Numerical Simulation of the Northwest Pacific Based on the MaCOM

Based on the GLORY reanalysis data, the simulation results of the two versions of MaCOM volume conservation and mass conservation in the Northwest Pacific Ocean for 8 years (1993-2000) are tested. In this paper, the sea surface height, sea surface flow field, sea surface temperature, salinity, vertical profile structures of sea temperature, sea salinity are evaluated respectively. The results show that the average absolute deviation of sea surface temperature is about 0.3 °C, the average absolute deviation of sea surface salinity is about 0.5 psu, the average absolute deviation of sea surface height is about 0.06 m, and the average absolute deviation of see surface current velocity is about 0.08 m/s. Among them, the difference between volume conservation and mass conservation is not large, about from percentile to thousandth percentile. In the vertical direction, the temperature profiles of the two versions are generally consistent, and The maximum deviation from GLORY data is about 0.5°C at depths of 100-400m. In terms of salinity profile, the deviation between the two versions and GLORY mainly exists in the depth range of 200-800 meters, and the deviation is 0.1-0.3 psu. On the whole, MaCOM runs stably and the results are good. It can well reproduce the temperature and salinity characteristics, ocean current field and dynamic environment of the ocean, and is consistent with the internationally recognized high-precision reanalysis data. It is suitable as another powerful tool for studying the ocean.


Introduction
Since Bryan [1] in the GFDL laboratory designed the first ocean circulation model in 1969, the ocean circulation numerical model has made many important progresses [2].Since the 1990s, with the continuous improvement of computer level, ocean numerical models have developed rapidly.Taking OGCM as an example, many models have been developed by different research institutions at this stage.MICOM (Miami Isopycnal Coordinate Model)[3], HYCOM (Hybrid Coordinate Ocean Model) [4], POP (Parallel Ocean Program) [5], MOM (Modular Ocean Model) [6].Institute of Atmospheric Physics, Chinese Academy of Sciences has developed a WOGCM with four layers in the vertical direction [7], which has been proved to have good simulation ability.This model has been further developed into LICOM (LASG / IAP Climate system Ocean Model) [8].With the rapid development of the ocean model, the ultimate goal of its research is to develop a three-dimensional numerical prediction system of ocean temperature and salinity, to provide reliable real-time prediction, and to develop marine reanalysis products that can provide information services for the majority of users [9].Although the numerical prediction models of Marine disasters such as storm surge, sea ice and tsunami have been completely independent in China, the operational prediction and scientific research of ocean circulation have been relying on the numerical models of the United States and the European Union [10].The international mainstream numerical models are large and complex, so the amount of model software code is also very large, which is not suitable for forecasters who want to fully control the numerical model and adjust the function at any time to apply to the forecast business.On the other hand, the current mainstream ocean numerical model is not responsible for R&D and maintenance by the business unit, so the upgrade is not oriented to the needs of the business unit.Often after a major upgrade, the functions of some business requirements disappear, many data interfaces or input and output specifications have changed, and even the numerical model basis of many assimilation systems has also changed.Forced the business system also had to make a series of unnecessary adjustments.In addition, the independent research and development model of the business unit can also improve the forecaster 's understanding of the numerical model, promote the closer combination of business work in different forecast fields, realize the positive feedback iteration between the numerical model and the forecast business, and better carry out the numerical forecast business work.In this context, we are required to independently develop the marine dynamic circulation model, develop the wave and sea ice model under the unified framework, and end the era of ' made by all countries ' for China 's marine core model.
The MaCOM (Mass Conservation Ocean Model) is an operational application-oriented ocean circulation numerical model independently developed by the National Marine Environmental Prediction Center of the Ministry of Natural Resources.It was officially released at the end of 2021.It has the characteristics of global and regional integrated numerical simulation capabilities, mass conservation dynamic framework, high-precision third-order nonlinear advection scheme, high-efficiency energysaving high-performance computing framework, and simple and easy-to-maintain software structure.The model mainly satisfies the numerical prediction ability of ocean circulation, and additionally takes into account the application ability of sea ice, tracer trajectory, coupling and so on.In addition, the model uses new technical means such as horizontal grid dimension reduction, graph virtual topology partitioning, and asynchronous I/O, which greatly improves the operation speed of the model and shortens the running time by more than 40% compared with the international mainstream global business system.At the same time, it supports GPU parallel acceleration, high efficiency and low consumption, and has excellent 'carbon friendly' attributes.
This paper consists of the following parts.The second part introduces the mode and configuration, the third part introduces the experimental results, and the last part is the summary.

