Accurate and up-to-date land cover information constitutes key environmental data for developing efficient policies in this era of resource scarcity and climate change. New Satellite Image Times Series offer new opportunities for detecting land cover class transitions. Nevertheless, the challenges of the "Big Data" have become imminent for the exploitation of this massive flow of data. Deep generative models are one of the most promising tools for big data analysis. The use of such models has just started to emerge in the remote sensing. In this project, Generative Adversarial Networks and Variational Autoencoders want to be explored to face common remote sensing challenges, which are the lack of reference data and the exploitation of complex and heterogeneous information. The originality of the project relies on the development of new online change detection methodologies by using generative models, which incorporate the temporal dynamics of the data and physical knowledge constraints

The overarching objective of HILIAISE is to better understand and model the human imprint on semi-arid energy and water cycles. To obtain this objective: 1) The project will have a long term field campaign with a 15-day Special Observing Period during summer 2020 over the Ebro basin in northeastern Spain when land surface heterogeneities are at their maximum. This campaign will focus on surface and boundary layer contrasts between irrigated and non-irrigated natural regions, and quantifying the water resource demand. 2) The project will use a multidisciplinary approach using a suite of hydrological, land-surface and meteorological models focusing on both using existing and improving parameterizations of anthropization and semi-arid surfaces. The improved representation of anthropogenic and semi-arid processes in models will form the foundation for improving the understanding and prediction of water resource changes for the recent past, present and under future climate change.

Agronomic, hydrologic, meteorological and climatic predictions rely on our ability to accurately represent soil evaporation (E) process, which is the boundary condition for the soil and atmosphere. For such wide range of applications, E should be modeled over extensive areas at multiple scales. Since the 60s many E models have been developed. Mechanistic theoretically-based models have been very useful to understand and describe the physical processes regulating E including gravity drainage, capillary rise, vapor diffusion, and the interplay with the atmospheric evaporative limitation. However, their regionalization has been a notorious challenge because of the unavailability and high uncertainty of soil hydraulic properties over extended areas (~100 m - 100 km) and the lack of data at such scales. Simplified models have been generally used across different application scales but their regionalization has been based on empiricism or ad hoc relationships with soil hydraulic properties or texture. In fact, none of the existing E formulations has been validated over an extensive range of soils and soil-atmospheric conditions and no consensus exists on the best way to parameterize E. The fast development of local, regional and global monitoring networks (e.g. Ameriflux, GHGEurope) now permits the advent of improved models. A related problem is that E is not directly observable using remote sensing platforms although our capability to monitor E-related quantities such as soil moisture is growing rapidly with new remote sensing technologies such as SMOS (Soil Moisture and Ocean Salinity) and SMAP (Soil Moisture Active Passive) L-band missions. Several studies have shown the potential of combining shortwave-derived vegetation cover, thermal-derived surface temperature, and microwave-derived surface soil moisture to partition evapotranspiration (ETR) into E and plant transpiration (TR) and to indirectly retrieve E. However, no remote sensing method has come up yet because spaceborne temperature and soil moisture products are readily available at different spatial resolutions, and because there is no quasi-instantaneous E model that combines consistently these data. The objective of this proposal is therefore to fill this gap in the E representation by developing an original phenomenological (intermediate between theory and experiment) model 1) that is valid over a wide range of soils and soil-atmospheric conditions while based on the data available at the application scales (scale-aware model) and 2) that can be coupled to readily-available remotely sensed vegetation cover, surface temperature and surface soil moisture (remote sensing-aware model). The different steps will be: 1) to develop a texture-based E model using an extensive in situ data set and compare it with state-of-the-art models, 2) to investigate three (mechanistic, data-based, mathematical) approaches to quantitatively evaluate the impact of soil moisture profile and soil surface state on E, 3) to implement the new model in a thermal-based disaggregation scheme of SMOS/SMAP soil moisture and to validate the E retrieved at high (~100 m) resolution using in-situ measurements collected under various pedo-hydro-climatic conditions in Chile, France, Morocco and Spain, and 4) to implement the new model in the CNRM (ISBA/SURFEX) and ECMWF (H-TESSEL) land surface models and to validate large scale simulated E using the remotely sensed estimates (previously validated at high resolution using in-situ measurements) aggregated at the corresponding resolutions.

