CTSM
CTSM copied to clipboard
ESCOMP
Add per-harvest outputs, incl. GRAINC_TO_FOOD
Description of changes
Previously, the recommended way to get annual crop yield was to sum up daily-average values of GRAINC_TO_FOOD over a year, then multiply by the number of seconds per day. Here, I introduce a new output variable, GRAINC_TO_FOOD_ACCUM_PERHARV, that gives the same information with just annual resolution and without needing to multiply by anything.
Specific notes
Whereas GRAINC_TO_FOOD has dimensions time x pft, GRAINC_TO_FOOD_ACCUM_PERHARV has dimensions time x mxharvests x pft.
mxharvests is hard-coded to the maximum number of sowings per year plus 1. For now, that means mxharvests is always 2. So whereas with GRAINC_TO_FOOD you need to sum over all days in each year, with GRAINC_TO_FOOD_ACCUM_PERHARV you need to sum over all harvests in a year. (If only one harvest happens in a year, the second harvest's value is zero.)
CTSM Issues ~~Fixed~~ Addressed: #1649
Are answers expected to change (and if so in what way)? No.
Any User Interface Changes (namelist or namelist defaults changes)? None mandatory. GRAINC_TO_FOOD_ACCUM_PERHARV is an optional output variable.
Testing performed, if any:
I did a test run for I2000Clm50BgcCrop at 10°x15° resolution (f10_f10_mg37) for 2000-2009. Processed outputs are the same for GRAINC_TO_FOOD_ACCUM_PERHARV as they were for GRAINC_TO_FOOD, although of course the processing scripts differ.
Test details
Case directory: /glade/u/home/samrabin/cases_ctsm/yield_perharv_f10_f10_mg37
Processing scripts in subdirectory outputs.
Raw outputs and processed files in subdirectory outputs/lnd/hist.
GRAINC_TO_FOOD
- Processing script:
CropPFT_GridRemap.ncl - Processed file:
yield_perharv_f10_f10_mg37_CropRemapped_GRAINC_TO_FOOD_200001-200912_Sum.nc
GRAINC_TO_FOOD_ACCUM_PERHARV
- Processing script:
CropPFT_GridRemap.ACCUM.ncl - Processed file:
yield_perharv_f10_f10_mg37_CropRemapped_GRAINC_TO_FOOD_ACCUM_PERHARV_199901-200912_Sum.nc(Note the extra year at the beginning is because that was included in my annual-resolution output file.)
Just realized it might also be useful to have another output variable that effectively does the "sum across mxharvests" step itself. I'll plan to add this in another commit; something like GRAINC_TO_FOOD_ACCUM_THISYR.
(The separate per-harvest output is useful for applications like ISIMIP-Agriculture/GGCMI, where it's important to be able to associate harvests with the year of their planting.)
Found some discrepancies between gridded/remapped results from GRAINC_TO_FOOD vs. GRAINC_TO_FOOD_PERHARV vs. GRAINC_TO_FOOD_ANN. Investigating…
Okay, in my test run, it looks like the new per-harvest and annual outputs are identical to within < 1 mgC/m2, which should be fine. But per-harvest differs from the original outputs by up to 444 gC/m2. Will investigate further.
Turns out it's all good; I just had forgotten to time-shift the daily outputs in my Python script.
Max diff, original vs. perharv: (150.0,-10.0) sugarcane (67) 2007: 0.000244140625 (1094.70556640625 vs. 1094.705322265625)
Max diff (tie), perharv vs. annual: (150.0,-10.0) sugarcane (67) 2000: 0.0001220703125 (1100.258056640625 vs. 1100.2579345703125)
Max diff (tie), perharv vs. annual: (150.0,-10.0) sugarcane (67) 2007: 0.0001220703125 (1094.705322265625 vs. 1094.7054443359375)
I think this is ready for review.
Commits above add more outputs, all with dimension mxharvests unless otherwise noted:
- SDATES_PERHARV: actual sowing date (day of year) for crops harvested this year.
- SYEARS_PERHARV: actual sowing year for crops harvested this year.
- GDDHARV_PERHARV: Growing degree days (gdd) needed to harvest.
- GDDACCUM_PERHARV: At-harvest accumulated growing degree days past planting date for crop.
- HUI_PERHARV: At-harvest accumulated heat unit index for crop.
- SOWING_REASON (dimension
mxsowings) - SOWING_REASON_PERHARV
- HARVEST_REASON_PERHARV