Model configuration
The study area is shown in Figure 1.The latitude and longitude range is (98 ° E ~ 165 ° E, -12 ° S ~ 52 ° N), the grid resolution is 1/24 °, and the number of vertical layers is 75.The topographic relief data earth_relief_01m used in the process of grid generation is derived from SRTM15+ [11], and the shoreline data is published by GSHHG (http://www.soest.hawaii.edu/pwessel/gshhg/).

Data
The initial field and boundary field of the two versions of MaCOM volume conservation and mass con servation contain six variables.The volume conservation is seawater temperature, seawater salinity, thr ee-dimensional flow field and sea surface height, and the mass conservation is seawater temperature, s eawater salinity, three-dimensional flow field and seabed pressure.The original files used to make the initial field and the boundary field are all from the GLOBAL_MULTIYEAR_PHY_001_030 reanalysi s field data of the Copernicus Ocean Data Center (https://data.marine.copernicus.eu/product/GLOBAL_MULTIYEAR_PHY_001_030/description).The data resolution is 1/12°, and it assimilates a variety of observation data sources including satellites and buoys.After comparison, it can well reproduce the sea surface height, water mass characteristics, and mesoscale changes of the ocean [12].About the forcing files, MaCOM can receive ten kinds of atmospheric forcing variables at present, including meridional and zonal sea surface wind speed (sbcu, sbcv), sea surface air temperature (sbct), specific humidity (sbcq), long and shortwave radiation (lwdn, swdn), precipitation (prec), snowfall (snow), sea surface atmospheric pressure (slp), and runoff (taken from a dataset presented in Suzuki et al. ( 2017) [13]), all of which come from the reanalysis data JRA55-do of Japan Meteorological Agency (JMA).On the basis of JRA55, it is corrected by using satellite and other atmospheric reanalysis data, and the spatial resolution is increased to about 55 km, and the temporal resolution is increased to 3 hours.The evaluation shows that JRA can well replace the current CORE/OMIP dataset [14].
The tidal part of MaCOM is currently composed of two parts: tidal potential and resonant tide.The tidal potential is directly added to the source code part.The resonant tide part is constructed based on the OTPS algorithm, and all twelve tides including ' M2 ', ' S2 ', ' N2 ', ' K1 ', ' M4 ', ' O1 ', ' MN4 ', ' 2N2 ', ' Q1 ', ' P1 ', ' K2 ' and ' MS4 ' are added to the model calculation in the form of side boundaries [15].The model will read the harmonic constant from the file and calculates the tidal current.

Experimental setting
With the continuous development and improvement of the ocean model and the need of climate research and application, there have been many conclusions about the model error caused by the Boussinesq approximation in the volume conservation model [16].Among the many factors that lead to global sea level rise, the contribution of thermal expansion effect is as high as 40 % [17].Although this process can be reflected by mass correction in the volume conservation model, it is only an equilibrium solution, ignoring the local nature of the thermal expansion effect of seawater, which has no essential significance for reducing the error caused by the Boussinesq approximation [18].The mass conservation model can directly simulate this process without correction.As Huang et al. pointed out, the model based on Boussinesq approximation may not be suitable for describing the adjustment process from one climate state to another, and the numerical model of energy conservation may be more suitable for climate simulation research [19].The numerical model MaCOM based on the mass conservation of pressure coordinates may be able to solve the above problems.This experiment is expected to run for 28 years (1993-2020) for both volume conservation and mass conservation versions (This paper only shows the results of the first 8 years), and the output frequency is the daily average field.At present, the output elements are sea surface height, potential temperature, salinity, meridional and zonal velocity, vertical velocity, temperature-salt turbulent mixing coefficient, dynamic turbulent mixing coefficient, seabed pressure (only mass conservation model), meridional and zonal wind stress, downward long-wave radiation, short-wave radiation, evaporation-precipitation-runoff difference.At the same time, this experiment plans to run the volume conservation version on CPU and the mass conservation version on GPU.The fastest time for GPU to calculate a model day can be 4 minutes faster than CPU (Details in table 1).

