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

TL;DR

This work tackles the challenge of predicting supraglacial lake evolution in Greenland under distribution shift by introducing RIC-TSC, which couples regionally informed causal discovery with a lightweight time-series classifier. Using J-PCMCI+ to identify lagged, invariant predictors from multi-modal remote sensing and reanalysis data, the authors feed causally grounded features into MiniROCKET followed by a RidgeClassifier, producing global and basin-specific models for four lake-fate classes. The approach yields robust generalization, achieving up to 12.59% accuracy gains in region-held-out (out-of-distribution) evaluations, and provides interpretable insight into immediate versus delayed drivers such as radar backscatter and optical water fraction, as well as climate variables. Overall, the study demonstrates that causal understanding can enhance both predictive performance and scientific interpretability for dynamic Earth surface processes under climate-related distribution shifts.

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.

Figures (6)

• Figure 1: Cumulative melt days (left) and melt day anomaly relative to the 1981--2010 average (right) across the Greenland Ice Sheet for April--October 2021. (Source: NSIDC/Thomas Mote, University of Georgia)
• Figure 2: Representative time series and imagery for four lake evolution classes. Grey lines: backscatter anomaly ($HV_{\text{anom}}$). Red dots: optical water fraction ($p_{\text{water}}$). Insets: Sentinel-2 (color) and Sentinel-1 (grayscale) before/after the event. Modified and extended from hossain2024time.
• Figure 3: Comparison of radar backscatter signal from within a lake (blue) and its surroundings (purple). A large contrast indicates surface water; signal convergence suggests refreezing or burial. Modified and adapted from hossain2024time.
• Figure 4: Causal discovery for multivariate time series. J-PCMCI+ recovers both lagged and contemporaneous causal links (black) while filtering spurious correlations (red), enabling interpretable modeling of SGL drivers across regions.
• Figure 5: Overview of the proposed causally-informed classification framework for supraglacial lake evolution. Multimodal remote sensing and reanalysis data are processed into daily per-lake time series. Using J-PCMCI+, we identify time-lagged causal relationships that reveal robust predictors of lake dynamics, both globally and by region. These causal features are then used in MiniROCKET-based sequence modeling to classify lake outcomes. The framework is evaluated under both in-distribution and out-of-distribution settings to assess generalization under spatial distribution shift.