Abstract:Accurately simulating glacier mass balance (GMB) data is crucial for assessing the impacts of climate change on glacier dynamics. Since physical models often face challenges in comprehensively accounting for factors influencing glacial melt and uncertainties in inputs, machine learning (ML) offers a viable alternative due to its robust flexibility and nonlinear fitting capability. However, the effectiveness of ML in modeling GMB data across diverse glacier types within High Mountain Asia has not yet been thoroughly explored. This study addresses this research gap by evaluating ML models used for the simulation of annual glacier-wide GMB data, with a specific focus on comparing maritime glaciers in the Niyang River basin and continental glaciers in the Manas River basin. For this purpose, meteorological predictive factors derived from monthly ERA5-Land datasets, and topographical predictive factors obtained from the Randolph Glacier Inventory, along with target GMB data rooted in geodetic mass balance observations, were employed to drive four selective ML models: the random forest model, the gradient boosting decision tree (GBDT) model, the deep neural network model, and the ordinary least-square linear regression model. The results highlighted that ML models generally exhibit superior performance in the simulation of GMB data for continental glaciers compared to maritime ones. Moreover, among the four ML models, the GBDT model was found to consistently exhibit superior performance with coefficient of determination (R2) values of 0.72 and 0.67 and root mean squared error (RMSE) values of 0.21 m w.e. and 0.30 m w.e. for glaciers within Manas and Niyang river basins, respectively. Furthermore, this study reveals that topographical and climatic factors differentially influence GMB simulations in maritime and continental glaciers, providing key insights into glacier dynamics in response to climate change. In summary, ML, particularly the GBDT model, demonstrates significant potential in GMB simulation. Moreover, the application of ML can enhance the accuracy of GMB modeling, providing a promising approach to assess the impacts of climate change on glacier dynamics.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to evaluate the performance of machine learning (ML) models in simulating glacier mass balance (GMB) data in High Mountain Asia (HMA), especially comparing the simulation effects of maritime glaciers and continental glaciers. Specifically, the research attempts to answer the following key questions: 1. **What are the differences in the performance of different machine learning models in simulating GMB data?** - The research selected four machine learning models: Random Forest (RF), Gradient Boosting Decision Tree (GBDT), Deep Neural Network (DNN), and Ordinary Least - Square Linear Regression (OLS), and compared their performance. 2. **How do these models perform on different types of glaciers (maritime and continental)?** - The research specifically focused on the continental glaciers in the Manas River Basin in the Tianshan Mountains and the maritime glaciers in the Niyang River Basin in southeastern Tibet to evaluate the responses of different types of glaciers to climate warming. 3. **What are the main factors affecting the effectiveness of machine learning models in GMB simulation?** - The research explored the impacts of topographic and climatic factors on the simulation of different types of glacier mass balance, revealing the different sensitivities of maritime and continental glaciers in response to climate change. By answering these questions, this research hopes to significantly improve the GMB simulation methods, make up for the limitations of physical - based models, and provide more accurate evaluation tools for monitoring glacier dynamics, water resource management, and dealing with the challenges brought by climate change. ### Research background The glaciers in High Mountain Asia are important freshwater reservoirs and are of great significance for regional socio - economic development and ecological protection. However, due to the increase in greenhouse gas emissions caused by global warming, these glaciers are shrinking rapidly, which not only threatens the risk of sea - level rise and glacial lake outburst floods but also affects the changes in the hydrological cycle. Therefore, accurately simulating glacier mass balance is crucial for evaluating the impact of climate change on glaciers. Traditionally, the simulation of GMB depends on process - based physical models, such as simple temperature - index models and complex energy - balance models. However, these models face problems such as the reliability of input data, the complexity of parameter calibration, and high demand for computational resources, which limit their wide application. In contrast, machine learning models perform well in handling complex data due to their strong non - linear fitting ability and flexibility, and can achieve good results even if the input data is biased. Nevertheless, the application of machine learning in GMB simulation in the HMA region is still in its early stage and requires further research to verify its effectiveness. ### Research significance This research provides new ideas and methods for improving the accuracy of GMB simulation by systematically evaluating the performance of different machine learning models in simulating the mass balance of different types of glaciers. This is of great significance for understanding glacier dynamic changes, predicting future trends, and formulating effective water resource management and disaster prevention strategies. In addition, the research results will also provide valuable data support for climate change adaptation, especially in those areas that are particularly sensitive to the impact of climate change.

Comparison of Machine Learning Models in Simulating Glacier Mass Balance: Insights from Maritime and Continental Glaciers in High Mountain Asia

Uncertainty-Aware Learning With Label Noise for Glacier Mass Balance Modeling

Seasonal Forecast of Winter Precipitation over China Using Machine Learning Models

Bridging the spatiotemporal ice sheet mass change data gap between GRACE and GRACE-FO in Greenland using machine learning method

Enhancing Flood Simulation in Data-Limited Glacial River Basins through Hybrid Modeling and Multi-Source Remote Sensing Data

Dynamic Glaciers Improve LGM Simulation in High Mountain Asia

Comparison of deep learning models and a typical process-based model in glacio-hydrology simulation

GlacierNet2: A hybrid Multi-Model learning architecture for alpine glacier mapping

Automated Machine Learning-Based Landslide Susceptibility Mapping for the Three Gorges Reservoir Area, China

Effective identification of debris-covered glaciers in Western China using multiple machine-learning algorithms

A comparative study among machine learning and numerical models for simulating groundwater dynamics in the Heihe River Basin, northwestern China

Glacier Mass Change in High Mountain Asia Through 2100 Using the Open-Source Python Glacier Evolution Model (PyGEM)

Development of a Water and Enthalpy Budget‐based Glacier Mass Balance Model (WEB‐GM) and Its Preliminary Validation

A Geodetic-Data-Calibrated Ice Flow Model to Simulate Historical and Future Response of Glaciers in Southeastern Tibetan Plateau

Climate warming effects in stratified reservoirs: Thorough assessment for opportunities and limits of machine learning techniques versus process-based models in thermal structure projections

Novel Machine Learning Method Integrating Ensemble Learning and Deep Learning for Mapping Debris-Covered Glaciers

Adaptability of machine learning methods and hydrological models to discharge simulations in data-sparse glaciated watersheds

A Physical Process and Machine Learning Combined Hydrological Model for Daily Streamflow Simulations of Large Watersheds with Limited Observation Data

Response of glacier modelling parameters to time, space, and model complexity: Examples from eastern slopes of Canadian Rocky Mountains

Reanalysis of the longest mass balance series in Himalaya using a nonlinear model: Chhota Shigri Glacier (India)

A hybrid machine-learning model to map glacier-related debris flow susceptibility along Gyirong Zangbo watershed under the changing climate