Potential morphological responses of an artificial beach to a flood in extreme events: field observation and numerical modelling

1. Introduction

Global warming places shallow water ecosystems at more risk than those in the open ocean as their temperatures may change more rapidly and dramatically (Belkin, 2009; Stramska and Białogrodzka, 2015). In the 21st century, offshore areas of the northern Pacific Ocean are expected to be some of the prominent and vulnerable climate change “hot spots” (Lima and Wethey, 2012; Frölicher and Laufkötter, 2018). As a part of the Northwest Pacific Ocean, the seas adjacent of Chinese mainland including the Bohai Sea, the Yellow Sea, the East China Sea and the South China Sea (Fig. 1) are also sensitive to the effects of natural and anthropogenic climate changes (Halpern et al., 2008; Cai et al., 2016). It is therefore necessary to give an accurate assessment of the trend of water temperature.

Figure  1.  Locations of the 26 coastal hydrological stations at the Chinese coast (black dots). Red triangles represent two gauge stations (Lijin and Datong) used in the study. The China coasts are divided into the coasts of Chinese mainland in the Bohai Sea, Yellow Sea, East China Sea and South China Sea. The full names of the stations are shown in Table 1.

DownLoad: Full-Size Img PowerPoint

The National Marine Observation System along the coast of China has been set up and gradually improved since 1960. Among them, there are 26 hydrological observing stations with long-term (exceeding 50 years) and continuous in situ sea surface temperature (SST) observations. Yet, studies have found that these long-term observations contain several spurious changes and artificial change points suffering from station relocation, instrument change, transformation of observing system and so on, which are prone to systematic errors or biases (Li et al., 2009; Xu et al., 2013; Hausfather et al., 2016; Minola et al., 2016; Yang et al., 2018). SST records from coastal hydrological stations of China are no exception (Li et al., 2016). Therefore, it is necessary to homogenize the historical observational data in order to monitor, detect and assess the long-term change in climate and environment in the Chinese coastal waters.

In the study, monthly mean SST series from 26 coastal hydrological stations for 1960–2015 were homogenized, using available metadata information and statistical homogeneity method. Then, the homogenized SST series were used to assess the oceanic climate change in the coastal waters at annual and seasonal scales. Results should potentially enable us to more accurately detect the magnitude of SST warming rate in this region. SST variability is governed by both atmospheric and oceanic processes (Deser et al., 2010). This work further studied the possible influence of large-scale atmospheric circulations on the observed SST variability across China.

The structure of the paper is arranged as follows. Data and metadata from coastal hydrological stations are described in Section 2. A brief introduction to methods of detecting and inhomogeneity adjustment is given in Section 3. Section 4 reports the statistics of detected change points and related causes, the inhomogeneity impacts on the long-term trends, and the trends of seasonal mean SST and possible influence of large-scale atmospheric circulation on the winter SST trends. At the end, discussion and conclusions are given in Section 5.

2. Data

2.1 Sea surface temperature, surface air temperature and metadata

Currently, more than 100 hydrological stations are being operated and run. Among these stations, only 26 stations have taking routine and continuous measurements since 1960 (Table 1), with the percentage of missing data lower than 4%. These observed data series have undergone the least relocation (no more than 5 relocations reported by the available metadata). All of these data are provided by the National Marine Data and Information Service of China and can be obtained in the National Science & Technology Resource Sharing Service Platform of China (http://mds.nmdis.org.cn). The locations of these stations are shown in Fig. 1. There are few stations at the southern Yellow Sea, because the coast area is a vast beach and salinized land which is not suitable for hydrological stations. Since the 2000s, there have been only some automatic stations. These measurements have been quality controlled, including the checks in climatological limit, gradient, the temporal and spatial consistency and so on, according to the codes from The Specification for Offshore Observations (People’s Republic of China National Bureau of Technical Supervision, 2006) (http://www.std.gov.cn/gb/gbQuery). Monthly mean SST was calculated from the average of each daily mean SST in a month. Following the method of Wan et al. (2010) and Azorin-Molina et al. (2014), monthly means have been calculated for months with at least 26 days of observations. Otherwise the whole month has been set as missing. Then the data in some missing months are reconstructed by the neighbor data using linear regression analysis method.

