Severe ice conditions in the Bohai Sea could cause
serious harm to maritime traffic, offshore oil exploitation, aquaculture,
and other economic activities in the surrounding regions. In addition to
providing sea ice forecasts for disaster prevention and risk mitigation, sea
ice numerical models could help explain the sea ice variability within the
context of climate change in marine ecosystems, such as spotted seals, which
are the only ice-dependent animal that breeds in Chinese waters. Here, we
developed NEMO-Bohai, an ocean–ice coupled model based on the Nucleus for
European Modelling of the Ocean (NEMO) model version 4.0 and Sea Ice
Modelling Integrated Initiative (SI
The Bohai Sea is the southernmost seasonal frozen sea in the Northern Hemisphere (Yan et al., 2017). The formation of sea ice in the Bohai Sea mainly depends on the geographical environment and hydrometeorological characteristics (Ding, 1999). Specifically, the Bohai Sea is located in the continental shelf area, and the average water depth is only 18 m (Su and Wang, 2012), which indicates low oceanic heat content in winter. Following a northern continental climate, the Bohai Sea is affected by the cold Siberian air every winter, which causes the sea surface temperature of the Bohai Sea to be significantly lower than that at the same latitude (Zhang et al., 2016; Donlon et al., 2012). In addition, the sea surface salinity of the Bohai Sea is about 30 PSU, which is the lowest in the entire coastal waters in China (Yan et al., 2020). It means that the Bohai seawater freezes before reaching the maximum density. Therefore, it even more easily convects and loses heat just before freezing. The sea ice season is generally from December to March. The sea ice in the Bohai Sea also exhibits significant interannual variability, which could cover half of the sea area during severe winters (Su and Wang, 2012).
With 4 of China's top 10 ports (Tianjin, Tangshan, Dalian, Yingkou), the
Bohai Sea is an important economic zone in China (Fu et al., 2017). Under
severe sea ice conditions, such as during the 2009/2010 winter, 296 ports in
this region were frozen. In particular, the 42 km artificial waterway of
Huanghua Port was covered by sea ice; thus, the maritime traffic was
severely restricted, and 7157 fishing vessels were damaged (Gu et al.,
2013). About
In addition, spotted seals (
Most research has focused on the polar region, however, sea ice at middle to high latitudes is also sensitive to climate change (Bai et al., 2011; Gong et al., 2007; Yan et al., 2017). Knowledge of regional sea ice variability is vital to studying how climate change will affect regional basins. The traditional view on the Bohai Sea ice is concentrated on extracting sea ice data from satellite imagery (Karvonen et al., 2017; Shi and Wang, 2012; Su et al., 2019). As the use of remote sensing data to retrieve sea ice data has limitations in providing a continuous record for a long time series, numerical modelling is an effective tool to understand the sea ice development from freezing to thawing (Hordoir et al., 2019; Pemberton et al., 2017).
Sea ice models can reasonably simulate sea ice conditions, which can
primarily supplement remote Earth observations. To our knowledge, there are
only very few studies on developing regional coupled ocean–ice models for
the Bohai Sea, and most of them have focused on short-duration simulations,
such as 1-week or 1-year case studies. Wang et al. (1984) reported the
first sea ice dynamic–thermodynamic model for the Bohai Sea, which simulated
the sea ice freezing–melting cycle at a resolution of 20 km
NEMO is a state-of-the-art numerical modelling framework designed for research activities and forecasting services in the ocean and climate sciences (Madec et al., 2016). The NEMO code, including its sea ice component, enables the investigation of ocean–ice dynamics and thermodynamics and their interactions with the atmosphere. NEMO offers a wide range of applications from short-term forecasts and climate projections (Drouard and Cassou, 2019; Obermann-Hellhund et al., 2018; Uotila et al., 2017; Voldoire et al., 2013) to process studies (Courtois et al., 2017; Declerck et al., 2016; Feucher et al., 2019). While global models offer a poor representation of regional ocean processes, regional models have been developed based on the framework of NEMO in recent years for various seas, e.g. the Atlantic marginal basins (Graham et al., 2018; O'Dea et al., 2017), the Baltic and North seas (Hordoir et al., 2019; Pemberton et al., 2017), the northwestern Mediterranean Sea (Declerck et al., 2016), the southern Indian Ocean (Schwarzkopf et al., 2019), the South China Sea (Thompson et al., 2019), and the Black Sea (Gunduz et al., 2020). However, a NEMO-based regional model for the Bohai Sea has not been attempted for long-term climate studies until now.
