Variability of the Red River Plume in the Gulf of Tonkin as Revealed by Numerical Modeling and Clustering Analysis

We study the daily to interannual variability of the Red river plume in the Gulf of Tonkin from numerical simulations at high resolution over 6 years (2011-2016). Compared with observational data, the model results show good performance. To identify the plume, passive tracers are used in order to (1) help distinguish the freshwater coming from different continental sources, including the Red river branches, and (2) avoid the low salinity effect due to precipitation. We show that the temporal evolution of the surface coverage of the Red river plume is correlated with the runoff (within a lag), but that the runoff only cannot explain the variability of the river plume; other processes, such as winds and tides, are involved. Using a K-means unsupervised machine learning algorithm, the main patterns of the plume and their evolution in time are analyzed and linked to different environmental conditions. In winter, the plume is narrow and sticks along the coast most of the time due to the downcoast current and northeast wind. In summer, the southwest monsoon wind makes the plume flow offshore. The plume reaches its highest coverage in September after the peak of runoff. Vertically, the plume thickness also shows seasonal variations. In winter, the plume is narrow and mixed over the whole water depth, while in summer, the plume can be detached both from the bottom and the coast. The plume can deepen offshore in summer, due to strong wind (in May, June) or specifically to a recurrent eddy occurring near 19N (in August). This first analysis of the variability of the Red river plume can be used to provide a general picture of the transport of materials from the river to the ocean, for example in case of anthropogenic chemical substances leaked to the river.