When you take a look at how we predict spread and height where necessary (G:\GOA\R\spread height\spread height.Rproj-see all of the various "predict" functions dependent on available data) one of the thing we're missing with this great tool Michael left for us is the model/variable selection logic/process (i.e., we have the equation but don't know how Michael settled on the variables used or whether he considered other than GAMs). At minimum, we (Gulf-Aleutian Team) need to document why this approach works and is good enough and at max this might be an area to do a little research and possible paper or TM writing formalizing and perhaps modernizing this technique. Of course, in between those two endpoints might be where someone else has already invented this wheel and we can adapt our approach to an existing one and cite it!
Why does the current equation work? Can we document it better? Should we produce a tech memo about how well it works?
Methods for predicting net spread and height with the GAM we use are in
G:\GOA\R\spread height\Instructions_Munro_Fall2020.docx
The two GAMs for predicting net spread and net height when there are missing data are:
From Ned: