Day co-sponsored by the GdR IASIS and the COMET Traitement du Signal et des Images of CNES, with the support of Polytech Annecy-Chambéry, Université Savoie Mont Blanc.
Description
Remote sensing is an essential tool for studying the evolution of the cryosphere: seasonal snow, glaciers, permafrost, sea, lake and river ice... Synthetic Aperture Radar (SAR) is a regular source of information, providing images day and night, regardless of weather conditions. On a local or regional scale, this data can be used to measure physical parameters such as glacier surface flow velocity, variations in snow depth, or to map wet snow. But the response of snow/ice/vegetation environments to electromagnetic waves is complex, and requires a combination of statistical signal processing, modeling and machine learning to extract the required information from SAR images.
The aim of this day is to present recent advances in the use of SAR images and their time series for hydrology and glaciology applications. These applications require major methodological developments for automated processing of colossal masses of data from space agency archives or periodically acquired by current radar satellites: the Sentinel-1 and Radarsat constellations in C-band, TerraSAR/Tandem-X, PAZ and CosmoSkyMed in X-band or by the future NISAR mission (USA/India) in S and L-band. By bringing together researchers from the fields of radar image processing and geosciences, this day will enable us to cross-reference the latest methodological advances with the knowledge and expectations of cryosphere specialists.
Guest speakers
- Praveen K. Thakur, Indian Institute of Remote Sensing (IIRS), Dehradun, guest researcher at LISTIC
- Fatima Karbou, Centre d'Etude de la Neige (CEN), Météo-France
- Florence Tupin (Télécom ParisTech)
Call for papers
We are launching a call for papers for those wishing to present their work in the form of an oral presentation. Proposals (title, abstract, authors, assignment) should be emailed to emmanuel.trouve@univ-smb.fr by Wednesday, September 4, 2024.
Organizers
- LISTIC: Emmanuel Trouvé, Ammar Mian, Yajing Yan, Abdourahmane M. Atto
- CNES : Nicolas Gasnier
Practical information
- Location: Annecy, Polytech Annecy-Chambéry, 5 chemin de Bellevue, Annecy-le-Vieux
- Room : B212 (Boardroom)
- Coffee breaks and lunch are provided and financed by COMET.
- Opening hours: 10 a.m. - 5 p.m.
- Teams link for remote monitoring: Click here
Schedule for the day
Abstracts of presentations :
- Florence TupinDeep learning methods for speckle reduction in multi-channel SAR images".
L. Denis, E. Dalsasso, C. Ulondu-Mendes, F. TupinIn the first part of this talk we will review some recent deep learning based methods for speckle reduction.Then we will focus on a new self-supervised approach called MuChaPro (Multi-Channel Projections)leveraging on single channel despeckling to handle multi-channel data.
- Praveen K. Thakur, "Current Status and Future Innovations in Space Based Observations for Cryosphere Research".
Satellite remote sensing continues to play an ever increasing, important and indispensable role in the regular mapping and monitoring of global cryosphere. The Polar ice-sheets, ice shelves, sea-ice, mountain glaciers and seasonal snow, and permafrost remains most sensitive and under constant scrutiny from the vagaries of global climate change. In this talk, I will highlight the main active and priority areas of research in cryosphere, and various physical principles and satellite sensors used for making observations of cryosphere components. Some of key research highlights from satellite based seasonal snow cover and snowpack monitoring and glacier dynamics of Himalaya shall be presented. The role of such observation in snow and glacier melt runoff models and inputs to water security of Himalayan Rivers will also be discussed. The use of SAR data for snowpack characterization and glacier dynamics studies along with role of machine learning shall be highlighted. The role of satellite observation in monitoring of snow/glacier hazards such as snow avalanches, rock/icefalls and glacier lake outburst floods shall be presented. The global space missions ongoing and planned relevant for cryosphere monitoring such as Scatsat-1, Sentinel-1, Radarsat, Icesat-1, 2, PLASR, NISAR, TRISHNA, CRISTAL shall be highlighted.
