Advanced statistics for model evaluation, simulation set-up and analysis (RA4)
The focus of RA4 is to provide and develop statistical models and methods that improve the analysis and evaluation of climate data and output from climate and earth system models on a global, regional, and local scale. The work includes both development of new statistical methodology that is widely applicable to spatial and spatio-temporal datasets and the use of modern statistical methods to analyse environmental data.
Environmental modelling and monitoring produces very large datasets of past, present and potential future climate and vegetation. The data, obtained from historic records and present day monitoring equipment, as well as from earth system models, include such diverse things as ordinary weather data (e.g. temperature and precipitation), atmospheric concentrations of greenhouse gases (e.g. CO2 and methane), land use, vegetation health, and many other measurements of important processes.
Land reconstructions by using historic pollen records
An example is work within RA2 to analyse historic pollen records. Since earth system models are developed and evaluated based partially on historic data, our ability to accurately describe past land-cover and human land use (i.e. farming) is important. Based on pollen records from lake sediments we can reconstruct the amount of land that, 200 years ago, was covered by coniferous forest, broadleaved forest or left unforested. However, these reconstructions are only possible around suitable lakes. The resulting maps provide only, as illustrated, a fragmented record of past land cover.
Using statistical methods for spatial interpolation developed by Behnaz Pirzamanbein, we are able to expand on these patchy measurements. The result is a coherent estimate of past land cover over all of North West Europe. An estimate that can be used both in the development and validation of regional earth system models, and to help us understand how past humans, through agriculture, may have affected regional climate.
Behnaz Pirzamanbein – maths.lth.se
Article on spatial and spatio-temporal datasets:
An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach – onlinelibrary.wiley.com
Authors: Finn Lindgren, Håvard Rue and Johan Lindström
Article on statistical methods to analyse environmental data:
Fast estimation of spatially dependent temporal vegetation trends using Gaussian Markov random fields – sciencedirect.com
Authors: David Bolin, Johan Lindström, Lars Eklundh and Finn Lindgren
johan [dot] lindstrom [at] matstat [dot] lu [dot] se