README

README concerning the preprocessed extract of the coarse-grained NARVAL data available via DKRZ resources 

Date: 08/16/2023

The sample dataset contains the input data (`X_all`), the output data (`Y_all`) and two arrays containing information which variable is displayed in which column for X and Y, these are called `X_expl` and `Y_expl`.
Per column there is an array in the explanation arrays where the first entry describes the variable and the second one the vertical level. For 2D variables this second entry is set to 0.

T I T L E O F T H E D A T A S E T
- NARVAL R2B10 coarse-grained to R2B5 and preprocessed 

P A T H T O T H E D A T A S E T
- /work/bd1179/b309215/heuer23_ml_convection_parameterization/local_data/narval_preprocessed

O W N E R / P R O D U C E R O F T H E D A T A S E T 
- NARVAL R2B10: Daniel Klocke and Deutscher Wetterdienst, Offenbach and Max-Planck-Institut für Meteorologie, Hamburg
Coarse-grained to R2B4 and R2B5: Helge Heuer and Deutsches Zentrum fuer Luft- und Raumfahrt e. V., Institut für Physik der Atmosphäre, Oberpfaffenhofen

D A T A U S A G E L I C E N S E 
- See LICENSE File in this directory

C O N T E N T O F T H E D A T A S E T
- To an R2B4 and an R2B5 grid coarse-grained data from the NARVAL (Klocke et al., 2017: 'Rediscovery of the doldrums in storm-resolving simulations over the tropical Atlantic') 

D A T A U S A G E S C E N A R I O S
- To study the link between the thermodynamic environment and convective subgrid fluxes using the output of storm-resolving model simulations on a climate model resolution
- To train ML-based parameterizations

M E T H O D S U S E D F O R D A T A C R E A T I O N
- See 'Machine Learning-Based Parameterizations of Convection for ICON' (Heuer et al., 2023)

I S S U E S
- The mixing length in the vertical diffusion scheme was mistakenly set to 1000m instead of 150m, causing unrealistically strong vertical diffusion in some situations.