In many important areas to society such as ecology, the environment and health sciences, advances in geographical information systems (GIS) and global positioning systems (GPS) now allow accurate geographical coding of the locations where data are collected, for example through latitude and longitude. Typically, the data will vary over space and may also have time dependency.
Often, these data sets are extremely large. For example, there may be a thousand monitoring stations each recording daily measurements on ten variables for air pollution over the last ten years. Data may have highly complex structure: different variables may be observed in different locations; some variables may be observed hourly whilst data on others are only recorded daily. The statistical challenge is to make valid inference about the true underlying spatio-temporal processes, recognising that such complications in the observed data introduce missingness, bias error and measurement error. Model-based inference methods, such as Bayesian hierarchical modelling, can adjust for these errors and are required in many contexts.
In S3RI, Sujit Sahu is leading the modelling work for space-time data. His research with Alan Gelfand (Duke University, USA) and David Holland (US Environmental Protection Agency, EPA) has been used by the EPA.
His recent projects and research papers are detailed on his personal webspace.
Professor Alan Gelfand, (Duke University USA)
David Holland, (US Environmental Protection Agency)