The features of the coastal fronts in the Eastern Guangdong coastal waters during the downwelling-favorable wind period

A coastal front was detected in the eastern Guangdong (EGD) coastal waters during a downwelling-favorable wind period by using the diffuse attenuation coefficient at 490 nm (Kd(490)). Long-term satellite data, meteorological data and hydrographic data collected from 2003 to 2017 were jointly utilized to analyze the environmental factors affecting coastal fronts. The intensities of the coastal fronts were found to be associated with the downwelling intensity. The monthly mean Kd(490) anomalies in shallow coastal waters less than 25 m deep along the EGD coast and the monthly mean Ekman pumping velocities retrieved by the ERA5 dataset were negatively correlated, with a Pearson correlation of − 0.71. The fronts started in October, became weaker and gradually disappeared after January, extending southwestward from the southeastern coast of Guangdong Province to the Wanshan Archipelago in the South China Sea (SCS). The cross-frontal differences in the mean Kd(490) values could reach 3.7 m−1. Noticeable peaks were found in the meridional distribution of the mean Kd(490) values at 22.5°N and 22.2°N and in the zonal distribution of the mean Kd(490) values at 114.7°E and 114.4°E. The peaks tended to narrow as the latitude increased. The average coastal surface currents obtained from the global Hybrid Coordinate Ocean Model (HYCOM) showed that waters with high nutrient and sediment contents in the Fujian and Zhejiang coastal areas in the southern part of the East China Sea could flow into the SCS. The directions and lengths of the fronts were found to be associated with the flow advection.

www.nature.com/scientificreports/ winds occurring in the northern South China Sea (NSCS) in winter. Downwelling is one of the most significant dynamic phenomena in the NSCS. It has a great influence on the nearshore circulation along the eastern Guangdong (EGD) coast. The EGD coastal fronts that occur in the EGD coastal waters not only have a pronounced influence on the hydrologic characteristics of the region but are also important for algal blooms, pollutant dispersion, seafloor sediments, and ship navigation in the NSCS. However, the features of coastal fronts in EGD coastal waters have not yet been investigated extensively. With the development of remote sensing techniques, satellite data have become an effective data source for studying the frequency and region of coastal front occurrences in EGD coastal waters. The purpose of the study is to analyze the locations, occurrence times, durations, intensities and convergence properties of coastal fronts in the EGD coastal region by applying remote sensing techniques. In this study, ocean color data and meteorological and hydrographic data were jointly utilized to illustrate the most important factors influencing the coastal fronts.

Materials and method
Study area. EGD coastal waters are known as subtropical coastal regions with high biological productivity and are located in the NSCS 14 . The EGD coastal area ( Fig. 1) is characterized by a complicated hydrodynamic system driven by many physical factors, including river discharge, bottom topography, monsoon winds, and coastal currents 15 . The region is influenced by the East Asian monsoon system, which is characterized by prevailing northeasterly and southwesterly winds in winter and summer, respectively 8 . In this study, the seasons refer to those of the Northern Hemisphere, for example, autumn refers to September, October, and November. With increasing human activity, EGD coastal waters have been contaminated by industrial pollution, agricultural runoff, and domestic sewage 16 .  www.nature.com/scientificreports/ Ancillary data. Level-1B moderate-resolution imaging spectroradiometer (MODIS) Aqua imagery from 2003 to 2017 were obtained from the online archive of the Level-1 Atmosphere Archive & Distribution System (LAADS) Distributed Active Archive Center (DAAC). The SeaWiFS data analysis system (SeaDAS, version 7.5.1) was applied to process the data. For the data processing procedure, an iterative f/Q BRDF correction and MUMM-based atmospheric correction were chosen [17][18][19] . In addition, sea surface temperature (SST) products at a 1-km resolution (Level 2) were obtained from the Physical Oceanography Distributed Active Data Center (PO. DAAC) of the NASA Jet Propulsion Laboratory; these data were derived by the NASA Ocean Biology Processing Group (OBPG) and can be accessed at the following site: https:// podaac. jpl. nasa. gov. The meteorological data were obtained from the latest climate reanalysis ERA5 product released by the European Centre for Medium-Range Weather Forecasts (ECMWF, https:// www. ecmwf. int/ en/ resea rch/ clima te-reana lysis). ERA5 was carried out by using the ECMWFS' Earth System model IFS, cycle 41r2. The ERA5 data used in this study, which are on regular latitude-longitude grids at a 0.25° × 0.25° resolution with atmospheric parameters at 37 pressure levels, are available in the climate data store (https:// clima te. coper nicus. eu/ clima tereana lysis) and cover the period from 1950 to the present. The hourly wind speeds at a height of 10 m above the sea surface were extracted from the MODIS pixels in EGD for the 2003-2017 period.
The hydrological data were acquired from the global Hybrid Coordinate Ocean Model (HYCOM) analysis dataset. The HYCOM-Navy Coupled Ocean Data Assimilation (HYCOM + NCODA) Global (1/12)° Analysis product was provided by the Global Ocean Forecasting System (GOFS). The data assimilation scheme is a threedimensional variational scheme (3DVAR) 20 . Three-hourly velocity fields obtained in 2005 were used in this study.
ETOPO1 bathymetry grids at a 1-arcminute resolution were obtained from the National Centers for Environmental Information and applied in this study (NCEI, https:// www. ngdc. noaa. gov/ mgg/ global/ global. html). The bathymetric grids were extracted to the same grid resolution as that of the MODIS imagery.
Ancillary software. To process the MODIS data, the SeaWiFS data analysis system (SeaDAS, version 7.5.1, available at https:// seadas. gsfc. nasa. gov/) developed by NASA was applied for image preprocessing. Several atmospheric correction algorithms exist, including Bailey's et al. 21 iterative correction model, which was used to remove nonnegligible NIR radiance from the NIR signals prior to aerosol determination, Wang and Shi's 22 model, and the Management Unit of the North Seas Mathematical Models (MUMM) 23 (more details can be found at https:// ocean color. gsfc. nasa. gov/ seadas/). In this study, the MUMM model was applied. The L1b data were processed to obtain the Rayleigh scattering corrected reflectance values by applying the default atmospheric correction algorithm during the first step. The ETOPO1 bathymetry grids were extracted to the same grid as the MODIS imagery. Then, the Rayleigh scattering corrected reflectance values were extracted over the clear water surface (a region with a water depth greater than 200 m). A regression line was fitted to the Rayleigh scattering corrected reflectance values at 748 nm and 869 nm. The slope of the regression line between the two bands was applied to determine epsilon, which is an input of the MUMM model used to correct for the aerosol scattering contribution. Moreover, the program R (version 3.6.3, available at https:// www.r-proje ct. org) and C Programming were applied for the statistical computing and creation of graphics.

