · Identify cost based on habitat value
    · Identify cost based on road values
    · Identify cost based on forest cover
    · Overlay roads and habitat to identify cost grid
    · Identify least cost path between mountain lion and source populations.

Reference to Related Works
    This work is related to many other similar projects continent wide that are using large carnivores as an indicator of ecological integrity.  “Continental Conservation” edited by Michael Soule and Jim Terbourgh (1999) examines many of these studies and others, providing strong evidence for the importance of landscape linkages.  Furthermore, the California Wilderness Coalition, in conjunction with the Wildlands Project, has performed a similar analysis for the South Coast Ecoregion.  Ventana Wildlands Project (VWP 2001) has similar performed a similar study for the Central Coast Ecoregion.  Finally, CWC and Nature Conservancy held a “Missing Linkages” conference in which the regional experts gathered to discuss and map the wildlife linkages of the state. The landscape linkages of the region were predicted based primarily on expert opinion.  A consensus called for more data and analysis.

Part 2:
Data Collection, Database Design/Management, Methodology
    Using a simple habitat model, a dictionary structure was set up for potential habitat to then be developed by overlaying a collection of three data layers.  Classified data types include the GAP Database, which was combined with the WHR to produce a mosaic suitable habitat layer.  Representing a “cost surface, ” a digital line graph from the USGS serves as a roads impact layer.  Costs were assigned to these roads based upon research assessing mountain lion preference of travel over different types of roads. The final dataset, The National Land Cover Dataset (NLCD) was used to generate a grid classifying where dense cover, thus displaying areas where mountain lions are most
likely to hang out.  Validation of the spatial data through analysis functions was the next step in the operation.  Most importantly, we needed to be critical of the spatial data while integrating it with attribute data and performing overlay operations.

GAP Database/WHR
The California Gap Analysis Project (1998) published a CD-ROM, digitally providing us with an entire GIS database, final report and an interactive atlas of California’s biodiversity.  Identifying Including information and maps describing California’s distribution of wildlife, major plant species, plant communities and wildlife habitats and their relationship to landownership patterns and designated areas managed for biodiversity conservation, the GAP database served as an atlas and reference for suitability ratings.

Wildlife Habitat Relationship (WHR) Model Process:
    The existing California Wildlife-Habitat Relationships (CWHR) database was used to assign suitability ratings of terrestrial species to habitat types.  The actual model for mountain lion is very complex.  About fifty different habitats are evaluated and for each habitat type, there are up to fifteen sub-classes, based on the average size of the individual plants, as well as their density within the habitat.  The importance of each one of these sub-classes to mountain lions is then determined.  CWHR ranks each habitat sub-type as high, medium, low, or unsuitable for breeding, feeding, and cover.  These rankings are based on the known habitat preferences of the species in question.  This database was compiled and revised by an interagency team of wildlife biologists to contain all available information on habitat requirements for terrestrial vertebrates (Davis et.al. 1998).
    There are many different ways that WHR could be combined with GAP database to produce a mosaic of mountain lion habitat quality across the landscape.  Because the habitat polygons are very large, they often contain several different habitats.  The researchers of the UCSB Biogeography Lab performed an analysis to combine these two databases, and they provided the output on the GAP CD.  Habitat suitability was determined by assigning the three most prevalent major CWHR vegetation types to each land cover polygon.  The aerial extent of the suitability rank in every habitat polygon is summed, accounting for the relative proportion of the primary, secondary, and tertiary habitat types present.  Thus a habitat polygon might be modeled as having 30% High suitability habitat, 10% as Medium, 20% as Low and the remaining 40% as unsuitable (Stoms 2001).  To visualize these complex, multivariate data, the output was categorized into a combination of area and suitability to give a single class per polygon that could be displayed and analyzed further (Table 1). Thus a polygon that contained a large proportion of the best habitat could be distinguished from one with only a small amount or one where an arbitrary threshold for habitat suitability is used to include or exclude polygons.  The categories are assigned in descending order, so that a polygon is assigned to category 4 only if it does not also qualify in category 5 (Davis et. al. 1998).
    For mountain lions, the above GAP/WHR combination was performed using just mountain lion breeding habitat suitability; feeding and cover suitability were ignored (Parisi, 2001).  If another iteration of this study is performed, it may be more exact to use cover as the primary feature, because this study is primarily about habitat connectivity.  Thus, vital corridors may exist that have low reproductive potential but do provide suitable cover.

