search

sdtaylor / PhenograssReplication

0 stars 0 forks source link

analysis expansion #8

Closed sdtaylor closed 3 years ago

sdtaylor commented 4 years ago

Look at site level errors per ecoregion on a MAT/MAP plot to potentially gain some insights.

Maybe look at the seasonality of precip too

sdtaylor commented 4 years ago

try splitting out C4/C3 grasses, especially for Deserts( rt: Rachel)

sdtaylor commented 4 years ago

Results for deserts and bleh, maybe try fitting with the observed soil moisture? My major critique is soil moisture is not modeled well. so...

sdtaylor commented 4 years ago

switch to fCover comparison. Thats what the original paper did. probably better to convert observed gcc -> fcover than the other way around.

That'll should drop most of the high values too.

need to extract these from model files...

sdtaylor commented 4 years ago

Note, need to refit everything after addressing some issues: see

https://github.com/sdtaylor/GrasslandModels/issues/39 https://github.com/sdtaylor/GrasslandModels/issues/37

sdtaylor commented 4 years ago

Cut out all the interscale model comparisons, just have:

Model Scale R2 rmse # Timeseries Site years
All sites 0.32 0.2 90 235
All Grasslands 0.32 0.2 90 235
All Shrubland 0.32 0.2 90 235
All Ag 0.32 0.2 90 235
E. Temperate
Grasslands 0.32 0.2 90 235
Ag 0.32 0.2 90 235
Gr plains
Grasslands 0.32 0.2 90 235
Shrub 0.32 0.2 90 235
sdtaylor commented 3 years ago

these things were implemented in the first minor "revision" to BG, b314cb696a00d6ecc1172c6a9371a76a848a0729