Long-term sea ice simulations could help improve our understanding of thermodynamic and dynamic sea ice processes in the Bohai Sea, crucial for sea ice disaster prevention, spotted seal habitat studies, and regional climate change studies. The paper aims to report the development of NEMO-Bohai and assess its performance. The paper is organized as follows: Sect. 2 introduces the model and observation dataset. Model comparison and validation are carried out in Sect. 3. The analysis of sea ice variation based on the 22-year hindcast simulations of NEMO-Bohai is presented in Sect. 4. A discussion and summary are provided in Sect. 5.
NEMO-Bohai is localized to the Bohai Sea based on the NEMO ocean engine
(Madec et al., 2016). We apply the NEMO 4.0 beta and Sea Ice Integrated Modelling
Initiative (SI
The global simulation is spun up in January 1987 with the rest state and
ended in 2017. The time step is set to 900 s. The ORCA025 configuration is
performed using 240 CPU cores on a Cray XC40 system on the Sisu
supercomputer and requires
The NEMO-Bohai domain consists of one central zone and three bays (see Fig. 1). The area covers 37–41
The location and bathymetry of the Bohai Sea. Black dots denote the coastal tide gauge stations. Red dots and black triangles indicate station locations with sea surface temperature and sea surface salinity observations, respectively. Black squares denote the oil platforms where the sea ice thickness observations were conducted. Line A–B represents the profile section along which temperature and salinity were presented (Figs. 6 and 7, respectively), while line C–D indicates the Bohai Strait. The straight line in the east denotes an open boundary for the regional ocean model.
The initial conditions of the numerical simulation in the Bohai Sea for 1 July 1995, including SST and SSS, are obtained by interpolating the
temperature and salinity fields from the global ocean simulations with the
ORCA025 configuration, as shown in Fig. 2. The regional model is forced with
lateral ocean boundary conditions, tidal forcing, atmospheric forcing, and
river runoff during the study period. Two kinds of boundary conditions are
used for the setting of lateral open ocean boundary. A flow relaxation scheme
is applied to baroclinic velocities and tracers (Engedahl, 1995), while a
Flather boundary condition (Flather, 1976) is used for barotropic dynamics,
such as SSH and barotropic velocities (u2d, v2d). NEMO-Bohai has one open
boundary located 100 km away from the Bohai Strait, and we set the
relaxation zone width to 1 (nn
Examples of
The NEMO-Bohai atmospheric forcing is derived from the ERA5 dataset, the
ECMWF's latest reanalysis product covering the period from 1979 to the
present, which has replaced the widely used ERA-Interim dataset. The data
cover the Earth with a horizontal spatial resolution of 30 km and represent
the atmosphere using 137 levels from the surface up to a height equaling
0.01 hPa. The forcing files consist of 1-hourly instantaneous fields of the 2 m air temperature, 10 m wind speed, downward short-wave and long-wave radiation,
sea level pressure, specific humidity, and precipitation. The equivalent
inverse barometer SSH is calculated from the atmospheric pressure
(ln
The bottom roughness impacts the dynamics of the tide, ocean circulation,
and storm surges in the Bohai Sea. A constant bottom roughness
(rn
Comparison of sea surface height (SSH) between the observations
and NEMO-Bohai simulations at
Key physical parameters in the ocean namelist
(namelist
The simulated sea surface temperature (SST) compared with in situ
observations at Laohutan, Dalian
SI
Simulated monthly mean current velocities at the surface and 16 m
depth in August 2012 and February 2013. The monthly mean current velocities
are calculated based on the outputs with hourly intervals. The black lines
and arrows represent the streamlines and directions of the current vector
field, respectively. The filled contours denote the current speed (in m s
In NEMO-Bohai, we selected and adjusted a series of sea ice model
parameters. We increase the number of ice categories (jpl) and the number of
ice layers (nlay
The ice initialization is activated (ln
Comparison of vertical profiles of water temperature (
In this section, we analyse how well NEMO-Bohai reproduces ocean properties (SSH, SST, SSS, temperature and salinity stratification, currents, and volume exchanges with the Yellow Sea) and sea ice properties (area, thickness, and volume). To evaluate the model's performance, we compare our model results to in situ and satellite observations from multiple sources.
