Sage Grouse Project
Below I provide the project results from ENVS 691, an advanced GIS course offered by Virginia Commonwealth University. My goals for this project were to take a species, the greater sage grouse (Centrocercus urophasianus), and a region, Montana, that I had no prior knowledge of and develop a habitat suitability model based on data from the public domain. Once the models were complete I would then compare the predicted suitable habitat with current, January 2015, known sage grouse populations in order to test the power of this modeling approach. Below are the simplified results with each step explained in detail. My secondary goal for this project was creating a project where I could apply my skillset to solve an ecological problem outside of my comfort zone, and one which I could then add to my portfolio.
I would like to thank the following organizations who contributed data to this project: Montana State Library, PRISM at Oregon State, Oregon Climate Service, Montana Heritage, and Montana Fish, Wildlife and Parks.
Step One: Pictured to the left is the state of Montana with county boundaries and largest city per county shown. From this point I drop the county boundaries but the cities will remain. Also pictured, shown with orange hash marks, is the core habitat for the sage grouse as defined as the area which contains the highest densities (25% quartile). Using this core habitat I isolated the values for the following variables: Population density (2000), Maximum and Minimum Daily Temperatures from 1971 - 2000, Average Precipitation from 1971 - 2000, LandCover (2000), and Elevation (2000, 1000m x 1000m). In order to make suitable comparisons datasets from 2000 were used, as that was the most recent year which was available across all layers. Cell size was standardized to 1000m2. The minimum and maximum temperature layer represents the average daily high and low temperature for every day, for the thirty year span indicated. Once value ranges for each of these variables was established using the core area, those values were projected state wide and a binary suitability model was created.
Step Two: Pictured to the left is the binary suitability model created with data from the core sage grouse habitat. Areas in green show the portion of the state where all variable thresholds were met, and the area in grey indicates where not all thresholds were met. This is an all or nothing approach, there is no "next best" habitat when using a binary model. A habitat either is or is not suitable according to this model. The individual variable thresholds are given below.
Step Three: Pictured to the left is the binary suitability model created with data from the core sage grouse habitat, as above, but this time with the current known sage grouse populations, in gold hatch. As expected the known area covers the core habitat area, however a good deal of the distribution is in areas that the binary model found unsuitable. Following these results I created a weighted suitability model, which differs from the binary in that it can give you those "next best" habitat areas.
Step Four: Pictured to the left is the weighted suitability model that was created. As the sage grouse was found in almost all of the land cover and precipitation values those variables were dropped from the model. I did run it with all variables and there was virtually no difference as land cover and precipitation had almost no influence. As sage grouses seemed most sensitive to population density (less than 1% of the range qualified as suitable) it was given the highest influence, followed by elevation, minimum temperature and maximum temperature. The suitability values range from not suitable (Red) to Ideal (Darkest Green).
Step Five: Pictured to the left is the weighted suitability model that was created with the current known sage grouse habitats, again in gold hatch. As the map shows, those areas that were deemed unsuitable by the binary model are now given in a gradient of suitability. The right half of the map which was deemed entirely unsuitable by the binary model now shows as halfway between unsuitable and suitable, indicating that the area can probably sustain some level of grouse density (and in fact does according to the known population layer).
Conclusion: The weighted suitability model, based off of the thresholds established by the core sage grouse area, was much more accurate in predicting areas that could host populations. Given I had no background in this species I think the weighted suitability model did a fine job of predicting suitable habitat and conversely the binary model was deemed "unsuitable". I think this approach could be very useful when trying to predict either potential habitats or potential population locations if population distributions are not known.
The full project write-up is available here