Over the past three years, we have contributed EO methodology sections to 14 successful grant proposals (Horizon Europe, USAID, World Bank GEF, UKRI). We have also reviewed proposals for two funding bodies. This post shares what we have learned about what passes and what fails.

What Technical Reviewers Check

1. Data access specificity. "We will use satellite data" fails. "We will use Sentinel-1 GRD Level-1 IW mode products accessed via Copernicus Data Space, Sentinel-2 L2A accessed via Planetary Computer, and Landsat Collection 2 Level-2 from USGS Earth Explorer — all free of charge" passes.

2. Accuracy claims with evidence. "Our AI model achieves high accuracy" fails. "Using pygeovision's ChangeFormer, tested on the LEVIR-CD benchmark (F1: 88.6%), validated against CSIR-GIMPA field data in 2023 (F1: 83.2% on 200 reference points)" passes.

3. Reproducibility commitment. No mention of reproducibility fails. "All analysis code will be deposited on GitHub under MIT licence within 6 months of project completion, with a Zenodo DOI. The specific input product IDs, software versions, and parameter values will be logged in a provenance JSON file for each analysis run" passes.

The Three Most Common Failures

Unspecified sensor. "Remote sensing data" is not enough. Name the satellite, sensor, band combination, product level, and spatial resolution.

Unreferenced accuracy claims. Every accuracy number needs a citation or a validation description. "Proven 95% accuracy" with no source is rejected by every technical reviewer we know.

No data management plan. EU funders (Horizon Europe) require a Data Management Plan. US funders (USAID, NSF) increasingly require open data commitments. Add a paragraph explaining where data will be stored, who will have access, and how long it will be retained.

Our Template

We offer proposal writing support for EO components — see the Services section of this site. We have turned around technically robust EO sections in 3 days for Horizon Europe proposals with 10-day submission windows.