Interpretable Machine Learning for Spatiotemporal Modeling of Land Surface Thermal Dynamics |
|
Shijie Jiang,
Christian Reimers
,
Gregory Duveiller
,
Alexander Brenning
,
Alexander Winkler
,
Markus Reichstein
|
Project descriptionUnderstanding the thermal dynamics and energy balance of the land surface is critical for effective climate change mitigation strategies, reliable climate risk assessment and preparedness, sustainable biodiversity conservation, and food security. Land surface temperature (LST) represents a key parameter in the physics of land surface thermal processes from local to global scales. The spatiotemporal dynamics and variations of LST provide essential insights into how different surfaces interact with solar radiation, which are strongly shaped by natural and anthropogenic changes in different regions. Despite significant advances, existing models often struggle to accurately capture the complex spatiotemporal dependencies, non-linear interactions, and feedback mechanisms (such as the relationship between vegetation and LST) that govern LST dynamics and are typically subject to uncertainties resulting from parameterization schemes. Advanced data-driven techniques capable of exploiting the now widely available satellite and in-situ observations offer the potential to deepen process understanding of these critical environmental processes and to complement insights from mechanistic models.The overarching goal of this PhD project is to elucidate the intricate spatiotemporal and non-linear patterns that drive LST variations from observational data, in particular the underlying mechanisms and feedback loops between LST, vegetation, and climate. The project will explore the potential of state-of-the-art data-driven methods to model and understand the complex interactions and dependencies between thermal dynamics and other environmental factors. For example, interpretable machine learning and causally/physically hybrid models can be developed to disentangle and quantify the feedback loops between LST and vegetation dynamics across space and time. Ultimately, the results of this research are expected to significantly improve our understanding of the complex signals that alter land surface energy dynamics. The methodological developments can help promote trustworthy, responsible AI practices across scientific domains. The prospective PhD student is encouraged to explore their own ideas in the context of causally/physically interpretable machine learning for the purpose of modeling and understanding LST spatiotemporal dynamics in land-atmosphere interactions, with a particular focus on the implications of global warming and greening. Working groupThe successful PhD candidate will work in the Department of Biogeochemical Integration at the Max Planck Institute for Biogeochemistry. The working group offers long-standing expertise and experience in the various fields relevant to this project: Machine Learning (Christian Reimers, Shijie Jiang), remote sensing (Gregory Duveiller), ecology and biophysics (Alexander Winkler, Gregory Duveiller), geospatial analysis (Alexander Brenning, Alexander Winkler), and environmental informatics (Shijie Jiang, Christian Reimers, Alexander Brenning). The PhD candidate will benefit from the European Lab for Learning and Intelligent Systems (ELLIS) and will work closely with the ELLIS Unit Jena.RequirementsApplications to the IMPRS-gBGC are open to highly motivated and qualified students from all countries. Prerequisites for this PhD project are:
|