Welcome to the MassBalanceMachine documentation!
MassBalanceMachine is global machine learning glacier mass balance model that assimilates all glaciological and remote sensing data sources.
Note
This project is under active development.
Check out the Installation instructions section for installation information. The notebooks will guide you through the data preprocessing and your first MassBalanceMachine training.
Getting started
Tutorials
After installing the massbalancemachine package, you can start exploring the tutorials. These notebooks are designed to walk you through using MassBalanceMachine with WGMS data, focusing initially on extracting data from the Open Global Glacier Model (OGGM). These data include comprehensive topographical information for nearly all glaciers worldwide.
Specifically, the example notebooks concentrate on glaciers documented in the WGMS database, particularly those in Iceland. They cover various topics, including:
Data pre-processing 🌍: If users have already their data formatted in the WGMS format, they can directly jump into the Data preparation tutorial. Alternatively, if the data are not properly formatted, the Data Preprocessing - convert to WGMS format notebook shows an example of how to convert the data. In the Data preparation workflow, topographical and climate data are fetched and aligned with the stake measurements. Subsequently, the data is aggregated to a monthly resolution, preparing it for use as training data for the model.
Note
If the OGGM cluster is shut down, users will be unable to retrieve topographical features for their region of interest. If you encounter a 403 error in your notebook while trying to retrieve these features, it likely means that the OGGM cluster is down. You can check the status of the cluster on their Slack channel.
Data Exploration 🔍: Users can gain deeper insights into their data by visualizing the time series of the available stake measurements, which are related to either the region-wide surface mass balance or the point surface mass balance. See the Data Exploration tutorial.
Model Training 🚀 & Testing 🎯: Two tutorials cover the use of two models: XGBoost and neural networks. Refer to the XGBoost Model Training and Neural Network Model Training tutorials.
Project information