Skill Assessment of SPM Transport Models using the Adjoint Technique

Abstract

The paper discusses the application of Goodness-of-Fit (GoF) criteria and adjoint modelling to assess the skill of an SPM transport model of the North Sea. Suspended Particulate Matter (SPM) is fine sediment with a grain size of less than 63μm. A GoF criterion that is used to quantify the model performance is a measure for the misfit between the model simulations and some pre-defined model output reference. A GoF -criterion should reflect the user's modelling objective in an appropriate way. In the case study, the user's modelling objective is the representation of the seasonal variation in SPM patterns. The GoF criterion used in this paper is built upon aggregation in space and time of both the observed and modelled SPM concentrations and their representativity given the user's modelling objective being the seasonal variation of SPM patterns. The model output reference is derived from SPM concentrations retrieved from Remote Sensing (RS) reflectance imagery. The paper illustrates the two-sided relation of observations and models for (1) the retrieval of information from remote sensing reflectance imagery using model data and (2) the analysis of an SPM transport model using observations as model output reference. By means of the adjoint SPM transport model the model's sensitivities for variations in model input parameters are determined in a spatially and temporally distributed way. The sensitivity analysis shows that the estimates of the loads and/or correctness of the local mass balance strongly determine the skill of the model. The paper concludes with some recommendations with respect to the rationalization of the integrated use of data and models in operational oceanography.