Back to overview

Fusing multiple sources of remote sensing data using textural information: high-resolution, high-frequency monitoring in alpine environments

Applicant Mariethoz Gregoire
Number 159756
Funding scheme Project funding (Div. I-III)
Research institution Institut des dynamiques de la surface terrestre Université de Lausanne
Institution of higher education University of Lausanne - LA
Main discipline Geomorphology
Start/End 01.09.2015 - 31.08.2019
Approved amount 261'464.00
Show all

All Disciplines (2)


Keywords (5)

remote sensing; lidar; topography; geostatistics; data fusion

Lay Summary (French)

Le monde moderne est caractérisé par une richesse d’information spatiale universellement disponible. Ceci est possible grâce à une diversité croissante des moyens d’observation (satellites, avions, drones, imagerie laser, etc). Les gigantesques jeux de données acquis régulièrement représentent une quantité d’information qui était inimaginable il y a seulement quelques décennies. Le revers de la médaille est que pour utiliser cette information à son plein potentiel, il est nécessaire de développer des méthodes permettant de fusionner des données provenant de sources hétérogènes.
Lay summary

Dans ce projet nous nous attaquons à la question de la fusion des données spatialisées. L’objectif principal est de développer une méthodologie pour la combinaison de données provenant de sources différentes, en incorporant des algorithmes provenant des domaines de la géostatistique et de l’imagerie numérique. Les résultats seront appliqués à la quantification des processus géomorphologiques de haute-montagne, en combinant des données de topographie provenant de différentes sources.

L’imagerie laser acquise par LIDAR est très précise (de l’ordre du centimètre), mais repose sur des campagnes de mesure ponctuelles qui sont limitées à une région spécifique. D’un autre côté, des images satellite très haute résolution (<50cm) sont disponibles quotidiennement. En fusionnant ces jeux de données, le but est d’obtenir une information topographique dont la précision est de l’ordre de celle du LIDAR, avec la couverture spatiale et la fréquence d’observation des mesures par satellite.

Direct link to Lay Summary Last update: 29.03.2015

Responsible applicant and co-applicants



Downscaling Images with Trends Using Multiple-Point Statistics Simulation: An Application to Digital Elevation Models
Rasera Luiz Gustavo, Gravey Mathieu, Lane Stuart N., Mariethoz Gregoire (2019), Downscaling Images with Trends Using Multiple-Point Statistics Simulation: An Application to Digital Elevation Models, in Mathematical Geosciences, 1-43.


Group / person Country
Types of collaboration
Matthew McCabe, KAUST Saudi Arabia (Asia)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Jim Chandler, Loughborough University Great Britain and Northern Ireland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication

Associated projects

Number Title Start Funding scheme
162754 Integrating spatial predictions of vegetation, soils, geomorphology and hydrology for improved assessment of ecosystem services under climate change 01.01.2016 Interdisciplinary projects
162882 Extracting new information from old satellite imagery by using n->m colorization 01.01.2016 Project funding (Div. I-III)


Today’s high-resolution remote sensing methods allow monitoring the Earth surface with an unprecedented level of detail. While the variety of available data sources opens avenues for high-resolution and continuous information, it also results in a series of issues related to data consistency, data homogenization and to the quality of Earth surface processes representation. On the one hand, important datasets have been produced historically by aerial photography, and more recently by satellites which can acquire images at increasingly high resolutions. At the same time, ground-based methods such as LIDAR (Laser Imaging, Detection and Ranging) allow rapid acquisition of very dense, millimeter to centimeter-scale datasets describing topographic surfaces. Such data are invaluable to understand various environmental processes, for example changes in mountain landscape, either to quantify erosion processes or climate-induced changes in snow and ice cover. However, several challenges remain before such data can be used to its full potential. One limitation is that most remote sensing measurements have gaps and sample the Earth surface in an uneven manner. Another limitation is that the frequency and resolution of measurements can vary depending on the measurement device. For example, satellite retrievals are recurrent with sometimes a daily overpass, whereas the frequency of ground-based measurements can depend on the timing of fieldwork campaigns. In order to optimize the information that can be obtained from such data, work is needed to fuse different data sources notably where deficiencies in one can be addressed with the help of others. This project addresses the issue of information fusion in remote sensing by developing a workflow for the combination of datasets coming from different sources, using algorithms recently developed in geostatistics and computer graphics. The data processing workflow will involve geostatistical methods based on training images, a family of algorithms commonly named multiple-point geostatistics (MPS). A training image is an analog of the spatial phenomenon studied, from which complex patterns are extracted and used to formulate a non-parametric model of spatial and temporal variability. In this project, we will test the idea that it is possible to derive multivariate training images from different remote sensing sources, and to use these training images to obtain spatially continuous, high-frequency datasets. The developed workflow will be applied to high-resolution topography change monitoring in high-altitude alpine areas. Topographic data from LIDAR, precise but infrequent and limited in extent, and aerial photogrammetry, less precise, also infrequent and often with data gaps but with larger area coverage, will be combined with frequent high-resolution (<50cm), high-frequency (daily) satellite retrievals, which will be for the first time available since 2015. The result will be a merged dataset that has higher resolution and precision than the original topographic data, and the retrieval frequency and spatial coverage of satellite images. Such rich and consistent datasets will inform geomorphological attributes at a high spatial and temporal resolution, allowing derivation of patterns of change that can be generalized to other regions in high-altitude alpine environments.