WP4 - supplementary data - Using deep learning to assimilate sun-induced fluorescence satellite observations in the ISBA land surface model: Datasets

The repository contains the data needed to reproduce the results and illustrations diplayed in the associated paper.

The weights of the neural network are stored in 10.5281/zenodo.18669437

The repository contains 6 files in the netCDF4 format,  here a brief description of each of it:

• "ebre_analyse_stats.nc" contains the LAI simulations results on the Ebro basin after assimilation of TROPOMI SIF, LAI, both or none. The names follows the one used in the article. The period covered start in June 2018 up to June 2020.
• "ebre_analyse_stats_gpp.nc" contains the GPP simulations results on the Ebro basin after assimilation of TROPOMI SIF, LAI, both or none. The names follows the one used in the article. The period covered start in June 2018 up to June 2020.
• "map_europe_sif_train.nc" contains the neural network RMSE, NRMSE, Pearson correlation with TROPOMI SIF, Number of TROPOMI SIF retrieval available per grid cell and the PFT fractions from ECOCLIMAP-II over the Europe domain computed per grid cell. The statistics are computed over the training period (ie. June 2018-May 2019).
• "map_europe_sif_test.nc" contains the neural network RMSE, NRMSE, Pearson correlation with TROPOMI SIF, Number of TROPOMI SIF retrieval available per grid cell and the PFT fractions from ECOCLIMAP-II over the Europe domain computed per grid cell. The statistics are computed over the test period (ie. June 2019-May2020).
• "ebre_lai_incr2021_SIF20.nc" provides the analyse increments on LAI after the assimilation of TROPOMI SIF with 20% of observation error for the year 2021 and per pft for the Ebro basin. The PFT fractions map from ECOCLIMAP-SG are also provided.
• "hyplant_px_ebre_tseries_sif20.nc" provides the timeseries on the cells grid located in 41.65°N, 1.05°E for the analyse SIF20 and the open-loop. The variables provided are in order: LAI [m2.m-2], GPP[mg[CO2].m-2.s-1], daily cumulated evapotranspiration[kg[H2O].m-2.s-1], Total soil moisture[m3[H2O].m-3] in the column and NN SIF estimate [mW.m-2.sr-1.nm-1].

For the PFT, here the full range is given contrary to what is displayed in the article for the sake of simplicity.

The 12 PFT are in order: Bare soil, Bare rock, Snow and Ice, Deciduous broadleaves forest, Coniferous needleleaves forest, Evergreen broadleaves forest, C3-crops, C4-crops, Irrigated/flooded crops, C3-Grass, C4-Grass, Peat and Wetlands.

In addition to the netcdf files, we provide text files which yield the Hyplant SIFO2A centered at 760nm measurement campaign. The raw data can be found in https://liaise.aeris-data.fr/about/.

Each file contains three columns of data: Longitude;Latitude;SIFA (no value =999) 

the files name is built as follow: date(fmt YYYYMMDD)-acquisition area(VER, FUL, GLO)-recording time(hhmm)-flight altitude(1150)-aircraft heading (S)-process used-dimension of the area.txt