Table  1.  The 26 coastal hydrological stations used in the study

Station	Abbreviation		Station	Abbreviation		Station	Abbreviation
Zhimaowan	ZMW		Lianyungang	LYG		Yunwo	YWO
Qinhuangdao	QHD		Shipu	SPU		Zhelang	ZLG
Xiaochangshan	XCS		Dachen	DCN		Zapo	ZPO
Laohutan	LHT		Kaimen	KMN		Beihai	BHI
Longkou	LKO		Sansha	SSA		Weizhou	WZU
Yantai	YTI		Pintan	PTN		Naozhou	NZU
Shidao	SDO		Congwu	CWU		Haikou	HKO
Qianliyan	QLY		Xiamen	XMN		Qinlan	QLN
Xiaomaidao	XMD		Dongshan	DSN

 | Show Table

DownLoad: CSV

Metadata information of each coastal hydrological station was used to verify the statistically detected change points, including the station coordinates, height, instrument change, station relocation, environment change, algorithms for calculating a daily mean, observing procedures, observation system change, etc. All of the metadata information was documented in standard files, called “Hydrological Station History Data Files”. The homogeneous monthly mean surface air temperature (SAT) series from National Meteorological Information Center of China Meteorological Administration were used to construct reference series which have significant relationship with SST series (Xu et al., 2013).

2.2 OISST

A globally gridded SST dataset with high resolution is used in this work for comparative analysis of SST series for the region in the present study. The 0.25°×0.25° NOAA Optimum Interpolation SST (OISST) version 2 using Advanced Very High-Resolution Radiometer infrared satellite SST data from the Pathfinder satellite combined with buoy data, ship data, and sea ice data, covering from 1982 to 2015 (Reynolds et al., 2007), are available at http://www.ncdc.noaa.gov/oisst.

2.3 Three atmospheric circulation indices

Three atmospheric circulation indices, (1) the Arctic Oscillation (AO) index, (2) the East Asian Trough (EAT) index, and (3) the North Pacific Oscillation (NPO) index, were used to analyze the possible influence of large-scale atmospheric circulation on the winter SST variability along the Chinese coast. The AO, EAT and NPO are considered the main modes for winter climate variability across the Chinese mainland (Hansen et al., 2010; You et al., 2013; Chen et al., 2013; Sun et al., 2018) and the SST variability over the China seas (Zhang et al., 2010; Yeh and Kim, 2010; Huang et al., 2016; Li et al., 2015; Pei et al., 2017; Cai et al., 2017); therefore, their possible influence on near-shore SST trends has been investigated in this study.

AO index is defined as the first leading mode from the empirical orthogonal function (EOF) analysis of monthly mean height anomalies at 1 000 hPa (Northern Hemisphere) (Thompson and Wallace, 1998). The AO index series has been obtained from the National Oceanic and Atmospheric Administration (NOAA) (available online at https://www.esrl.noaa.gov/psd/data/correlation/ao.data). The definition of EAT index adopted here is the one defined by Sun and Li (1997); that is, the EAT index is the normalized 500 hPa geopotential heights averaged over the area (25°–45°N, 110°–145°E). NPO index is defined as the principal component of the second mode of analysis for the standardized winter sea level pressure (SLP) over the North Pacific (20°–80°N, 120°E–120°W) (Linkin and Nigam, 2008). Monthly mean 500 hPa geopotential heights and SLP used to calculate EAT index and NPO index in the study is obtained from the National Centers for Environmental Prediction/National Center for Atmospheric Research reanalysis (downloaded from http://www.cdc.noaa.gov) (Kalnay et al., 1996).

3. Homogenization method and statistical analysis methods

3.1 Homogenization method

It is a common practice to use a statistical method to detect break points in long-term series, and rely on metadata for break point determination. Several homogenization algorithms have been identified and assessed in recent years (Reeves et al., 2007; Ribeiro et al., 2015). Among them, the PMTred algorithm is based on the penalized maximal t (PMT) test for searching for the most probable position of shift in a segment of the series being tested and tests its statistical significance using a penalized maximal t test statistic, in which the lag-1 autocorrelation is also accounted for (Wang, 2008a). PMTred methods have three advantages comparing to other techniques: (1) PMTred accounts for the effects of unequal lengths of the two segments before and after a shift and thereby has higher detection power (Wang et al., 2007); (2) PMTred has a re-cursive testing procedure for detecting multiple change points in a single time series, and (3) PMTred accounts for lag-1 autocorrelation which are often non-negligible for climate data time series and thereby reduces the number of false alarms (Wang, 2008b). The PMTred algorithm is implemented using the RHtest V4 software package (Wang and Feng, 2013). This package includes plots of the relevant time series and the resulting estimate of shifts and trends in the candidate series, which are convenient for its user. RHtest V4 and its previous versions developed have been widely used in homogenizing climate data (Wan et al., 2010; Kuglitsch et al., 2012; Vincent et al., 2012; Xu et al., 2013). Thus, PMT test based algorithm is also applied to detect non-climatic change points in monthly mean SST time series. Because the time series being tested is assumed to have no temporal trends, the PMTred algorithm needs to use a reference series to represent trends and low frequency variations in the candidate series. There are two steps to detect these non-climatic change points.

