One key limitation of this analysis was the use of CIMP5 climate projection data, which has since been rendered obsolete by CIMP6 (CIMP Phase 6, 2020). However, the use of CIMP6 data, which replaces the Registered Concentration Pathway model with the more flexible and holistic Share Socioeconomic Pathway model, was not an option as CIMP6 data was not readily available in 30 arc second resolution. The Maxent software requires that all environmental variable layers be inputted in the same projection and resolution. As such, we were unable to use a coarser resolution when the current distribution model was already completed using 30 arc second resolution rasters.

​

Another limitation present in our study was the lack of land use type variables in the Maxent models. As we were unable to source land use rasters at acceptable resolution scales, the Maxent was constructed using only climate variables from WorldClim. For example, Kalboussi and Achour (2018) found in their Maxent analysis of snake species that the most important variable was “Distance to streams”, which accounted for nearly 60% of prediction importance. As a result, there are areas where the output rasters show habitable zones in what appear to be bodies of water. While rough skin newts far prefer wetlands and certain individuals may spend the entirety of their lives underwater (Marks and Doyle, 2020), they tend to dwell only in very shallow areas, and would normally only be found near the edge of streams, ponds, and lakes. Furthermore, our variables do not account for potential sea level rise, which may eliminate many potential habitat cells along low-lying coastlines.

​

A third limitation lies within our sample data. Because it is volunteer contributed citizen science data, newt samples from BISON are subject to high degrees of sampling bias due to certain locations being more popular for people and more frequently visited, leading to higher observation counts. This leads to skewed presence data where presence counts are higher not necessarily because newts are more likely to be found in an area, but rather because more observers are there to record them. There is also the distinct possibility that the same individuals have been counted multiple times by different observers, which inflates count values. This in turn leads to flaws in the resulting Maxent model (Kalboussi and Achour, 2018). One potential way to rectify this would have been to filter out repeat samples found in the same area. However, this would have been difficult to do accurately, as different samples had differing levels of specificity and significance with regards to coordinate georeferences. Furthermore, imperfect spatial filtering would run the risk of skewing data even more. As a consequence of being volunteer contributed data, the Taricha granulosa observations may have been impacted by observers mistakenly recording sightings of other similar Taricha newts that occupy somewhat different habitat niches in overlapping geographical ranges.

​

​With regards to our land cover analysis, the only covered Washington and Oregon due to a lack of data availability. BC land cover data sourced from DataBC was not as precise and lacked the land cover classification details present in the USGS data, so we only focused on the Washington and Oregon states. As such, it is unknown to what extent human land use will impact newt habitat niches in Canada.