Results
The results of temperature, salinity, sea surface current and sea surface height in the first 8 years (1993-2000) of the model operation are compared with the GLORYS12V1(mentioned above) data.The comparison results of each physical quantity are as follows:    It can be seen that the MaCOM simulation results basically restore the "sandwich" structure of the sea surface height in the Northwest Pacific Ocean, that is, the distribution pattern of " low, high and low " from high latitude to low latitude.At the same time, the difference between the two versions of volume conservation and mass conservation can be preliminarily seen in the region of Kuroshio Extension where the dynamic environment is more complex.

Sea surface results
Figure d, h and I show the simulation results of the sea surface flow field.The North Equatorial Current (NEC), South Equatorial Current (SEC), North Equatorial Counter current (NECC), New Guinea Coastal Current (NGCC), Kuroshio, Subtropical Counter current (STCC), etc [20].are mainly distributed in the Northwest Pacific Ocean.It can be seen from the figure that MaCOM basically simulates these flow systems, as well as Mindanao Vortex (ME) and Halmahera Vortex (HE).
Figure 3 shows the deviation between the two versions of MaCOM volume conservation and mass conservation and GLORY data.
In terms of salinity, compared with GLORY, the average absolute deviation of volume conservationis about 0.58 psu, and the average absolute deviation of mass conservation is about 0.57 psu.The main deviations of the two are concentrated in the Yellow Sea, Bohai Sea and Southeast Asian Seas, with the size of −1.5 ~ −0.5 psu.This may be due to the fact that MaCOM has not added the sea ice module for the time being, and the runoff data used in the MaCOM forcing field is inconsistent with the runoff data used by GLORY.
Judging from the situation of sea surface temperature, from a local perspective, the deviation between MaCOM and GLORY data is mainly concentrated in the Kuroshio Extension region, the Hamachira vortex and the New Guinea coastal current region.From the perspective of the whole region, the average absolute deviation between volume conservation and GLORY is about 0.28 °C, and the average absolute deviation between mass conservation and GLORY is about 0.30 °C.
Let's take a look at the height of the sea surface, the deviation between MaCOM and GLORY is small, basically between ± 0.3 m.The average absolute deviation between volume conservation and GLORY is about 0.064 m, and the average absolute deviation between mass conservation and GLORY is about 0.065 m.The larger deviation occurs in the Kuroshio Extension and the North Equatorial Counter current.
Finally, in terms of sea surface current, the deviation between MaCOM and GLORY data is small.The average absolute deviation between volume conservation and GLORY is about 0.082m / s, and the average absolute deviation between mass conservation and GLORY is about 0.08m / s.The main flow velocity deviations are distributed in the Kuroshio Extension, the North Equatorial Current and the South Equatorial Current, which are 0.4 ~ 0.5m / s, −0.2 ~ 0.1m / s, −0.35m / s ~ −0.25m / s, respectively.

Vertical result
The vertical temperature and salinity profile of the upper 2000 meters is shown in Figure 4.In terms of sea temperature, the results of mass conservation and volume conservation are basically the same.The deviation between them and GLORY data is mainly in the upper 800 meters, and the maximum deviation from GLORY data is about 0.5°C at depths of 100-400m.In terms of salinity, the difference between mass conservation and volume conservation is large, and it is mainly distributed in two depth layers of about 200 meters and about 600 meters.At the same time, the deviation between the two and GLORY also mainly exists in the depth range of 200-800 meters, and the deviation is 0.1-0.3psu.

