search

cmip6dr / CMIP6_DataRequest_VariableDefinitions

Definitions of variables in the CMIP6 Data Request
7 stars 0 forks source link

soil moisture variable name inconsistencies #413

Open taylor13 opened 2 years ago

taylor13 commented 2 years ago

The CORDEX project noted:

In the CORDEX-CMIP6 Atmos Variable List there are 3 variables with Soil Moisture Content (water in all phases):

These 3 variables are taken from the CMIP6 Data Request, exactly as they are (https://clipc-services.ceda.ac.uk//dreq/mipVars.html). If we follow the same logic, a similar triplet for Frozen Water Content should be mrfso, mrfsos and mrfsol and for Liquid Water Content - mrlso, mrlsos and mrlsol. There are mrfso and mrlso in the CMIP6 DR but instead of mrfsol/mrlsol one can find only mrsfl/mrsll. Both variables (mrsfl and mrsll) are only defined in the Palaeoclimate MIP and I don’t know how names for them were created. It should be checked with CMIP.

I replied:

The rules for naming MIP variables originated at the time of AMIP about 30 years ago (formulated primarily by Peter Gleckler and me). You can find the guidance at https://pcmdi.llnl.gov/mips/amip/DATASTDS/VARNAMES/main.html . According to those rules, we should have:

                           Liquid+Frozen         Liquid               Frozen
                          ----------             ------               -------
column total       |      mrso                    mrsol                mrsof
top 10 cm          |      mrsos                   mrsols               mrsofs
by layer           |

In the early 90's we didn't anticipate needing to record soil moisture by layer.

In CMIP3 we had 

                           Liquid+Frozen         Liquid               Frozen
                          ----------             ------               -------
column total       |      mrso                                         mrfso
top 10 cm          |      mrsos                  
by layer           |

We didn't exactly follow the AMIP rule for frozen water (mrfso) because we put the "f" before "so", rather than after it, which we should have done to be consistent with mrsol in the first table above.

In CMIP5 we had 

                           Liquid+Frozen         Liquid               Frozen
                          ----------             ------               -------
column total       |      mrso                                         mrfso
top 10 cm          |      mrsos                  
by layer           |      mrlsl

We appear to have mistakenly used the name "mrlsl" to represent the total moisture content in each layer. It should have probably been named "mrsol": moisture (content)-(of) soil-(by) layer. This got corrected in CMIP6.

In CMIP6 we have:

                           Liquid+Frozen         Liquid               Frozen
                          ----------             ------               -------
column total       |      mrso                    mrlso                mrfso
top 10 cm          |      mrsos                  
by layer           |      mrsol                   mrsll                mrsfl

In CMIP6 we also have mrsosLut and mrsoLut which have the additional dimension of "land use type" but are otherwise defined like mrsos and mrso.
Note that mrsll and mrsfl should probably have been mrlsol and mrfsol (without dropping the "o"). for consistency with mrlso and mrfso and the other soil moisture variables.

I would recommend that CORDEX adopt the names from CMIP6 despite any flaws.

I would further note that we are reconsidering the rules for the naming of variables in CMIP7. Most names will be unchanged, but we will use additional descriptors to modify the root names in a standardized way. Distinguishing between variables representing the "top 10 cm" or "by layer" from the "column totals" would be left up to these additional descriptors and all the 3 roots (mrso, mrlso, and mrfso) would apply to to all rows in the tables above.

Finally, if it were up to me, I would never ask anyone to record "liquid+frozen", "liquid", and "frozen" soil moisture. Two of the three should be sufficient because given two, the third can be computed from subtraction or addition.

For Frozen Water, I guess either mrsofs (consistent with the original rules and not inconsistent with mrsos) or mrfsos (building on mrfso and not inconsistent with mrsos) would be acceptable, but my slight preference would be mrfsos, since the original rule wasn't followed in creating mrfso.

jesusff commented 7 months ago

May I bring your attention also to soil temperature?

It is currently coded as tsl. This is a 3D var stored by layer (thus, ends in l). In the context of the urban CORDEX FPS, we are planning to store the temperature of the top layer. Shall this be called then tss? What would you recommend?

Luckily, column totals (or averages) for soil temperature are not very useful, otherwise, following the rules above we would have ts, and this would collide with the skin temperature. All this arises because of not using consistently so to refer to the soil. In the tsl name, is seems that sl was meant for soil, rather than being soil layers. A consistent use of so for the soil would lead to:

column total  | tso  (not useful)
top 10 cm     | tsos             
by layer      | tsol
taylor13 commented 7 months ago

Just to note that in CMIP6, "tsl" can either be a function of model layer or it can represent the mean over the uppermost 10 cm of soil. The variable name normally does not need to indicate how the data have been sampled (what dimensions are covered) either spatially or temporally. So, it would be o.k. to use "tsl" to report the temperature of a model's top soil layer. The variable should be a function of lon, lat, and the appropriate vertical coordinate defined in the model (presumably z ... the depth below the surface). The z-coordinate would be a scalar with a value usually set to the middle depth of the uppermost layer and with bounds set to 0 (top of soil layer) and the layer's lower bound (which presumably will vary from one model to another). If soil layers vary with location, I'm not sure what one would do.

Would it be easier to compare models if everyone reported the temperature of the top 10 cm of soil, as in CMIP6?

jesusff commented 7 months ago

Yes, sorry, I meant we would save the temperature of the top 10 cm, which might not be a soil layer for all models. Saving just the top model layer is not very useful, since this top layer depth in general depends across models. One would need the rest of the layers to compute some comparable diagnostics. Our idea was to compute this top 10cm temperature diagnostic in advance and save it as an output field.

taylor13 commented 7 months ago

I would name the temperature of the top 10 cm as in CMIP6: "tsl". I think you're right that "sl" is an abbreviation for SoiL. If soil temperature as a function of depth is also collected it could also be called "tsl", but it would have to appear in a different CMOR table.