Bison Ecology & Management education module - Page 9

Next page or back to Table of Contents

 

Project activity. Part 3 - “Use / Availability” Quantitative Analysis & Model Comparison Activity

MAIN SCIENCE & POLICY ISSUE:

Bison leaving Yellowstone National Park in certain winters & the related Culling Policy.

MAIN RESEARCH QUESTION:

“What type of habitat do bison select as their preferred place to live?”
Or, put in terms that help us better understand why bison leave the Park boundaries in certain winters, “What type of habitat facilitates bison movement?”

OVERVIEW OF METHODS FOR ANSWERING QUESTION:

  1. Navigate to http://ynp.csumb.edu/mapper/mapperV2.htm
  2. Pose hypotheses that address the question “What type of habitat do bison select as their preferred place to live?” Ultimately, this question will be answered for 3 seasons: average winters in Yellowstone, harsh winters in Yellowstone, and summers in Yellowstone
  3. Develop “use / availability” models based on each hypothesis (“use / availability” models are used to compare habitat conditions in areas that have been used by bison – the “use” part – against all areas that could have been used by bison – the “availability” part)
  4. Perform quantitative analysis of each model and display results in graphical format
  5. Compare model results and determine which model is the best “fit”, given the available data (e.g. which model is most likely to explain habitat selection by bison?)

EXAMPLES OF HYPOTHESES (MODELS):

  • H1: In an average winter in Yellowstone, bison prefer to occupy low elevation areas over high elevation areas
  • H2: In an average winter in Yellowstone, bison prefer to occupy areas that are influenced by geothermals over areas that are not influenced by geothermals
  • H3: In an average winter in Yellowstone, bison prefer to occupy areas that have light snow cover over areas that have heavy snow cover
  • H4: In an average winter in Yellowstone, bison prefer to occupy areas that have both light snow cover and are near waterways, over areas that have heavy snow cover and are not near waterways

DETAILED METHODS FOR SINGLE-COVARIATE MODELS:

  1. Navigate to http://ynp.csumb.edu/mapper/mapperV2.htm
  2. List the 9 possible covariates (as identified by our data) on bison habitat selection (E.g. Meadow Cover [MC], Elevation [EL], Slope [SL], Snow Average [SNAVG], Snow Heavy [SNHVY], etc)
  3. Think about these covariates and visually look at the data to determine which covariates (or combination of covariates) you think are most important to bison habitat selection in: an average winter, a harsh winter, and summer
  4. State list of 5 candidate, plausible hypotheses / models associated with one covariate, e.g.
    a. H1: EL “In an average winter in Yellowstone, bison prefer to occupy areas that are at low elevation over areas that are at higher elevation”
    1. H2: GT “In an average winter in Yellowstone, bison prefer to occupy areas that are influenced by geothermals, over areas with no geothermal influence”
  5. Draw a step function for each model listed above that predicts the probability of bison to be located in that habitat
  6. Perform quantitative analysis of each model by following steps listed under “Quantitative Analysis”
  7. Compare model results to determine which model is best (called “best likelihood”), given the available data (e.g. which model is most likely to explain habitat selection by bison?)

QUANTITATIVE ANALYSIS:

  1. Navigate to http://ynp.csumb.edu/mapper/mapperV2.htm
  2. Identify model for analysis (in this example, we will use H1: EL “In an average winter in Yellowstone, bison prefer to occupy areas that are at low elevation over areas that are at high elevation”
  3. Display the selected raster image and associated legend (e.g. elevation) on the Mapper webpage (both left and right images)
  4. Overlay the “RAND 50 Sites” on the left raster image
  5. Overlay the “RAND MONTH YR” on the right raster image (e.g. RAND Feb 93 for an average winter in Yellowstone”) || Click here for notes on data ||
  6. Estimate a quantitative value for each data point in the “RAND 50 Sites” dataset and enter into Excel spreadsheet (use the color bar legend to estimate the quantitative value of that point)
  7. Estimate a quantitative value for each data point in the “RAND MONTH YR” dataset and enter into Excel spreadsheet (use the color bar legend to estimate the quantitative value of that point)
  8. Sort both columns (RAND 50 Sites & RAND MONTH YR) so the values are listed in ascending numerical order (e.g. 0 to highest number)
  9. Choose 4 values to analyze as “cutoff” values (e.g. above what value will you consider elevation to be “high”?)
  10. For each cutoff value, count the number of times the “availability” data fall above that cutoff value (e.g. this is true for elevation, but for Meadow Cover you might reverse the rule and indicate that higher values are better than lower values – b/c higher values mean more Meadow Cover. So, for Meadow Cover you would count the number of times the “availability” data falls below the cutoff value). || Click here for notes on data ||
  11. For each cutoff value, count the number of times the “use” data falls below that cutoff value (e.g. this is true for elevation, but for Meadow Cover you might reverse the rule and indicate that higher values are better than lower values – b/c higher values mean more Meadow Cover. So, for Meadow Cover you would count the number of times the “use” data falls above the cutoff value)
  12. Add the “availability” and “use” counts (from #10 and #11) for each cutoff value (e.g. you should have 4 cutoff values and a total count for each the falls between 0 and 100. We call this total the “likelihood” of the model)
  13. Note – if your “availability” and “use” data do not have the same number of visible data points (e.g. some of the “use” data points are stacked on top of each other, making it impossible to count all 50 of them), then you must normalize your data before adding them up (to make all likelihood models comparable).
  14. Of your 4 selected cutoff values (from #9), which one has the highest total count? This value is the best fit (or best likelihood) for this Covariate model (e.g. for Elevation, which cutoff value gives you the highest count?)
  15. Address your Hypotheses in terms of the model analysis (e.g. H1: EL “In an average winter in Yellowstone, bison prefer to occupy areas that are at low elevation over areas that are at higher elevations.” Through our model analysis, we have found that bison prefer to occupy areas that are no higher than 2300m.)
  16. Note that some analyses may not produce a clear answer (and therefore this Covariate may not be as important to bison as other Covariates)
  17. Repeat steps #2 - #13 for each Hypothesis / Model
  18. Compare your model best fit (also called best likelihood) with the best likelihood of a model based on a different Covariate. Which Covariate is most important?
  19. Summarize your results
 
 

 



29 Nov 2006 Webmaster