How to treat pixels that never reach theshold GDD?

usa-npn / cales-thermal-calendars

Estimating trends in phenology in the northeastern US

https://usa-npn.github.io/cales-thermal-calendars/spatial-trends-report.html

MIT License

0 stars 0 forks source link

How to treat pixels that never reach theshold GDD? #31

Closed Aariq closed 1 month ago

Aariq commented 2 months ago

Currently these are NAs, but they aren't really "missing" per se. Also, points that have these NAs for any year are currently excluded from the pixel-wise regressions (#27).

Things to try/consider:

Code these as Inf (this will only make a difference in maps of DOY by allowing plotting of all values ≥max DOY as the same color; Inf is treated the same as NA by lm())
Code these as 365
Do regressions with at least x number of non-NA values

Aariq commented 2 months ago

If I choose the third option (only keep pixels were at least x years reach the threshold), it turns out my choice of "x" is very consequential to the range of slopes calculated for the region.

Aariq commented 2 months ago

From Theresa:

let's only fit a line if we have DOYs in at least half of the years with data, does that seem reasonable? I don't think we should convert them to 365, I think years where a threshold isn't met should be left as NA for those pixels/years.

Aariq commented 1 month ago

Going to close for now. We can always tweak the % of pixels that need to be non-NA to fit a line.