Ocean Bottom Seismometer (OBS) deployed in the seafloor can record information of ocean ambient noise, and typhoon can generate elastic waves propagating through the seafloor directly or indirectly. These factors cause great changes on OBS recording data to some extent. The OBS data during the process of typhoon was analyzed, and the methods of optimum filter were used in order to separate signals. After those work, we found that wind wave and ground swell were well recorded by short-period OBSs for the first time, and a new mode which typhoon affected seafloor ambient noise was put forward. We get three preliminary conclusions from above analysis: (1) Wind wave and ground swell caused by typhoon have their own distinctive modes to affect seafloor ambient noise；（2）The range and extent of seafloor ambient noise are obviously different which have been strengthened by two above waves;（3）Short-period OBSs can clearly record the information of ground swell, whose dominating period is 6—8 seconds and its energy is generally steady (we call it “8-second phenomenon); These results will have great significance for the future research on marine seismology and other aspects of oceanography.

Nowadays research device and monitoring system are power supplied by underwater lithium batteries. Underwater lithium batteries powering method has so many disadvantages that constrains the development of deep-sea scientific research. So a new mixed transmission technology based on the armored coaxial cable is proposed and finally realized. This new technology can mixed transfer direct-current high power source, real-time color image and data of both host and slave at the same time. This technology made the superposition of direct-current power source and SHDSL data signal mainly by capacitance coupling and Low-pass filtering. It adopts coaxial cable as the media for the data signal and power signal mixture transmission. Thus the goal of high-speed and long-distance transmission can be achieved.

Selecting an effective and convenient simplified model for dual-probe seafloor heat-flow meter is the theoretical basis for optimizing dual-probe structure, which is very important to improve heat-flow calculation accuracy. Basing on the finite element numerical model about seafloor pulsing dual-probe, the authors analyzed in detail how the dual-probe’s heat pulse duration, heat generation rate, thermal properties, length and radius affect model errors in three line source simplified models. The selected effective and convenient simplified model is the one whose model errors are the least. The results are as follows: 1) in dual-probe heat pulse method, Pulsed Finite Line Source (PFLS) model is a more practical simplified model, in which the model errors from pulse duration and probe length can be avoided; 2) in the PFLS model, model errors from probe thermal conductivity can be neglected. In addition, model errors can recede with probe spacing increasing, probe radius decreasing and probe volume heat capacity approaching the surrounding medium. When surrounding medium temperature change can be detected and recorded effectively by the sensor probe, probe heat pulse power cannot affect model errors.

 The authors assimilate cruise data in the northern South China Sea using the Princeton Ocean Model (POM). The results show that the model assimilation improves the simulation effectively by reducing model errors, namely making the model results closer to the observations. The assimilation results, however, are not quite ideal in the regions where cruise observations are absent, indicating the method needs to be improved. This assimilation method once improved can provide re-analysis dataset for studying the South China Sea.

 The noise from dynamic positioning system of a vessel interferes with the data of ultra short base line (USBL) underwater positioning system in hydrothermal sulfide field investigation, while in situ decision-making and continuing research demand much better underwater positioning data. It is therefore necessary to eliminate abnormal positioning data rapidly and effectively for the existing Posidonia 6000 USBL. This study takes the USBL procedure data (x, y, z) as elimination objects, sets up (x, y, z) elimination models according to field water depth and the block angle, and uses reasonable data structure and algorithm to realize interactive elimination according to maps between (x, y, z) and time series. With the models and interactive elimination, abnormal USBL positioning data can be eliminated effectively, and much better underwater positioning data can be provided for in situ decision-making of hydrothermal sulfide investigation and continuing research.

OPeNDAP stands for “Open-source Project for a Network Data Access Protocol.” The OPeNDAP protocol provides a discipline-neutral means of requesting and providing data across the World Wide Web. The goal is to allow all end users to access immediately whatever data they require in a form they can use, all while using applications they already possess and are familiar with. This article introduces the available OPeNDAP servers, and elaborates on the design and building of the OPeNDAP-based Sharing Platform of Physical Oceanographic Data of the South China Sea, which uses the GDS (GrADS Data Server), Dapper and THREDDS (Thematic Realtime Environmental Distributed Data Services) data servers to realize OPeNDAP services of wind, current, wave, Argo, and tide data. It also presents the OPeNDAP Services of the Sharing Platform. With the help of OPeNDAP, the service system achieves the objective of seamless access to physical oceanographic data of the South China Sea in a highly heterogeneously distributed environment over the network.

