PredSL: prediction of snow cover and lake area change using cellular automata

Article Type

Research Article

Publication Title

Georisk

Abstract

An alarming effect of climate change is the accelerated glacier melting, leading to unstable glacial lakes. Rapid thawing may cause sudden lake overflow and discharge due to glacial ice or moraine dam failure. Therefore, monitoring snow coverage, glacier-fed lake surface area, and atmospheric variables is crucial for early glacial lake outburst flood warnings. ”PredSL” is proposed as a novel method for forecasting changes in glacial lake surface area and snow cover using stochastic cellular automata (CA). This study notably advances the field by integrating snow-cover changes with lake surface area expansion, addressing an important limitation of current work. A mathematical model is introduced that accounts for elevation, surface air temperature, and pixel geo-referenced coordinates. The model improves prediction accuracy by modifying thresholds based on Latin Hypercube Sampling (LHS) coefficients. The Mean-Squared-Error (MSE) of the predictions, compared to remote sensing images, ranged from 0.08 to 0.1. This performance surpasses that of existing methods, confirming its efficacy in disaster prediction. Satellite data for model development was gathered every 15-20 days, optimising data availability under clear conditions, facilitating predictions up to 15–20 days ahead. Another key aspect of the proposed method is its strong performance across diverse geographical contexts, demonstrated by its effectiveness in various GLOF incidents at different locations.

DOI

10.1080/17499518.2025.2567474

Publication Date

1-1-2025

Link to Full Text

Share

COinS