2021 Vol. 40, No. 7

2021, 40(7): .
2021-7 Contents
2021, (7): 1-2.
Physical Oceanography, Marine Meteorology and Marine Physics
Three-dimensional circulation in northern South China Sea during early summer of 2015
Huiqun Wang, Yaochu Yuan, Weibing Guan, Chenghao Yang, Dongfeng Xu
2021, 40(7): 1-14. doi: 10.1007/s13131-021-1815-1
Using the hydrographic data obtained during two nearly simultaneous surveys in June 2015, we carried out semi-diagnostic calculations with the help of a finite element model and a modified inverse method, to study the circulation in the northern South China Sea (NSCS) during the early summer of 2015. A number of new circulation features were found. (1) In most of the observation region, a large, basin-scale anticyclonic gyre appeared south of the 50-m isobath, which contained anticyclonic eddies. One anticyclonic eddy existed from the sea surface to 50-m depth, whose center showed no tilt, while the center of another eddy tilted eastward from the sea surface to 500-m depth. In the eastern part of the observation region, which is west of the Dongsha Islands, there was a sub-basin-scale cyclonic gyre containing a cyclonic eddy whose center tilted southward from the sea surface to 200-m depth. (2) There was a cross-continental slope current (CCSC) in the area southwest of the Dongsha Islands. Its volume transport was about 2.0×106 m3/s. (3) From the estimated order of magnitude of the stream function equation, the joint effect term of the baroclinity and relief (JEBAR) and β-effect term are two important dynamic mechanisms affecting the variation of the circulation in the NSCS. (4) The JEBAR, as a transport-generating term, resulted in the dynamic mechanism determining the pattern of the depth-averaged flow across the contours of potential vorticity fH–1. Furthermore, we show that the negative values of the JEBAR were the most dominant dynamic mechanism, causing the CCSC southwest of the Dongsha Islands to deflect from the isobaths and veer toward the deep water. The CCSC around the Dongsha Islands was located further southwest during the early summer of 2015 than during the fall of 2005 (revealed by a published study), which suggests that the location of the CCSC around the Dongsha Islands may vary with season.
A Lagrangian study of the near-surface intrusion of Pacific water into the South China Sea
Gaolong Huang, Haigang Zhan, Qingyou He, Xing Wei, Bo Li
2021, 40(7): 15-30. doi: 10.1007/s13131-021-1766-6
Satellite-tracked Lagrangian drifters are used to investigate the transport pathways of near-surface water around the Luzon Strait. Particular attention is paid to the intrusion of Pacific water into the South China Sea (SCS). Results from drifter observations suggest that except for the Kuroshio water, other Pacific water that carried by zonal jets, Ekman currents or eddies, can also intrude into the SCS. Motivated by this origin problem of the intrusion water, numerous simulated trajectories are constructed by altimeter-based velocities. Quantitative estimates from simulated trajectories suggest that the contribution of other Pacific water to the total intrusion flux in the Luzon Strait is approximately 13% on average, much smaller than that of Kuroshio water. Even so, over multiple years and many individual intrusion events, the contribution from other Pacific water is quite considerable. The interannual signal in the intrusion flux of these Pacific water might be closely related to variations in a wintertime westward current and eddy activities east of the Luzon Strait. We also found that Ekman drift could significantly contribute to the intrusion of Pacific water and could affect the spreading of intrusion water in the SCS. A case study of an eddy-related intrusion is presented to show the detailed processes of the intrusion of Pacific water and the eddy-Kuroshio interaction.
The mechanism of the banded structure of drifting macroalgae in the Yellow Sea
Yan Li, Fangli Qiao, Hongyu Ma, Qiuli Shao, Zhixin Zhang, Guansuo Wang
2021, 40(7): 31-41. doi: 10.1007/s13131-021-1771-9
At the end of May 2008, a massive bloom of macroalgae occurred in the western Yellow Sea off China and lasted for nearly two months, and annual blooms have occurred since then on. During bloom period, the surface-drifting macroalgae have showed an interesting pattern dominated by a banded structure, and the distance between neighboring bands ranged from hundreds of meters to about 6 km with a peak at 1–1.5 km, which is an order of higher than the scale of Langmuir circulation of 50–100 m. In order to explain this new phenomenon, ocean current data obtained from a Doppler current profiler off Qingdao was used to implement stability analysis. By numerically solving the resulting differential Orr-Sommerfeld equation, the secondary circulation induced from the instability of the Emkan current was found to fit well with the observed spatial scale of the surface-drifting macroalgae’s banded structure. As the wind driven Emkan current exist universally in the global ocean, it is reasonable to conclude that the banded structure with kilometers distance between adjoining bands is ubiquitous. We found a new circulation in the upper ocean which is important for exchange of energy, materials and gas between the upper ocean and subsurface layer.
