AI4GHEObs:

Ground Heat Flux Estimates from Earth Observations and Hybrid Artificial Neural Networks

Duration: 2023-10-01 to 2025-09-30
Team: Francisco José Cuesta-Valero, Jian Peng
Funding agency: European Space Agency

The project aims to provide global estimates of ground heat flux and ground heat storage using Earth Observation datasets. Previous analytical models based on land surface temperatures and the solution of the one-dimensional heat diffusion equation cannot account for the influence of surface processes, such as vegetation cover, soil moisture and snow cover. Therefore, machine learning models will be used to derive ground heat flux based on satellite remote sensing products of land surface temperature, soil moisture, land cover, leaf area index and snow cover. Machine learning techniques are flexible enough to predict variables taking into account non-linear relationships between target and feature variables, allowing the combination of categorical data (e.g., land cover type) and numerical data (e.g., land surface temperatures).

This research uses Earth observation datasets from the European Space Agency (ESA) Climate Change Initiative (CCI) to derive long-term daily estimates of ground heat flux at near-global scale, filling the gap for observation-based ground heat flux data at daily temporal scales in the recent past. The proposed approach uses machine learning algorithms to derive ground heat flux from four ESA CCI products, consisting of land cover type, soil moisture, snow coverage, and land surface temperature, together with additional Earth observation datasets and in situ measurements. The main objectives of the proposed research are:

The AI4GHEObs global dataset of ground heat flux will allow us to study the influence of ground heating on other processes, such as soil respiration and extreme temperature events.