GitLab

Updated plot scripts. Long titles are shorter now. Added units for y-axis.

See merge request !23

Release clean

See merge request !21

Streamlined the process from simulation output to predicted errors. Added...

See merge request !20

# Conflicts:
#   scripts/merge.py

Running examples for comparison of cosmo vs our simulation now stores everything in one folder. Fixed a few bugs in microphysics: Particle constant lambda was not stored and always zero. Added clipping for every process such that when adding the change for any hydrometeor or water vapor, we take the total amount of available mixing ratio or particle number into account and whatever happened in previous functions. This does not avoid negative values in all cases but it is better than before.

Changed parcel height to 50m. Added better clipping in microphpysical processes. Added asstertion for changes larger than existing mixing ratio.

Fixed a bug where heterogenous freezing would have a bigger impact by a scale of time step than other processes.

BUGFIX! Rain evaporation has been using diffusitivity when it should have been mean volume diameter.

Working ensemble code

See merge request !18

Updated job scripts. Running ensembles was not long enough for some trajectories, mostly convective 600 hPa trajectories.

Added more input parameters, scripts, segment detection...
Added more input parameters for sensitivity analysis and moved those to
model_constants as it should be.
Also added black as linter for python. Added segment identifier. All
scripts are properly documented. README and docs are updated.

Added a script to plot statistics (aka histograms) of segment starts. Also fixed how sliding windows are created by padding values and centering the window. Getting the confusion matrix is now corrected for the predicted segment and not just the predicted window if feasible and wanted.

Added adaboost as additional classifier. Changed classification from multiclass (multiple, mutual exclusive classes) to multilable (multiple classes with possible multiple labels) by wrapping classifier to one vs rest classifier.

Deleted some functions that are already available with better documentation in segment_identifier.py.