Characteristics of oceanic mesoscale variabilities associated with the inverse kinetic energy cascade
Mengmeng Li, Zhiliang Liu, Jianing Li, Chongguang Pang
2021, 40(7): 42-57. doi: 10.1007/s13131-021-1814-2
Oceanic geostrophic turbulence theory predicts significant inverse kinetic energy (KE) cascades at scales larger than the energy injection wavelength. However, the characteristics of the mesoscale variabilities associated with the inverse KE cascade in the real oceans have not been clear enough up to now. To further examine this problem, we analyzed the spectral characteristics of the oceanic mesoscale motions over the scales of inverse KE cascades based on high-resolution gridded altimeter data. The applicability of the quasigeostrophic (QG) turbulence theory and the surface quasigeostrophic (SQG) turbulence theory in real oceans is further explored. The results show that the sea surface height (SSH) spectral slope is linearly related to the eddy-kinetic-energy (EKE) level with a high correlation coefficient value of 0.67. The findings also suggest that the QG turbulence theory is an appropriate dynamic framework at the edge of high-EKE regions and that the SQG theory is more suitable in tropical regions and low-EKE regions at mid-high latitudes. New anisotropic characteristics of the inverse KE cascade are also provided. These results indicate that the along-track spectrum used by previous studies cannot reveal the dynamics of the mesoscale variabilities well.
Exact solution and approximate solution of irregular wave radiation stress for non-breaking wave
Liangduo Shen, Zhili Zou, Zhaode Zhang, Yun Pan
2021, 40(7): 58-67. doi: 10.1007/s13131-021-1809-z
Wave radiation stress is the main driving force of wave-induced near-shore currents. It is directly related to the hydrodynamic characteristics of near-shore current whether the calculation of wave radiation stress is accurate or not. Irregular waves are more capable of reacting wave motion in the ocean compared to regular waves. Therefore, the calculation of the radiation stress under irregular waves will be more able to reflect the wave driving force in the actual near-shore current. Exact solution and approximate solution of the irregular wave radiation stress are derived in this paper and the two kinds of calculation methods are compared. On the basis of this, the experimental results are used to further verify the calculation of wave energy in the approximate calculation method. The results show that the approximate calculation method of irregular wave radiation stress has a good accuracy under the condition of narrow-band spectrum, which can save a lot of computing time, and thus improve the efficiency of calculation. However, the exact calculation method can more accurately reflect the fluctuation of radiation stress at each moment and each location.
Application of deep learning technique to the sea surface height prediction in the South China Sea
Tao Song, Ningsheng Han, Yuhang Zhu, Zhongwei Li, Yineng Li, Shaotian Li, Shiqiu Peng
2021, 40(7): 68-76. doi: 10.1007/s13131-021-1735-0
A deep-learning-based method, called ConvLSTMP3, is developed to predict the sea surface heights (SSHs). ConvLSTMP3 is data-driven by treating the SSH prediction problem as the one of extracting the spatial-temporal features of SSHs, in which the spatial features are “learned” by convolutional operations while the temporal features are tracked by long short term memory (LSTM). Trained by a reanalysis dataset of the South China Sea (SCS), ConvLSTMP3 is applied to the SSH prediction in a region of the SCS east off Vietnam coast featured with eddied and offshore currents in summer. Experimental results show that ConvLSTMP3 achieves a good prediction skill with a mean RMSE of 0.057 m and accuracy of 93.4% averaged over a 15-d prediction period. In particular, ConvLSTMP3 shows a better performance in predicting the temporal evolution of mesoscale eddies in the region than a full-dynamics ocean model. Given the much less computation in the prediction required by ConvLSTMP3, our study suggests that the deep learning technique is very useful and effective in the SSH prediction, and could be an alternative way in the operational prediction for ocean environments in the future.
