Comparison Of Three Methods For Spatial Distribution Of Error-Correction Algorithms

Data assimilation is a useful tool to correct the discrepancies of numerical model results by extracting reliable information from observed data. One of popular data assimilation techniques is the spatial distribution based on error-correction, since it can address the challenge when number of monitoring stations is limited. Current research only focuses on the estimation of spatial distribution pattern, or the improvement of the competence of different spatial distribution methods, but lacks the comparison either in their characteristics or in the performances. In this study, we compared three different approaches, Kriging, Artificial Neural Network (ANN) and inter-model correlation inspired by Kalman Gain, for spatial distribution on error correction. Based on the application in a real case of Singapore Regional model, the performance and adaptive capabilities of these methods are analyzed through testing the sensitivity in response to different observation points and hydrodynamic regimes. The results suggest that the performance varies among different methods and changes with various scenarios, indicating that an appropriate selection of algorithms under different environmental condition is necessary