Table 1: Wildlife Habitat Values of GAP Polygons
 

The Habitat Quality Layer
    The first step in the Habitat Quality Layer was to generate a coverage using the “Convert Interchange file to Coverage” tool.  Next, we needed to make a study area clipping coverage, or “cookie cutter” in order to clip the above coverage to the study area.  For this, we used the county1 coverage from the Biogeography CD, which also needed to be converted from an interchange file, again using the “Convert Interchange file to Coverage” tool.  After we had the file in the form of a coverage, it was then brought into ArcView and opened in the attribute table. Sorted by “countyname” in order to select the 5 counties (San Luis Obispo, Santa Barbara, Kern, Ventura and Los Angeles) that make up the study area, we then used the “convert to shapefile” feature in the Theme menu to make a new layer, which contained only the 5 counties.  Next, we used the “Convert shapefile to coverage” tool in ArcToolbox to make a coverage so that we could use it to clip the land cover coverage.  Next, we used the “Clip” tool in ArcToolbox to clip the Landcover coverage using Aicookiecut as the clip coverage.  The resulting coverage was named Landcov1cc.  (See appendix 1, data dictionary)
    The next step was to make whr habitat classification for mountain lions and cost a part of the attribute table for Landcov1cc.  The CD contains text files for 455 different vertebrate species which each have a list of every polygon ID for the landcover layer and the corresponding habitat value for each particular polygon for each animal.  A list of the codes for each animal can be found at UCSB’s biogeography website.  The code for mountain lion is “m165.”  Once you have the code for the animal you want, you can go onto the CD in the Vertspec folder, which is in the Data folder on the CD, and select the text file for the animal you want.  This file can be brought into Excel as a table and then joined to the PAT for the landcover layer using the Polygon-ID attribute.  The first step for this is to get the delimited text file from the CD into Excel.  The “m165” text file on the CD contained polygon_ID numbers for Landcov1cc and the corresponding WHR mountain lion habitat rank of 0-5 for each polygon.  The text file was then brought into Excel making sure to specify that it was a text file delimited by commas.  After the table was in Excel, we saved it as a DBASE IV file into the working directory.  We then went into ArcCatalog and “right-clicked” on the DBASE IV file and selected the “Convert DBASE to INFO” option.  This resulted in an INFO table, which could be joined to the PAT for Landcov1cc.  We then used the “join” tool in ArcToolbox to make the WHR habitat classifications a permanent part of the PAT.  We used the PAT as BOTH the input and output tables, the m165 table as the join table and Poly-ID as the join attribute.
    Next, we needed the cost of each habitat rank to be a part of the PAT.  In Excel a new look-up table (LUT) was made, which contained the 6 habitat ranks (0-5) and the corresponding cost for each rank.  The information for this LUT was obtained from the Ventana Wildlands Project Report (VWP 2000).  The following are the costs associated with each WHR habitat rank:

    Once again, the “join” tool in ArcToolbox was used to make the cost attribute a permanent part of the PAT for Landcov1cc.  We used the PAT as both the input and output table, the new LUT as the join table and “m165” (the WHR habitat value) as the join attribute.  The next step was to rasterize the layer based on cost.  To do this, we used the “Convert Polygon Coverage to Grid” tool in ArcToolbox.  We used the clipped landcover layer as the input coverage, the costs as the value item, no Look-up table, no weight table and a cell size of 200 (meters).  This resulted in a raster where each cell had a “cost” between 1 and 10 based on the quality of mountain lion habitat found in that cell.

