Open lawinslow opened 9 years ago
EDIT: This was the secchi-only subset. I take these values back.
:100: :+1: !!
Oh, just saw the next comment..
Ok, either way these runs are pretty good. Using "best available" secchi for all the lakes with validation data, we get:
Metric | Value |
---|---|
RMSE | 2.57 |
Bias | 0.05 |
Quite reasonable.
Metric | Value |
---|---|
RMSE | 3.17 |
Bias | 0.80 |
Covers only 1980-1999. This does not include any driver bias corrections. Not too bad.
Trends actually look pretty good with NLDAS wind fixed (which may have been the major culprit) and new, with-ice continuous runs. Here are the trends for surface temperature, JAS for the modeled northern LTER lakes. This compares favorably to the overall ~0.04 degC/yr we see in Wisconsin.
site_id Name Abbreviation WBIC nldas_mod_JAS_trend
1 WBIC_1835300 Big Muskellunge Lake BM 1835300 0.03412556
2 WBIC_1842400 Crystal Lake CR 1842400 0.03598052
3 WBIC_1881900 Sparkling Lake SP 1881900 0.03610157
4 WBIC_2331600 Trout Lake TR 2331600 0.03564253
5 WBIC_2332400 Allequash Lake AL 2332400 0.03351620
Median overall surface JAS trend of modeled lakes
A little muted perhaps, but definitely better than before.
Remove bias. Remove wind issue. Use new ice model.