Environmental Sensors

To advance our understanding of environmental change and improve prediction we must find smart ways to combine satellite data, field sensors and results from environmental simulators that take into account physical principles (e.g., climate models). To reconcile the spatial scales and diverse modalities of these data, this project aims to build toolkits driven by real-world case studies. The results contribute to building further Uk research capability in the development of Digital Twins of the natural environment.

The project consists of three working packages:

As part of my postdoc in the project, my role is transversal in each of the lines. In particular, my main contributions are in the WP1 where in collaboration with other researchers from British Antarctic Survey and Met Office we are developing a generalizable tool based on probabilistic modeling capable of combining data from different modalities (grid data, historical in situ data and auxiliary data) to have finer and local observations of environmental variables. In addition, with greedy algorithms, the tool is proposed to suggest where to install local sensors to reduce the uncertainty of the modelled environmental variable. Within the proposed case studies, my environmental variable to be modelled is soil moisture. The input data are ERA5, 2013-2019 observations from COSMOS-UK stations and the 1 km digital multi-resolution terrain model (GMTED2010).

Finally, motivated by the diversity of projects and researchers with whom I collaborate in the project, as well as communities active in the development of open source software such as Pangeo I have also developed the Environmental Data Science community. To this end, I am maintaining a resource developed in Jupyter Book which hosts executable content in Jupyter notebooks format as well as records of monthly community meetings in which we discuss topics related to environmental data science.