The Remote Sensing Gazette

Nov 2 – Nov 8, 2026

An all-weather microwave-based AI retrieval of pan-Arctic melt pond fraction

The first all-weather method uses radar and microwave signals to track Arctic sea ice melt ponds continuously through clouds and dark winter. A continuous pan-Arctic view is valuable for understanding and predicting rapid Arctic climate change, where melt ponds play a significant role in accelerating ice loss.

An all-weather microwave-based AI retrieval of pan-Arctic melt pond fraction · Sara Aparício, Simon Driscoll, Mark R. Drinkwater et al. · arXiv:2611.00940

New Wine in an Old Bottle -- NeuralPOM: A Hybrid Ocean Forecast Model with Differentiable Dynamics and Staged Learning

NeuralPOM combines a physics-based ocean simulator with a neural network that periodically corrects its output, keeping heat and salt conserved. The hybrid approach offers a blueprint for robust, trustworthy AI for complex Earth system modeling and forecasting.

New Wine in an Old Bottle -- NeuralPOM: A Hybrid Ocean Forecast Model with Differentiable Dynamics and Staged Learning · Ruiqi Shu, Guangliang Liu, Ruijian Gou et al. · arXiv:2611.02460

Shared-Pattern Collaborative Network for Multi-Source Remote Sensing Data Classification

The method learns common patterns across hyperspectral, radar and laser scans, and chooses which source helps most in each area. Combining multiple satellite perspectives can overcome individual data limits and give a more robust understanding of the planet.

Shared-Pattern Collaborative Network for Multi-Source Remote Sensing Data Classification · Chenfang Yang, Chuanzheng Gong, Feng Gao et al. · arXiv:2611.03069

Machine Learning Sensor Fusion for Water Vapor Radiometry based Tropospheric Delay Calibration

A machine learning approach combines water vapor sensors, weather models and signal measurements to correct atmospheric interference that distorts radio telescope observations. The method is vital for next-generation telescopes to achieve unprecedented clarity in high-precision observations.

Machine Learning Sensor Fusion for Water Vapor Radiometry based Tropospheric Delay Calibration · T. K. Sridharan, Yoshiharu Asaki, Brian E. Svoboda · arXiv:2610.41052

Original Insights Into Rock Slope Damage Processes Until Collapse From Passive Seismic Monitoring

Passive seismic sensors listened to an unstable rock slope for over a decade and detected subtle changes in internal stiffness and damage before collapse. Combining these deep internal insights with surface deformation data may offer earlier warning for dangerous landslides as climate change increases slope instability.

Original Insights Into Rock Slope Damage Processes Until Collapse From Passive Seismic Monitoring · P. Bottelin, L. Baillet · arXiv:2611.02863

DeepWaveRL: Self-Supervised Full Waveform Inversion via Reinforcement Learning

An AI uses reinforcement learning to map what lies beneath the Earth from seismic waves by iteratively guessing subsurface properties and refining guesses to match real data. It offers a flexible way to solve challenging inverse problems without detailed gradient information, possibly enabling more complex physical models.

DeepWaveRL: Self-Supervised Full Waveform Inversion via Reinforcement Learning · Peng Jin, Yinan Feng, Yinpeng Chen et al. · arXiv:2611.03264

Spectral-spatial vision test-time training for hyperspectral image classification

A machine learning method identifies materials and objects in hyperspectral images and adjusts its understanding to each image, improving accuracy with very little pre-labeled training data. It may help apply remote sensing AI to real-world problems where labeled training data are scarce.

Spectral-spatial vision test-time training for hyperspectral image classification · Xiang Zhao · arXiv:2611.02364

Characterizing landslide dynamics from time-lapse time domain induced polarization and ground-based imaging: a case study of the MontGombert landslide (French, Alps)

An electrical imaging technique called Induced Polarization peered inside landslides to reveal how water moves through the ground and where unstable zones are. Understanding internal water flow is key to predicting dangerous landslide movements and may improve monitoring as climate patterns change.

Characterizing landslide dynamics from time-lapse time domain induced polarization and ground-based imaging: a case study of the MontGombert landslide (French, Alps) · Aurore Carrier, Ombeline Meric, Pierre Bottelin · arXiv:2611.02868

Extended Field-of-View Passive mmWave Imaging Using Implicit Neural Representations

A technique uses Implicit Neural Representations to turn ghost images in passive millimeter-wave cameras into useful data and reconstruct a wider scene. It could improve all-weather autonomous driving or security screening by making existing sensors more capable without new hardware.

Extended Field-of-View Passive mmWave Imaging Using Implicit Neural Representations · Sanjaya Herath, Matthew R. Ziemann, Christopher A. Metzler · arXiv:2611.03396

All Gazettes · Remote Sensing