Global Sea Surface Temperature Analysis Based on Domestic Ocean Satellite Data. Part II: Comparative Analysis of Multiple Sets of Global Sea Surface Temperature Products in 2022

With more and more global gap-free fusion products of the sea surface temperature, understanding the consistency and discrepancy of the different SST fusion products will not only help data providers to improve their algorithms, but also help them to select the one that may better suit their applications. In this article, we have compared and analysed 10 sets of global gap-free fusion products of sea surface temperature in 2022, with different fusion techniques and related configurations. It is found that each SST analysis product has the same spatial distribution, with the minimum NMEFC mean value (20.11) and the maximum MGDSST mean value (20.31). Compared with ARGO in-situ data, the RMSE ranges are from 0.3233 (OSTIA) to 0.5180 (MGDSST). The RMSE between NMEFC fusion products and Argo in-situ data is 0.3861, ranked fifth out of 10 sets of fusion products. Compared with GMPE analysis among 9 sets of fusion products, the RMSE ranges are from 0.1579 (CMC) to 0.3199 (K10), and the NMEFC fusion product has a RMSE of 0.3040, which is at the intermediate level, ranked sixth out of 9 fusion products.


Introduction
Sea surface temperature plays a key role in heat, freshwater, and momentum flux exchange between ocean and atmosphere.SSTs are widely used as boundary conditions in weather and climate forecast systems [1] and as initial conditions in ocean forecast systems [2].Therefore, evaluating the quality of SST data is critical from operational to climate studies and related services.The Group for High-Resolution Sea Surface Temperature (GHRSST) is aimed at coordinating the provision of SST products developed and distributed by different operational agencies and research institutes [3].Level 4 data (L4) among GHRSST products provide gap-free SST at regional and global scales, based on different algorithms that merge and interpolate satellite-based SST data, which are acquired by different sensors, sometimes also including in-situ observations [4].Different fusion techniques and related configurations (e.g., interpolation grid size and bias correction) induce a significant diversity between L4 SST products [5].Understanding the consistency and discrepancy of the different SST L4 products will not only help data providers to improve their algorithms, but also represents an important step to inform users about the characteristics of the different products, helping them to select the one that may better suit their applications [6].
With the increasing variety of sea surface temperature and the development of data fusion technology, research institutions and operational departments around the world have continuously developed globally gap-free and high-quality fusion products of the sea surface temperature that merge in-situ and multi-satellite remotely sensing data.The widely used and high-quality global fusion products of the sea surface temperature internationally include OISST products from NOAA in the United States [7,8], MGDSST products from Japan [9], and OSTIA products from Met Office in the UK [10,11].OISST and MGDSST are both daily average and have a horizontal resolution of 1/4 degree at global scales using optimal interpolation.OSTIA is based on multiple infrared and microwave satellite remotely sensing data and in-situ data, with a horizontal resolution of 1/20 degree , and adopts multi-scale variational technology.
In this article, we have compared and analysed 10 sets of global gap-free fusion product of sea surface temperature in 2022, with different fusion techniques and related configurations.Session 2 of the paper is data and methods, and results are in session 3. The session 4 is the conclusion.
(2)The global gap-free fusion product of sea surface temperature in the National Marine Environment Prediction Center merged foreign satellites remotely sensing AVHRR and AMSR, domestic satellites remotely sensing H1C and H2B, and in-situ GTS data (referred to as NMEFC), which was based on optimal interpolation and covered the period from January to August 2022.
(3)The comparative analysis products compared basically covered internationally operational fusion products.The comparative SST analysis fields were shown in Table 1.The 8 sets of products used for comparison were integrated with the National Marine Environment Forecast Centre's products, totalling 9 sets, which were compared with GMPE products.
GMPE products were proposed the multi product ensemble product in GHRSST in 2012 [4].GMPE had a horizontal resolution of 0.25 degree, and different SST fusion products were interpolated into the same resolution grid.At each grid point, the ranking of different fusion products was taken as the median value, rather than the mean value.GMPE products had advantages such as smaller errors among different fusion products [6].

