We are currently just dropping "frozen" samples or running the model separately.
Ideas to reconcile this, in order of time to implement (easy to hard):
(A) add DOY including sin(DOY) and cos(DOY) as predictors
(B) add air temperature or radiation (but met station has not been around the whole monitoring period - could get from nearby Greenbay airport)
(C) add "% vegetation cover" variable by creating a function that estimates vegetation between known dates of planting and harvest. Look in literature to see if one already exists.
(D) add snowmelt variable by pulling weather service(?) snow depth data, and getting a difference in this value from day to day as an indication of snowmelt
In the Otter Creek report, they analyzed "vegetated" and "non-vegetated" periods separately, and threw out winter samples. This is another idea for how to analyze the data.
Currently, we are not accounting for:
We are currently just dropping "frozen" samples or running the model separately.
Ideas to reconcile this, in order of time to implement (easy to hard): (A) add DOY including sin(DOY) and cos(DOY) as predictors (B) add air temperature or radiation (but met station has not been around the whole monitoring period - could get from nearby Greenbay airport) (C) add "% vegetation cover" variable by creating a function that estimates vegetation between known dates of planting and harvest. Look in literature to see if one already exists. (D) add snowmelt variable by pulling weather service(?) snow depth data, and getting a difference in this value from day to day as an indication of snowmelt
In the Otter Creek report, they analyzed "vegetated" and "non-vegetated" periods separately, and threw out winter samples. This is another idea for how to analyze the data.