Taiwan Island is at the joint of Eurasian Continent and Pacific Plate, and is always threatened by typhoons and northeasterly winds, which may cause enormous loss of human life and property every year. Therefore it is necessary to de-velop a coastal sea-state monitoring system. The COMC (Coastal Ocean Monitoring Center, National Cheng Kung University) built the Taiwan coastal sea-state monitoring system, which is modern and self-sufficient, consisting of moored buoy, pile station, tide station, coastal weather station, and radar monitoring station. To assure the data quality, Data Quality Check Pro-cedure and Standard Operation Procedure were developed by the COMC. For data analysis, some new methods are introduced to make more detailed analysis, such as EMD (empirical mode decomposition) method that is used in the analysis of storm surge water level, wavelet transform that is used to discuss the near-shore wave characteristics from X-band radar images, and data assimilation technique that is applied in wave nowcast operation. The coastal sea-state monitoring system has a great potential in providing ocean information to serve the society.

Sensitivity of sea surface temperature (SST) to wave energy factor a and Charnock parameter b is discussed using Mellor-Yamada 2.5 turbulence closure model in which wave breaking is considered. The upper-ocean temperature data in summer from OWS Station Papa is assimilated to estimate a and b via a variational approach optimally. It shows that the cost function reaches minimum when a =167 and b =4.1×10⁵. Both monthly and daily SSTs at OWS Station Papa can be successfully reconstructed with the optimal a and b , and the simulated temperature and depth of surface mixed layer are also consistent with the observation. The equation of turbulent kinetic energy is diagnosed utilizing the optimal parameters, from which the effect of wave breaking on the turbulent energy budget is revealed.

A new hybrid data assimilation scheme based on ensemble adjustment Kalman filter (EAKF) and three-dimensional variational (3D-Var) analysis is developed. In this assimilation scheme, the perturbation of ensemble from EAKF is applied to the background field by using a transformation matrix, thus the perturbation of the analysis field can be obtained by taking advantage of a sequential filter, which will then be optimized by being combined with observations under the framework of 3D-Var. The data assimilation experiment in a perfect case is carried out by using Lorenz-63 model. The results demonstrate that the hybrid data assimilation scheme performs better than EAKF.

The observational data from the International Nusantara Stratification and Transport (INSTANT) Program is used to investigate the variation of the Indonesian Throughflow (ITF). Specifically, the power spectrum of the flow in surface and thermocline layers in the three major outflow passages is analyzed. The results show that the flow in the surface layer of the Lombok and Ombai straits has an obvious annual cycle. During the monsoon transition period, the flow reverses at all depths at the Lombok and Ombai straits, whereas the current reveres only below 300 m in the Timor Passage. The variations of the currents in both surface and thermocline layers are high at intraseasonal time scale. The annual cycle is prevalent in surface layer, while the semiannual cycle dominates in thermocline layer. At tidal frequencies, each strait shows that the semi-diurnal tide M₂ is dominant expect for the surface layer of the Lombok Strait where the diurnal tide K₁ is dominant.

The seasonal characteristics of water masses and the monthly variation of mixed layer depth are studied using Argo profiling floats in 2006 east of the Luzon Strait. The temperature-salinity relation indicates the seasonal variation of water masses is not obvious. Compare to those in summer and fall, the surface temperature is lower and the surface salinity is higher in spring and winter; they change little in deeper layers. Seasonal variation of the mixed layer depth is obvious; it is the deepest in winter with a value over 160 m, whilst it is the shallowest in summer with a value of 20 m. Based on the P -vector method, the current field in fall is calculated using Argo seasonal mean data between 2002 and 2009, Levitus data and Argo data in 2006, respectively. The wind-driven Ekman drift current is also computed. The result obtained by Argo seasonal mean data is significantly better than that by the Levitus data; the Argo data can show the structure of Kuroshio and eddies. The coupled current structure obtained by Argo data in 2006 and Ekman drift current is similar to that by the altimeter data, except that the velocity of the former is less, which may be related to the coarse and unevenly distribution of Argo profiling floats and the subsequent interpolation error. However, a three-dimensional structure of flow field could be obtained by the Argo data, whilst only the surface current field can be obtained by the altimeter data.

