Use Case 3: Diurnal Cycles

One main advantage of RCAT is that it can run analyses of large data sets through the use of parallelization (using the dask module). In principle it involves splitting the data into several “chunks” (in the space and/or time dimensions) and then run a set of operations on each of the chunk in parallel. Read more about it on dask homepage.

Depending on the analysis you want to run on your data, you might consider chunking your data differently. If, for example, you would like to calculate a quantile value for the data over all time steps then you should do the chunking in space only so that each chunk has all time steps available. Here, RCAT will be applied to calculate diurnal cycles of some model output using different statistical measures and how the splitting/chunking of data matters.

Similar to Use Case 1 most changes will be done in the configuration file, src/config/config_main.ini.

Calculate diurnal cycles of mean CAPE and plot the results

STEP 1: Data input

Under section MODELS specify the path to model data and set start and end years as well as months to analyze.

arome = {
     'fpath': '/nobackup/rossby21/rossby/joint_exp/norcp/NorCP_AROME_ERAI_ALADIN_1997_2017/netcdf',
     'start year': 1998, 'end year': 2002, 'months': [5,6,7,8,9]
     }
aladin = {
     'fpath': '/nobackup/rossby21/rossby/joint_exp/norcp/NorCP_ALADIN_ERAI_1997_2017/netcdf',
     'start year': 1998, 'end year': 2002, 'months': [5,6,7,8,9]
     }

If you would like to include observations as well, set accordingly in the OBS section.

STEP 2: Variables

Set output directory under the SETTINGS section.

In the key variables we specify in this example pcape (a specific model version of CAPE) available on 3 hourly time resolution. If only models will be analyzed set ‘obs’ to None. ‘regrid to’ is set to the coarser grid of the two models and data is interpolated using the bilinear method.

Specify region(s) in the regions key for which statistics will be selected, and finally plotted, for.

output dir = /nobackup/rossby22/sm_petli/analysis/test_dcycle_analysis

variables = {
    'pcape': {'freq': '3H', 'units': 'J/kg', 'scale factor': None, 'accumulated': False, 'obs': None, 'regrid to': 'aladin', 'regrid method': 'bilinear'},
    }

regions = ['Sweden', 'Denmark', 'Norway', 'Finland']

STEP 3: Select statistics

The statistics, diurnal cycle, is specified under the stats key in the STATISTICS section. Default options for diurnal cycle is found in the default_stats_config function in the statistics module. In default settings, hours is set to all 24 hours in a day. Since the data here is on 3 hourly resolution we specify these hours. The stat method (the statistical measure) for each hour is mean in default and it is kept here, and the data is chunked in the time dimension (also default so not specified here).

stats = {
   'diurnal cycle': {'hours': [0, 3, 6, 9, 12, 15, 18, 21]}
    }

STEP 4: Plotting

Under PLOTTING, validation plot should be set to True to enable plotting. Plotting of diurnal cycles will be both maps (for each hour) and line plots for specified regions.

validation plot = True

map configure = {}
map grid setup = {}
map kwargs = {}

line grid setup = {}
line kwargs = {'lw': 2.5}

STEP 5: Configure cluster

The number of nodes to be used in the selected SLURM cluster is set to 10 (increase if needed) and a walltime of 2 hours.

cluster type = slurm
nodes = 10
cluster kwargs = {'walltime': '02:00:00'}

STEP 6: Run RCAT

To run the analysis run from terminal (see Run RCAT in RCAT Configuration):

python $HOME/git/rcat/src/RCAT_main.py -c config_main.ini

Output statistics and image files will be located under the user-defined output directory in the stats and imgs sub-folders respectively

Calculate diurnal cycles of 99th percentile CAPE values

Instead of the mean value for each hour in the diurnal cycle (at any grid point in the domain) it could be meaningful to use another statistical measure, for example the 99th percentile. To do this, in addition to changing the stat method value, one will need to have all time steps available for the calculation and thus the chunk dimension should be changed from ‘time’ (default) to ‘space’:

stats = {
   'diurnal cycle': {'hours': [0, 3, 6, 9, 12, 15, 18, 21], 'stat method': 'percentile 99', 'chunk dimension': 'space'}
    }

When set, run RCAT again.