The objective of LabCom RT-Twin is to improve the work in the field of Earth observation (i.e., remote sensing), such as the preparation of space missions and the exploitation of measurements from existing observation systems. The approach is based on the physical modelling of remote sensing measurements, in particular with the three-dimensional (3D) radiative transfer (RT) model DART (https://dart.omp.eu) developed at CESBIO. DART has now reached a level that allows accurate and efficient simulation of remote sensing measurements of natural and urban landscapes, for any observation configuration (viewing direction, spatial resolution, spectral resolution, any altitude, etc.), from ultraviolet to thermal infrared, including LiDAR signal and solar-induced chlorophyll fluorescence (SIF). In addition, CNRS certified DART on July 2023, and Toulouse III University already delivered more than 600 DART licences to scientists. The aim is to take advantage of the scientific progress in TR modelling made by the DART team at CESBIO, to consolidate DART, adapt it to the needs of ACRI-ST and thus improve the potential of ACRI-ST in the field of Earth observation. ACRI-ST and CESBIO have collaborated several times for the optimization of Earth Observation missions with DART. This joint work has confirmed the validity and the need to create a common laboratory to progress in the transformation of the DART model into an operational tool for a wider audience (scientists and technicians of Earth observation). This approach ensures the valorisation of DART. It is a two-way process: ACRI-ST will benefit from CESBIO's expertise on the DART tool to jointly enhance its use, and CESBIO will benefit from an ideal context to maintain and "keep alive" DART in its scientific excellence, in a very competitive environment. These joint benefits are naturally built on an iterative approach between the activities of the two entities. Two areas in particular are being considered for this joint laboratory: 1) DART enhancement: - TR modeling in DART for operational simulation (i.e., computation time, memory space) of realistic and physically correct images of natural and urban landscapes. - Landscape models and landscape elements (e.g., house, tree, plant) used by DART, with in particular georeferencing of DART models and products and coupling with plant growth models, and operational parameterization of the atmosphere. - Propagation of uncertainties within DART due to unavoidable uncertainties concerning in particular the optical properties of landscape elements. 2) Valorization of research work: - Preparation / pre-dimensioning of new space missions, with evaluation of their performance, based on DART simulations. - Qualification of new processing algorithms on reference landscape simulations, and support for validation activities and space applications. - Promotion of developments with an on-demand image generation and expertise service for space industry players. - Promotion and valorization of developments within the framework of institutional space missions and/or the NewSpace ecosystem.

The MAESTRIA project (Multi-modAl Earth obServaTion Image Analysis) aims to solve the methodological challenges related to the fully automatic analysis of the massive amount of images acquired by Earth Observation platforms. MAESTRIA targets to generate land-cover and land-use descriptions at country scale at various spatial resolutions and sets of classes. The ultimate goal is to provide a continuum of spatially and semantically consistent products, that are relevant for many end-users and applications. Both public policies at local or national levels and scientific models would benefit from such kinds of products for climate modelling, urban planning, crop monitoring or impact assessment of surface changes. The output of the MAESTRIA project will be two-fold: (i) methods that leverage current challenges in Earth Observation image analysis; (ii) a large range of precise and up-to-date land-cover maps available over very large scales from 2m to 100m. Both will be made freely available so as to stimulate research and commercial services built upon such maps. Many global and land-cover geodatabases have been established during the last two decades. However, they do not meet the current requirements in terms of semantic and spatial accuracy and updateness. In parallel, a large body of literature has tackled automatic EO data exploitation. However, most of the existing papers are limited to a specific environment, site or sensor, and a specific need. They are not flexible and not adapted to the new paradigm in EO with the advent of satellite missions with short revisit time and increased spatial resolutions. What makes the analysis task challenging now is the heterogeneous physical nature of such images. One has now to design adequate methods to optimally exploit the complementary information provided by images acquired from a large variety of sensors. The current situation is exacerbated when addressing the upscaling issue, i.e., when classifying this amount of images at large scales while trying to guarantee a homogeneous accuracy in all areas of interest. Three main methodological challenges will be tackled. First, it deals with multiple sensor fusion so as to extract the most meaningful knowledge from the huge amount of heterogeneous data and subsequently its sparse representation. This will ensure to limit the amount of information required to generate a consistent land-cover map in a timely manner, by facilitating and improving the underlying supervised classification task. Secondly, no very large scale learning techniques have been proposed so far so as to deal with noisy information. Noise may come from the images themselves but also from the input labels. Two main tasks are considered: the design of (i) semi-supervised learning strategies that rely both on unlabelled and labelled samples to alleviate the needs of large/well-balanced and accurate training set, and (ii) efficient optimization procedures over millions of samples with thousands of features. Thirdly, we aim to develop methods that derive automatically new land-cover products with different spatial and semantic resolutions out of those produced in the two first steps. As a consequence, we target to obtain a continuum of adapted land-cover layers, both in terms of spatial scales (2->50-100m) and semantics. The core of this task relies on multi-scale semantic segmentation that can handle uncertainties and inconsistencies between scales. In addition, we will link our generic national land-cover maps with various end users' needs, both at scientific and institutional levels, and eventually, integrate the developed algorithms into a common open-source framework for further dissemination and data exploitation. MAESTRIA is embedded into the Theia Land Data Centre initiative: we will benefit from the existing infrastructures, datasets, institutional partners, and first connections with private companies through regional animation networks and booster initiatives.