Handling raster data with sdmTMB

[seananderson]

Question by email:

The environmental variables that I have are currently in raster form, can sdmTMB handle raster data or do I need to convert it to another form? In addition, the raster data that I have is in a different spatial resolution than the catch data I have, can the package handle this difference or is this something that needs to be handled before running the model?

[seananderson]

sdmTMB takes a data frame for model fitting where predictors are in columns and matched to an observation within the same row, similar to lm() or glm() etc. There’s no infrastructure on the backend to relate the raster data to your observations. sdmTMB can certainly be used with that type of data, but you’ll need to interpolate your raster data to your observation locations first and you’ll need to do the same thing for any new data frame you want to use in the predict function. Then you’d pass these data frames to sdmTMB() and predict(). There are functions in the raster package to do that. Take a look at raster::resample() for bilinear interpolation and raster::rasterize() if you need to get one of the data frames into a raster format first.