Image and Signal Processing • ISP

A Flag Decomposition for Hierarchical Datasets Read the full article here: https://tr.ee/FlagDecompositionFlag manifolds encode hierarchical nested sequences of subspaces and serve as powerful structures ...

❔ How can we represent hierarchies inside data?

🗞️ Read "A Flag Decomposition for Hierarchical Datasets" to find out! 

🔗 https://tr.ee/FlagDecomposition

💻 https://github.com/nmank/FD

#DataScience #MachineLearning

Invertible Neural Networks for Probabilistic Aerosol Optical Depth Retrieval Read the full paper here: https://tr.ee/InvertibleNeuralNetworksSatellite remote sensing is the primary source of global aerosol observations, providing esse...

❔How do we estimate aerosol levels when a single satellite signal can match multiple physical scenarios?

🗞️"Invertible Neural Networks for Probabilistic Aerosol Optical Depth Retrieval"!

🔗https://tr.ee/InvertibleNeuralNetworks

#EarthObservation

Want to play with the idea?? greenwave.earth

Accelerated north–east shift of the global green wave trajectory | PNAS Viewed from space, a “green wave” seasonally traverses Earth’s surface, from the north in boreal summer to the south in austral summer. This wave r...

Very happy to finally share a paper that has been in my mind for a long time 🌍 🔗 www.pnas.org/doi/10.1073/...

Big news from the IPL, University of Valencia!🚨Two new MSCA projects joining the lab!
📈 Dr. Katerina Giamalaki (STOCKCLIM): Bridging climate extremes and financial risk with integrated AI.
🛰️ Dr. Dimitri Gominski (GEODE): Bringing transparency and explainability to satellite-based object detection.

GitHub - kipoju/udl4fl Contribute to kipoju/udl4fl development by creating an account on GitHub.

💻 Access the Code & models at

Can flood detection models really work everywhere? #floods #RemoteSensing #machinelearning "Understanding flood detection models across Sentinel-1 and Sentinel-2 modalities and benchmark datasets"Read the full article here: https://tr.ee/Understand...

❔Can flood detection models really work everywhere?

Find out in "Understanding flood detection models across Sentinel-1 and Sentinel-2 modalities and benchmark datasets"!

🔗https://tr.ee/UnderstandingFloodDetectionModels

#Floods #RemoteSensing

Our colleagues Roberto Fernández Moran, Moritz Link, Andrés Terrer, and Maria Piles recently joined the ESA CIMR L2PAD workshop to refine the algorithms behind the Copernicus Imaging Microwave Radiometer.

🔗Details: ir.uv.es/j0oIUJr

#CIMR #ESA #Climate #RemoteSensing #Science

📢 Call for Papers: PGM 2026

Join us in Valencia this September for an incredible program:

• ✨ Two world-class keynote speakers.
• 🎓 Pre-conference workshop for young researchers.
• 🤝 Great networking with the Probabilistic Graphical Models community.

🔗 More details: uv.es/pgm2026/cfp.html

📢 Call for papers: AI4Earth

Join us at ICANN 2026 (Padua, Italy) for our special session on ML & Signal Processing for Earth Systems 🌍

📄 Accepted papers published in Springer LNCS.
🗓️ Deadline: 16 March 2026

📌Details & topics: links.uv.es/ipluv/icann2...

#ICANN2026 #AI4Earth #ML

Near-Real-Time Turbidity Monitoring at Global Scale Using Sentinel-2 Data and Machine Learning Techniques HighlightsWhat are the main findings?We have developed a machine-learning algorithm able to quantify high-turbid environments.We use open-source and free databases to train the model and open-source and free tools to develop it, which makes it easily replicable and transferable.What are the implications of the main findings?The social impact of our work implies improved monitoring of areas with high turbidity, which can lead to a better understanding and use of forecasting.Ocean color and water quality communities can take advantage of the lessons learned for developing new products or services.

❔What if water turbidity could be monitored globally every few days?

🗞️Read more in "Near-Real-Time Turbidity Monitoring at Global Scale Using Sentinel-2 Data and Machine Learning Techniques"!

🔗 https://tr.ee/Near-Real-TimeTurbidityMonitoring

#EarthObservation #MachineLearning

Do you want to dive deeper into #AI, #Earthobservation, statistics or color vision? Learn about dimensionality reduction, radiative transfer models, explainable AI, hyperspectral imaging, spatial information, and more.isp.uv.es/courses

🧑‍🏫👩‍🏫 Do you want to dive deeper into AI, Earth observation, statistics or color vision?

🔍 Learn about dimensionality reduction, radiative transfer models, explainable AI, hyperspectral imaging, spatial information, and more.

🔗 https://isp.uv.es/courses

Sub-Seasonal Forest Carbon Dynamics Lose Persistence Under Extremes Read the full paper here: http://tr.ee/Sub-SeasonalForestCarbonDynamicsExtreme weather events pose a growing threat to the stability of terrestrial ecosystem...

❔ How do extreme weather events disrupt the stability of forest carbon fluxes?

🗞️ Learn about it in "Sub-Seasonal Forest Carbon Dynamics Lose Persistence Under Extremes"!

🔗 http://tr.ee/Sub-SeasonalForestCarbonDynamics

#ExtremeWeather #ClimateChange

Can we make machine learning models more resilient to distribution shifts? #machinelearning "Out-of-distribution robustness for multivariate analysis via causal regularisation"Read the full article here: http://tr.ee/Out-of-DistributionRobustness

❔ Can we make machine learning models more resilient to distribution shifts?

🗞️ Find out in "Out-of-distribution robustness for multivariate analysis via causal regularisation"!

🔗 http://tr.ee/Out-of-DistributionRobustness

#ML #CausalInference

You can still take part!

