The aim of this page is to provide a brief and somewhat informal description of our research in environmental informatics to researchers in the field. All research items below are linked to descriptive pages that contain within them references to published articles, should further technical details be required.
Statistical remote sensing
There are a number of observational programs that measure atmospheric carbon dioxide (CO2), a leading greenhouse gas. Some of these are land-based and are spatially sparse but temporally rich. Satellite observations solve the spatial-sparseness problem while giving up some temporal richness. The ultimate goal is to use these remote sensing data, which typically represent spatio-temporal column averages of CO2 (in ppm) from Earth's surface to the satellite, to solve the inverse problem of estimating surface CO2 fluxes. This is key, since state-of-the-art climate models (or earth system models) now include an interactive carbon cycle. The most recent satellite launched to measure CO2 is NASA's Orbiting Carbon Observatory-2. Uncertainty quantification starts with the retrieved radiances at individual soundings, progresses to dynamical spatial modelling of column-averaged CO2, and finishes with spatially resolved flux fields. This quantification is crucial to understanding the uncertainties in projected climate, and gives us the ability to see for which regions climate change will be real. Click here for details on our research efforts in this area.
Regional methane flux mapping
The European Union has stringent methane emission targets in place that are challenging for several of its member countries to comply with, especially the United Kingdom. In order to know whether or not targets are being met, methane flux estimates first need to be obtained. However, assessing a country's emission portfolio is no simple task. Click here for details on our research on atmospheric gas inversion that focuses on regional flux estimation from ground data.
Multivariate stochastic modelling
Multivariate geostatistics is based on modelling all covariances between all possible combinations of two or more variables at any locations in a continuously indexed domain. Multivariate spatial covariance models need to be built with care, since any covariance matrix that is derived from such a model has to be nonnegative-definite. In our research, we are exploring techniques for model construction that require only univariate covariance functions yet yield valid multivariate models. The approach, based on conditioning, has a vast array of applications since it features conditional dependencies on a potential causative process. The project on regional methane flux mapping given above is one such application. Click here for details on our research efforts in this area.
Exceedances and their Uncertainty Quantification
A compelling reason for environmental monitoring, and the purpose of many environmental studies, is to predict the location and the extent of extreme events, such as floods and drought, and to quantify the probability of such an event. Whilst statistical models are routinely used for spatial prediction, exceedance probabilities for spatial extremes are not as straightforward: Dependent spatial predictions at many locations must be assessed simultaneously, and it is often the extremes of spatially aggregated regions that is of interest. Important applications of this work include predicting the locations of carbon dioxide sources and sinks, and predicting the regions that may be affected by global climate change. Click here for details on our research efforts in this area.
Reproducibility and visualisation in environmental modelling and inference
In the digital age, when data and algorithms are implemented in a computer, reproducibility has never been easier. However, many researchers still consider reproducibility as an `afterthought'. At CEI we spend a considerable amount of time exploring ways in which to make our results reproducible through versioning and packaging. In this page we outline the protocols that we have found most useful in ensuring permanence of the results. After ensuring reproducibility, we then explore ways the results can be disseminated, visually, to scientists and members of the general public.