[Apologies for cross-posting]

==============================================================
Reminder: One-day meeting on Environmental and Spatial Statistics.
=============================================================
June 19, 2009: University of Southampton.

The emphasis of the meeting is on environmental applications
where methodology is drawn from spatial (and possibly temporal)
statistics. The meeting is partially sponsored by the Environmental
Statistics Section of the Royal Statistical Society and
the Southampton Statistical Sciences Research Institute (S3RI).
Please note you must book a place for this meeting.
For further information please visit:
http://www.s3ri.soton.ac.uk/courses/environmental/


Here is the programme outline followed by the abstracts.
===========================================================
09:30 - 10:00: Coffee and Registration
10:00 - 11:20: Session 1 : Biodiversity (Species Abundance)
10:00 - 10:40: Alan Gelfand. Latent Spatial Modeling for Species Abundance.
10:40 - 11:20: Mark Brewer. Accounting for observer effort in
               the spatial modelling of African bird data.

11:20 - 11:40: Coffee

11:40 - 13:00: Session 2 : Data Assimilation (Downscaling)

11:40 - 12:20: Li Chen. Spatial prediction, data assimilation and
               ensemble adjustment Kalman filter
12:20 - 13:00: Serge Guillas. Statistical correction and downscaling
               of chemical transport model ozone forecasts.
13:00 - 14:00: Lunch
14:00 - 15:20: Session 3 : Spatial Modelling of Air Pollution
14:00 - 14:40: Adrian Bowman. Additive models for environmental applications
14:40 - 15:20: Gavin Shaddick. Issues of bias in spatial studies
               of environmental factors.
15:20 - 16:00: Panel Discussion.
16:00: Close of Meeting
================================================
Abstracts
==================================================


Latent Spatial Modeling for Species Abundance

ALAN GELFAND (Duke University)


Predicting biodiversity, i.e., the abundance of species, in response to
environment is a primary goal of much ecological research.
This talk attempts to explain observed abundance using spatial
modelling incorporating local environmental features.
Through latent variable specifications, we are able to infer
about potential abundance in the absence of landscape
transformation (useful for planning and conservation decision
making) as well as to explain observed abundance in the presence
of transformation (useful for understanding the spatial variation
in abundance). To study abundance for six species of proteaceae
we analyse a data set covering 37,000 minute
by minute pixels in the Cape Floristic Region in South Africa.


Accounting for observer effort in the spatial modelling of African bird data

Mark J Brewer (Biomathematics and Statistics Scotland)

Forecasts of rapid climatic change highlight the importance of gaining
an understanding of the relationship between animal populations
and climate. Given spatially-referenced data on observer effort and recorded
presence of 154 species of bird living in Tanzania's semi-arid habitats
between 1960 and 2007, we explore the spatial pattern of distribution
change via a Bayesian hierarchical model while modelling observer effort
directly. We estimate the probability of detection and consider
methods for relating changes in population locations over time to
environmental factors.



Spatial prediction, data assimilation and ensemble adjustment Kalman filter
            Li Chen (University of Bristol)

This talk gives a brief overview of spatial prediction, data
assimilation and Kalman filter type techniques. Particularly, a new
approach is presented for data assimilation using the ensemble
adjustment Kalman filter for surface measurements of carbon monoxide in
a single tracer version of the community air quality model. Three
different sets of numerical experiments were performed to test the
effectiveness of the procedure and the range of key parameters used in
implementing the procedure. In each case the proposed method provided
better results than the method without data assimilation.



Statistical correction and downscaling of chemical transport
model ozone forecasts

Serge Guillas (University College, London)

The Regional Air Quality forecAST (RAQAST) model is a regional chemical
transport modeling system for ozone and its precursors over the
United States. Since the grid size is 70 by 70km, forecasts cannot
be made for a specific surface site. We use EPA monitoring stations
to downscale and improve local forecasts using
RAQAST outputs. We first use a time series regression approach to
correct deficiencies. Evaluation using measurements for a different
period confirms that the methods reduce forecast errors by up to 25%.
Then, we implement a spatio-temporal approach with a non-separable
covariance structure which enables us to describe the deficiencies of RAQAST.




Additive models for environmental applications

    Adrian Bowman (University of Glasgow)

   Additive models extend standard regression methods by allowing
   very flexible, but smooth, relationships between variables of
   interest. The role of these models in environmental applications,
   where there is a need to model complex forms of spatial and temporal
   trends, as well as spatial and temporal correlation, will be discussed.
   The technical aspects of the talk will focus on computational
   strategies for spatiotemporal smoothing and on the construction of
   appropriate models of spatial variation over river networks.
   Applications will include the modelling of SO2 pollution over Europe
   and water quality in the River Tweed.



Issues of bias in spatial studies of environmental factors

Gavin Shaddick (Bath University)

In this talk, I will discuss some problems encountered
when analysing the long-term effects of air pollution on health.
The problems discussed will include examining association between
aggregate disease counts and environmental exposure measured, for example via
air pollution monitors. There may also be problems when
using spatial models where the monitoring network might be sparse relative to
the study area, in which case the use of predicted concentrations
can produce serious bias in effect parameters because the number of monitors
is not sufficient to characterize the concentration surface.
A simulation study and an example on association between
sulphur dioxide and respiratory mortality are provided.