⌛ Just 5 minutes of your time is all it takes! ⬇️

🎨 Database based on board games for testing vision models: huesandcues-d76ee.web.app

Considerable yet contrasting regional imprint of circulation change on summer temperature trends across the Northern hemisphere mid-latitudes Abstract. Rising summer temperatures and more frequent heat extremes are well-documented outcomes of anthropogenic climate change. However, the extent to which atmospheric circulation changes contribu...

Summer temperatures have strongly been influenced by circulation changes in the northern mid-latitudes.

In our new study we evaluate and compare 4 statistical and ML methods that decompose trends into a "thermodynamical" and a "circulation induced" part.
wcd.copernicus.org/articles/7/8...

Generative networks for spatio-temporal gap filling of Sentinel-2 reflectances Read the full article here: https://tr.ee/GapFillingSentinel2Earth observation from satellite sensors offers the possibility to monitor natural ecosystems by...

❔ How can AI help us fill the gaps in satellite imagery?

🗞️ More about it in "Generative networks for spatio-temporal gap filling of Sentinel-2 reflectances"!

🔗 https://tr.ee/GapFillingSentinel2

#EarthObservation #RemoteSensing #Sentinel2

Machine Learning for Climate Science session at EGU26

Submit your abstract to our #EGU26 session: Machine Learning for Climate Science

Details: www.egu26.eu/session/57569

With @blankabalogh.bsky.social, Tom Beucler, Gustau Camps-Valls and @dwatsonparris.bsky.social

#ML #climateAI #ML4climate #ESM

@isp-uv-es.bsky.social @unibremen.bsky.social

ISP - 2025 Recap (II) Here's a short recap of what has happened at the ISP during the second half of this year!#research #machinelearning #sentinel2 #earthobservation #geoscience ...

✨ A dynamic second half of the year has come to an end!

👏 Our team has been busy attending conferences, workshops and presentations. We have also welcomed new researchers and celebrated our colleagues' PhD thesis defenses.

⏩ Thanks to everyone who engaged with us!

🤝 Take part and help us!

🎨 Our team is developing a database based on board games for testing vision models. It only takes five minutes to complete it and works best on a PC.

🔗 Instructions and link: https://huesandcues-d76ee.web.app

👏 Congratulations to Gustau Camps-Valls on receiving the Blaise Pascal Medal in Earth and Environmental Sciences at the #EurASc 2025 Symposium!

🏅 This award recognises his pioneering contributions to integrating AI into Earth and climate sciences.

🔗 https://www.eurasc.eu/2025-award-recipients/

⬇️ Kai-Hendrik Cohrs sharing his most recent work:

📗SHRUG-FM: Reliability-Aware Foundation Models for Earth Observation
🔗https://tr.ee/SHRUG-FM

📙Leveraging a Fully Differentiable Integrated Assessment Model for RL and Inference
🔗https://tr.ee/LeveragingDifferentiableIntegratedModel

 #EurIPS #ELLIS

⏪ Last week, the ISP attended the six-day @euripsconf.bsky.social conference in Copenhagen!

📷 Jordi Cerdà ("Causal Effects of Price Spikes on Food Insecurity") and Gustau Camps ("Causality in Earth Science") during their presentations.

 #EurIPS #ELLIS

Near-Real-Time Turbidity Monitoring at Global Scale Using Sentinel-2 Data and Machine Learning Techniques HighlightsWhat are the main findings?We have developed a machine-learning algorithm able to quantify high-turbid environments.We use open-source and free databases to train the model and open-source and free tools to develop it, which makes it easily replicable and transferable.What are the implications of the main findings?The social impact of our work implies improved monitoring of areas with high turbidity, which can lead to a better understanding and use of forecasting.Ocean color and water quality communities can take advantage of the lessons learned for developing new products or services.

🚀 This study offers strong generalisation across ecosystems, provides interpretable machine-learning insights aligned with optical physics.

🔗 Read it here:

ℹ️ The authors present a novel, global-scale turbidity estimation model built on Sentinel-2 data and machine learning, trained using two harmonised open datasets (GLORIA and MAGEST), covering lakes, rivers, estuaries, and coastal oceans across 17 countries and turbidity levels from 0 to 2200 FNU.

🌍 MSI enables more detailed, operationally relevant monitoring, yet existing algorithms still lack generalisation and fail in extreme turbidity scenarios.

🛰️ This article delves into how satellite technologies have transformed turbidity monitoring, particularly through the capabilities of Sentinel-2’s MultiSpectral Instrument (MSI).

🌊 Turbidity is a critical indicator of aquatic ecosystem health and water resource sustainability. It reflects a wide mix of natural processes and human activities, and its elevation can impair photosynthesis, disrupt oxygen distribution, and trigger harmful algal blooms.

Near-Real-Time Turbidity Monitoring at Global Scale Using Sentinel-2 Data and Machine Learning Techniques HighlightsWhat are the main findings?We have developed a machine-learning algorithm able to quantify high-turbid environments.We use open-source and free databases to train the model and open-source and free tools to develop it, which makes it easily replicable and transferable.What are the implications of the main findings?The social impact of our work implies improved monitoring of areas with high turbidity, which can lead to a better understanding and use of forecasting.Ocean color and water quality communities can take advantage of the lessons learned for developing new products or services.

🗞️ 🆕 New paper published which explores the environmental and scientific context behind turbidity monitoring: "Near-Real-Time Turbidity Monitoring at Global Scale Using Sentinel-2 Data and Machine Learning Techniques".

#RemoteSensing #EarthObservation

⏩ It advances the understanding of how artificial and biological systems differ and lays the foundation for developing vision models that not only perform well but also “see” the world in a more human-like way.