remote-sensing

This paper uses 1,744 Landsat and Sentinel-2 satellite images of Robert Smithson's Spiral Jetty to compute multi-feature complexity signatures, revealing that image complexity acts as a leading indicator of lake elevation and climate metrics over 40 years.

Planet Labs has announced an open satellite feed, providing public access to its satellite imagery data.

This research paper examines worst-case scenarios for glacial lake outburst floods in transboundary Himalayan basins, assessing current and future hazards through integrative approaches and multiple scientific studies.

KnowChange leverages vision-language models to guide flexible and diverse change data synthesis, improving remote sensing change detection training by generating cost-effective synthetic data.

This paper introduces a scalable, modular geospatial machine learning framework for asset-level power-line risk modelling from lightning and vegetation hazards using remote sensing data, enhancing utility asset management and climate resilience.

This paper evaluates machine learning models using free satellite data (radar interferometry and weather) to predict which Himalayan glacial lakes are at risk of outburst floods and when triggers occur, achieving ROC scores up to 0.89 in Nepal and providing a ranked watchlist.

Presents a globally trained CNN for forest above-ground biomass estimation using multi-sensor satellite data, with a sparse field calibration workflow to adapt predictions locally. Achieves improved accuracy over uncalibrated global models and ESA CCI products.

This paper proposes an unsupervised fire-zone segmentation method combining watershed detection with K-means clustering to improve short-term wildfire prediction, showing consistent gains over grid-based approaches across multiple French departments and forecasting models.

GeoArbiter proposes a training-free pipeline that selectively injects image-unverifiable geographic facts into remote-sensing multimodal LLMs to reduce knowledge hallucinations while preserving retrieval accuracy gains.

Introduces Obshazard-bench, a real-time, observation-driven benchmark for evaluating multimodal foundation models on disaster intelligence from raw Earth observation streams, spanning 8 disaster categories across 60+ countries.

RRS-10K is a benchmark dataset for evaluating vision-language models on rare remote sensing image interpretation, containing over 10,000 military-related images and multiple task formats. Evaluation of 52 models reveals moderate zero-shot performance and weaknesses in visual grounding and complex reasoning.

Introduces RSMeM, a knowledge-enhanced memory evolution mechanism for remote sensing agents that bootstraps LLMs with domain knowledge and iteratively integrates failure experience to improve multi-step tool execution, achieving 6% accuracy gain on DeepSeek-V3.2 with minimal additional tokens.

Introduces OVEarth-Bench, a benchmark for open-vocabulary Earth observation that broadens category coverage and query diversity, revealing that current methods remain limited and MLLM-based approaches perform best.

This paper presents a human-in-the-loop bootstrapping method for detecting PFM-1 mines in UAV hyperspectral imagery, showing that ACE with bootstrapping can find all targets in 2 rounds of inspection, while aggregate ROC-AUC scores hide large operational differences between detectors.

The paper proposes FUSAR-R1, a large-scale reasoning model for SAR image interpretation that uses chain-of-thought reasoning and reinforcement learning to achieve better performance than existing models.

Delineate Anything v2 is a globally scalable foundation model for agricultural field boundary mapping, outperforming state-of-the-art by 103.3% relative gain in [email protected], using a 73-million-instance multi-resolution dataset spanning 61 countries and a manually curated 100-country evaluation benchmark.

This paper surveys the emerging paradigm of Geospatial Foundation Models (GeoFMs), which are pre-trained on massive geospatial datasets to enable rapid fine-tuning and zero-shot analysis of satellite and aerial imagery. It covers the paradigm shift, model adaptation strategies, and a forward-looking vision of Agentic Geospatial Reasoning using LLMs as orchestrators.

This chapter reviews design principles and current landscape of foundation models for Earth observation, highlighting the need for domain-specific adaptation, physically plausible representations, and consistent evaluation benchmarks. It includes case studies on harmful algal bloom prediction and adaptive monitoring station selection.

The paper proposes Geo-Anchored Cloud Removal (GACR), a framework that uses Observation-Anchored Residual Flow and Geo-Contextual Prior Alignment to remove clouds from optical remote sensing images while preserving semantic structures for downstream tasks.

Satellite interferometry from ESA's Sentinel-1 shows ground displacement up to 30 centimeters after Venezuela's twin earthquakes, revealing deformation along the San Sebastián fault system.