I’ve written two previous posts about visualizing iNaturalist data with R. Both of these posts dealt strictly with vector data (coordinate points and polygons). Now, I want to take a moment to show a little of what R can do to rasterize iNaturalist data. The focal species for this post is Argiope aurantia, the Yellow Garden Spider, a large and colorful orb-weaver found across much of the continental United States. For simplicity, I’ve narrowed in on iNaturalist data from Georgia.
Code can be found on my GitHub iNaturalist repository: https://github.com/Deitsch-John/iNaturalist/blob/main/scripts/RasterMaps_blogpost.R
The value of rasterized data is that we can do ~math~ with it. The main goal for today is a simple abundance map of A. aurantia in Georgia. To begin, here is a map of A. aurantia observations. Not surprisingly, most of the observations are from the Atlanta metro region or other parts of the state with greater population density (Athens, Savannah, etc.).
A map of observations of all spiders in Georgia shows a similar distribution.
Two things immediately come to mind.
1) All those points blur together, making it hard to see how many observations there are in certain regions.
2) How do we disentangle A. aurantia abundance from the bias of ‘all the observers observing in the same places’?
Rasterizing the observation data can help us with both of these considerations!
Here is a raster of A. aurantia observations in Georgia. Each grid square is color-coded to reflect the total number of observations within each grid square. This makes it a bit easier to see where most A. aurantia are being reported. The red grid cells are in Atlanta, Athens, Augusta, and Macon – confirming our initial suspicions that most observations are coming from cities.
Next, where are A. aurantia really most common in Georgia? Can iNaturalist data tell us this, or is it doomed by the biases of where observers go? Additionally, iNaturalist observations don’t tell us number of individuals seen at a location, every observation has to be treated like a ‘1’, and there are no negative observations. This limits what can be done to measure relative abundance of A. aurantia across Georgia.
I decided to scale the number of observations of A. aurantia in a grid cell against the number of observations of all spiders in that grid cell. This may offset the bias of more observations where there are more observers, but it does not save us from the other biases mentioned above. The map below shows a raster of spider observations on iNaturalist from Georgia.
By dividing the number of A. aurantia observations with the number of all spider observations in each cell, we end up with a raster showing the proportion of spider observations that are A. aurantia. This can serve as a proxy for relative abundance of Yellow Garden Spiders.
It is immediately apparent that the new map is quite different from the first. The Atlanta metro now has many of the lowest abundance values for A. aurantia.
Whereas the raw data exaggerates abundance in urban areas, I think this approach tends to overestimate the abundance in rural areas. The true abundance map is probably somewhere in the middle.
Where do we go from here? More elegant visualization? Trends over time?