Spatial Modeling: The Basics

My week in the lab has been lots of double checking calculations, entering new information into the model, and considering how data sources may bias model results. However, these topics aren't information I can currently share, so instead I thought I'd close out 2024 by talking through spatial modeling and explaining the basics in a way that should make sense to everyone. For this we are going to use relevant examples to explain how the modeling works and what our program evaluates as it completes a single model run.

A spatial model is an analysis tool that uses simplified representations of reality to analyze model inputs or data. These simplified representations of reality come as generally two aspects for the model, which I will refer to as the model domain and the spatial maps. The model domain is the geographic representation of the area you are interested in understanding. Here, let's pretend like we want to know something about our neighborhood or our apartment complex. The geographic layout of the neighborhood or complex is the model domain. This includes spatial information about the shape of the domain so that we can best represent what the area looks like. That means if one of the neighbors is installing a fence, we could draw in that fence perimeter into our model domain, especially if that increases the size of our area of interest. Additionally, as part of creating our model domain, we need to decide the level of detail for our domain. This level of detail is known as grid size or model resolution and describes how many small areas we divide our domain into. The smaller each area, the more detailed our model outputs become (though the tradeoff is that increasing the detail means increasing the computer processing power required). So here, imagine that you go around with a piece of sidewalk chalk and draw small squares to divide up your entire neighborhood space. You will need to visualize some of these squares because it would be impolite to draw on someone's house with chalk. What's really important here is that you aren't changing the size of your square areas just because there are items in your neighborhood or apartment complex in the way. Houses, fire hydrants, cars, etc. may be split across multiple of these little squares, and that's where the spatial maps are important.

Spatial maps are information we provide to the model about features within the model domain, like temperature, depth of the water column, and salinity. These features inform where certain model groups thrive and where they don't survive. The spatial maps also help inform the movement of model groups because if a model group can move itself from an area it doesn't enjoy, it will try to move to a better area within the model domain. In our neighborhood example, we can envision spatial maps as knowing some information about our neighbors, places in the neighborhood we don't visit (like the spooky abandoned house on the end of the street), places in the neighborhood where the kids really like (the bump in the sidewalk that makes a good skateboarding ramp), and perhaps areas that the animals like (the dog park, for instance). From the small chalk squares we drew, we might have one of these areas of interest partially in one chalk square and partially in another. These spatial maps, though, just represent the location of features across the model domain and do not represent how the model groups will respond to those features. The response curves that I wrote about a few weeks ago provide the information about how the model groups respond to different features of the model. Here we could generate a response curve for the kids where they find the spooky house unsuitable, the bump in the sidewalk the best place ever, the dog park also quite nice, and then the school down the road as unsuitable space.

Finally, we can put all these modeling aspects together for our model run. We now have the model domain or the area that we are evaluating, the spatial maps, which describe the features of the model domain, and the response curves, which describe the suitability of different spatial features to the different model groups. When we run our model, we add our model groups (animals, plants, etc.) to each of our chalk squares and then we evaluate how the size of the model groups change over time as they find and thrive in suitable space or fail to find suitable space based on their starting location. We can also evaluate where the model groups are moving to, so we will know if our friend Jimmy is at the dog park or is hanging out somewhere else, and where we might best create an area of protection for model groups (in the real world where we might create protected habitat for birds or fish or large mammals).

I hope this basic explanation gives you a better understanding of how spatial modeling works. Our modeling is more complicated because we also have time factors built in, which I will explain in an upcoming blog. I will not be posting new updates until the new year, so happy holidays and you'll hear from me again in 2025!