Modeling waves over the Changjiang River Estuary using a high-resolution unstructured SWAN model

Abstract

A high resolution unstructured SWAN model has been implemented for the Changjiang River Estuary (CRE). Five different winds are adopted to assess their quality over the CRE, including ERA5, CFSv2, GFS, CCMP-NRT and a newly available high resolution (9 km) wind product from APRCP (Asia-Pacific Regional Coupled Prediction System). The performance of the five winds, together with four wind input source functions is evaluated by comparing with satellite altimeter observations and multiple in-situ observations. Systemic difference is presented in the four input source functions, with the JANSSEN and KOMEN packages outperformingWST in terms of significant wave height (Hs) over the CRE. The default ST6 has comparative skill as JANSSEN during fair weather conditions but tends to predict higher Hs during extreme weather events. All winds have the tendency to underestimate the altimeter observed Hs during fair weather conditions. Part of the negative bias attributes to the overestimation of altimeter observations, while the boundary conditions only play a minor role, given a relatively large model domain extending to about 135° E. The CFSv2/GFS wind has the best performance during fair weather conditions, but overestimates the extreme wave height during typhoon events. The ERA5, on the other hand, underestimates Hs during both fair and extreme weather conditions. In terms of extreme waves, CCMP-NRT generally agrees best with observations. The high resolution APRCP behaves significantly different from other wind forcings and shows poorer performance in wave simulation. But high resolution winds have the advantage in resolving the detail wind structure and show better performance in reproducing the high frequency wave fluctuations adjacent to the CRE. In addition, the deviation due to different wind forcings is more distinct than the different input source functions. Nevertheless, given a certain wind forcing, calibrations would further improve the model performance. This study provides a good start to build a hindcast and forecast system for the CRE region.