Conclusion
In this paper, the MaCOM simulation results of 8 years (1993-2000) are processed and evaluated with GLORY data.The operation of MaCOM is stable and the simulation effect is good.From the results of this paper, MaCOM can simulate the basic characteristics of sea temperature and salinity.At the same time, it can accurately simulate the change of sea surface height and the main flow systems in the Northwest Pacific.In terms of there are still deviations in salinity in the Yellow Sea, Bohai Sea and Southeast Asia, mainly due to the fact that the MaCOM model does not currently couple the sea ice module, and the inconsistency with the use of runoff files in the GLORY forcing field file.At the same time, it can be seen that in the case of high resolution, the difference between volume conservation and mass conservation is relatively significant in the area with complex dynamic environment, and how this difference is generated remains to be explored in detail in the future.In the vertical direction, there are also differences between the two in the subsurface area, which is not completely consistent with the previous results in the case of low resolution [21].Overall, in the case of only 8 years of operation, MaCOM has been able to reflect the difference between volume conservation and mass conservation at high resolution, and the simulation results are highly consistent with the internationally recognized high-precision reanalysis data.At present, the assimilation of MaCOM and the coupling of sea ice and ecological modules are in progress.The data assimilation method based on A-4DEnVAR [22] has been applied to MaCOM.At the same time, the R & D team hopes to build an ice-sea coupling system that can reflect the characteristics of mass conservation.In addition, they also hope to establish a coupling with the CoSiNE model [23,24].It is believed that MaCOM will become another powerful tool to study the ocean after improving the data assimilation work and the coupling of sea ice and biological modules in the future.

Fig 1 .
Fig 1. Regional and topography of the Experiment.

Fig 2 .
Fig 2. The results of a-d are the sea surface salinity, sea surface temperature, sea surface height, and sea surface flow field (the colour diagram is the flow velocity, and the black arrow is the flow direction) of the MaCOM volume conservation version, e-h are the result of mass conservation, and i-l are the result of GLORY.Fig 2 shows the results of the eight-year mean field of four quantities of sea surface temperature, sea surface salinity, sea surface height and sea surface current.

Figure
Figure a, e and i show the annually average results of sea surface salinity of MaCOM and GLORY in 8 years.It can be seen that MaCOM basically simulates the distribution characteristics of temperature in the northwest Pacific Ocean.On the whole, affected by solar radiation, sea surface temperature shows a decreasing distribution pattern from low latitude to high latitude.At the same time, it simulates that the temperature is affected by ocean currents, and the isotherms in some areas have changed greatly, especially in the Kuroshio area.

Fig 3 .
Fig 3.As shown above, a-d are the deviation results of sea surface salinity, sea surface temperature, sea surface height, sea surface flow field (colour map is flow velocity, black arrow is flow direction) of the MaCOM volume conservation version and the GLORY, and e-h is the deviation result of mass conservation version and GLORY.Figs.b, f and j show the annually average results of sea surface temperature in 8 years.MaCOM simulation results show that the salinity of the Northwest Pacific Ocean is regulated by the high salinity of the North Equatorial Current.Similar to the distribution of sea temperature, the sea surface salinity also shows a gradual decrease from low latitude to high latitude.At the same time, because it is the Figure c, g and k show the annually average sea surface height of MaCOM and GLORY in 8 years.

Fig 4 .
Fig 4. The vertical structure diagram of seawater temperature and salinity, in which figure a is the result of temperature and figure b is the result of salinity.The blue curve is the result of GLORY, the black curve is the result of volume conservation, and the red curve is the result of mass conservation.

Table 1 .
Comparison table of operation rate of volume conservation and mass conservation.