Digital Line Graph: Roads
The Road Impact Layer
    The road impact layer is significant part of the final analysis.  This layer represents a “cost surface” where the value of each cell reflects the length and type of roads within a search radius of the cell.  Obtaining a USGS Roads DLG, which had been previously clipped to the study area, specifications were kept in consideration and documented in a data dictionary.  For more information on this layer, please see the data dictionary.
    The class attribute was assigned by the USGS to every road segment, reflecting the type of roads, ranging from dirt trail to divided highway.  Previous work conducted by the VWP assigned a series of “costs” ranging from 1-10 for individual road classes to be used in a least cost analysis.  Based on our research, a decision was made to run one analysis using the VWP classification and another using our own scheme.
    Research concerning mountain lion interaction with different types or intensities of roads affected the weights used in determining likeliness of occurrence for the least cost path analysis.   “Track surveys on dirt roads are convenient because mountain lions often travel along them for long distances (simple road crossings are rare), and many of their tracks remain visible despite vehicle traffic” (Smallwood 1995) Thus, we revised the weighting used by VWP to give cost value of 0 instead of 1.  Table 2 displays the USGS road classes along with their descriptions and cost in the VWP scheme and our scheme.

Table 2: Roads Weighting
 

    The first step in processing the roads layer was to convert the interchange file that we received to a polyline coverage.  Next, we needed to make the costs a permanent part of the AAT of the roads coverage.  To do this, we first created the table above.  Opening a new Excel document, a LUT was generated containing the USGS road classes and the VWP cost as well as our cost for all of the classes.  We then saved this table as a DBASE IV file in which we later converted to an INFO table (see description of this process above).  To join the tables, we used the “Join” tool in ArcToolbox with the roads coverage AAT as the input and output tables, our LUT as the join table and “class” as the join attribute.
    After joining the tables, the LINEDENSITY function in ArcInfo Workstation was used to create a grid out of the roads layer.  The reason that Workstation was used instead of ArcInfo 8 was that, no matter what we tried, running Density function in ArcInfo 8 always resulted in a grid where all of the cells had a value of zero.  The LINEDENSITY function converts a line coverage into a grid and assigns each cell a value depending on the number and attributes of lines around it within a given search radius.  When we ran the function, we used the cost as the attribute that we wanted the output grid based on (population field), a search radius of 500m, and an output cell size of 200m.  We chose a kernel density because it weighs lines that are closer to the center of the search radius more heavily.  We used a scaling factor of 125, which divided all of the cell values by 125 because we wanted the values in our output grid to be between 1 and 10 so that they could have the right amount of weight in the final analysis.  The LINEDENSITY function was the final step in processing the roads layer because it rasterized and smoothed the layer all in one step.

Identifying Source Areas
    We needed to identify source areas where mountain lions lived in order to study them.  We defined a source area as any area with habitat quality of 5 on the WHR scale that was over 100,000 acres.  The first step was to dissolve the landcover polygons in our clipped landcover layer by habitat quality.  To do this, we used the dissolve tool in ArcToolbox with our clipped landcover layer as the input coverage, polygon as the feature class and habitat quality as the dissolve attribute.  Next, we brought the resulting dissolved coverage into ArcView and selected by attribute (under the Theme menu) all of the polygons which had an area greater than 100,000 acres and a habitat quality of 5. Next, we used the convert to shapefile option in the Theme Menu to make a new layer out of our selection.  These shapefiles were originally meant to be used in the analysis but they covered far too much of the study area so smaller areas were selected ad hoc for the analysis.

