Machine Learning for Remote Sensing and Modelling of Mountain Snow Patches

Project Description

Scientific background and motivation

Snow in the mountains provides many services as a store of water, a habit and a playground. It also poses threats as a flood and avalanche risk, and snow is highly sensitive to climate variability and change. The high spatial variability of snow cover in mountains, compared with the resolutions of satellite sensors and models, makes measurement and prediction of changes in snow cover particularly difficult in the very environments for which they are most needed. Even in the maritime climate of Scotland, snow patches can persist throughout the summer in favourable mountain locations; the fine balance between preferential snow deposition in winter and sheltering in summer would make predicting the distribution of these snow patches a stern test for the kind of physically based snow models used in climate projection and impacts studies. The motivation for this project is to use newly available high resolution remote sensing and meteorological modelling with machine learning to improve understanding of the climate sensitivity of mountain snow.

Aims and objectives of the PhD project

Artificial Neural Network (ANN) methods will be applied to three problems:

  • remote sensing of mountain snow cover;
  • downscaling meteorological variables over mountain topography;
  • constraining physical models of snow mass and energy balances driven with meteorological variables and evaluated with remote sensing.

Fieldwork in the Cairngorms in support of the PhD objectives will allow rapid responses to favourable conditions and will reduce the project’s environmental footprint compared with international travel.

Methodology

Remote sensing of snow cover is most often achieved using differences between visible and near-infrared channels to distinguish snow from cloud, normalized to compensate for variations in illumination and viewing conditions, and thresholded to distinguish snow from snow-free ground. Multispectral imagers and snow reflectance models actually provide much more information on snow structure and contaminants that can be exploited by machine learning. Training sets will be developed by visual inspection of images and used to train an ANN to discriminate snow specifically for snow conditions occurring in the Cairngorms. The principle sensors to be used will be MODIS and Sentinel-2 at moderate and high resolutions, respectively.

Weather stations that could provide inputs for snow models are sparse and difficult to maintain in the mountains. Numerical Weather Prediction (NWP) at 1 km resolution and Computational Fluid Dynamics (CFD) models at higher resolutions are now available but limited in spatial and temporal coverage. Hydrological forecasting and impacts models generally use simple elevation lapse rates to downscale meteorological variables, but some variables – precipitation and wind, in particular – clearly do not behave in such simple ways. Machine learning will be used to identify patterns in NWP and CFD model outputs that can be used for rapid downscaling.

A high-resolution, physically-based snow model will be driven with downscaled meteorology and evaluated with remote sensing products developed in this project. Some of the processes required in models of this type, such as shading of snow on slopes, are well understood but require expensive simulations at high resolutions. Other processes, such as wind erosion, turbulent transport and re-deposition of snow, are poorly understood and heavily parametrized. Physical constraints of mass and energy conservation will be incorporated into machine learning of snow patterns, using meteorological variables and topographic metrics as inputs to gain insight into the relative importance of factors contributing to the preservation of late-lying snow patches.

Fieldwork for this project will take advantage of connections with existing hydrometeorological measurements in the Feshie and Coire Cas catchments of the Cairngorms. Measurements that could be made for ground truthing and model evaluation using NERC equipment loans include field spectroscopy, topographic laser scanning and structure-from-motion photogrammetry.

The student will be based at the University of Edinburgh. Applicants should have strong programming skills and an interest in environmental remote sensing.

This PhD is part of the NERC and UK Space Agency funded Centre for Doctoral Training “SENSE”: the Centre for Satellite Data in Environmental Science. SENSE will train 50 PhD students to tackle cross-disciplinary environmental problems by applying the latest data science techniques to satellite data. All our students will receive extensive training on satellite data and AI/Machine Learning, as well as attending a field course on drones, and residential courses hosted by the Satellite Applications Catapult (Harwell), and ESA (Rome). All students will experience extensive training on professional skills, including spending 3 months on an industry placement. See http://www.eo-cdt.org

Funding Information

This 3 year 9 month long NERC SENSE CDT award will provide tuition fees (£4,500 for 2021/22), tax-free stipend at the UK research council rate (£15,609 for 2021/22), and a research training and support grant to support national and international conference travel. View Website

Eligibility Requirements

We expect  most  applicants to have a  degree equivalent to a UK first class honours, or a high upper second class, in an environmental science, maths, physics, computer science, or engineering discipline. However, we will consider applications from a wide range of backgrounds, including those with non-traditional  qualifications or from industry – please feel free to contact us to have a chat about your suitability for the programme.

Application Process

Please visit our application page.

To apply for this PhD, please use the following application link: https://eo-cdt.org/apply-now/

Overview
Location