- Adina Racoviteaunu, "Spatio-temporal evolution of Glacier Facies in Himalaya using SAR Remote Sensing: examples from the Khumbu Himalaya"
Adina Racoviteanu, Baptiste Georjon, Aravind Manimaran, Jean-Pierre Dedieu
Understanding glacier behavior with respect to climate patterns and its impact on water resources required a detailed study of glacier surfaces (snow, ice, firn etc.). Such measurements are difficult to obtain in the field in remote and difficult to access regions such as the Himalayas, particularly in areas frequently covered by clouds such as the monsoon influenced part where optical remote sensing methods fail. Here we use Synthetic Aperture Radar (SAR) time series data from Sentinel-1 for the period 2017 to present explore glacier facies and transitional snow lines and their spatio-temporal evolution in the monsoon-monsoon-influenced area of Khumbu Himalaya, with a focus on the field-monitored Mera Glacier. We apply the Nagler ratio method to differentiate the various glacier facies and explore the seasonal variability of the ouputs and compare these with transitional snow lines detected from Planet/Venµs imagery (5 m spatial resolution) and spectral unmixing of Sentinel-2 data. We discuss current challenges associated with the use of SAR imagery in C-band to separate glacier facies in monsoon-dominated areas.
- Fatima KarbouMethods for monitoring snowmelt using SAR imagery".
This presentation will review some of the work undertaken to explore the potential of image segmentation methods applied to Sentinel-1 SAR images to improve snow detection. This includes addressing the criteria for selecting snow-covered pixels, selecting the reference image, and introducing new methodological developments that better account for topography.We will also discuss the value of monitoring snow elevations derived from satellite observations, particularly in monitoring snow retreat elevations/dates during the melt season and minimum and maximum wet snow elevations at the scale of mountain ranges. New approaches to derive this information directly, without snow detection, will be presented and their implications for merging images of different types and resolutions will be discussed.
- Abdourrahmane Mr. Atto, " Deep learning for snow analysis from multimodal remote sensing images. "
Thu Trang Lê, Abdourrahmane M. AttoEmmanuel TrouvéFatima Karbou
With the increasing accessibility of snow products derived from Synthetic Aperture Radar (SAR) and optical data, like Sentinel-1 wet snow and Sentinel-2 total snow, users have benefited from improved snow mapping and monitoring. However, snow mapping in the mountainous areas remains challenging due to the difficulty of obtaining reliable ground truth data on steep mountain terrain. Deep semantic learning frameworks are analyzed in this study to help in detecting wet snow in mountainous areas.
- Laurane Charier, "Glacier surges observed using SAR backscattering and image correlation".
Laurane Charier, Amaury Dehecq, Elise Colin, Lei Guo, Fanny Brun, Romain Milan and Luc Beraud
- Swann Briand, "Merging time series of SAR and optical remote sensing images of different resolutions for snow cover monitoring".
Monitoring snow cover using optical remote sensing images is a widely used and documented method, but is often limited by the presence of clouds. SAR, on the other hand, despite a longer revisit time and complex geometry, is not sensitive to weather conditions and can retrieve snow cover properties (height, presence/absence of wet snow). We propose to use this wealth of information to produce snow cover maps through a SAR image segmentation task using a neural network trained with data derived from MODIS optical images. In this context, we study the quantity and quality of data required for training. We propose different methods of interpolation (linear interpolation, Kalman smoothing) of optical data in order to increase the quantity of data available, at the expense of quality. We are also studying the influence of different combinations of SAR channels at the network input. In particular, we want to understand whether channels directly representative of snow physics are more useful than the raw channels from which they are derived.
- Yajing Yan, "Glacier thickness estimation at a regional scale with deep learning".
L. Lopez-Uroz, Y. Yan, A. Benoit, A. Rabatel, S. Giffard-Roisin
Mountain glaciers play a critical role for mountain ecosystems and society with major concerns related to their future evolution and related water resources. Modeling glacier future evolution allows anticipating climate change impacts and informing policy decisions. It relies on accurate ice thickness estimation at regional scales. We propose a deep learning based approach in a supervised learning framework for ice thickness estimation at a regional scale from surface ice velocity measurements and a digital elevation model. A neural network model built upon a ResNet architecture is proposed based on the trade-off between the model complexity and the prediction efficiency. Promising results are obtained from data including 1400 glaciers in the Swiss Alps, highlighting the potential of deep learning based approach for large scale ice thickness estimation.