Methods
The validation methods include bias, root mean square error (RMSE) and correlation coefficient (CC), with the corresponding calculation formulas as follows: where   is the fusion products of SST,   is the ARGO in-situ SST or GMPE SST, and  is the number of the matchup data between them.

Comparative analysis with ARGO data
Through the analysis of bias and the RMSE distribution diagram (Figure 1) between SST fusion products and Argo in-situ data, the RMSE ranges are from 0.3233 (OSTIA) to 0.5180 (MGDSST).
The RMSE between NMEFC fusion products and Argo in-situ data is 0.3861, which is at the middle level of comparative analysis of SST products, ranking fifth out of 10 fusion products (Table 2).

Comparative analysis of different SST fusion data
(1) Distribution of the Mean Value Through the analysis of the mean value distribution map of each SST analysis product (Figure 2), it is found that each SST analysis product has the same spatial distribution, with the minimum NMEFC mean value (20.11) and the maximum MGDSST mean value (20.31).The range of mean distribution is similar to the results of [16].By analysing the daily variation of the mean values of each SST fusion product (Figure 3), it is found that the daily variation trend of each SST analysed product is consistent.The NMEFC fusion analysis field has a minimum value on May 24th, and the reason will be needed to further analysis.(2) Distribution of the Bias and RMSE By analysing the bias distribution map of each SST (Figure 4), it is found that the distribution areas with high deviation were basically similar.The deviation ranges are from -0.0507 (K10) to 0.0538 (OISST), which is similar to the results of [16].Through the analysis of the daily variation of deviation (Figure 5), the NMEFC fusion product is mainly characterized by negative deviation.The difference in daily variation of deviation between different fusion products is mainly caused by various fusion methods, resolutions, and data sources.Through the analysis of the RMSE distribution map of each SST analysis product (Figure 6), the distribution of the RMSE of each SST analysis product is basically similar.The RMSE ranges are from 0.1579 (CMC) to 0.3199 (K10), and the NMEFC fusion product has a RMSE of 0.3040, which is at the intermediate level, ranked sixth out of 9 fusion products (Table 3).Analysing the daily variation of the RMSE of each SST analysis product (Figure 7), it is found that the daily variation of the RMSE of NMEFC fusion products is at the intermediate level.
Furthermore, it reveals that these products exhibit significant errors in the western boundary current regions of the Northern Hemisphere (Figure 6).The resolution is the main reason, with low-resolution (NMEFC, DMI_OI, MGDSST and OISST) corresponding to higher RMSE and high-resolution (NOAA_GEO and OSTIA) corresponding to lower RMSE.(3) Distribution of the Correlation Coefficient Through the analysis of correlation coefficient distribution maps for each SST (Figure 8), the distribution of correlation coefficients for each SST analysis is basically similar.However, the areas of low CC, especially in the tropical Pacific and the Polar regions in the NMEFC and NAVO-K10 products, need to be further studied.

Figure 2 .Figure 3 .
Figure 2. Distribution Map of the Mean Value of each SST Fusion Data

Figure 4 .Figure 5 .
Figure 4. Distribution Map of the Bias of each SST Fusion Data

Figure 6 .Figure 7 .
Figure 6.Distribution Map of the RMSE of each SST Fusion Data

Figure 8 .
Figure 8. Distribution Map of the Correlation Coefficient of each SST Fusion Data

Table 1 .
SST Fusion Products (The citations following the abbreviation are the references)

Table 2 .
Statistical Table for the Bias and RMSE of SST Fusion Data and ARGO In-situ Data (Ranked by RMSE)

Table 3 .
Statistical Table for the Bias and RMSE of Each SST Fusion Data Compared to GMPE Analysis (Ranked by RMSE)