Causal Time Series Modeling of Supraglacial Lake Evolution in Greenland under Distribution Shift

2025-10-16

AI Summary

To address poor out-of-distribution generalization in predicting Greenland supraglacial lake evolution under distributional shift, this paper proposes RIC-TSC: a framework that embeds lag-aware causal discovery (J-PCMCI+) into time-series modeling. It integrates multi-source remote sensing data—including Sentinel-1/2 and Landsat-8—and CARRA meteorological reanalysis to jointly identify both region-specific and temporally invariant causal relationships. By explicitly modeling causal mechanisms rather than statistical correlations, RIC-TSC achieves both mechanistic interpretability and lightweight deployment. On a benchmark of 1,000 labeled lakes, it improves out-of-distribution accuracy by up to 12.59% over correlation-based baselines. This work presents the first causally driven temporal modeling of supraglacial lake dynamics, advancing Earth system modeling from purely statistical association toward mechanism-grounded, transferable prediction.

Technology Category

• Machine Learning: Causal Learning
• Computer Vision: Remote Sensing / Geospatial AI
• Reasoning under Uncertainty: Causality

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Causal modeling offers a principled foundation for uncovering stable, invariant relationships in time-series data, thereby improving robustness and generalization under distribution shifts. Yet its potential is underutilized in spatiotemporal Earth observation, where models often depend on purely correlational features that fail to transfer across heterogeneous domains. We propose RIC-TSC, a regionally-informed causal time-series classification framework that embeds lag-aware causal discovery directly into sequence modeling, enabling both predictive accuracy and scientific interpretability. Using multi-modal satellite and reanalysis data-including Sentinel-1 microwave backscatter, Sentinel-2 and Landsat-8 optical reflectance, and CARRA meteorological variables-we leverage Joint PCMCI+ (J-PCMCI+) to identify region-specific and invariant predictors of supraglacial lake evolution in Greenland. Causal graphs are estimated globally and per basin, with validated predictors and their time lags supplied to lightweight classifiers. On a balanced benchmark of 1000 manually labeled lakes from two contrasting melt seasons (2018-2019), causal models achieve up to 12.59% higher accuracy than correlation-based baselines under out-of-distribution evaluation. These results show that causal discovery is not only a means of feature selection but also a pathway to generalizable and mechanistically grounded models of dynamic Earth surface processes.

Problem

• Model supraglacial lake evolution in Greenland under distribution shifts
• Identify causal predictors using multi-modal satellite and reanalysis data
• Improve generalization accuracy for dynamic Earth surface processes

Innovation

• Embeds lag-aware causal discovery into sequence modeling
• Uses Joint PCMCI+ to identify region-specific predictors
• Supplies causal graphs to lightweight classifiers globally