Global Flood Partnership Meeting is in Provo, Utah, this August 4-6. its free to go! its a nice small meeting of end users of flood data and the scientists, companies, and government agencies that produce flood data www.globalfloodpartnership.org/events/gfp-a...
Consider signing this for the continuation of the EM-DAT disaster database
openletter.earth/the-worlds-c...
A worldmap of the spatial distribution of extracted flood events in the Groundsource dataset. The map displays the total number of flood events extracted by the LLM-based pipeline aggregated per grid cell. The data are visualized using a Robinson projection, with event counts represented by a logarithmic color scale. Red points indicate the spatial centroids of reference flood events from the GDACS database.
Excited to announce Groundsource - an open-source dataset of historic flood events! This has easily been one of the coolest projects I've worked on recently!
Thread 🧵 for details and all relevant links. 1/n
#Methane Hunters Track Swamp Gas That Is Driving Climate Warming
www.nytimes.com/2026/02/18/c...
our work, showing that only 65%–75% of the Planet Basemap #water area was also classified by #Landsat was featured in #Planet's monthly science newsletter:
tinyurl.com/37rcadvp
Article here: doi.org/10.1029/2025...
NASA EarthRISE Developers Academy Summer 2026 Paid Research Opportunities: appliedsciences.nasa.gov/what-we-do/c...
The Academy builds the next generation of science and technology leaders through a 10wk opportunity applying NASA Earth observations to real-world challenges
Deadline: March 6
Our latest contribution in Geophysical Research Letters examines spatial-scale effects on surface #water detection with important implications for multi-sensor water monitoring.
doi.org/10.1029/2025...
Thank you, NASA FINESST & Commercial Satellite Data Acquisition program for funding.
#FloodMapping #DisasterResponse #Hydrology #FloodRisk #UrbanFlooding #CoastalFlooding #EarthObservation #FoundationModels
🌊 Cross-sensor comparisons using public and commercial satellite data for coastal and urban floods
💨 Methane estimates from small water bodies
It’s the 1st AGU in many years sans moi, but I’m incredibly proud of the team’s work and excited to see these advances shared with the community.
Headed to #AGU25 next week?
Check out the work my group is presenting on #flooding and climate-driven hazards:
🌧️ Using geospatial foundation models for global flood mapping
🏙️ Satellite- and process-based comparisons of flood extent & depth in urban envs
🌪️ Hurricane-induced flash floods #Helene
• We need a shift from model-centric development to impact-centric downstream tasks.
• As with early deep learning, GFMs are advancing fast—but with limited attention to societal, environmental, & economic implications.
• Responsible AI in EO depends on real-world deployment—not just benchmarks.
Key takeaways:
• Freely available, complementary radar + optical (S1 & S2) data remain essential for global flood applications.
• GFMs like TerraMind can reduce compute and labeled-data requirements for accurate flood mapping.
• GFM-based maps show higher precision but lower recall than U-Net.
• Models trained on FloodsNet outperformed the Sen1Floods11-trained TerraMind example in recall while maintaining similar accuracy.
• U-Net achieved the highest recall overall, with slightly lower accuracy and precision.
Key results:
• Base–unfrozen offered the best balance of accuracy, precision, and recall at lower computational cost.
• Large–unfrozen achieved the highest recall.
We compared results with the TerraMind Sen1Floods11 example and a U-Net trained on FloodsNet + Sen1Floods11.
🚀 Using TerraMind, we fine-tuned four backbone configurations (base vs. large; frozen vs. unfrozen) on FloodsNet, our harmonized global multimodal flood dataset (Sentinel-1, S1 radar + Sentinel-2, S2 optical), and evaluated generalization across 85 events.
🛰️ Geospatial Foundation Models (GFMs) such as ESA–IBM’s TerraMind promise better transferability through large-scale self-supervised pretraining. Yet their real-world performance for flood mapping remains underexplored.
🌊 Floods are among the most damaging weather-related hazards, and in 2024—the warmest year on record—extreme events hit communities across five continents. EO satellites provide critical coverage, but operational accuracy still depends heavily on labeled data and model generalization.
🌍 Geospatial data science in 2025 still comes down to two words: foundation models. And over the past months, we’ve had the chance to put that idea into practice for global flood mapping.
Summary figures & discussion on Hugging Face:
👉 lnkd.in/euR9DXfX
Full technical preprint:
👉 lnkd.in/eYMCZd9e
Free access to Google Colab Pro for academics and students:
#python
blog.google/outreach-ini...
With NISAR successfully launched on July 30, 2025, the mission is opening up an extraordinary new era of global L-band SAR observations. Look forward to using the data for #floods
The NISAR Science Users’ Handbook Second Edition can be accessed here: dataverse.jpl.nasa.gov/dataset.xhtm...
#NISAR
I look forward to seeing how this can support sustainable water management & inform modelling of reservoirs more broadly.
#Hydrology #RemoteSensing #Irrigation #Reservoirs #SWAT #WaterResources
- The framework is designed to be replicable in other regions, supporting water-resource agencies with better data on how small waterbodies influence watershed hydrology in an era of increasing irrigation demand & climate stress.
- Effects varied spatially and temporally; smaller channels showed the largest percent changes, and the largest monthly impacts occurred from January to May when storage is highest and rainfall greatest.
- In our study watershed in eastern Arkansas, the presence of these reservoirs was associated with a 14–24 percent reduction in annual flow and a 43–60 percent reduction in peak flows.
Key points:
- We developed a novel framework combining a remote-sensing time-series algorithm and the SWAT+ hydrological model to estimate how hundreds of on-farm reservoirs alter surface water flows.
It's been a minute on this one (in review for ~ 1.5 yrs), but congrats to Vini on his 4th PhD chapter, thx to Sankar for support & to me for pushing it over the line 😊
"Assessing the cumulative impact of on-farm reservoirs on modeled surface hydrology" is out: hess.copernicus.org/articles/29/...
Likewise, I look forward to seeing your GFM results!
join us with the Global Flood Partnership next Wednesday am for our ALL WOMEN remote sensing and flooding webinar! @mirelagtulbure.bsky.social @ekta-aggarwal.bsky.social and I discuss AI and flood mapping +applications re: migration + emergency response www.globalfloodpartnership.org/events/gfp-2...