National Land Cover Dataset (NLCD, EPA MRLC 2000)
    The National Land Cover Database (NLCD) is a dataset concerning landscape classifications.  Created in a partnership by the EPA and USGS, the NLCD is a 30meter grid taken from Landsat thematic Mapper (TM) imagery.  Due to the coarse resolution of the input data to create the cost surfaces, discrepancies lie in the possibility of identified potential links sufficing as high quality core habitats.   This high resolution NLCD data acts as a good balance to the coarse GAP data.
    We obtained the NLCD image from Ethan Inlander at Conception Coast Project in a zipped ‘*.gz’ format.  We unzipped it using WinZip and changed the resulting ‘.bin’ file extension to ‘.bil’ (band interleaved by line) so that Arc/Info and ArcView could recognize the image.  An ASCII text header (.hdr) file must be created because both Arc/Info and ArcView require a header file in order to recognize, ingest, and display the image.  Instructions for header file construction should be included with the associated readme file.  The header file must be located in the same directory as the ‘.bil’ file.  We were provided with the proper header file.
    Using ArcView with Spatial Analyst, we converted the .bil file to a grid.  We then used ArcToolbox to project the new NLCD grid to our other datasets.  (Initially, the new grid image has no defined projection, so it is necessary to define the projection according to the projection information contained in the associated readme file.  This step is necessary for successfully changing projections using ArcToolbox.)  During this step we had the option of resampling the cell size, and we did so to 200m.
    Using Arc/Info 7.x we made a binary grid of Forest/Non-Forest by reclassifying all values to 1 and 6 (1 = Forest, 6 = Non-Forest).  Again using Arc/Info 7.x we smoothed the grid using a 400m circular filter to account for edge effects and misclassification.  The result was a floating-point grid, which means the values were scaled from 1-6 as the range of forest/non-forest distinction because this would refine the existing low-high quality habitat costs. All steps were done in accordance with VWP’s procedures.

Implementation, hardware/software
    In order to design a GIS database suitable for analysis, the correct software for the task needed to be appropriately identified. Primarily ArcGIS, Arc 8.0 on the Windows NT machines of the Star lab in UCSB’s Ellison Hall was used the for processing and data display.  Technical problems such as the line density function were taken to ARC/INFO 7 and permanent join carried out in Arc Toolbox as opposed to ArcMap.  ArcView 3.2 was used in converting a shapefile to a grid with full extent, and the least cost path analysis was run using a single “to” area.

Analysis/Discussion
    The purpose of the final analysis was to make a grid where the value of cells created the “cost” for a mountain lion.  By bringing the refreshed layers of habitat, road impact, and forest cover rasters into ArcMap, the calculator in Spatial Analyst could then be used to overlay the three layers.   A habitat layer based on the WHR model was assigned a value from 1 to 10, with 1 representing a low cost for movement (Figure 1).  Similarly, a roads impact layer based on the quality and type of roads in an area was weighted using the same assignment (Figure 2).  Based on quantity of forest cover in an area, the dataset was then re-sampled to 200m, reclassified as a binary grid and smoothed using a focal mean operation to account for edge effects and miscalculation (Figure 3).  Here, a range from 1 to 6 typified the habitat, with 1 representing the most suitable or densely covered area.
    After the overlay was performed, one “to” area and several source areas were identified.  Running the cost weighted analysis in Spatial Analyst, the CostDistance and CostDirection were obtained. By summing the roads, habitat and non-forested cover cost layers, a movement “cost” was generated to indicate the level of danger a mountain lion might face by taking a given path (Figure 4). Finally, the shortest path analysis was run to reveal a high quality habitat with a low cost for movement. Ultimately, a grid where the value of connected cells is the “cost” for a mountain lion was generated by an overlay operation, completing the shortest path analysis (Figures 5).