Identification of the sensitive area for targeted observation to improve vertical thermal structure prediction in summer in the Yellow Sea
Huiqin Hu, Jingyi Liu, Lianglong Da, Wuhong Guo, Kun Liu, Baolong Cui
2021, 40(7): 77-87. doi: 10.1007/s13131-021-1738-x
The sensitive area of targeted observations for short-term (7 d) prediction of vertical thermal structure (VTS) in summer in the Yellow Sea was investigated. We applied the Conditional Nonlinear Optimal Perturbation (CNOP) method and an adjoint-free algorithm with the Regional Ocean Modeling System (ROMS). We used vertical integration of CNOP-type temperature errors to locate the sensitive areas, where reduction of initial errors is expected to yield the greatest improvement in VTS prediction for the selected verification area. The identified sensitive areas were northeast−southwest orientated northeast to the verification area, which were possibly related to the southwestward background currents. Then, we performed a series of sensitivity experiments to evaluate the effectiveness of the identified sensitive areas. Results show that initial errors in the identified sensitive areas had the greatest negative effect on VTS prediction in the verification area compared to errors in other areas (e.g., the verification area and areas to its east and northeast). Moreover, removal of initial errors through deploying simulated observations in the identified sensitive areas led to more refined prediction than correction of initial conditions in the verification area itself. Our results suggest that implementation of targeted observation in the CNOP-based sensitive areas is an effective method to improve short-term prediction of VTS in summer in the Yellow Sea.
Seasonal variation in the three-dimensional structures of coastal thermal front off western Guangdong
Yan Zhang, Lili Zeng, Qiang Wang, Bingxu Geng, Changjian Liu, Rui Shi, Na Liu, Weiping Wang, Dongxiao Wang
2021, 40(7): 88-99. doi: 10.1007/s13131-021-1739-9
The seasonal structure and dynamic mechanism of oceanic surface thermal fronts (STFs) along the western Guangdong coast over the northern South China Sea shelf were analyzed using in situ observational data, remote sensing data, and numerical simulations. Both in situ and satellite observations show that the coastal thermal front exhibits substantial seasonal variability, being strongest in winter when it has the greatest extent and strongest sea surface temperature gradient. The winter coastal thermal front begins to appear in November and disappears after the following April. Although runoff water is more plentiful in summer, the front is weak in the western part of Guangdong. The frontal intensity has a significant positive correlation with the coastal wind speed, while the change of temperature gradient after September lags somewhat relative to the alongshore wind. The numerical simulation results accurately reflect the seasonal variation and annual cycle characteristics of the frontal structure in the simulated area. Based on vertical cross-section data, the different frontal lifecycles of the two sides of the Zhujiang (Pearl) River Estuary are analyzed.
The long-term trend of Bohai Sea ice in different emission scenarios
Donglin Guo, Rui Li, Peng Zhao
2021, 40(7): 100-118. doi: 10.1007/s13131-021-1703-8
Based on a coupled ocean-sea ice model, this study investigates how changes in the mean state of the atmosphere in different CO2 emission scenarios (RCP 8.5, 6.0, 4.5 and 2.6) may affect the sea ice in the Bohai Sea, China, especially in the Liaodong Bay, the largest bay in the Bohai Sea. In the RCP 8.5 scenario, an abrupt change of the atmospheric state happens around 2070. Due to the abrupt change, wintertime sea ice of the Liaodong Bay can be divided into 3 periods: a mild decreasing period (2021–2060), in which the sea ice severity weakens at a near-constant rate; a rapid decreasing period (2061–2080), in which the sea ice severity drops dramatically; and a stabilized period (2081–2100). During 2021–2060, the dates of first ice are approximately unchanged, suggesting that the onset of sea ice is probably determined by a cold-air event and is not sensitive to the mean state of the atmosphere. The mean and maximum sea ice thickness in the Liaodong Bay is relatively stable before 2060, and then drops rapidly in the following decade. Different from the RCP 8.5 scenario, atmospheric state changes smoothly in the RCP 6.0, 4.5 and 2.6 scenarios. In the RCP 6.0 scenario, the sea ice severity in the Bohai Sea weakens with time to the end of the twenty-first century. In the RCP 4.5 scenario, the sea ice severity weakens with time until reaching a stable state around the 2070s. In the RCP 2.6 scenario, the sea ice severity weakens until the 2040s, stabilizes from then, and starts intensifying after the 2080s. The sea ice condition in the other bays of the Bohai Sea is also discussed under the four CO2 emissions scenarios. Among atmospheric factors, air temperature is the leading one for the decline of the sea ice extent. Specific humidity also plays an important role in the four scenarios. The surface downward shortwave/longwave radiation and meridional wind only matter in certain scenarios, while effects from the zonal wind and precipitation are negligible.