Based on MVC ( Client Middleware Server) concept and VisualDB technology, we designed and implemented the Service System of the South China Sea Science Data Products (SCSPD). Users can query, search and access data products in the SCSPD visually via Web browser. In this paper, the application of VisualDB was introduced. Software architecture design of the system , visualized process of data management and release were described in detail. In the last section , we discussed the system expansibility of the SCSPD according to the technical characteristics and application of VisualDB. The SCSPD improves data service forms of the South China Sea Ocean Database, and is instrumental in the construction of the South China Sea Data Sharing and Service System.

Three-dimensional (3D) ocean bottom seismometer (OBS) survey provides a significant foundation for the deep crustal and upper mantle structure of the hydrothermal field (49°39′E) (Area A) in the Southwest Indian Ridge. OBS data processing is the basic step of obtaining the 3D seismic velocity structure. The flow steps for data processing of three types of OBS (domestic, French and Germanic OBS) were firstly introduced, containing the decompilation, cutting and seismic signal visualization. Taking the shot 2790th for example, waveforms and frequency spectrums of three types of OBSs were then analyzed, which were related to frequency band, sensor and seismograph for different OBSs. Domestic and French OBSs recorded long-periodic and short-periodic noises, and Germanic OBS only recorded short-periodic noises. However, air-gun signals were highlighted and noises were suppressed for all the OBSs after using a band-passed filter. Moreover, several seismic phases, e.g., Pg, PmP and Pn, were clearly revealed in the recorded seismic sections of three types of OBSs (OBS04, OBS08 and OBS23) along the profile X1X2. These phases will provide a strong data base for 3D tomography for studying Area A.

In order to obtain raw and accurate data processing section of ocean bottom seismograph (OBS) in marine geophysical exploration, negative effects of positional time and positional accuracy in acquisition and processing must be eliminated. Through analyzing the effects of the accuracy of positional time in OBS data processing and analyzing the timing error sources, we developed a gun controlling chronograph to improve positional precision and to eliminate conventional timing error; the accuracy can reach 0.01 ms, to realize high accuracy of navigation shotpoint time positioning and timing and to ensure the accuracy in OBS data acquisition. The results of stability testing and production testing show the time fixing capacity of chronograph can meet the demands of OBS multi-component acquisition perfectly, as well as in other marine geophysical prospecting areas that need high accuracy timing.

The moderate resolution atmospheric radiance and transmission (MODTRAN), with the input parameters obtained from Terra/MODIS (moderate resolution imaging spectroradiometer), was used to calculate the atmospheric effect parameters, including atmospheric transmittance, upward radiance and downward radiance. Combined with the sea surface emissivity provided by the Jet Propulsion Laboratory (JPL) and spectral response functions for five bands of advanced spaceborne thermal emission and reflection radiometer (ASTER) sensor, the sea surface temperatures (SST) were solved from radiative transfer equations pixel by pixel for each band. If the difference between any two bands’ SSTs for the same pixel is less than 0.5K, the average of five bands’ SSTs of the pixel was taken as true SST, which was then used to derive the undetermined coefficients of Split-Window algorithm. Finally, the Split-Window algorithm was applied for the SST retrieval. The results showed the algorithm has good accuracy. The SST values retrieved from ASTER data acquired on October 20, 2007 were compared to the MODIS SST, which gives the mean deviation of 0.35℃ and the root-mean-squared error (RMSE) of 0.49℃. The same procedure was repeated for ASTER data acquired on May 15, 2008, which gives the mean deviation of 0.38℃ and RMSE of 0.52℃.

This study assesses the accuracy of Aquarius/SAC-D satellite sea surface salinity (SSS) data in the South China Sea using Argo buoy data, and analyzes the influencing factors. The results indicate that the linear relationship of co-located data is not significant. The SSS inversion accuracy in the South China Sea and northeastern waters is 0.62‰ and 0.70‰, respectively. Located at low latitude, the sea surface temperature (SST) in the South China Sea is high, which has a small effect on the accuracy. The factors that influence the SSS accuracy may be the strong breeze, rainfall and land radio frequency interference (RFI), etc. The results show that the SSS retrieval error increases with wind speed when the wind speed is greater than 7 m·s^-1, and that the error has a clear ascendant trend. Meanwhile, the South China Sea is seriously contaminated by the land RFI. Even if the Aquarius/SAC-D SSS product has corrected the land RFI in its new Version 1.3. The RFI is not eliminated, and may still have an influence on the accuracy.