Similar to Fig. 6 but for salinity (PSU).
To evaluate the SSH, the modelled tide amplitude and phase were compared to
the tide tables with hourly intervals of Yantai (37.550
Comparison of daily sea ice area (DSIA) between NEMO-Bohai simulations (black line) and satellite-derived data (blue circles) from 1996 to 2017.
The sea ice model is evaluated against a series of observational datasets. A satellite-derived dataset covering the winters from 1988 to 2017 was retrieved. Sea ice area was extracted from two datasets: the first was based on the zonal threshold method for Advanced Very High Resolution Radiometer (AVHRR) during 1988–2000, and the second was based on an object-based feature extraction method for Moderate Resolution Imaging Spectroradiometer (MODIS) from 2001 to 2017. A detailed description can be found in Yan et al. (2017). The shapefiles derived by the aforementioned methods are further modified by visual interpretation to build a more accurate sea ice area dataset. For the sea ice thickness evaluation, we use the in situ observations from four offshore oil platforms (see Fig. 1) on 2, 6, 9, 16, and 26 January and 2, 9, 12, and 16 February during 2013 (Zeng et al., 2016; Karvonen et al., 2017).
Comparison of the spatial distribution of sea ice from NEMO-Bohai
simulation and remote sensing inversion in freeze, severe freeze, and thaw
periods for light
Reliable precision of the tidal model is a prerequisite for subsequent
simulations by NEMO-Bohai. The tidal model is validated by in situ hourly
tidal observations at Bayuquan and tide table data at Yantai, Dalian, and
Jingtanggang. Figure 3 provides a robust description of the comparison of SSH
time evolution at the four stations. The tidal range in the Bohai Strait
(Yantai and Dalian) is larger than that at other tide gauge stations. It
is clear that the modelled water elevation at Bayuquan and Jingtanggang
stations agrees less with the tide table data than at the other two
stations, as shown in Table 2. Nevertheless, the model reproduced the
semidiurnal M2 tidal cycle well with exact phase and amplitude at four stations
with a mean correlation of 0.92, mean absolute error of 0.17 m, and
root-mean-square error of 0.22 m. The model reproduces SSH standard
deviation well, and the difference is within 6 cm in Yantai, Dalian, and
Jingtanggang (Table 2); the largest deviation at Bayuquan is approximately
11 cm. Simulated SSH depends on the model's bathymetry, open boundary
forcing, and freshwater flux (Kärnä et al., 2021). The errors are
partly due to the deviation of positions between the tidal gauges and the
model grid points, as the model has a horizontal grid size of
The comparison of sea ice thickness between the simulated and observation data in the Bohai Sea. The vertical red bars represent the range of the observations, while the blue triangles denote the simulations based on NEMO-Bohai.
Sea surface height representation, in terms of standard deviation (SD, metres), correlation, mean absolute error (MAE, metres), and root-mean-square error (RMSE, metres), made by NEMO-Bohai for four Bohai Sea stations.
As shown in Fig. 4, the modelled SST followed the seasonal cycle well, and
the mean absolute error was typically less than 1
Comparison of daily sea ice volume between satellite-derived data (circles) and NEMO-Bohai simulations (grey bars) from 1996 to 2017.
