This is a repository for diagnostics of temperature extremes in climate data. For now, it focuses on analysis of daily minimum and maximum temperatures.
The spectral element files are on the unstructured mesh, so variables have coordinates of (ncol, time). For convenience, I have regridded selected files and put them in the Regridded folder in the data directory.They are regridded to the 0.5degree grid that the observations use. To do the regridding, I used the NCO command ncremap, invoked as
ncremap -4 -L 4 -m map_ne30_to_halfdegree.nc
-s /project/amp/brianpm/TemperatureExtremes/ne30np4_091226_pentagons.nc
-d /project/amp/jcaron/CPC_Tminmax/tmax.1980.nc
/project/amp/brianpm/TemperatureExtremes/TSMX/f.e13.FAMIPC5CN.ne30_ne30.beta17.t3.cam.h1.TSMX.19650101-20051231.nc /project/amp/brianpm/TemperatureExtremes/Regridded/f.e13.FAMIPC5CN.ne30_ne30.beta17.t3.cam.h1.TSMX.1965\
0101-20051231.regrid.nc
On subsequent files, I just use the mapping file map_ne30_to_halfdegree.nc that is produced on the first file.
We have developed a series of tools that can be chained together to produce data files (netCDF) and (potentially) some summary statistics and (potentially) some simple figures.
Each script has been developed separately, and can be used independently from the others. They have been modified to be run in a prescribed order based on minimal input to a command line interface called extreme_wf. The input to extreme_wf is a YAML file and must include several pieces of information.
The workflow is very straight-forward. Call extreme_wf with the YAML file as the only argument. The YAML file contains all the input and output specification. Within extreme_wf, the YAML file is validated (i.e., it looks for required inputs) and then runs Save_Quantile_Dataset.py which saves a file with quantile information from the input. That file is then used in prototype_event_detection.py along with the input files to detect the events above the threshold quantile (right now that's the 90th percentile; see notes). The next stage is to derive the event attribute file from the event detection file using event_attributes.py.
It is clear that some inefficiency is built into this system if the code in extreme_wf does not preserve the objects being created along the way. We have modified the original scripts so that each has a run() method that returns an Xarray DataSet object; when the scripts are run as standalone programs, they save this dataset, and when run is called directly, extreme_wf is in control of saving the intermediate files (could be optional in YAML) and for providing the (in memory) dataset objects to the next step.