Methodology
The first stage of this project consisted of data wrangling, and getting everything in line to work with in R Studio and then in Maxent. All data was downloaded and looked at in ArcGIS pro. The data for Clark’s Nutcracker needed to be formatted correctly (from tab delimited to comma delimited) and the projection had to be changed, as all data needed to be in the same projection (WGS 84). Additionally, only the data in the study area need to be selected (see flowchart). All this was done in ArcGIS (see flowchart) and then columns not required for analysis were deleted in Excel. For the whitebark pine data, the projection needed to be changed, and we had to select for data only within our study area. This was all done in ArcGIS (see flowchart). Columns not needed were then deleted in Excel. The bioclimatic data was also looked at initially in ArcGIS. Then, all layers were made a uniform extent (using coordinates, that roughly approximate study area), projections all made the same, and converted from .tif to .asc in R Studio (Figure1).
After the data was all in order, the second stage of the analysis was necessary to decide what settings to use in Maxent. For this, ENMeval, a package for R was used (Figure 2, code and explanation from Banta (primarily) and Watt, 2018). This step is essential for creating meaningful results. As demonstrated in a case study (Muscarella et al., 2014) there is a major difference between using the default Maxent settings and optimized settings. One important step of this process is choosing a data partitioning method. This choice “...depends on the research objectives and the characteristics of the study system” (Muscarella et al., 2014, p. 1200). One method in particular, the block method, is “...desirable for studies involving model transfer across space or time, including the possibility of encountering non‐analog conditions (e.g. native versus invaded regions, climate change effects)” (Muscarella et al., 2014, p. 1200-1201). Therefore, the block method was chosen. It is worth noting, “[t]he ‘block’ method partitions data into four bins based on the lines of latitude and longitude that divide occurrence localities as equally as possible. The amount of geographic (and environmental) space corresponding to each bin, however, is likely to differ” (Muscarella et al., 2014, p. 1200).
Looking at the results of this ENMeval procedure (Figure 4), we can identify the lowest delta.AICc value, and by extent the settings that should be used for Maxent. For the Clark’s Nutcracker, the settings “LQHP_1.5” were identified (Figure 4). This means the linear, quadratic, hinge, and product features should be used and a regularization multiplier of 1.5 (for smoothing). Additionally, a bias file was also created (Figure 2; Figure 3), because the sighting data was heavily biased in some areas of the range.
After the optimized settings had been decided upon and the bias file was created, the analysis in Maxent was performed. Maxent uses the respective points of the species distribution, the 20 environmental layers (19 bioclimatic and elevation), and the bias file (which is optional).
The .asc layer that Maxent creates to represent the species distribution (Figure 5) was then converted into a .tif format in R Studio, so it could be used in ArcGIS pro.
