Machine Learning Applied to the Analysis of Glacier Masses

Abstract

Glaciers play a vital role as climate change indicators, offering valuable insights into global climate evolution and the wide-ranging impacts of glacier melting on nearby cities, including water supply and ecosystems. The hydrology of Antisana glacier, which provides high-quality drinking water to Quito and its surrounding region, is of critical research importance. In this context, this work aims to explore the potential of machine learning in predicting the mass balance of glaciers in Ecuador, specifically in Antisana Glacier 12 $\alpha$. To achieve this, a comprehensive dataset of climatic variables from a region of Antisana was collected and processed using TerraClimate and ERA5 datasets. Many ARIMA models, were developed and compared. The ARIMA(0,0,1) configuration provided reliable predictions. Precipitation and surface pressure were identified as significant variables, with precipitation having a substantial effect on glacier mass balance, as confirmed by the forecast results. This study emphasizes the importance of machine learning to improve our understanding of glacier dynamics and support informed decision-making in the face of climate change. Collecting additional specific data to enhance accuracy and comprehensive experience of volcano dynamics is recommended. Incorporating these additional data into the analysis will allow for model refinement and more accurate forecasts. Furthermore, considering alternative machine learning techniques alongside traditional statistical approaches can capture complex interactions, reveal non-linear relationships, and further improve prediction accuracy.