1.Tick and Deer data
Tick and deer data came in the form of a CSV table, but with XY coordinates. Thus, the points were georeferenced and represented spatially.
Deer data was originally obtained in the form of point data. To include it in the MaxEnt model, it needed to be converted to a raster format. Using the Kernel Density tool, the deer point data was converted to a kernel density format. All parameters were left to default.
2. Reprojection and resampling
All climate data was not reprojected and left in their original WGS 1984 coordinate system. Again, previous iterations of this analysis found that modifying the climate data in any way overloaded the program and caused it to run out of memory and crash. For ease of completion of this project, it was left unprojected.
The rest of the data was reprojected to NAD 1983 Canada Atlas Lambert projection. The Lambert conformal conic map projection is widely used for general maps of Canada at small scales and is the most common map projection used at Statistics Canada. The Lambert conformal conic projection provides good directional and shape relationships for mid-latitude regions having a mainly east-to-west extent (Statistics Canada).
Polygon data (soil order data) was reprojected using the Reproject tool and raster data was reprojected using the Reproject Raster tool.
Soil and Kernel Density data were both resampled to a resolution of 3000m in order to obtain a finer resolution but still coarse enough not to overload the program and extend processing times. Land cover data already came at a resolution of 30m and so was left at this resolution.
3. No specific study area outline was defined for this analysis as previous iterations of this analysis found that it completely overloaded the program and did not run. The general areas of focus were Ontario, Quebec and Nova Scotia (with the addition of New Brunswick which happened to be in the same area).
4. The MaxEnt Program (Model 1)
ArcGIS’s MaxEnt software: Presence-only Prediction (MaxEnt) (Spatial Statistics) models the presence of a phenomenon (in this case, ticks) given known presence locations of ticks and explanatory variables (climate, land cover, soil, host and data) using a maximum entropy approach (MaxEnt). The principle of maximum entropy states that the probability distribution which best represents the current state of knowledge about a system is the one with largest entropy (this will be the system with the largest remaining uncertainty), in the context of precisely stated prior data. The tool provides output features and rasters that include the probability of presence and can be applied to problems in which only presence is known and absence is not known (ESRI).
Parameters:
Input: Tick data (projected)
Explanatory variables:
- All climate data (23 variables unprojected)
- Deer kernel density data (projected)
- Soil data (projected, specified as categorical data)
- Landcover data (projected, specified as categorical data)
I did NOT check “Contains background points” as I did not have data for this.
Explanatory Variable expansions used:
- Original (Linear)
- Squared (Quadratic) – Transforms each explanatory variable value by squaring it, resulting in a quadratic relationship between the explanatory variable and the presence response. In some domains, such as species distribution, species’ responses to environmental conditions are often nonlinear and unimodal, and a quadratic form may best represent the relationships (ESRI).
- Pairwise interaction (Product) – Performs a pairwise multiplication on explanatory variables. These transformed variables are commonly known as interaction terms and may be useful representations of complex relationships that depend on conditions among multiple variables (ESRI).
- Smoothed Step (Hinge) – Converts the continuous explanatory variable into two segments, a static segment (all zeros or ones) and a linear function (increasing or decreasing), separated by a threshold called a knot. This can be performed using a forward hinge (start with zeros between the minimum and the knot, and then apply an increasing linear function between the knot and the maximum) or a reverse hinge (ESRI).
Number of knots was left at the default of 10.
The Study Area was set to the default Convex Hull . The convex hull of a set of points is defined as the smallest convex polygon, that encloses all of the points in the set (Sommer, 2016). Spatial thinning was applied in order to equalize the spread of data and weightings, with a minimum nearest neighbour distance set to 10 km.
Advanced model options were left as default.
Training Outputs were: Trained features, a Trained Raster, a Response Curve Table and a Sensitivity Table.
Prediction outputs were specified as an Output Prediction Raster and data was automatically matched. Predictions outside of Data ranges was checked as allowed.
Under Validation Options, a Random resampling scheme was set with 10 groups for cross validation.
5. Models 2 and 3
The parameters were set for Models 2 and 3, however: Model 1 used only current climate data (no deer, soil or land cover data). Model 2 used only current climate data for the training inputs and then used future climate data as the predicted raster inputs.
6. Tables and maps were outputted from all 3 models
7. Park layers for Quebec, Ontario, New Brunswick and Nova Scotia were then overlaid onto the predicted raster outputs for Models 1, 2 and 3.