Table 3 displays the comparisons of multi-year average sea surface
salinities at eight ocean stations in the Bohai Sea between simulations from
NEMO-Bohai and observations reported by Yuan et al. (2015). Generally
speaking, NEMO-Bohai is able to capture the main variations of the sea
surface salinity in the Bohai Sea. Values at six ocean stations agree with
observations with a relative deviation of less than 5 %, while the modelled
values at Huludao and Tanggu are less salty than observed. This demonstrates
that the model faces more significant challenges in the low-salinity areas.
Primarily, the freshwater river runoff leads to lower salinity in the
coastal regions. In NEMO-Bohai, the river runoff is based on climatological
estimates without interannual variability. The river salinity is assumed to
be 0 PSU, and the river temperature is set to the same value as the SST at
the closest grid point. All these reasons might cause underestimations. In
addition, the biases of river runoff and shallow water depth (generally
Spatial distributions of multi-year monthly average sea surface temperature in the Bohai Sea for the period of 1996–2017.
Comparison of multi-year average sea surface salinity between the observations and simulations.
The simulated monthly mean current velocities at the surface and 16 m depth
in February and August are shown in Fig. 5. The monthly mean current
velocities are calculated based on hourly model output during August 2012
and February 2013. Figure 5 shows that both the sea surface and 16 m depth
currents are usually less than 0.4 m s
Spatial distributions of multi-year monthly average sea surface salinity (PSU) in the Bohai Sea during 1996–2017.
The monthly mean water volume exchange at the Bohai Strait (see Fig. 1) based
on hourly model simulations during August 2012 and February 2013 are also
calculated to evaluate the model's performance. The Bohai Sea water exchange
with the Yellow Sea is weak due to its half-closed shape and a relatively
independent circulation system. The model results show that the inflow from
the Yellow Sea to the Bohai Sea in August reaches
NEMO-Bohai and observed water temperature and salinity profiles along the transect A–B (see Fig. 1) are shown in Figs. 6 and 7, respectively. Observations are from the atlas by Chen (1992), which is based on data from the 1950s to 1990s. The temporally closest 5-year period from 1995 to 2000 of NEMO-Bohai simulations was selected for model–observation comparisons. Common features are found between the model and observations. The Bohai Sea waters are mixed well vertically in autumn and winter, and they have a remarkable homogeneous vertical distribution for both temperature and salinity. In spring and summer, thermal stratification occurs with a significant cold-water core at depth that is eventually eroded in autumn. As apparent in Figs. 6 and 7, the stratification in shallow coastal waters is generally homogeneous. Similar features were reported by Wang et al. (2008), who analysed the seasonal variations of the vertical profiles in the Bohai Sea.
The annual average sea ice area
The model results, however, show some discrepancies compared to the atlas. Although the model reproduces the summer saline stratification, it is weaker than in the atlas. Nonetheless, Li et al. (2015) reported that the summer salinity stratification in the Bohai Sea is possibly weaker than in the atlas, with an observed top-to-bottom salinity difference of 0.6 PSU. The modelled salinity stratification in summer is weaker compared to the atlas, which is possibly caused by the vertical mixing setting with the used TKE closure scheme, and the high setting of vertical diffusivity in the model. In the northern part of the transect, which corresponds to northern Liaodong Bay, a negative salinity bias is visible compared to the atlas. In addition to the reasons mentioned in Sect. 3.2.2, inaccuracies in the ETOPO1 bathymetry, especially in the low-water-depth region seen from Figs. 6 and 7, may also cause these underestimations.
Seasonal cycles of daily sea ice area
Sea ice is the focus of our study and is, in this section, directly validated against observations with respect to the area, thickness, volume, and their variations.
To evaluate the performance of NEMO-Bohai in simulating the characteristics
of sea ice area, a series of variables were validated, including the daily
sea ice area (DSIA), annual maximum sea ice area (AMSIA), and spatial
patterns. Figure 8 shows that the model-estimated DSIA agrees well with the
satellite-derived observations (
Comparison of the occurrence dates of the annual maximum sea ice area between the observations and model simulations.
