Translating images into maps…
As many volunteers know, one of the primary purposes of Snapshot Wisconsin is to get a better handle on where species are located throughout the state. We now have enough pictures classified (and a reasonable handle on the effects of error within the classification process) to be able to start the process.
The essence of this exercise is to model associations between 1) locations where animals are and are not found on trail cameras, and 2) environmental or spatial characteristics at those camera locations. As noted in a previous blog post, in order to effectively map predictions about animal distribution, we need to have spatially explicit environmental variables, many of which are derived from satellite imagery. Examples of environmental variables include land cover, seasonal patterns in plant productivity/greenness, snow cover, and the intensity of night-lighting, which is a good index for human activity.
Below are a collection of maps resulting from our first attempt at mapping statewide species distributions based on Snapshot Wisconsin data.
A couple things to note:
- How we think about commonness or rarity depends on the species being considered. For example, “less common” for turkey may mean an area is visited by only a few turkeys over a couple weeks, while “less common” for bear may mean that the area has not had a bear visit for a year or more.
- There are some imperfections. In particular, the tip of the Bayfield peninsula tends to exhibit some patterns that are probably wrong. Mapping will continue to be an iterative process based on our best metrics.
The below maps are accurate enough to be useful, but there’s always room for improvement. We hope that volunteers find this first round of analysis interesting, and maybe even useful for classifying.