SEED-FD project precipitation data merging satellite and reanalysis products

The dataset contains two precipitation products obtained from the integration of satellite and reanalysis precipitation products for five study areas: the Paraná, Bhima, Danube, Juba, and Niger river basins. 

Precipitation data are obtained from 3 sources: ERA5-Land (Hersbach et al., 2020) provides reanalysis estimates of all surface variables, SM2RAIN-ASCAT (Brocca et al., 2019) contains soil moisture-based rainfall estimates obtained through the application of SM2RAIN to ASCAT retrievals, and IMERG-LR (Huffman et al., 2019) provides satellite precipitation estimates from microwave and infrared information. Using these sources, two distinct products are generated:

• Merged IMERG-SM2A: This product is characterized by the integration of satellite soil moisture-derived rainfall and IMERG Late Run precipitation estimates; therefore, it is based solely on satellite estimates. The two products are merged with a triple collocation technique (Gruber et al., 2016; Chen et al., 2021; with ERA5-Land precipitation used as the third product from reanalysis) following the procedure of Filippucci et al. (2025) (without spatial downscaling).
• Merged ERA5-IMERG-SM2A: This product is characterized by the integration of ERA5-Land precipitation, satellite soil moisture-derived rainfall, and IMERG Late Run precipitation estimates. The three products are merged with a triple collocation technique (Gruber et al., 2016; Chen et al., 2021) following the procedure of Filippucci et al. (2025) (without spatial downscaling).

Prior to merging, the original precipitation data are downsampled to the GloFAS 5 km grid using bilinear interpolation. The merging technique allows the generation of two high-resolution products (5 km, 24-hour) over the desired areas.

For research applied to developed countries, where ERA5-Land estimations are more accurate, we recommend using the ERA5-IMERG-SM2A product, thanks to the additional value provided by reanalysis data over areas where satellite precipitation estimates are inaccurate (e.g., topographically complex areas). For research applied to areas where observed data density is lacking and/or ERA5-Land precipitation is less accurate, we recommend using the IMERG-SM2A product.

The products have been developed within the framework of the SEED-FD project for the period 2007–2024. The dataset is potentially available worldwide upon request.

-------

SEED-FD (Strengthening Extreme Events Detection for Floods and Droughts) aims to advance global forecasts of extreme hydrological events by enhancing the Copernicus Emergency Management Service (CEMS). To achieve this, the project has developed improvements to CEMS’s underlying hydrological model, LISFLOOD, integrated Earth observation and in-situ micro-sensor data, and developed new tools to detect a wider range of extreme flood and drought events. These efforts contribute to more accurate and reliable forecasting, with a particular focus on data-scarce and high-risk regions in the Global South.

SEED-FD follows a phased approach, with developments first tested in data-rich river basins, including the Danube and Bhima, and then validated in the hydrologically diverse basins of Juba-Shebelle, Niger and Paraná. This process ensures the testing, refinement and assessment of the project results under different regional conditions, providing a basis for their potential global applicability. The improvements are intended to support future integration into operational CEMS services such as GloFAS and GDO, as well as potential uptake by other flood and drought forecasting and early warning systems.

By strengthening the scientific and technical basis for flood and drought forecasting, SEED-FD supports civil protection authorities, humanitarian actors and other decision-makers in better preparing for and responding to extreme hydrological events.

Funded by the European Union, Grant agreement ID: 101135110

Further information related to the project can be found here: https://www.seed-fd.eu/

The data have been used as input for SEED-FD use cases: https://www.seed-fd.eu/use-case-datasets-06-2026/

A dataset of satellite-derived NIR reflectance indices, proxy of river discharge (Filippucci et al., 2025b), was also developed within the SEED-FD project for 34 virtual stations distributed across 4 of the 5 basins, and is accessible via DOI: https://zenodo.org/records/20759160