Climatological monthly mean sea ice concentration in the Bohai Sea during 1996–2017, calculated as the average of all daily sea ice concentrations for each month during the ice period.
The comparison of the spatial distribution of sea ice from the NEMO-Bohai simulation and remote sensing inversion during the freeze, severe freeze, and thaw periods in a light ice year (2007), normal ice year (2009), and heavy ice year (2010) is shown in Fig. 9. The simulated spatial distributions reflect general characteristics of sea ice evolution in the Bohai Sea except for the bias at the ice edge zones. Sea ice is mainly located in Liaodong Bay in light and normal ice years with extension to Bohai Bay and Laizhou Bay in heavy ice years. It is worth mentioning that NEMO-Bohai has reproduced the development cycle of the sea ice well but shows a relatively slow thawing process.
Figure 10 shows that the modelled sea ice thickness based on NEMO-Bohai mostly
lies in the range of in situ observations with a slight overestimation. The
mean relative bias of sea ice thickness between the simulations and
observations is 4.6 cm (
Similar to Fig. 16 but for sea ice thickness.
Sea ice volume is defined as the total ice over the whole Bohai Sea,
calculated through sea ice concentration multiplied by ice thickness in all
grids. As demonstrated in Fig. 11, modelled daily sea ice volume is in
reasonable agreement with satellite-derived data between 1996 and 2017. The
modelled sea ice volume is higher than the observed, with the mean relative
bias of
Frequency distribution of daily sea ice thickness larger than 10 cm
Results of model comparison with in situ data and satellite-derived data confirm the robustness of the developed model, which allows us to use it in a more detailed evaluation of the spatial and temporal changes of Bohai Sea ice and study the continuous processes of freezing and melting at daily scales.
Figure 12 shows the monthly seasonal cycle of average SST in the Bohai Sea.
There is an obvious “warm tongue” starting from the Bohai Strait to the
central Bohai Sea in winter (December, January, and February) due to the
warm current from the Yellow Sea. It turns into a “cold tongue” in summer
due to faster temperature increases in shallow coastal waters as the
influence of the warm current from the Yellow Sea also weakens. During
January and February, the SST is below 0
As shown in Fig. 13, except for estuaries, the salinity isolines of the Bohai Sea are roughly parallel to the coast, as the salinity in the coastal water is generally low due to river runoff. The average salinity of the whole Bohai Sea is low in summer (lowest in August at 29.5 PSU) and high in winter (highest in February at 30.0 PSU). The Bohai Strait in the east, which connects the Bohai Sea with the Yellow Sea, exhibits the highest salinity in the entire Bohai Sea, with an average of about 32.8 PSU.
Similar to Fig. 16 but for sea ice velocity.
The variation of annual average and maximum sea ice area in the Bohai Sea
based on NEMO-Bohai from 1996 to 2017 are shown in Fig. 14a and b,
respectively. The annual average sea ice area (AASIA) is an average of the
DSIAs during the ice period for each year. The sea ice area exhibits
apparent interannual and decadal variability in the study period. The mean
AASIA during 1996–2017 is
The length of the ice period is defined as days with a sea ice area greater
than 100 km
Pearson correlations between annual average sea ice area (AASIA)
The climatological seasonal cycles of the sea ice area and volume in the
Bohai Sea averaged over 1996–2017 show unimodal variations (Fig. 15). Ice
usually starts to form in mid-December and reaches the maximum in early
February. It then starts to melt until the Bohai Sea becomes free of ice by
mid-to-late March. The climatological mean of the length of the ice period
is about 3.5 months, and the freezing period (
This section will mainly discuss the spatial features of Bohai Sea ice. As shown in Fig. 16, there are substantial seasonal and spatial differences in sea ice concentrations. To be specific, seawater freezes first in Liaodong Bay, then in Bohai Bay, and finally in Laizhou Bay, showing significant variation with latitude; the melting process happens in exactly the reverse order. Similar to the aforementioned temporal changes in the last section, the sea ice concentration reaches its peak in February. Seawater in Liaodong Bay freezes most severely, and the high concentration zone moves from the north in January to the east in February, where it survives the longest until March.
