search

NIEHS / PrestoGP

Penalized Regression on Spatiotemporal Outcomes using Gaussian Processes a.k.a. PrestoGP
https://niehs.github.io/PrestoGP/
0 stars 0 forks source link

EPA Ecoregions Indicator - Covariate Calculation #6

Open kyle-messier opened 11 months ago

sigmafelix commented 9 months ago

HUC-8 and -12 indicators are available in xlsx format in EPA data library; Covariates need to be calculated in HUC-10 manually.

kyle-messier commented 9 months ago

@sigmafelix Wow! I had no idea about this resource. Having HUC8 and 12 would be all good in my opinion. Can you take a loook at the indicators and make a recommendation regarding if we can drop the other planned covariates and use these in their place? It would likely simplify and speed up our lives tremendously. Off-hand, the only one I see as potentially needed in addition to these are the USGS Pesticide applications, but perhaps they even have something similar already done here.

sigmafelix commented 9 months ago

I am comparing the metadata of HUC8 (2020) and HUC12 (2023) indicators. Some climate variables can be dropped, but it needs to be determined according to temporal alignment. Also, NLCD-based indicators in HUC8 and HUC12 data were calculated from 2016 and 2019 NLCD data, respectively. Do we use assign these single-year values as covariates in all years?

kyle-messier commented 9 months ago

Yes, a purely spatial variable can be assigned the same value across all years in the spatiotemporal outcome. Some temporal misalignment isn't a deal breaker.

sigmafelix commented 9 months ago

I understand. I will make a full list of covariates and its sources and share it here.