Speaker
Description
G. Rivolta, C. Orrù, M. Iesué, C. Camporeale, A. Mujeeb, M. Zribi, E. Ayari, N. Baghdadi,N. Sami, J. Cohen, J. Jorge Ruiz, J. Lemmetyinen, A. Fiengo, F. Ticconi, and D. Comite
Forward and inverse models developed by the Earth Observation (EO) scientific community describe, respectively, the relation between electromagnetic waves interacting with natural surfaces and the methodologies for retrieving bio-geophysical variables from remotely sensed data. Yet the current landscape – marked by the development of an extensive and heterogeneous suite of models – reveals some limitations. These encompass models not systematically implemented; models tested and validated only on small amounts of data; and limited integration of the models with the emerging Artificial Intelligence (AI)-based inversion techniques.
In this poster, we introduce the Wave Interaction Models EXploitation (WIMEX ) framework, which aims at addressing these challenges, in the frame of an ESA-funded project.
The framework, leveraging the unprecedented amount of EO data today available, creates a systematic approach to the development, validation, and use of existing and future forward and inverse models for increased efficiency and flexibility. Unlike existing solutions, WIMEX is designed to be model-independent, accommodate, manage, and operate any forward and inverse model independently of sensor and model objectives, supporting the use of cloud infrastructure for improving performance. This versatility reinforces the framework role in supporting existing and next-generation EO missions.
Moreover, being designed to interface with different EO data sources (EO data, in situ data, ad-hoc datasets, etc.), it supports the design, development and validation of forward models, the generation and storing of look-up tables and datacubes in a flexible way. It assists with the use of these outputs for testing and calibrating inverse models, through processes like neural networks training and use, as well as for performing sensitivity analyses across broad ranges of input parameters. It also facilitates the application of inverse models over extensive volumes of EO data, hence improving the statistics of the retrieved bio-geophysical variables and enhancing their interpretation. WIMEX also aims to boost the design of new and/or more accurate inverse models seamlessly combining their implementation with AI-based inversion techniques.
This added value helps improving the overall performance of the models and aligns with the evolving needs of the research community.
The prototype version of WIMEX, set to be released in the second half of 2024, demonstrates its efficacy and flexibility for remote sensing over land applications, specifically through the management and execution of forward and inverse models targeting soil moisture and snow water equivalent.
In the near future, upcoming versions will incorporate support to additional sensors and variables, as well as diverse missions, to enhance user experience and address a wider spectrum of requirements.