As shown from Fig. 17, sea ice thickness shows quite similar seasonal and
spatial features to sea ice concentration. The monthly mean Bohai Sea ice
thickness simulated by NEMO-Bohai usually reaches its maximum in February,
with a monthly mean thickness of 16.9 cm, following the second-highest
monthly mean thickness of 15.8 cm in January. Sea ice is thicker on the eastern
coast of Liaodong Bay than that on the western coast in January and February.
The maximum sea ice thickness appears near Bayuquan on the eastern coast of
Liaodong Bay, where the thickness reaches up to
Spatial correlation between daily sea ice concentration in the
Bohai Sea and
According to the guideline for risk assessment and zoning of sea ice disaster issued by the State Oceanic Administration in 2016, a high-risk level is reached when sea ice thickness becomes greater than 25 cm, while the risk level is low when sea ice thickness is lower than 10 cm. When sea ice thickness is between 10 and 25 cm, the risk level is moderate. Accordingly, in this study, the thresholds for moderate-risk and high-risk levels of sea ice disaster are set at 10 and 25 cm, respectively. In Fig. 18, sea ice risk maps are calculated based on daily sea ice thickness from 1996 to 2017, and they clearly show that the high-risk area is mainly located at Liaodong Bay, with the highest risk area in the eastern Liaodong Bay. Thus, the seas around Yingkou, Bayuquan, and Wafangdian bear high exposure to sea ice disasters.
As the drift ice is a major component in Bohai Sea ice, studying its motion
characteristics is also essential. As shown in Fig. 19, the direction is
mainly northeast–southwest, and the drift exhibits a high spatial
variability. The zone with the highest sea ice velocity (
Synoptic forcing may play an essential role in the changes of Bohai Sea
ice, which is primarily influenced by the Eurasian continental climate. To
explore the potential regional climate drivers on the evolution of sea ice,
correlations between sea ice and various indices based on detrended air
temperature, air pressure, wind speed, and precipitation were examined.
These indices were calculated based on daily meteorological measurements
during the winter (DJFM) from 1996 to 2017 at 12 weather stations (Dalian,
Wafangdian, Xiongyue, Yingkou, Jinzhou, Xingcheng, Suizhong, Qinhuangdao,
Laoting, Tanggu, Huanghua, Dongying) surrounding the Bohai Sea. These
measurements were obtained from the China Meteorological Data Service Center
(
Figure 21 illustrates the spatial correlations between daily sea ice concentration with vertically integrated ocean heat and salt content from the surface to the bottom of the mixed layer during the ice period for each grid point. Ocean heat content strongly correlates with sea ice concentration in Liaodong Bay and moderately correlates near the coastal areas. The spatial pattern corresponds quite nicely with climatological monthly mean sea ice concentration (shown in Fig. 16), with the highest values appearing in the eastern part of Liaodong Bay. On the other hand, a very weak positive or negative correlation between daily sea ice concentration and integrated ocean salt content was found in most sea areas, indicating a more complicated relationship throughout the whole Bohai Sea. Therefore, the interannual variability of sea ice is more dominated by ocean heat content than salt.
In this study, we provided a detailed description of NEMO-Bohai, a newly developed setup of an ocean–ice model for the Bohai Sea. The primary intent of our study was to test how NEMO-Bohai represents the ocean characteristics and sea ice properties.
