Closed matthew-mizielinski closed 6 years ago
I also found discrepancies between 1.0.21 and 1.0.22 for the following experiments / MIPS when trying to load our relational database with the latest data included in the dreqpy library:
Did all these experiments get updated (renamed or removed) from the 1.0.22 version?
+-----+------------------------+-------------+--------------+
| id | name | design_mip | dreq_version |
+-----+------------------------+-------------+--------------+
| 5 | ssp434 | ScenarioMIP | 01.00.21 |
| 9 | spinup-1950 | HighResMIP | 01.00.21 |
| 26 | land-crop-grass | LUMIP | 01.00.21 |
| 28 | land-cCO2 | LUMIP | 01.00.21 |
| 38 | highresSST-4co2 | HighResMIP | 01.00.21 |
| 40 | control-1950 | HighResMIP | 01.00.21 |
| 47 | highresSST-future | HighResMIP | 01.00.21 |
| 51 | omip1-spunup | OMIP | 01.00.21 |
| 71 | amip-TIP-nosh | GMMIP | 01.00.21 |
| 88 | land-crop-noIrrig | LUMIP | 01.00.21 |
| 94 | amip-TIP | GMMIP | 01.00.21 |
| 95 | lfmip-pdLC-wfdei | LS3MIP | 01.00.21 |
| 99 | highresSST-LAI | HighResMIP | 01.00.21 |
| 100 | omip2-spunup | OMIP | 01.00.21 |
| 107 | land-hist-altLu1 | LUMIP | 01.00.21 |
| 110 | land-hist-altLu2 | LUMIP | 01.00.21 |
| 116 | land-cClim | LUMIP | 01.00.21 |
| 121 | amip-hld | GMMIP | 01.00.21 |
| 124 | lfmip-initLC | LS3MIP | 01.00.21 |
| 125 | land-noPasture | LUMIP | 01.00.21 |
| 141 | lfmip-pdLC-princeton | LS3MIP | 01.00.21 |
| 147 | highresSST-smoothed | HighResMIP | 01.00.21 |
| 155 | ssp460 | ScenarioMIP | 01.00.21 |
| 157 | land-crop-noIrrigFert | LUMIP | 01.00.21 |
| 168 | ssp534-over | ScenarioMIP | 01.00.21 |
| 171 | lfmip-rmLC-wfdei | LS3MIP | 01.00.21 |
| 174 | land-hist-altStartYear | LUMIP | 01.00.21 |
| 182 | highres-future | HighResMIP | 01.00.21 |
| 185 | hist-noLu | LUMIP | 01.00.21 |
| 186 | lfmip-rmLC-princeton | LS3MIP | 01.00.21 |
| 193 | land-noShiftCultivate | LUMIP | 01.00.21 |
| 201 | highresSST-p4K | HighResMIP | 01.00.21 |
| 202 | highresSST-present | HighResMIP | 01.00.21 |
| 212 | omip1 | OMIP | 01.00.21 |
| 213 | omip2 | OMIP | 01.00.21 |
| 219 | land-noWoodHarv | LUMIP | 01.00.21 |
| 225 | lfmip-rmLC-cruNcep | LS3MIP | 01.00.21 |
| 230 | land-noFire | LUMIP | 01.00.21 |
| 234 | lfmip-pdLC-cruNcep | LS3MIP | 01.00.21 |
| 237 | land-crop-noFert | LUMIP | 01.00.21 |
| 242 | land-noLu | LUMIP | 01.00.21 |
| 243 | 1pctCO2Ndep-bgc | C4MIP | 01.00.21 |
+-----+------------------------+-------------+--------------+
42 rows in set (0.00 sec)
Thanks for posting this @bertinia. I know that the HighResMIP experiments should still exist, and there are now 222 experiments in the data request (rather than 248 at the last version and in the CVs). This suggests to me that 26 experiments have been removed and the remaining 16 in your list have been renamed.
222 experiments is now what I see has been loaded into our relational database using the dreqpy v1.0.22 data. Do we know if there is a time limit for when the list of experiments and MIPS in the data request is fixed or will it continue to evolve with time?
And perhaps a somewhat unrelated question... is there a reason why the data requests aren't stored in a relational database? It seems like if we used a normalized relational database that we would be able to avoid some of the problems associated with name changes. I have built some tools for NCAR's use to ingest at least the experiments, MIPS, UID's, and descriptions from the dreqpy data to help us manage and associate the metadata surrounding a CESM case with a MIP experiment. The database also has a web interface for viewing.
-- Alice Bertini NCAR - CSEG Software Engineer
Sorry, the 1.00.21 versions were correct, but due to a coding oversight version 01.00.22 reverted to names used in 01.00.20. I'll issue a correction this week.
Concerning the format and structure of the database: it is distributed as XML because that is a robust format for sharing information between platforms. I provide a static web interface (http://clipc-services.ceda.ac.uk/dreq/ ) because it is the only way that I can at least guarantee stability of the interface. The request document is constrained by a considerable number of consistency checks, but there are no community requirements. In the future, it might make sense to establish requirements in advance and design something around those. This is obviously not a new idea, but even obvious first steps can get lost in the excitement of the CMIP process. From my perspective, there are two many change requests coming from too many directions.
Experiment names corrected in 01.00.23
@bertinia identified an issue with version 01.00.22 in cmip6dr/CMIP6_DataRequest_VariableDefinitions/issues/315 which I've explored a little further;
The labels for ScenarioMIP experiments have changed from
in version 01.00.21 to
in version 01.00.22. This can be seen clearly here too. Looking at the new SSP437 experiment, its description now refers to
3.7 W/m2 by 2100
rather than the3.4 W/m2 in 2100
for ssp434 at dreq version 01.00.21. I'm presuming that this change is in error?