RSSeNEWS to be determined...

The Environmental Statistics Section and Southampton Statistical Sciences Research Institute jointly hosted a meeting on environmental and spatial statistics in June. The meeting was addressed by six speakers in three themed sessions and ended with a panel discussion in the style of the BBC's Question Time programme.

Alan Gelfand of Duke University started the first session on biodiversity and species abundance. This joint work with Avishek Chakraborty provided spatial analysis of abundance for six species of proteaceae in the Cape Floristic region of South Africa. Through latent variable specifications in the form of a bivariate spatial process, inference methods were developed for understanding spatial variation of abundance that are useful for planning and conservation decision making.

Mark Brewer of Biomathematics and Statistics Scotland presented his joint work with Colin Beale of the Macaulay Institute on the recorded presence of 154 species of birds living in Tanzania's semi-arid habitats between 1960 and 2007. Using important environmental covariates, the work explored the spatial pattern of changes in habitat via a Bayesian hierarchical model.

Data assimilation and downscaling

The second session on data assimilation and downscaling was shared by Li Chen of the University of Bristol and Serge Guillas of University College London who had previously collaborated at the University of Illinois. Li gave an overview of spatial prediction, data assimilation and Kalman filter techniques. He then presented an improved approach using the ensemble adjustment Kalman filter and applied this to a computer simulation of surface measurements of carbon monoxide.

Serge outlined a downscaling method to reduce errors in the forecasts of ozone concentration levels made by a computer simulation model known as the regional air quality forecast. He described the deficiencies of the model by using ground level ozone monitoring data collected by the United States environmental protection agency.

The air-pollution modelling session discussed several aspects of analysing sulphur dioxide pollution levels in Europe and particularly in London. Adrian Bowman of the University of Glasgow presented additive models which provided very smooth relationships between variables of interest for analysing such pollution data. He also discussed computational strategies for spatio-temporal smoothing and the construction of appropriate models of spatial variation over river networks and illustrated the results with data on water quality in the River Tweed.

Gavin Shaddick of the University of Bath discussed some problems often encountered when analysing the long-term effects of air pollution on health. Results from a simulation study showed bias in environmental exposure parameters on aggregate disease counts.

Discussion session

The final discussion session was chaired by Sujit Sahu with the speakers plus John Haslett in the panel. Frank discussions took place regarding the context of the papers presented in the meeting and the statistical issues surrounding them. The panel very carefully discussed each of the questions from the floor and warned of the dangers of using off-the-shelf statistical methods which may not be appropriate for the practitioner's particular problem.

The meeting concluded by noting that statistical methods need to emerge as a series of dialogues between practitioners and specialist statisticians over a period of time.

The slides of the presentations are available for download.