Comparisons with observational data confirm that NEMO-Bohai is able to reproduce the ocean properties reasonably well, including ocean surface information (e.g. SSH, SST, SSS), currents, and temperature and salinity stratification. However, the ongoing development of the NEMO ocean engine, which will soon provide an updated version with the inclusion of wetting and drying processes, could improve the model performance in terms of sea surface height in shallow areas (Hordoir et al., 2019), especially for the Bohai Sea, which has an average depth of only 18 m. A higher-precision bathymetry could also improve the SSH performance. In this study, the monthly climatology of coastal runoff flux from Dai et al. (2009) was recorded only for 1948–2004, which is different from the study period. More importantly, the river runoff is based on climatological estimates without interannual variability, and we assume that the river salinity is set to 0 PSU and that the runoff temperature is the same as the ocean surface. Major future development would be using gauge records with observations of temperature and salinity assimilated into the NEMO-Bohai model, which would reduce uncertainties in temperature and salinity. Furthermore, a possible implementation of multiple embedded methods (Hvatov et al., 2019; Schwarzkopf et al., 2019), such as Global Ocean–West Pacific Ocean–East China Sea–Bohai Sea rather than the current Global Ocean–Bohai Sea direct nesting, should be investigated in the future. In addition, in order to carry out more accurate estimations of vertical mixing, it is worth implementing the experiments of turbulent vertical mixing options (Reffray et al., 2015) for the Bohai Sea for further development of NEMO-Bohai. Moreover, interactive feedback from the Bohai Sea to the global ocean could also be considered in a two-way coupling method compared to current one-way nesting.
For the sea ice component, NEMO-Bohai satisfactorily reproduced the seasonal and interannual variabilities of sea ice area compared to the satellite remote sensing data for the period 1996–2017. The modelled dates of the annual maximum sea ice area were 0.9 d earlier than the observed ones. Spatially, the simulation results realistically reflect the main characteristics of Bohai Sea ice evolution compared to the satellite data. Therefore, NEMO-Bohai can reliably be used to provide ice information during the dates without satellite-derived data. Nevertheless, it is worth mentioning that the simulated sea ice area tends to be somewhat overestimated, which is also reported for other seas in earlier NEMO-related publications (Blockley et al., 2014; Massonnet et al., 2011; Rjazin et al., 2019). In spring, the NEMO-Bohai melting process is delayed, particularly in the land-fast ice zone in eastern Liaodong Bay. It is likely that thick ice melts away slower than thinner observed ice. The applied parameterization of the surface albedo may cause a slow melting process, which does not take realistically into account the relevant physical processes, such as surface melt ponds, as the surface albedo continues to decrease until sea ice completely disappears (Mortin et al., 2016). Thus, regional atmospheric forcing data with higher accuracy and resolution can be used for further development. In addition, the space discrepancy between modelled sea ice thickness and extremely limited in situ observations makes their comparison difficult and introduces significant uncertainties.
In conclusion, NEMO-Bohai can simulate ocean and sea ice properties with reasonable skill in a broad spatiotemporal context, especially in terms of seasonal evolution and long-term interannual variations of sea ice. This finding implies that NEMO-Bohai complements the discontinuous satellite data well in sea ice hazard risk analysis. Therefore, NEMO-Bohai is a valuable tool for long-term ocean and ice simulations and climate change studies.
NEMO-Bohai is built upon the standard NEMO code (NEMO 4.0 beta, revision
10226). The reference code is available at the NEMO website
(
YY, WG, and PU designed the study. YY and AMUG ran the experiments. YY and YX analysed the model and observational data. YY and PU wrote the manuscript. All authors provided scientific input.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors acknowledge CSC – IT Center for Science, Finland, for HPC computational resources. The authors also gratefully thank Juha Lento, Robinson Hordoir, Yongmei Gong, Yu Zhang, and Xiaoqiao Wang for sharing the scripts for data processing.
This research has been supported by the National Key Research and Development Program of China (No. 2021YFA0719104), the Academy of Finland (grant 322432), the National Natural Science Foundation of China (No. 41977406), the China Scholarship Council (No. 201806040130), the 111 project (No. B20011), and the Fundamental Research Funds for the Central Universities (No. 2-9-2020-007).Open-access funding was provided by the Helsinki University Library.
This paper was edited by Andrew Yool and reviewed by two anonymous referees.