Changes in water reaching the base of glaciers strongly affect their sliding speed by altering water pressure beneath the ice. High water pressure is often thought to develop during summer melt, enlarging cavities at the glacier base and increasing sliding speed. Our observations show instead that the cavities mainly grow during winter, when water pressure builds up in isolated cavities. As the cavities become connected during the melt season, water pressure decreases and the glacier slows down.

An initial estimate of sea ice conditions are required to start modern weather forecasting models. To get these we have to combine observations, typically from satellites, with previous estimates of the sea ice conditions. This paper talks about how we do this in the latest version of the model, where the sea ice state has to be specified for ice of different thicknesses. We describe the method that is used when we produce an estimate for every hour for the past 50 years.

Predicting Greenland's surface melt is critical for understanding sea-level rise, but traditional firn models are too slow for exploring many climate scenarios. We developed a neural network optimized through systematic input selection and network tuning to identify the necessary information to accurately emulate surface melt. This approach cuts computation costs by orders of magnitude and can be retrained for different climate forcings or extended to other surface mass balance properties.