Using Geostatistics to Analyze Prediction Errors from a Simulation Model of Sediment Particle Sizes across Tokyo Bay

Abstract

Orton, T.G.; Lark, R.M.; and Sasaki, J., 2013. Using geostatistics to analyze prediction errors from a simulation model of sediment particle sizes across Tokyo Bay. Buildup of sediment on the seabed of Tokyo Bay has had an impact on water quality and ecosystems in the bay. Process models have been built to simulate the mechanisms governing sediment buildup, but such models are useful only if they can be validated against measurements obtained from the study area. Visual comparison of maps produced by the process model with measurements can provide some information about the ability of the process model to simulate the spatial variation of sediment buildup, but we consider a more rigorous geostatistical approach to perform this validation task. We consider predictions by process simulation models of the median particle diameter (MPD) of sediment across Tokyo Bay; we use a geostatistical approach to test whether the process model provides useful information for explaining the spatial distribution of 95 MPD measurements from the bay. Furthermore, we use the geostatistical approach to investigate the spatial distribution of model errors and to test whether there is evidence that errors could be explained by other environmental variables. The aim of the study is to demonstrate the kind of information we can get from such a geostatistical analysis of model output and errors, without looking directly inside the process simulation model. This information could subsequently be used to assist research into further simulation model developments.