3.1.1   Construction of reference series

Generally, the most common way to construct reference series is to use a data series from nearby stations for a candidate series (a series to be tested and homogenized). However, this is not appropriate for SST series, because the coastal hydrological stations along the coast of China are sparse and not evenly. Besides, all of the stations have experienced a transformation from manual observation system to automatic observation system, during the early 2000s. These facts increase the difficulty of establishing reference series based on SST series from nearby hydrological stations. Thus, homogeneous SAT series was used to construct a reference series for each SST series, because SAT series from neighboring meteorological stations well represent the same climatic variations as the SST in the coastal areas of China (Sun, 2006). The proposed method above has been used in previous studies. For instance, Stephenson et al. (2008) systematically detected the inhomogeneities in the SAT series of Caribbean and adjacent Caribbean using neighboring SST series. The procedure of reference series construction in this article is referred. These homogenized SAT series from meteorological stations have been selected because they met all of the following selection criteria in Table 2. Then, the homogenous monthly SAT series which are well correlated with the SST series are used to construct reference series by correlation coefficient weighted average method.

Table  2.  Criteria for selecting the reference meteorological stations

Number	Criteria
1	The horizontal distance between candidate coastal hydrological station and reference meteorological stations should be within 100 km, in term of the geographical distribution
2	Correlation coefficient between the annual mean original SST series and the annual mean homogeneous SAT series should be equal to or higher than 0.7, following with the previous research (Malcher and Schönwiese, 1987). And only in this case, the reference series can represent at least 50% the variability in the candidate series (Malcher and Schönwiese, 1987; Stephenson et al., 2008)
3	In order to reduce the urban heat island effect on SAT series, the selected meteorological observing stations are located at rural, towns or middle-size cities, rather than big cities according to Ren et al. (2010)

 | Show Table

DownLoad: CSV

3.1.2   Verification of the change points and the adjustmen

Homogeneities of the 26 monthly SST series have been tested by applying PMT test. PMT test detects all the possible inhomogeneities at a statistically significance level of 5%—these detected break points which are confirmed by metadata have been homogenized. These detected break points are considered to be a real one only when metadata indicate a documented change within 1 year before or after the date of change point detected by PMT and identified to be a documented break point. The documented time of change is used to replace the estimated time of change when they are not identical (due to estimation error). The break points which cannot be validated by the metadata are kept as they are.

The objective of the quantil-matching (QM) adjustment is to adjust the series to that the empirical distributions of all segments of the de-trended base series match each other, the adjustment value depends on the empirical frequency of the datum to be adjusted. As a result, the shape of the distribution is often adjusted, although the tests are meant to detect mean-shifts. Importantly, the annual cycle, lag-1 autocorrelation, and linear trend of the base series were estimated in tandem while accounting for all identified shifts (Wang, 2008a); and the trend component estimated for the base series is preserved in the QM adjustment algorithm (Wang et al., 2010).

3.2 Statistical analysis method

The linear trends in the raw and homogenized SST during 1960–2015 are estimated via the least-squares linear fitting method. The significance of each trend also has been test. Here the Mann-Kendall test for trends and Sen’s slope estimates are used to detect and quantify trends (Sen, 1968), with magnitudes of trends and slopes assessed at the 0.05 significance level (p<0.05).

4. Results

4.1 Detected break points and their causes

Figure 2 shows an example of applying PMT test to the monthly mean SST series of BHI hydrological station, which is located at the coast of the South China Sea. March 2005 was identified as a highly significant break point (PT_max=6.92>3.71, the upper bound of the corresponding 95-th percentile of PT_max), which is easily visible in Fig. 2a. According to the historic metadata information of BHI station, the observing practice was transformed from manual observation system to automatic observation system and the instrument was changed from SWY1-1 Thermometer to YZY4-1 Thermohaline Sensor on May 18, 2005. The estimated time (March 2005) of breakpoint in the monthly mean series is 2 months earlier than the actual time documented in the metadata. Then, the estimated time was replaced by the actual time to calculate the final adjustment for this breakpoint. After change points are identified, homogenized monthly mean SST series can be obtained by adjusting the significant shifts caused by documented break points. The QM adjustment method which also combined in RHtest V4 was used to adjust these shifts in the 26 SST series. The SST series of BHI hydrological station was adjusted to the most recent segment by adding −0.35 to all data before May 2005. Here the value −0.35 was calculated from QM method. The homogenized monthly mean series is shown in Fig. 2b.