Method. Model for K d (490) retrieval.
In aquatic systems, the light diffuse attenuation coefficient K d (λ) is defined by the exponential decrease in irradiance that occurs with depth 24,25 . Traditionally, K d (λ) is measured by the ocean color scientific community at 490 nm, and this metric is expressed as K d (490). To estimate K d (490), a neural network inversion that was trained using a combination of simulated and in situ data sets by Jamet et al. 26 was applied in this study. The type of neural network was the Multi-Layer Perceptron. The remote sensing reflectance values (sr −1 ) at 443, 488, 531, 547, and 667 nm (R rs (443), R rs (488), R rs (531), R rs (547), and R rs (667)) were applied to train the neural network.
Model for C tsm retrieval. For the retrieval of the concentration of suspended sediments (C tsm ), the two-band ratio exponential algorithm was applied, as follows 16 : where Ekman pumping. The Ekman pumping velocity was computed directly using the following equation 27,28 : where ρ is the mean density of seawater, f is the Coriolis parameter, k is a unit vector, and − → τ is the hourly wind stress vector.

Results and discussion
Temporal and spatial pattern of K d (490). Figure 2 shows the mean attenuation coefficients at 490 nm (K d (490)) in the EGD coastal water in four seasons from 2003 to 2017, as retrieved from the MODIS data. K d (490) is higher in the EGD coastal water in autumn and winter than in spring and summer. In particular, a well-defined front was observed extending from the southeastern coast of Guangdong Province to the Wanshan  Downwelling-favorable wind. To determine the dynamics of the coastal fronts, the monthly mean 10-m wind speeds in November and December from 2003 to 2017 were used to calculate the mean Ekman pumping velocity (Fig. 6). Figure 6 shows that a strong northeasterly monsoon wind prevailed in the NSCS. Ekman transport created onshore flows in the surface Ekman layer, and the surface waters near the shore were pushed against the shore, then piled up and sunk as a consequence of undercurrents. Vertical mixing was very strong due to the strong monsoon winds. In particular, in areas with coastal water depths of less than 25 m, K d (490) and www.nature.com/scientificreports/ C tsm were higher in autumn and winter than in spring and summer. The corresponding fronts observed in the K d (490) and C tsm series were identified in the EGD coastal waters. The sediments in the Fujian-Zhejiang coastal waters (ZFC) of the East China Sea (ECS) could be advected to the EGD coast by the strong coastal currents. For areas with depths of less than 25 m, the water is mixed well from top to bottom due to downwelling. For areas with water depths between 25 and 60 m, thermoclines and haloclines could be found at depths of 15 m in the EGD 29 . The thicknesses of these thermoclines and haloclines were found to be approximately 5 ~ 10 m. These thicknesses increased with the water depth. The thermocline intensity was approximately 0.2 °C m −1 , and the halocline intensity was more than 0.05 m −129 . Figure 6 shows  www.nature.com/scientificreports/ K d (490) anomalies in shallow coastal waters less than 25 m deep along the EGD coast and the monthly mean Ekman pumping velocities retrieved from the ERA5 datasets was − 0.71. Intense northeasterly winds parallel to the coast induced Ekman transport that pushed surface water to the coast. The onshore flow in the surface Ekman layer created a compensating seaward-submerged subsurface front. An overturning cross-shelf circulation induced by a downwelling jet together with breaking waves and other nonlinear wave effects triggered enhanced vertical mixing. The effects of tides, waves and downwelling on bed shear stress can cause sediment erosion in shelf seas. In addition, the shelf topography in the longshore direction is also an important factor in cross-shore bottom transport processes. Cross-shelf circulation can induce sediment resuspension when a critical bed shear stress occurs. The coastal suspended sediments were mixed well by downwelling currents and turbulence. A geostrophic coastal jet was generated when downwelling-favorable winds prevailed. The secondary circulation process transports coastal suspended sediments in the subsurface and bottom waters to the open ocean. The convective instabilities led to the development of