For the data characteristics of interferometric multi-beam echo sounder (MBES), a single ping filtering method of MBES based on dynamic clustering is proposed. Considering the continuity of real terrain, the problem of outlier detection is transformed into clustering of real terrain data. Through continuous clustering of real terrain data, the outliers in data are eliminated. For the large data size in the process of clustering, dynamic clustering is adopted after partitioning clustering sets. Simultaneously, a trend adjusted factor is introduced for the feature domain selection, which is helpful for the decision of clustering direction. At last, the improved k-means method is utilized for output of clustering object. The results from processing sea test data of GeoSwath MBES show that the algorithm has good adaptability for different terrain characteristics, and is simple for implementation, which can be used for real-time filtering and post-processing of MBES data.

Stirring is an important part of mixing, which can be quantified using Finite Time Lyapunov Exponents (FTLE) based on Lagrangian view. In this paper, we calculated the FTLE of surface ocean derived from satellite altimeter from 2002 to 2011, and then analyzed spatial and temporal variation of horizontal stirring in the South China Sea (SCS). Results show that FTLE in the SCS is not uniform, with high values southeast of Vietnam indicating strong stirring, and low values in the northwest and southeast of the SCS. A slowly increasing trend of stirring in the SCS is observed during the 10 year period. FTLE also displays a seasonal fluctuation, strong in summer but weak in winter. Furthermore, we found that FTLE has a similar spatial distribution with Euler-based eddy kinetic energy (EKE) and strain rate, with high and low values of these three quantities locating roughly in the same areas. A comparison with Okubo-Weiss parameter reveals a strong relationship between vortices and FTLE ridges, referred to as Lagrangian Coherent Structures (LCS). Low values of FTLE are mainly present inside rotation-dominated vortices, while high values occur in strain-dominated regions surrounding the vortices.

First arrival seismic tomography (FAST) is one of the most widely used seismic tomography tools to achieve complicated three-dimensional (3D) deep crustal structure. The selection of optimal inversion parameters for FAST is a key to obtain a reasonable velocity model effectively. Based on the 3D seismic survey data collected from the central sub-basin of the South China Sea (SCS), the selection process of inversion parameters was illustrated in detail in this paper, using the control variate method, namely, selecting one parameter while fixing the other parameters, and then selecting the combination of all parameters comprehensively. The result showed that the combination of parameters (damping parameter λ=2, smoothness weighting parameter s_z=1, inversion number I=4) was an assembly of optimal inversion parameters for the 3D seismic structure of the central sub-basin of the SCS. The preferred preliminary velocity structure acquired by this combination indicated that the central sub-basin could be classified as a typical slow-spreading oceanic crust. The research of choosing inversion parameters not only lays a foundation for further modeling of detailed 3D seismic structure of the Zhenbei-Huangyan seamount chain but also provides reference and experience for the application of FAST software in other potential areas.

In this paper, sea surface salinity data of the equatorial Pacific Ocean in 2011~2012, acquired by the Soil Moisture and Ocean Salinity (SMOS) satellite, was processed for quality control and then analyzed for the first time in terms of dynamic process factors that may impact salinity retrieval. The neural network method was introduced to improve the quality-controlled salinity data of the same time period. It was found that precipitation and surface wave induced by precipitation can increase salinity error substantially in a negative trend. Sea surface roughness caused by wind also increased salinity errors. There was a weak positive correlation between wind speed and salinity error. Changes in sea surface temperature had little effect on salinity retrieval. Considering rainfall, wind speed and other major marine dynamic processes, the neural network method was used to revise sea surface salinity data of the equatorial Pacific in December 2011. The results showed that the RMS (root mean square) of salinity reduced from 0.383 7 to 0.244 1. It was found that not only was the salinity error caused by precipitation and other dynamic processes eliminated, but also the high salinity tongue in the equatorial Pacific was revealed, which the Level-3 data failed to.

According to the types of South China Sea marine datasets, especially the cruise data format, based on the marine data specifications at home and abroad, and on the references of a large number of relevant specifications, we designed “Collection and Management Guide of the South China Sea Data”, which is suitable for the South China Sea Ocean Database. It includes six data management specifications. In terms of data collection, the guide regulates data factor, data information, data classification, and data format, in terms of data management and guide proposes of data format conversion specification. This achieves various marine data collection and management effectively, promotes the specification system of the South China Sea marine data and improves the sharing of marine data and application services.