Rapid decrease in Antarctic sea ice in recent years
Guanghua Hao, Hui Shen, Yongming Sun, Chunhua Li
2021, 40(7): 119-128. doi: 10.1007/s13131-021-1762-x
A 41-year Antarctic sea ice concentration (SIC) dataset derived from satellite passive microwave radiometers during the period of 1979–2019 has been used to analyze sea ice changes in recent decades. The trends of SIC and sea ice extent (SIE) are calculated during the periods of 1979–2019, 1979–2013, and 2014–2019. The trends show regionally dependent features. The SIC shows an increasing trend in most of the regions except the Bellingshausen Sea and Amundsen Sea (BA) during 1979–2019 and 1979–2013. The SIE trend shows a decreasing or decelerating trend in the period of 1979–2019 ((6 835±2 210) km2/a) compared with the 1979–2013 period ((18 600±2 203) km2/a). In recent years (2014–2019), the SIC and SIE have exhibited decreasing trends (–(34 567±3 521) km2/month), especially in the Weddell Sea (WS) and Ross Sea (RS) during summer and autumn. The trends are related to regionally dependent causes. The analyses show that the SIC and SIE decreased in response to the warming trend of 2 m air temperature (Ta-2m) and have exhibited a good relationship with Ta-2m in summer and autumn in recent years. The sea ice decrease in the Antarctic is mainly caused by increases in absorbed energy and southward energy transportation in recent years, such as the increase in gained solar radiation and moist static energy from the south, which demonstrate notable regional characteristics. In the WS region, the local positive feedback from the additional absorbed solar radiation, resulting in warmer air and reduced sea ice, is the main reason for the sea ice decrease in recent years. The increase in southward energy transport has also favored a decrease in sea ice. In the RS region, the increase in southward-transported moist static energy has contributed to the decrease in sea ice, and the increases in cloud cover and longwave radiation have prevented sea ice growth.
Modelling the annual cycle of landfast ice near Zhongshan Station, East Antarctica
Jiechen Zhao, Tao Yang, Qi Shu, Hui Shen, Zhongxiang Tian, Guanghua Hao, Biao Zhao
2021, 40(7): 129-141. doi: 10.1007/s13131-021-1727-0
A high resolution one-dimensional thermodynamic snow and ice (HIGHTSI) model was used to model the annual cycle of landfast ice mass and heat balance near Zhongshan Station, East Antarctica. The model was forced and initialized by meteorological and sea ice in situ observations from April 2015 to April 2016. HIGHTSI produced a reasonable snow and ice evolution in the validation experiments, with a negligible mean ice thickness bias of (0.003±0.06) m compared to in situ observations. To further examine the impact of different snow conditions on annual evolution of first-year ice (FYI), four sensitivity experiments with different precipitation schemes (0, half, normal, and double) were performed. The results showed that compared to the snow-free case, the insulation effect of snow cover decreased bottom freezing in the winter, leading to 15%–26% reduction of maximum ice thickness. Thick snow cover caused negative freeboard and flooding, and then snow ice formation, which contributed 12%–49% to the maximum ice thickness. In early summer, snow cover delayed the onset of ice melting for about one month, while the melting of snow cover led to the formation of superimposed ice, accounting for 5%–10% of the ice thickness. Internal ice melting was a significant contributor in summer whether snow cover existed or not, accounting for 35%–56% of the total summer ice loss. The multi-year ice (MYI) simulations suggested that when snow-covered ice persisted from FYI to the 10th MYI, winter congelation ice percentage decreased from 80% to 44% (snow ice and superimposed ice increased), while the contribution of internal ice melting in the summer decreased from 45% to 5% (bottom ice melting dominated).
