Terms of References (ToR)

Gravity field data provide invaluable insights into various aspects of Earth’s dynamics, ranging from monitoring changes in water storage to understanding mass redistribution and geophysical phenomena. These data play a crucial role in fields such as hydrology, geophysics, and climate science. However, traditional methods of gravity field determination and analysis often face challenges in terms of data processing, spatial resolution, and filling temporal gaps.

This working group seeks to address these challenges by leveraging the power of AI and machine learning techniques. AI/ML algorithms have demonstrated remarkable capabilities in extracting patterns from large-scale datasets, automating data processing tasks, and quantifying uncertainties. Recent generative AI algorithms have further pushed the envelope of learning from foundation models. Moreover, ML techniques offer the potential to address the issue of spatial resolution in gravity field analysis. The exploration of downscaling techniques will contribute to improving the spatial resolution of gravity field data, allowing for more detailed analyses of water dynamics at localized levels.

Additionally, machine learning models can fill the temporal gaps in gravity field datasets by learning from existing data and predicting missing values. By training these models on historical data from missions such as GRACE, we can generate reliable estimates during periods when data collection is unavailable. This capability ensures a more complete and continuous understanding of gravity field variations and facilitates more accurate analyses of temporal trends and anomalies.