Tian Wei, ZOU Shuguang, LIU lizhao, ZOU Yan

1. Information Service Center of Offshore oil production plant, SINOPEC Shengli Oilfield Company,Dongying 257000, Shandong province, China；

2. Yantai Oceanic Environmental Monitoring Central Station, Rongcheng 264321, Shandong province,China

Abstract: The observational sea surface temperature(SST) data from 1960 to 2017 of the Chengshantou marine station has been adjusted by the Penalized Maximal Test (PMT)developed by the Climate Research Center of the Environment Ministry of Canada, based on the metadata archive. In this study, the homogenous surface air temperature (SAT)data from neighboring meteorological observation stations are used to construct the reference series by correlation coefficient weighted averaged method. The climate change characteristics of the Chengshantou SST were analyzed using the homogenized data. Results show that the annual average SST trend has changed significantly before and after the homogenization. The warming trend increased from 0.04 °C/10 a before revision to 0.15 °C/10 a. The warmest five years occurred mostly after 1980, that is,1973, 1989, 2002, 2007 and 2017. SST generally showed a significant upward trend and significant inter-decadal fluctuations. From the 1960s to the end of the 1980s, it was a colder stage, and then began to warm up. It was a warmer period from the 1990s to the present. From 1960 to 2017, the jump point occurred in 1987, which is very consistent with the temperature characteristics of China's mainland.

Keywords: seawater temperature, climatic variability, changing trend, climatic tendency

1 Introduction

Temperature is an important physical element reflecting the thermal state of sea water. Heat flux(including solar radiation, long wave radiation, sensible heat flux and latent heat flux) is one of the main factors affecting sea temperature(Li, 2010; CAI, 2015).Ocean currents(such as coastal upwelling, Taiwan Warm Current, Yellow Sea Warm Current) also have certain influence on the local sea temperature. In the open ocean, the distribution of the surface water isotherm is roughly parallel to the latitude. In the nearshore area, due to the influence of current, the isotherm moves to the north-south direction. The vertical distribution of sea water temperature generally decreases with the increase of depth and shows seasonal variation(Jin, 2011). Sea water temperature is often used as a basic index to study the properties and movement of water masses. The temporal and spatial distribution and variation of sea water temperature is an important part of oceanography, which is of great significance to marine fishing, aquaculture, and maritime operations. It is also particularly important to meteorology, navigation and underwater acoustics. Due to its unique regional factors, the sea water temperature in the China coastal waters is warming faster than that in the open ocean. At the same time,further studies show that the increase of sea water temperature will also cause the changes of marine environment and marine ecosystem, such as the change of marine productivity, the frequent occurrence of red tide, and the northward migration of biological species. However, currently, there are few studies on the characteristics of the local small-scale sea water temperature climate change. Therefore, this paper mainly analyzed and discussed the characteristics of sea surface temperature(SST) change monitored by Chengshantou marine environment monitoring station, reveals the scientific facts of local climate change, so as to support marine climate research and aquaculture in this area,and the service demand of scientific response to marine climate change.

2 The study area

Chengshantou, also known as "the end of the sky"in Chinese language, is located in Chengshan Town, Rongcheng, Weihai City, Shandong Province. Chengshantou is three sides surrounded by the sea . It is only 5 nautical miles away from the North-South international main channel and 94 nautical miles away from South Korea. Chengshantou is 200 m above sea level, 1.5 km wide from east to west, 2 km long from north to south,and covers an area of 2.5 square km. Chengshantou goes straight into the Yellow Sea,with the cliffy mountain. This area is controlled by temperate marine monsoon climate,which is always impacted by strong winds, huge waves and storm surges. The maximum wave height in Chengshantou is more than 7 m which is less affected by human activities.It is a valuable scientific research base for marine meteorology, physical oceanography and marine energy researches in China.

3 Data and method

The SST observations from coastal hydrological station Chengshantou employed in this study are collected and processed by Chengshantou marine environment monitoring station, Ministry of Natural Resources, China. Avariety of quality control(QC) procedures were conducted on these data, including checks for physically unreasonable values,temporal and spatial consistency, climatological limits, extreme values, etc., according to the“Coastal Observation Specification”(GB/T14916). In addition, the PMFT method (Wang et al., 2007, 2008) of RHtestV4 software package developed by the Climate Research Center of Environment in Canada was used to detect and adjust the inhomogenities in the SST series, together with the metadata of this station. Homogenized monthly mean SST series were obtained by adjusting all significant change points which were caused by the non-climate effects, such as instrument change, station relocation, and environment change(Li Yan, et al., 2018). The SST series covers the period from January 1960 to December 2017.