Marine Geology
Possible triggering relationship of six Mw > 6 earthquakes in 2018–2019 at Philippine archipelago
Qiu Zhong, Yangfan Deng, Zhigang Peng, Lingyuan Meng
2021, 40(7): 142-158. doi: 10.1007/s13131-021-1813-3
Philippine archipelago (PA) has strong background seismicity, but there is no systematic study of earthquake triggering in this region. There are six earthquakes (Mw > 6) occurred between 2018/12/29 and 2019/09/29 in PA, which provides an excellent opportunity to investigate the triggering relationship among these events. We calculate the static Coulomb stress changes of the first five events, and find that the local seismicity after the 2018/12/29 Mw 7.0 earthquake is mostly associated with positive Coulomb stress changes, including the 2019/05/31 Mw 6.1 event, suggesting a possible triggering relationship. However, we cannot rule out the dynamic triggering mechanism, due to increased microseismicity in both positive and negative stress change regions, and an incomplete local catalog, especially right after the first Mw 7.0 mainshock. The dynamic stresses from these Mw > 6 events are large enough (from 5 kPa to 3532 kPa) to trigger subsequent events, but a lack of seismicity and waveform evidence does not support delayed dynamic triggering among these events, even the shortest time interval is less than 24 hours. In the past 45 years, the released seismic energy shows certain peaks every 5–10 years. However, earthquakes with Mw > 6.0 were relatively infrequent between 2004 and 2018 at PA. Hence, it is possible that several regions are relatively late in their earthquake cycles, which would enhance their susceptibility of being triggered by earthquakes at nearby and regional distances.
Modeling calving process of glacier with dilated polyhedral discrete element method
Lu Liu, Ji Li, Qizhen Sun, Chunhua Li, Sue Cook, Shunying Ji
2021, 40(7): 159-169. doi: 10.1007/s13131-021-1819-x
Mass loss caused by glacier calving is one of the direct contributors to global sea level rise. Reliable calving laws are required for accurate modelling of ice sheet mass balance. Both continuous and discontinuous methods have been used for glacial calving simulations. In this study, the discrete element method (DEM) based on dilated polyhedral elements is introduced to simulate the calving process of a tidewater glacier. Dilated polyhedrons can be obtained from the Minkowski sum of a sphere and a core polyhedron. These elements can be utilized to generate a continuum ice material, where the interaction force between adjacent elements is modeled by constructing bonds at the joints of the common faces. A hybrid fracture model considering fracture energy is introduced. The viscous creep behavior of glaciers on long-term scales is not considered. By applying buoyancy and gravity to the modelled glacier, DEM results show that the calving process is caused by cracks which are initialized at the top of the glacier and spread to the bottom. The results demonstrate the feasibility of using the dilated polyhedral DEM method in glacier simulations, additionally allowing the fragment size of the breaking fragments to be counted. The relationship between crack propagation and internal stress in the glacier is analyzed during calving process. Through the analysis of the Mises stress and the normal stress between the elements, it is found that geometric changes caused by the glacier calving lead to the redistribution of the stress. The tensile stress between the elements is the main influencing factor of glacier ice failure. In addition, the element shape, glacier base friction and buoyancy are studied, the results show that the glacier model based on the dilated polyhedral DEM is sensitive to the above conditions.
The crustal nature of the northern Mozambique Ridge, Southwest Indian Ocean
Nelta David Matsinhe, Yong Tang, Chun-Feng Li, Jiabiao Li, Estevão Stefane Mahanjane, He Li, Yinxia Fang
2021, 40(7): 170-182. doi: 10.1007/s13131-021-1747-9
The Mozambique Ridge (MOZR) is one of the basement high structures located in the Southwest Indian Ocean, parallel to the Southeast African continental margin. It was formed as a result of the tectono-magmatic evolution of the Gondwana breakup. The origin of the MOZR has been highly debated, with models suggesting either continental or oceanic origin. With new free-air gravity anomaly and multichannel seismic (MCS) reflection data, we present results of 2D density modeling along two seismic profiles acquired by R/V Xiangyanghong 10 at the northern Mozambique Ridge (N-MOZR) between 26°S and 28°S. We observed high free-air gravity anomaly and strong positive magnetic anomaly related to the emplaced seaward dipping reflectors (SDR) and high density lower crustal body (HDLCB), and high Bouguer gravity anomaly associated with the thinning of the continental crust underneath the N-MOZR over a distance of ~82 km. This suggests a thinned and intruded continental crust bound by the Mozambique Fracture Zone (MFZ) that is characterized by gravity low and negative magnetic anomaly. This fracture zone marks the continent-ocean boundary (COB) while the N-MOZR is the transform margin high, i.e., marks the continent-ocean transition (COT) of the Southern Mozambique margin, following the definition of transform margins. We suggest that the N-MOZR was formed by continental extension and subsequent breakup of the MFZ, accompanied by massive volcanism during the southward movement of the Antarctica block. The presence of SDR, HDLCB, and relatively thick oceanic crust indicates the volcanic nature of this transform margin.