vigorous turbulence, which induced sediment resuspension. From a remote sensing point of view, the prominent fronts observed in the K d (490) and C tsm series can be considered evidence confirming the existence of resuspension induced by the cross-isobath transport of downwelling circulation on the shelf. The fact that the intensities of the fronts in the K d (490) and C tsm series were clearly associated with the strength of the downwelling further illustrated this point. The monthly mean SSTs between November and December from 2003 to 2017 are shown in Fig. 8. The cold currents from Fujian-Zhejiang flowed southwestward, driven by the northeast monsoon. The SSTs gradually increased from north to south along the coast of Fujian to Guangdong. The temperature gradient was very large during the occurrence of downwelling-favorable winds. The boundary between the warm and cold waters was prominent and smooth in the EGD region. The intensity of the cold tongue was relatively stronger in 2005 and weaker in 2009 from north to south (it is worth noting that the color bar varies from 15.5 to 28.5 °C in Fig. 8c). This indicates that K d tended to increase as the cold water mass from the ECS increased. The nutrients and sediments in the ZFC waters could be advected southward to the EGD coast by the strong coastal currents. The lower the SST values along the EGD coast were, the more cold water was transported from the ZFC region to the EGD region.
The fronts' development and maintenance. The distribution of the Ekman pumping velocities was shown to increase from September 2005 to December 2005 (Fig. 9). The corresponding fronts observed in the K d (490) and C tsm series also exhibited similar increasing trends (Figs. 10 and 11). September and October are the autumn monsoon transitional periods over the NSCS. Figure 12 shows that the northeasterly monsoon pre-  With the influence of downwelling-favorable winds and Guangdong coastal currents, the fronts became stronger and were most pronounced in December 2005. The fronts extended from the southeastern coast of Guangdong Province to the Wanshan Archipelago in the SCS. Then, the fronts became weaker and gradually disappeared after January 2006. The spatial distribution of K d (490) appeared to indicate a decrease in the offshore direction due to the nearshore sediment suspension. However, in January and February 2006, there appeared to be a location where the sediments rose to the surface offshore. This could be attributed to the intense southwestward currents that brought sediments from the northern area to this region. The monthly average global surface current velocities from the HYCOM data were calculated and are given in Fig. 12. The results showed that waters with high nutrient and sediment contents in the ZFC in the southern part of the ECS could flow into the SCS. However, the influence of the ECS coastal waters  The peak tended to narrow as the latitude increased. The fronts extended southwestward from the southeastern coast of Guangdong Province to the Wanshan Archipelago in the South China Sea. The intensities of the fronts were found to be highly associated with the downwelling intensity. The Pearson correlation between the monthly mean K d (490) anomalies in shallow coastal waters less than 25 m deep along the EGD coast and the monthly mean Ekman pumping velocities retrieved from the ERA5 datasets was − 0.71. The fact that the intensities of the fronts in the K d (490) and C tsm series were clearly associated with the downwelling strength illustrated this point. The prominent fronts in the K d (490) and C tsm series can be considered evidence confirming the existence of resuspension induced by the cross-isobath transport of downwelling circulation on the shelf. The monthly average surface coastal current velocities estimated using HYCOM were large in February 2006, at which time the coastal fronts had already completely vanished. The sediment fronts near the Wanshan Archipelago were enhanced by the resuspension that occurred in this downwelling region. The intensity of the fronts was affected by the secondary circulation induced by this downwelling region. The direction and length of the fronts were largely determined by coastal currents. The resuspension that occurred in the downwelling region was proven by long-term satellite data analysis conducted in this study.