Outlook/Conclusion
    Technical development of GIS is an important tool in creating a new understanding of ecological awareness.  However, in order to realize the potential of data and reach a high accuracy of modeling, additional datasets with more rigorous parameters to identify source areas could be used.  In the future, examining the differences between using GIS based analysis and recommendations of experts in a workshop setting could potentially be an interesting comparison to our study.  Furthermore, additional analysis could include the following.  Additional land use costs, such as urban and sub-urban development as well as land use information such as ownership and Williamson Act data could provide a more strict analysis by generating areas of higher feasibility and for conservation a “lower” cost.  Another approach could involve creating a database of mountain lion sightings and signs; cross validating WHR model and “source populations.”  Distinguishing the difference between WHR using reproduction as opposed to ‘cover’ values, in addition to data describing how well mountain lions actually travel through chaparral weighting in the forest layer, would be interesting introductions.   Through analysis of data and a better vision of how to sustain ecological integrity through conservation priorities such as habitat connectivity, we look forward to a proactive conservation plan.

Bibliography

Davis, F.W., D.M. Stoms, A.D. Hollander, K.A. Thomas, P.A. Stine, D. Odion, M.I. Borchert, J.H. Thorne, M.V.  Gray,
        R.E. Walker, K. Warner, and J. Graae. 1998. The California Gap Analysis Project—Final Report. University of
        California, Santa Barbara, CA. http://www.biogeog.ucsb.edu/projects/gap/gap_rep.html.

Harveson, LA; Route, B; Armstrong, F; Silvy, NJ; Tewes, ME. Trends in Populations of Mountain Lion in Carlsbad
        Caverns and Guadalupe Mount National Parks.  Southwestern Naturalist, DEC, 1999, V44 (N4):490-494.

Hunter, Rich, California Wildlands Project: A Vision for Wild California South Coast Regional Report, California
        Wilderness Coalition, Talon Associates, Bodega Bay, April 1999.

Living with California Mountain Lions, Resource Agency California Fish and Game.
        http://www.dgf.ca.gov/lion

Parisi, Monica. CWHR Director, Department of Fish and Game. PersonalCommunication Discussing WHR and WHR/GAP.
        May, 2001.

Pierce BM, Bleich VC, Wehausen JD, Bowyer RT, Migratory patterns of mountain lions: Implications for social
        regulation and conservation, Journal of Mammalogy 80 (3): 986-992 AUG.

Pray, Leslie A., Habitat Lost: Imbreeding Depression and Extinction, Wildearth Carnivore Ecology and Recovery,
        Summer 1999, pages 12-14.

Rocky Mountain Institute Homepage
        http://www.rmi.net/genesee/html/body_mountain_lion_habitat.htm

Smallwood, K. Shawn; Fitzhugh, E. Lee.  A track count for estimating mountain lion Felis concolor californica
        population trend.  Biological Conservation 1995. 71 (3): 251-259.

Smallwood KS. Trends in California Mountain Lion Populations. Southwestern Naturalist, MAR, 1994, V39
        (N1):67-72

Stoms, D. 2001.  Meeting about WHR and GAP database. May 1.

Torres SG; Mansfield  TM;  Foley JE; Lupo T; Brinkhaus A. Mountain Lion and Human Activity in California-Testing
        Speculations.  Wildlife Society Bulletin, FAL, 1996 V24 (N3): 451-460.

University of California Santa Barbara Biogeography
        http://www.biogeog.ucsb.edu/projects/gap/data/meta/vert-meta.html#section9

Ventana Wildlands Project Website
        http://www.greeninfo.org/HTML/clients/cwc_ccoast/draft_cwc_rev2.pdf
 

Wehausen JD. Effects of Mountain Lion Predation on Bighorn Sheep in the Sierra Nevada and Granite Mountains
        of California.  Wildlife Society Bulletin, FAL, 1996, V24 (N3): 471-479.

Wild Earth; Carnivore Ecology and Recovery, V9, number 2, Summer 1999.

Figures 1-5

    Figure 1  Habitat Cost
    Figure 2  Roads Impact
    Figure 3  Forest Cost
    Figure 4  Movement Cost
    Figure 5  Least Cost Path Analysis

Related Links

Conception Coast Project
The Wildlands Project