With the emergence of cloud computing, great changes have taken place in data sharing. Cloud computing is considered to be a scientific data sharing platform. Through the network, software and hardware based on resource intensive operations will improve the stability and scalability of the server, and the efficiency for data sharing. This paper builds scientific investigation data sharing platform of cloud computing, using VDB3.0 based on the enterprise cloud computing platforms (ECCP). Integrated with phpMyAdmin and other tools, the unified management of heterogeneous data sources is realized, transforming data into flexible services. This study improved the security of data storage, organization and management efficiency.

In this study, we conducted barotropic detiding of the summertime shipboard ADCP (SADCP) dataset in the southern Taiwan Strait (TWS) from June to September during 2004-2013, employing the widely used spatiotemporal fitting by the least squares method (STF-LSM). The results show that the mean flow derived from STF-LSM and that derived from averaged vectors by dataset gridding both flowed northeastward. Moreover, these two mean flows had similar spatial distribution patterns of strong and weak flows, but their maximum velocities were slightly different: 0.46 and 0.34 m·s^-1 for the former and the latter, respectively. Besides, the strong currents (about 0.3 m·s^-1) were characterized by the narrow mainstream confined around the depth of 30 m off the sea at Dongshan and Longhai as well as the Taiwan Bank (TWB) with drastic topographic change, whereas the weak currents (less than 0.1 m·s^-1) were mainly distributed in the central TWB. Estimation of the volume transport based on the above mentioned mean flows revealed similar results, up to (0.71±0.24) ×10⁶ m³·s^-1. The strong M₂ tidal currents derived from SADCP were near the TWB and extended to the offshore of Dongshan, with a maximum M₂ current amplitude as large as ~0.64 m·s^-1, which was located in the TWB. The cotidal chart for M₂ currents demonstrates the characteristics of progressive wave, which was delayed successively from the southwest to the northeast along the direction of the tidal wave as the wave propagated (about 2 h). Besides, there was a boundary line along northwestward to southeastward direction with ellipticity equal to zero at the offshore of Dongshan. The M₂ currents rotated counterclockwise in the southern region of this boundary line and clockwise in the northern region.

The dynamic characteristics of time-mean meridional overturning circulation in the Indian Ocean was examined using six reanalysis datasets. The results showed consistent time-mean features of the deep meridional overturning circulation, which is an anticlockwise overturning cell with inflows in the bottom and deep layers and outflows in the intermediate and upper layers. Dynamic decomposition of meridional overturning circulation was applied to examine the similarities and differences of every dynamic component. The structure of Ekman component is an anticlockwise overturning cell in the South Indian Ocean with maximum strength at ~10°S. In the region south of 10°S, geostrophic and external components show clockwise and anticlockwise overturning cells, respectively, they both reach maximum strengths at ~27°S. Based on different products of heat and momentum fluxes used, the dynamic components resulted from the six datasets show some inconsistent features as follows. The overall structures of Ekman component are similar since the wind fields of the six datasets have few differences. The discrepancies of the geostrophic component in the six datasets are due to the strength of the baroclinic flows in the interior ocean and the structure of the western boundary current: the greater the baroclinic flows in the interior ocean, the stronger the strength of the geostrophic component; the wider the western boundary current, the greater impact on the baroclinic flows in the interior ocean, and then on the strength of the geostrophic component. The strength of the external component is affected by the intensity of the western boundary current: the greater the western boundary current, the stronger the strength of the external component.

By regarding the moderate resolution imaging spectroradiometer (MODIS) remote sensing data as data sources and the southwest coastal waters of Hainan Island as the study area, the improved spilt-window algorithm is used to inverse the SST of the study area from 2005 to 2014. To verify the retrieval accuracy of the SST, we calculate the correlation coefficient between the inversion and measured results using the measured values of 12 observation points; the coefficient is 0.9. Based on the inversion results of SST, the data of four seasons in a year and a decade are analyzed to show the spatial and temporal variations of SST distribution, respectively. Results show that using the MODIS multi-channel improved spilt-window algorithm we can inverse Hainan Island offshore SST accurately. The spatial and temporal variations of SST distribution thus obtained can provide important reference values for marine fishery, change of marine environment and meteorological monitoring.