TPR and Bayesian methods a re the best choice for homogenization of long-term climate series. Wang et al in Canada put the penalty factor into the TPR methods, and developed PMF and PMT methods. These two methods empirically consider the lag first-order autocorrelation of time series, and embed multiple linear regression algorithm,which can be used to test and correct multiple discontinuities of data series containing first-order autoregressive error, so that the false alarm rate and test ability are poor at both ends of the sequence The problem has been improved. The reference sequence needs to be established in the process of PMT method test, and the difference between the test sequence and the reference sequence is the object to be tested. The PMT method was introduced as follows (Wang et al., 2007, 2008; Zhang, et al., 2012).

The PMT method is based on the penalty maximum T-test. Firstly, it is assumed that the sequence {Xt}( t = 1，…，N) is normal distribution, that is, the original hypothesis is satisfied {Xt}～IIDλ（μ，σ2）. Hypothesis:

where μ1≠μ2, t = k is called a discontinuity if it is assumed to be true. {Xt}～IIDλ（μ1，σ2）represents the Gaussian distribution, the mean value is and the variance is σ2. This method is called Maximum Likelihood Estimation. The criterion of maximum likelihood estimation in logarithmic form can be transformed into the maximum value of the following formula:

Tmax= max T（k）（1≤t≤N - 1）can be calculated. On this basis, in 2006, Wang Xiaolan constructed a statistical function PT as a significant criterion:

where P（k）is the empirical function obtain ed through a series of tests. The practice shows that this method is much better than SNHT test method, and the use of homogenized reference sequence can effectively improve the judgment of real nonclimatic break points.

Because of the integrity and ease operation (Wang and Feng, 2013), RHtest V4 and its previous versions developed have been widely used in homogenizing climate data（Li et al., 2016, 2018; Xu et al., 2013）. The PMT algorithm needs to be used with good reference series to diminish the trend and periodic components that may exist in the data series. Homogeneous SAT series from neighboring meteorological observing stations are used to construct a reference series for the SST series. In conjunction with detailed metadata of Chengshantou station, these statistical change points from the PMT test are further validated. In general, only one type of change points is retained for adjustment,which were supported by metadata and were named as “documented change points”(caused by instrument change, environmental change, station relocation, etc.). In this case, the documented time of change is used to replace the PMT-estimated time of change for the adjustment. The change points which cannot be identified in the metadata are kept as they are, without adjustment. Then, the quantile-matching (QM) adjustment method which was also combined in the RHtest V4 is used to make adjustments after the change points were found (Wang and Feng, 2013).

Here, the reference meteorological observing stations are selected by the following criteria: the distance between the reference meteorological observing station and the candidate coastal hydrological station should be nearby. The correlation coefficient between the original SST series and the homogeneous SAT series should be equal or higher than 0.70.

The homogeneous monthly SAT series which are well correlated with the candidate SST series are used to construct the reference series by correlation coefficient weighted average method:

where i denotes the time, j represents the number of the reference meteorological observing station, ρ is the corr elation coefficient between yearly SST series and SAT series, x is the monthly mean SAT series, and y is the final monthly reference SAT series for the candidate SST series. The correlation coefficients between the SST series from the Chengshantou station and the SAT series of the neighboring meteorological stations(ID: 54751, 54774, 54776) are 0.70, 0.74, 0.72, all exceeding the 99% confidence level).

The division of seasons adopts the conventional division standard: March to May is spring, June to August is summer, September to November is autumn, and December to February of the next year is winter (Cheng, 2017). Linear regression method (Guo, 2013)was used to study the inter-annual, monthly and seasonal variation of SST in Chengshantou. In this paper, the least square method was used to estimate the variation trend of SST series, and the significance level of linear trend is determined by t-test. The sequence of cumulative anomaly statistics is used to judge the climate break point, that is, when the maximum value (or minimum value) appears in the curve, the continuous change of cumulative anomaly in the two periods before and after the corresponding year will change from the continuous positive (or negative) trend to the negative (or positive)trend, which may be the break point year (Wei, 2007).

