Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Survey of Foundation Models for Environmental Science

Runlong Yu, Shengyu Chen, Yiqun Xie, Xiaowei Jia

TL;DR

This survey analyzes how foundation models can transform environmental science by integrating diverse data sources, modeling spatiotemporal dynamics, and adapting to multiple tasks. It surveys applications across forward prediction, data generation, data assimilation, downscaling, model ensembling, and decision-making, and details end-to-end development, including data collection, architecture, training, tuning, and evaluation. Key contributions include a structured overview of current methods, delineation of methodological design choices, and discussion of challenges such as trust, uncertainty, and computational efficiency, with guidance for future open-data and collaborative research. The work highlights practical implications for resource management and policy by enabling more robust, scalable, and interpretable environmental analytics through unified, pre-trained representations. It aims to catalyze interdisciplinary collaboration and accelerate the adoption of cutting-edge knowledge to support sustainable development and resilience.

Abstract

Modeling environmental ecosystems is essential for effective resource management, sustainable development, and understanding complex ecological processes. However, traditional methods frequently struggle with the inherent complexity, interconnectedness, and limited data of such systems. Foundation models, with their large-scale pre-training and universal representations, offer transformative opportunities by integrating diverse data sources, capturing spatiotemporal dependencies, and adapting to a broad range of tasks. This survey presents a comprehensive overview of foundation model applications in environmental science, highlighting advancements in forward prediction, data generation, data assimilation, downscaling, model ensembling, and decision-making across domains. We also detail the development process of these models, covering data collection, architecture design, training, tuning, and evaluation. By showcasing these emerging methods, we aim to foster interdisciplinary collaboration and advance the integration of cutting-edge machine learning for sustainable solutions in environmental science.

A Survey of Foundation Models for Environmental Science

TL;DR

This survey analyzes how foundation models can transform environmental science by integrating diverse data sources, modeling spatiotemporal dynamics, and adapting to multiple tasks. It surveys applications across forward prediction, data generation, data assimilation, downscaling, model ensembling, and decision-making, and details end-to-end development, including data collection, architecture, training, tuning, and evaluation. Key contributions include a structured overview of current methods, delineation of methodological design choices, and discussion of challenges such as trust, uncertainty, and computational efficiency, with guidance for future open-data and collaborative research. The work highlights practical implications for resource management and policy by enabling more robust, scalable, and interpretable environmental analytics through unified, pre-trained representations. It aims to catalyze interdisciplinary collaboration and accelerate the adoption of cutting-edge knowledge to support sustainable development and resilience.

Abstract

Modeling environmental ecosystems is essential for effective resource management, sustainable development, and understanding complex ecological processes. However, traditional methods frequently struggle with the inherent complexity, interconnectedness, and limited data of such systems. Foundation models, with their large-scale pre-training and universal representations, offer transformative opportunities by integrating diverse data sources, capturing spatiotemporal dependencies, and adapting to a broad range of tasks. This survey presents a comprehensive overview of foundation model applications in environmental science, highlighting advancements in forward prediction, data generation, data assimilation, downscaling, model ensembling, and decision-making across domains. We also detail the development process of these models, covering data collection, architecture design, training, tuning, and evaluation. By showcasing these emerging methods, we aim to foster interdisciplinary collaboration and advance the integration of cutting-edge machine learning for sustainable solutions in environmental science.

Paper Structure

This paper contains 19 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Historical and Conceptual Overview
  3. Environmental Modeling
  4. Emergence of Foundation Models
  5. Application-Centric Objectives
  6. Forward Prediction
  7. Data Generation
  8. Data Assimilation
  9. Downscaling
  10. Model Ensembling
  11. Decision-making
  12. Method Design
  13. Data Collection
  14. Foundation Model Architecture
  15. Training
  16. ...and 4 more sections

Figures (2)

  • Figure 1: Application-centric objectives and advancements enabled by foundation models.
  • Figure 2: Model design workflow for foundation models in environmental science.