Multi-beam and seismic investigations of the active Haima cold seeps, northwestern South China Sea
Bin Liu, Jiangxin Chen, Li Yang, Minliang Duan, Shengxuan Liu, Yongxian Guan, Pengcheng Shu
2021, 40(7): 183-197. doi: 10.1007/s13131-021-1721-6
To confirm the seabed fluid flow at the Haima cold seeps, an integrated study of multi-beam and seismic data reveals the morphology and fate of four bubble plumes and investigates the detailed subsurface structure of the active seepage area. The shapes of bubble plumes are not constant and influenced by the northeastward bottom currents, but the water depth where these bubble plumes disappear (630–650 m below the sea level) (mbsl) is very close to the upper limit of the gas hydrate stability zone in the water column (620 m below the sea level), as calculated from the CTD data within the study area, supporting the “hydrate skin” hypothesis. Gas chimneys directly below the bottom simulating reflectors, found at most sites, are speculated as essential pathways for both thermogenic gas and biogenic gas migrating from deep formations to the gas hydrate stability zone. The fracture network on the top of the basement uplift may be heavily gas-charged, which accounts for the chimney with several kilometers in diameter (beneath Plumes B and C). The much smaller gas chimney (beneath Plume D) may stem from gas saturated localized strong permeability zone. High-resolution seismic profiles reveal pipe-like structures, characterized by stacked localized amplitude anomalies, just beneath all the plumes, which act as the fluid conduits conveying gas from the gas hydrate-bearing sediments to the seafloor, feeding the gas plumes. The differences between these pipe-like structures indicate the dynamic process of gas seepage, which may be controlled by the build-up and dissipation of pore pressure. The 3D seismic data show high saturated gas hydrates with high RMS amplitude tend to cluster on the periphery of the gas chimney. Understanding the fluid migration and hydrate accumulation pattern of the Haima cold seeps can aid in the further exploration and study on the dynamic gas hydrate system in the South China Sea.
Holocene sea level trend on the west coast of Bohai Bay, China: reanalysis and standardization
Jianfen Li, Zhiwen Shang, Fu Wang, Yongsheng Chen, Lizhu Tian, Xingyu Jiang, Qian Yu, Hong Wang
2021, 40(7): 198-248. doi: 10.1007/s13131-021-1730-5
Using 110 newly revised Holocene sea level indicators categorized into three types, sediments (67), shelly cheniers (27) and oyster reefs (16), this paper firstly provides a Holocene relative sea level curve, based on multiple approaches of litho- and biostratigraphies and sedimentary faces analysis, for the west coast of Bohai Bay, China. Following considerations, including indicative meaning, the paleo tidal pattern and range and conversion from mean tidal level to mean sea level, an apparent relative mean sea level (RMSL) curve was further reconstructed. After systematical calibration using CALIB, those of the 48 reworked samples were further corrected for the residence-time effect. Similarly, the younger ages for another 35 samples were chosen at the subsample level. These result in a younger-oriented shift for about 0.5 ka. Three local spatial factors, including neotectonic subsidence (average rate about 0.1 mm/a), self-compaction of unconsolidated sediments (between a few decimeters to about 6 m) and subsidence due to groundwater withdrawal (between a few centimeters to about 2.5 m), were quantitatively corrected. Finally, the amended RMSL curve after eliminating all these local temporo-spatial factors is very likely to show non-existence of mid-Holocene highstand and imply potential influences of both ice-volume equivalent sea level and regional glacial isostatic adjustment. Although it is still unable to divide both global and regional factors, the slowdown of sea level rise, in 7.5–6.8 ka with a maximum height less than +1 m, may constrain the model effort in the near future.
Marine Information Science
Evolving a Bayesian network model with information flow for time series interpolation of multiple ocean variables
Ming Li, Ren Zhang, Kefeng Liu
2021, 40(7): 249-262. doi: 10.1007/s13131-021-1734-1
Based on Bayesian network (BN) and information flow (IF), a new machine learning-based model named IFBN is put forward to interpolate missing time series of multiple ocean variables. An improved BN structural learning algorithm with IF is designed to mine causal relationships among ocean variables to build network structure. Nondirectional inference mechanism of BN is applied to achieve the synchronous interpolation of multiple missing time series. With the IFBN, all ocean variables are placed in a causal network visually, making full use of information about related variables to fill missing data. More importantly, the synchronous interpolation of multiple variables can avoid model retraining when interpolative objects change. Interpolation experiments show that IFBN has even better interpolation accuracy, effectiveness and stability than existing methods.