PFT distribution has straight lines across Australia and Africa

[mchen0312]

Hi all,

There are two PFTs, "broadleaf deciduous tropical tree" and "broadleaf deciduous temperate shrub" that are distinguished by a straight line across the entire Australia and the south corner of Africa. The surface maps I used are: surfdata_1.9x2.5_simyr1850_c150626.nc, surfdata_360x720cru_simyr1850_c150626.nc and surfdata_360x720cru_simyr2000_c160307.nc. All three have the same issue.

Best, Ming

[bishtgautam]

@mchen0312 , All the datasets you mentioned were generated by the ACME team. Can you check if the PFT map shows similar issue for the surfdata_1.9x2.5_simyr1850_c141219.nc, which generated by NCAR? I have added surfdata_1.9x2.5_simyr1850_c141219.nc in ACME svn inputdata repo.

[mchen0312]

@bishtgautam I checked a raw data from NCAR's repository, mksrf_pft_0.5x0.5_simyr2005.c090313.nc, and that one is good without the issue. I will take a look at surfdata_1.9x2.5_simyr1850_c141219.nc. Thanks.

[bbye]

Looks like it might be a cutoff between tropical and temperate vegetation.

[mchen0312]

That could be. When looking at PFT derived plant trait maps, the straight line is really distinguishing. Plant leaf N/P will have a step change across that latitude....

[mchen0312]

@bishtgautam I just checked a file surfdata_360x720cru_simyr1850_c141219.nc, with the same time stamp as you suggested file, it has the issue. There is the latitude-line.

[bishtgautam]

Thanks for checking the data from NCAR. This gives me confidence that ACME team didn't make a mistake in the surface dataset generation procedure. From the surfdata_1.9x2.5_simyr1850_c141219 available from NCAR, it seems pftlandusedyn.0.5x0.5.simyr1850-2 005.c090630/mksrf_landuse_rc1850_c090630.nc is used to create the PFT map. Based on the attached plot for PFT=07 (broadleaf deciduous tropical tree) and 11 (broadleaf deciduous temperate shrub), the straight line is obviously present in the surface dataset but not in the raw dataset.

pft_07_and_11.pdf

@bbye is correct and there is a cutoff logic embedded in mksurfdat.F90#L1251.

[dmricciuto]

Looks like this specific piece of code was written to avoid a high LAI problem in earlier versions of CLM. We could try running an offline experiment with a modified surface data file (which keeps the information in the raw data) to see if the LAI problem still persists.

[bishtgautam]

@dmricciuto : That is a good idea. But, I believe this would be an expensive simulation as one would have to use a BGC compset to spinup at 1850 and a transient run to the present day. Correct? Maybe a coarser 2deg run would be to explore the issue.

[thorntonpe]

The issue as I remember it has to do with phenology. There was a daylength threshold for leaf senescence in deciduous temperate types that was not being met if the type occurred below a certain latitude. So those types were arbitrarily reassigned as tropical, meaning that they would experience drought deciduousness, and then roll over gradually to an evergreen habit if the drought signals were not present. The relevant phenology routines have not been modified by ACME, but changes may have been made between CLM4 and CLM4.5.

[dmricciuto]

@bishtgautam : We could cut out a relevant region (e.g. Australia) and run much faster. Peter's explanation makes sense, so I guess the problem may still exist. It is something to keep in mind since if we tune the daylength parameter, we could end up re-introducing the problem even with the current surface data. One possibility could be to do the boreal/temperate/tropical categorization internally within ALM (and so could be varied according to certain parameters).

[bishtgautam]

There was a daylength threshold for leaf senescence in deciduous temperate types that was not being met if the type occurred below a certain latitude

Maybe the daylength threshold (crit_dayl) could be made dependent on latitude.

@dmricciuto I like your suggestion of doing a regional simulation (👍 ). But, I didn't follow the internal categorization suggestion.