4 Analysis ofsea surface temperature variation characteristics in Chengshantou

4.1 Inter-annual variation characteristics of annual mean sea surface temperature

The purp ose of homogenization and adjustment of the observational data is to remove the artificial non-climatic breakpoints in the series, and obtain a homogenized and reliable climate series. This work is the most basic and key work in the study of climate change. Fig. 1 shows the annual sea surface temperature series before and after the homogenization at the Chengshantou marine station. It can be seen that the annual sea surface temperature before 1980 was significantly higher(black dotted line in Fig. 1),which made the SST warming rate much lower, about 0.04 °C/10 a. The homogenized SST series corrected the high SST caused by early artificial observation and instrument changes, and the annual sea surface temperature increased significantly, with the rate of 0.15 °C/10 a. For this tendency, we calculate the correlation coefficient between time and variable, that is, the correla tion coefficient between time t and variable is calculated as follows:

For certain significance levels α = 0.05, α = 0.01, the critical value of correlation coefficient is obtained and searched by the critical value table of correlation coefficient.Before the adjustment, the warming trend did not pass the significance level test,however, after the adjustment, the significance level increased significantly, r = 0.43,exceeding the significance level of 0.01. This result indicated that the significant warming trend in recent 58 years.

The statistical characteristics of the homogenized annual mean SST show that the climatic annual mean temperature in Chengshantou is 11.5 °C and the standard deviation is 0.6 °C. The annual mean SST values are mostly concentrated between 11.0 °C and 12.0 °C, accounting for 62.1% of the total. The highest annual mean SST was 12.8 °C in 1973, and the lowest was 9.6 °C in 1969. The difference between the highest and lowest annual mean SST was 3.2 °C(Fig. 2). The five coldest years in Chengshantou occurred before 1980, namely, 1977, 1963, 1968, 1976 and 1969. The five warmest years occurred after 1980, namely, 1973, 1989, 2002, 2007 and 2017 (Fig. 1).

Fig. 1 Annual mean SST series before (black dotted line) and after homogeinziation (red solid line) at Chengshantou marine station (unit: °C)

4.2 Analysis of abrupt breakpoint characteristics of annual SST

Fig. 3 shows the inter-decadal variation of annual SST of Chengshantou from 1960 to 2017. It is found that the SST in Chengshantou has a significant upward trend and obvious inter-decadal fluctuation. From 1960s to the end of 1980s, it was cold stage, and then it began to warm up. Since 1990s, it has been in warm stage. In the past 58 years,the coldest decade occurred in the 1960s, and the warmest decade occurred in the first decade of the 21st century.

Fig. 2 Climatological statistical characteristics of homogenized annual mean SST at the Chengshantou marine station (unit: °C)

Fig. 3 Low frequency filtering curve of homogenized annual mean SST from 1960 to 2017 at Chengshantou marine observation station (That is to remove the decadal fluctuation of time scale changes below 10 years)

The periodic characteristics of SST in Chengshantou are much consistented with those in China Mainland(China Climate Change Blue Paper, 2018). In our study, the cumulative anomaly values of annual SST after homogenization from 1960 to 2017 are calculated,and the cumulative anomaly curve is shown in Fig. 4. Although the mean cumulative distance of SST is negative, the change pattern of the curve clearly shows that the SST in Chengshantou has experienced a significant fluctuation in the past 58 years. From the 1960s to the end of 1980s, SST showed a cooling trend. It began to increase in the 1990s, and the warming trend has not stopped. Its break point appeared in 1987.Previous study showed that the abrupt break point of temperature warming in China was in 1987(You, et al., 1998). It is clear that the break point of the SST in Chengshantou has consistent in the break point of temperature in China Mainland.

Fig. 4 Cumulative anomaly curve of homogenized annual mean SST at Chengshantou marine station from 1960 to 2017

5 Conclusion

(a) The annual mean SST of Chengshantou marine station from 1960 to 2017 was checked and homogenized. The results show that the SST data of this station has the problem of inhomogenization. The early artificial observation and instrument change make the SST higher to a certain extent.

(b) The trend of annual mean SST at Chengshantou station before and after homogenization changed significantly, showing that the warming tendency is much stronger than that before, from 0.04 °C per decade up to 0.15 °C per decade. The five warmest years occurred after 1980, that is, 1973, 1989, 2002, 2007 and 2017.

(c) The SST in Chengshantou has a significant upward trend and obvious inter-decadal fluctuation. From 1960s to the end of 1980s was a cold stage, and then it began to warm up. Since 1990s, it was in a warm stage. During 1960-2017, the SST change point occurred in 1987, which is consistent with the change point and the phased characteristics of air temperature in Chinese Mainland.