RECON Moisture Flows Dataset Elena De Petrillo
[email protected] Luca Monaco [email protected] 18 November 2025 Dataset: 10.5281/zenodo.14191919 Examples: https://github.com/elenadepetrillo/RECON-globally-reconciled-moisture-flows How to cite: De Petrillo, E., Monaco, L., Tuninetti, M., Staal, A., & Laio, F. (2025). Cell-scale atmospheric moisture flows dataset reconciled with ERA5 reanalysis. Scientific data, 12(1), 629. This repository presents the RECON dataset, a global atmospheric moisture connections NetCDF dataset, together with its usage tutorial for data handling and a snapshot of the data generation process. The RECON dataset is a post-processed version of the Lagrangian (forward trajectory-based) tracking model UTrack dataset (DOI UTrack dataset: 10.1594/PANGAEA.912710, DOI UTrack support paper: 10.5194/essd-123177-2020). Data Overview The RECON dataset provides moisture flow volumes, in cubic meters, from evaporation sources to precipitation targets and vice-versa. It offers global coverage at a resolution of 0.5°for an average year based on the period 2008–2017. The RECON dataset is available at 10.5281/zenodo.14191919 in a compressed .7z format, along with a data download and treatment guide. Further, the associated GitHub repository RECON-globally-reconciled-moisture-flows provides Python scripts in Jupyter notebooks to retrieve moisture flow volumes from the dataset. •RECON_moisture_flows_0.5.nc.7z : RECON dataset packed in .7z format, to be unpacked as explained in the next subsection •RECON_ERA5_avgYear_0.5_volumes.nc : our own edited version of yearly averaged ERA5 data, which have been used to postprocess moisture volume flows retrieved from the UTrack dataset in order to get the RECON dataset, as explained in our official GitHub repository. These average yearly volumes are in cubic meters, and the annual balance between precipitation and evaporation is reconciled. •ERA5_m_0.5_volumes_corrected.nc (where m denotes the month): a post-processed version of the monthly averaged ERA5 precipitation and evaporation data. These files were used to reconstruct monthly moisture-volume flows from the UTrack dataset prior to applying the IPF procedure. •README.pdf : a guideline document covering dataset description, data handling instructions, relevant repositories, and details on data generation. By sharing these post-processed ERA5 files, we provide means for reproducibility of both the IPF algorithm and the possibility to normalize volumes (cubic meters) to the fraction of evaporation to precipitation. 1
Data Handling •The dataset supports the retrieval of source-to-target and target-to-source sheds using a 4-tuple of coordinates (source_lat, source_lon, target_lat, target_lon): – Latitudes (lats) are in the range [90, -90]. – Longitudes (lons) are in the range [0, 360]. • To retrieve evaporation sheds (downwind region receiving precipitation) from an evapotranspiration point, users must specify the source coordinates. Conversely, to retrieve precipitation sheds (upwind region contributing to precipitation), users must specify the target coordinates. • Data Format: Moisture flow volumes in the dataset are stored as integers [0, 255] and must be converted to cubic meters. The data conversion formula y= 10z−1 254 ·[log10 (ymax)−log10 (ymin )]+log10 (ymin) is included in the scripts described in the next section. y is the converted volume in m3 , z is the volume retrieved from RECON, ymax ≈ 122079329 m3 is the maximum volume in m3 contained in RECON and ymin = 10−3m3is the minimum threshold we chose to consider a moisture volume. GitHub repository contents •example_scripts\: –Rec_evaporation_shed.ipynb : This notebook retrieves and plots the reconciled moisture flows originating precipitation in a downwind area from the source of interest based on the RECON NetCDF data. –Rec_precipitation_shed.ipynb : This notebook retrieves and plots the reconciled moisture flows contributing to precipitation at the sink of interest from upwind evaporation sources based on the RECON NetCDF data. •requirements.txt : The dependencies include essential Python libraries needed to run the provided scripts. Versions are specified as used by the authors, but they’re not mandatory. You can install them with pip install -r requirements.txt Downloading and extracting the dataset The file RECON_moisture_flows_0.5.nc.7z is a compressed/packed version of the NetCDF4 RECON dataset. To get the NetCDF dataset file, you can use a general-purpose unpacking software such as WinRar in Windows, or the 7z tool in Linux. Instructions for Ubuntu Here are the steps to download and unpack the .7z dataset in Ubuntu via terminal: Before we start, go to the location where you want to store the dataset, for example /home/user/Downloads . Downloading For version v3 the URL of the dataset is: https://zenodo.org/records/15025813/files/RECON_moisture_flows_0.5.nc.7z. Update this URL in the command if you want to download another version. Download the packed RECON_moisture_flows_0.5.nc.7z file with wget: wget https://zenodo.org/records/15025813/files/RECON_moisture_flows_0.5.nc.7z 2
Unpacking If p7zip is not already installed, install it with: sudo apt install p7zip Unpack RECON with the following command: 7z x RECON_moisture_flows_0.5.nc.7z How to use the notebooks 1. Set Up the Environment: Ensure you have a Python environment with the required packages as specified in requirements.txt. 2. Run the notebooks: Use the notebooks to retrieve and analyze reconciled moisture flows. You can specify either source or target coordinates based on whether you want precipitation or evapotranspiration sheds. 3. Visualization: The notebooks also generate plots that visualize global moisture flows, focusing on evaporation-toprecipitation or precipitation-to-evaporation pathways. Example of visualizations Evaporation shed map 3
Precipiration shed map Data generation process 1. Monthly Averages: We calculated the monthly averaged (2008–2017) moisture flow volumes from the UTrack dataset, using a customized version of ERA5 monthly average data. In this preprocessing step, we ensured that: •Negative evapotranspiration values were set to zero. • The global hydrological cycle was balanced, ensuring global evapotranspiration equals global precipitation. 2. Yearly Averages: The monthly averaged moisture flows were integrated to obtain yearly averages. 3. Reconciliation with the IPF algorithm: • The yearly averaged moisture flows were reconciled with our processed version of ERA5 yearly average data. • This reconciliation was performed using the Iterative Proportional Fitting (IPF) method to ensure consistency with the annual hydrological balance. Here follows an animated gif showing the IPF correction coefficient at every iteration at global scale 4
4. Data conversion: To ensure continuity with the Utrack dataset data format and reduce the RECON dataset weight, we converted the moisture volumes into integer values [0, 255] using the following formula y= 10z−1 254 ·[log10(ymax)−log10 (ymin )]+log10(ymin ) where: •y: Converted volume in m3, •z: Volume retrieved from RECON, •ymax ≈122079329 m3: Maximum volume in m3, •ymin = 10−3m3: Minimum threshold chosen for moisture volume. Update Note – Version v3 Version v3 standardizes the variable naming for greater clarity, ensuring that the location receiving precipitation is identified using targetlat and targetlon, consistent with version v1. No modifications have been made to the underlying data. Elena De Petrillo1orcid.org/0000-0001-7398-5742
[email protected] Luca Monaco2orcid.org/0000-0001-7701-5954 luca.monac[email protected]g Romain Thomas1orcid.org/0009-0008-9900-1930 1Politecnico di Torino, Department of Environment, Land and Infrastructure Engineering, Turin, Italy 2CIMA Research Foundation, International Centre for Environmental Monitoring, Savona, Italy 5