A two-dimensional ocean bottom seismometer (OBS) survey line (OBS2015-2) was carried out in 2015, which was an E-W trending and located in the front of the Manila Trench subduction zone. In this study, we take two OBS stations (OBS04 and OBS08) as examples to introduce the data processing flow for this short-period OBS type of data. The process includes data format transforming from original RAW data to SAC format and then to SEGY format; and it contains shot relocation and OBS position relocation. The preliminary results of data processing demonstrate that the OBS data are of high quality in this experiment. The seismic phases, such as PsP, Pg, PmP, etc. are very clear along the OBSs’ seismic record sections. They are further identified by the forward modeling using the Rayinvr software. Especially, the reflective seismic phases from the Moho interface are abundant and obvious in the incoming plate along the Manila Trench subduction zone, which will provide good constraints for the crustal bottom of the incoming plate.

It is of great significance to fully exploit the oceanic application potential of the GaoFen-1 data, which is the first domestic high-resolution satellite of Earth observation in China. In this study, a GaoFen-1 multi-spectral image is used to retrieve the island shallow water depth of Jinqing Island in Xisha sea area by adopting the dual-band linear model and log-ratio model, after serials of image processing including image geometric correction, atmospheric correction and sun-glint correction. The accuracy of the two models is evaluated by using actual water depth data. Meanwhile, possible influence factors of retrieval accuracy of island shallow water depth from satellite data are discussed. It is shown that the result from the dual-linear model is obviously better than that from the log-ratio model, and the dual-band linear model is more suitable for the retrieving shallow water depth of Jinqing Island. For water depth less than 20 m, the overall root mean square (RMS) error is 1.8 m with RMS error of 1.14 m within 5 m, which reaches the level of accuracy of water depth inversion by satellite remote sensing.

Using offshore airguns and onshore explosions as seismic sources, along with portable seismic stations, permanent seismic stations and Ocean Bottom Seismometer (OBS) as a synchronously receiving array, a onshore-offshore 3D seismic exploration experiment was carried out for the first time in the Zhujiangkou area of the northern South China Sea. This paper introduces in detail the seismic data acquisition, observation system, data preprocessing method, and phase identification. A variety of seismic phases, including Pg, PcP, PmP, Pn, and Sg, can be clearly identified in most data collected by the permanent, portable seismic stations and OBS. Jointly using airgun and explosion sources, and dense stations receiving seismic signals simultaneously both on land and at sea, it can highly improve the ray coverage and cross overlap beneath the onshore-offshore transitional belt, which lays a foundation for obtaining fine crustal structure beneath such onshore-offshore area.

This study was based on the raw eddy-covariance (EC) data measured in February 2016 from the observation tower on Yongxing Island. EddyPro software was used for data post-processing, and the influence on flux calculation was quantitatively analyzed before and after each processing step. Furthermore, flux data quality classification, footprint analysis and comparison with calculated results using easyflux_dl real time programme were conducted. The results showed that despiking, coordinate rotation, frequency response correction and density correction all influenced the flux calculation result to different degrees. The influences of coordinate rotation on momentum flux, latent heat flux and carbon dioxide flux were evident, which revealed increases of 11.28%, 7.36% and 18.53%, respectively. The density correction had a significant effect on the carbon dioxide flux calculation, which was reduced by 7.87% after density correction. Comprehensive quality evaluation results revealed that the ratios of high quality data (grades 1~3) of momentum flux, sensible heat flux, latent heat flux, and carbon dioxide flux were 90.29%, 90.22%, 88.46%, and 85.41%, respectively. Footprint analysis showed that 72.04% source area of the flux was scattered in the first interesting area (10~80°, the underlying surface is the ocean surface with an average water depth of more than 10 m). Comparison of flux calculation results between EddyPro and EasyFlux showed that the two flux time series were fundamentally consistent; the results obtained by EddyPro were smaller but more reliable and accurate due to more strict despiking adopted to get rid of outliers’ impacts.

Based on the Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) data in 2015, we analyzed the seasonal variation and genesis mechanisms of thermal fronts in Guangdong coastal zone. The Guangdong coastal thermal fronts are strong in winter and weak in summer. In the eastern Guangdong coastal zone, thermal fronts are seen all year round, while in the western Guangdong coastal zone, thermal fronts disappear in summer. In the Pearl River estuary zone, thermal fronts are associated with the Pearl River plumes, which extend east northward in summer and southward in autumn. We suggest wind speed play important roles in the genesis and disappearance of these thermal fronts. Using a simple model, we find that monsoon influences the front strength through air-sea turbulent heat fluxes. Under southwesterly (northeasterly) wind condition, warm and wet (cold and dry) air mass could decrease (increase) the latent heat release near the cold coastal zone, leading to a temperature increase in the cold zone, and finally the front disappears.