be smart about number of layers in models

[wdwatkins]

Code is already there in R/get_base_lake_nml.R

[jordansread]

Does this issue need more context? It might be written in a way that only makes sense to you.

[wdwatkins]

referring to this, in order to minimize unneeded layers and keep the output files small:

min_thick <- get_nml_value(nml, arg_name = "min_layer_thick")
  max_depth <- get_nml_value(nml, arg_name = "lake_depth")
  max_layers <- ceiling(max_depth/min_thick * 1.1)
  nml <- set_nml(glm_nml = nml, arg_name = "max_layers", arg_val = max_layers)

@jread-usgs this was working fine for Mendota, but I'm getting errors from many lakes similar to:

Array bounds error - too many layers. NumLayers = 96, M = 5
i = 95 V = 371165.207014160230756 VMax = 4133007.668881853111088 D =    0.505414803317936 DMax =    0.30000000000000

I'm guessing this has to do with the lake being shallow? Max depth is 1.2 for this lake. Should there be an alternative minimum hundred layers maybe?

[jordansread]

yes, good idea. We hadn't run into this before because we always had way too many layers.

[wdwatkins]

200 seems to be a safe number so far.

[wdwatkins]

I suppose this should theoretically work now that the precipitation units are correct, although Luke pointed out we could just convert the netcdf files to netcdf4 so they use compression — this probably saves more space than messing with layers. Although I suppose if wanted to maximize disk space efficiency we could use both.

nccopy -knc7 -d9 output.nc output_4.nc adds max compression through the -d option. Converted a 1.7GB file to 48M. Might make reading a bit slower, but doesn't seem particularly noticeable.

[jordansread]

nice! That is a huge gainer

[wdwatkins]

After fixing the precip data units, there were still a few lakes that failed (shallow ones again). I was able to get them to run by increasing the layer threshold to ceiling(max_depth/min_thick * 1.5).

[jordansread]

We might want to propagate that change back into the functions that create the base nmls for each lake.