An empirical comparison of kriging methods for nonstationary spatial point prediction
This thesis compares the performance of geostatistical and geostatistical nonparametric hybrid models for providing accurate predictions together with relevant measures of prediction confidence. The key modelling theme is nonstationarity, where models that cater for nonstationary second-order effect...
| Main Author: | |
|---|---|
| Published: |
University of Newcastle Upon Tyne
2009
|
| Subjects: | |
| Online Access: | http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.492440 |
| Summary: | This thesis compares the performance of geostatistical and geostatistical nonparametric hybrid models for providing accurate predictions together with relevant measures of prediction confidence. The key modelling theme is nonstationarity, where models that cater for nonstationary second-order effects nave the potential to provide more accurate results over their stationary counterparts. A comprehensive review and comparison of this particular class of nonstationary predictors is considered missing from the literature. To facilitate this model comparison, models are calibrated to assess the spatial variation in freshwater acidification critical load data across Great Britain, which is shown to be a heterogeneous process requiring a nonstationary modelling approach. |
|---|