This repository contains the data used in the manuscript "Cell tracking -based framework for assessing nowcasting model skill in reproducing growth and decay of convective rainfall" by Ritvanen, J., Pulkkinen, S., Moisseev, D., and Nerini, D. submitted to the Geoscientific Model Development journal. A description of the data files is provided below. ## observations.zip The observation data is provided as a single hdf5 file `swiss_composite_bbox_26012024.h5`, with the structure ```txt ├YYYY-mm-dd HH:MM:SS │ └measurements │ ├data [uint8: 640 × 704] │ │ └2 attributes: │ │ ├CLASS: b'IMAGE' │ │ └IMAGE_VERSION: b'1.2' │ └what │ └1 attributes: │ ├quantity: 'RATE' │ ├units: 'mm/h' ``` The values are packed using a lookup table, which is provided in the metranet_lookup.py file, where the variable `METRANET_RR_8BIT` contains the rain rate values in units of mm/h corresponding to the packed values from 0 to 255. The data is saved in a metric coordinate system with the following projection string: `+proj=somerc +lat_0=46.95240555555556 +lon_0=7.439583333333333 +k_0=1 +x_0=2600000 +y_0=1200000 +ellps=bessel +towgs84=674.374,15.056,405.346,0,0,0,0 +units=m +no_defs` The corner coordinates of the data are: "UR_lon": 12.463400010413896, "UR_lat": 49.36540000898943, "LL_lon": 3.2432298672374404, "LL_lat": 43.63301754190871, ## nowcasts.zip The zip file contains the nowcast data used in the study. NOTE: the uncompressed size of the files is approximately 53GB. The following files contain the nowcast data: - `extrapolation_swiss_composite_bbox_26012024.h5`: the advection nowcast - `lcnn_diff_rmse_swiss_31012024.h5`: the L-CNN nowcast - `linda_domain_swiss_composite_bbox_26012024.h5`: the LINDA nowcast - `sprog_swiss_composite_bbox_26012024.h5`: the S-PROG nowcast The files have the following structure: ```txt ├YYYY-mm-dd HH:MM:SS <model-name> ├<leadtime-index> │ ├data [uint8: 640 × 704] │ │ └2 attributes: │ │ ├CLASS: b'IMAGE' │ │ └IMAGE_VERSION: b'1.2' │ └what │ └13 attributes: │ ├LL_lat: 43.63301754190871 │ ├LL_lon: 3.2432298672374404 │ ├UR_lat: 49.36540000898943 │ ├UR_lon: 12.463400010413896 │ ├description: <description> │ ├institution: 'MeteoSwiss' │ ├lookup: 'METRANET_RR_8BIT' │ ├projdef: '`+proj=somerc +lat_0=46.95240555555556 +lon_0=7.439583333333333 +k_0=1 +x_0=2600000 +y_0=1200000 +ellps=bessel +towgs84=674.374,15.056,405.346,0,0,0,0 +units=m +no_defs' │ ├qty: 'RATE' │ ├timestep: 5 │ ├unit: 'mm/h' │ ├xsize: 704 │ └ysize: 640 ├<leadtime-index> │ ├data [uint8: 640 × 704] │ │ └2 attributes: │ │ ├CLASS: b'IMAGE' │ │ └IMAGE_VERSION: b'1.2' │ └what │ └13 attributes: ... ``` where `<model-name>` is the name of the nowcasting model and `<leadtime-index>` is the index of the lead time of the nowcast starting from 1 (e.g. 1=5min, 2=10min, 3=15min, ...). The model names are: - `extrapolation`: the advection nowcast - `linda`: the LINDA nowcast - `sprog`: the S-PROG nowcast - `lcnn-diff-rmse-31012024`: the L-CNN nowcast The values are packed using a lookup table, which is provided in the `metranet_lookup.py` file, where the variable `METRANET_RR_8BIT` contains the rain rate values in units of mm/h corresponding to the packed values from 0 to 255. The data is saved in a metric coordinate system with the following projection string: `+proj=somerc +lat_0=46.95240555555556 +lon_0=7.439583333333333 +k_0=1 +x_0=2600000 +y_0=1200000 +ellps=bessel +towgs84=674.374,15.056,405.346,0,0,0,0 +units=m +no_defs` The corner coordinates of the data are: "UR_lon": 12.463400010413896, "UR_lat": 49.36540000898943, "LL_lon": 3.2432298672374404, "LL_lat": 43.63301754190871

2024