This repository is dedicated to the Great Lakes Summer Fellows Program project focused on optimizing the placement of temperature sensors across the Great Lakes using advanced machine learning techniques. By leveraging the DeepSensor framework, we aim to improve the spatial network design for environmental monitoring, providing valuable insights into the Great Lakes' surface temperature variability. This project will serve as a proof-of-concept, allowing for expansion into other key variables, such as nutrients.
The Great Lakes are a critical natural resource, providing drinking water, transportation routes, recreational opportunities, and supporting a diverse ecosystem. However, monitoring such a vast area is challenging due to logistical constraints and resource limitations. It is crucial to make the most efficient use of available observing platforms (e.g. buoys, research vessels). This project seeks to utilize convolutional Gaussian neural processes, as formulated in the DeepSensor tool, to propose a strategic placement of temperature sensors, thereby optimizing the observation network.
Our overall objective is to develop a quantitative framework for strategic placement of the next generation of Great Lakes observing stations in order to best capture surface temperature variability. To phrase this goal as a research question, “where should the next generation of temperature measurement sensors be placed in order to most efficiently improve our quantitative understanding of Great Lakes surface temperature variability?”
The summer fellow will use DeepSensor, an open source Python package for probabilistically modeling environmental data with neural processes, to characterize Great Lakes surface temperature and to make informed suggestions for future temperature sensor locations. Specifically, the student will:
Please refer to the environment setup guide for detailed instructions on how to set up deepsensor for different environments and purposes.
Instructions on how to train the DeepSensor model, analyze the data, and propose sensor locations will be provided in the docs directory or as separate markdown files within this repository.
Contributions to this project are welcome! Please see our CONTRIBUTING.md file for guidelines on how to contribute effectively.
We've set several Milestones to organize the progress of the project, including a review of the literature and other resources, data preparation and visualization, model training and runing, analysis and site recommendation, reporting and documentation, and final presentation.
The detailed timeline is available under the repository's Projects tab.
For a full list of contributors, please see the contributors page.
The Alan Turing Institute hosts the development of this software, in part by maintaining a Slack channel. For info on how to join the Slack channel, visit the DeepSensor Repository and check out the README.
This project is licensed under the MIT License - see the LICENSE file for details.
Thank you to the NOAA SOAR funding initiative for supporting this research work.
For details on the methodologies and algorithms used in this project, refer to the following paper:
Andersson, T., Bruinsma, W., Markou, S., Requeima, J., et al. (2023). Environmental sensor placement with convolutional Gaussian neural processes. Environmental Data Science, 2, E32. doi:10.1017/eds.2023.22
Further resources included in the DeepSensor repository:
https://github.com/alan-turing-institute/deepsensor/resources.html