search

coreytcallaghan / cs_sampling_effort

0 stars 0 forks source link

Predictor variables from GEE #5

Closed coreytcallaghan closed 3 years ago

coreytcallaghan commented 3 years ago

So next step as outlined in #4 is to extract some variables that are likely to influence the amount of 'sampling' necessary to sample SR.

I'm happy to consider whatever really, as I think here the important thing is to get some relatively strong relationships to be able to make spatial predictions out for unsampled grids/regions/sites. I think the important thing is that they have to be able to be mapped globally (i.e., calculated from underlying global maps) - as for this to be operationalized, it would be most important in regions with little data anyway! So, kind of starting from scratch.

Things I was thinking about are 1) habitat heterogeneity, 2) percent urban cover, 3) percent water cover. Others? @mitchest has some sweet code to calculate spectral diversity so it would be good to bust that out as well.

@mitchest the spatial grids are here: https://github.com/coreytcallaghan/cs_sampling_effort/tree/main/Spatial%20data. I'm happy to ingest the shapefiles into GEE if you want - just let me know. There are four spatial resolutions uploaded over the study extent.

I'm not sure how to treat the temporal change in remotely-sensed products though... For something like NLCD we could just use 2016 or something which sites in the middle of the extent of the study period (2014-2019). NDVI and EVI for example, could calculate yearly. But my gut is it isn't worth doing this yet, as it likely won't make a big difference at large spatial scales (10 or 20 km) anyway.

Any thoughts appreciated.

bowlerbear commented 3 years ago

we want stuff that (1) predicts species richness and/or (2) predicts rare species Maybe NPP? Elevation? Human footprint?

do we also want some spatial grouping terms too? like ecoregion/biome WWF maps?

I would just take the mean of some 30 year average for climate, NDVI etc..

mitchest commented 3 years ago

Sounds good - i'll start by getting a few different things so we can get the analyses cranking. We can revisit the variables and re-run if/when required.

mitchest commented 3 years ago

Sounds good - i'll start by getting a few different things so we can get the analyses cranking. We can revisit the variables and re-run if/when required.

Stats will land here when they're done (within a few hours probably): https://drive.google.com/drive/folders/1q5AEmn7vNxYUvbkoIRdxKIP6QCDYc952?usp=sharing

coreytcallaghan commented 3 years ago

Yeah, I think ecoregion is important at a larger scale, but part of the motivation with picking a small area (i.e., peninsular Florida) was to avoid that for now, so the assumption is that they are all in the same Bird Conservation Region, or biogeographical bird zone. But human footprint, NPP are good ideas, so thanks!

And thanks @mitchest - I'll have a look in the morning to see what's there. :)

coreytcallaghan commented 3 years ago

closing this for now.