Figure  2.  Monthly mean de-seasonalized SST series with its mean annual cycle subtracted BHI coastal hydrological station (black line; Red line represents the estimated mean response along with the estimated mean shift) (a), and monthly mean homogenized SST series of BHI coastal hydrological station (b).

DownLoad: Full-Size Img PowerPoint

To summarize, a total of 68 break points have been detected at the 26 monthly SST series for 1960–2015 by PMT method at the significance level of 5%. The statistical break points were identified by metadata. Results show that 62 statistical break points coincide with dates of metadata within ±1 year and are referred to as documented break points. Only 10% statistical break points cannot be identified by available metadata. Since 2001, the coastal hydrological stations have experienced a transformation from artificial observing system to automatic observing system, and 15 change points are detected in the year of 2002 (Fig. 3a). Instrument change (including the observing times change caused by the observing system change) is the main cause of break points in monthly SST series. It is responsible for about 61.3% of the total documented break points. Station relocation is found to be the other main cause for discontinuities, accounting for 24.2%. Environmental change and human errors or instrumental malfunctions are the other causes, accounting for 6.4% and 8.1%, respectively (Fig. 3b). Clearly, one or more documented break points are common in the SST series of each station. Series with more than 3 break points are mainly at the coast of the Bohai Sea and the southern East China Sea (Fig. 3b).

Figure  3.  The total number of identified break points in the 26 SST series of each year from 1960–2015 (a) and the identified break points in SST series of each station (b). See Table 1 and Fig. 1 for the full names and locations of the 26 coastal hydrological stations.

DownLoad: Full-Size Img PowerPoint

Whereas instrument change, station relocation and other changes at observing sites can cause either a rise or a drop in the observed temperatures, the magnitude of all artificial changes is not necessarily symmetrical about zero. For each series, the series to the most recent segment was adjusted by adding ∆SST to the data before the documented break points in each series. Here, ∆SST is the shift size estimated by QM method. Statistical results show that the ∆SST values range from −1.2°C to 1.0°C. The larger portion is the negative adjustments, about 62%. Positive adjustments are responsible for the remaining 38%.

4.2 Inhomogeneity impacts on estimated annual mean SST trend

In this subsection, XCS hydrological station, which belongs to the Bohai Sea, was taken as an example to show the influence of break points on estimation of annual mean trend. According to metadata, XCS hydrological station has changed the type of thermometers for three times since 1960. The first change was on January 1, 1965. Only the instrument was changed. The second change was on January 1, 2002. The observing practice was transformed from artificial observation system to automatic observation system. In the artificial observing period, bucket was used to sample and SST was recorded three times at 08:00, 14:00 and 20:00 UTC by SWY1-1 Thermometer. In the automatic observing period, the observation instrument YZY4-1 Thermohaline Sensor was installed in the thermohaline recorder wells and fixed at the height of 0.5 m below sea level. The SST was recorded every hour. The last change was on July 1, 2004 and the only change is from YZY4-1 Thermohaline Sensor to SWL1-1 Thermometer. The three change points at January 1965, January 2002, and July 2004 introduced three SST shifts of about −1.2°C, −0.5°C, and 0.36°C in the monthly mean series, respectively. Compared to the warming rate of the homogenized SST series for 1960–2015, the raw SST series underestimated the warming rate (0.27°C per decade vs. 0.18°C per decade). This example illustrates that inhomogeneities caused by artificial reasons can weaken the long-term trends in this region. The result is similar to the conclusions in previous studies of SAT series in China (Cao et al., 2013; Xu et al., 2013; Zhang et al., 2014), despite the causes for the impact of inhomogeneities of the SAT and SST data may be different (Zhang et al., 2014).

For each of the SST measuring sites, there is at least one meteorological station within 100 km distance, according to Table 2. Such a meteorological station which SAT series had the highest correlation with the SST series of hydrological station was selected, and form 26 pairs of SST/SAT data. Figure 4 displays the regionally averaged raw SST, homogenized SST and the homogenized SAT series (Xu et al., 2013) from 1960 to 2015. The linear trends for 1960–2015 with 95% confidence interval of raw SST series was (0.09±0.06)°C per decade, while the linear trend of the homogenized SST are increased to (0.19±0.06)°C per decade. The daily and monthly mean SST in the automatic observing period was cooler than that in the artificial observing period (Li et al., 2020). Also, in the study, the correction values of manual transfer were mainly negative. For SST series of each station, the series to the most recent segment was adjusted by adding ∆SST to the data before the documented change points. Finally, the homogenized monthly SST series was obtained. And the homogenized SST largely corrects the warm biases caused by earlier artificial observing. Consequently, the trend from adjusted data is much higher than that of the raw data and the original SST observation is underestimated. Thus, as shown in Fig. 4, the main differences between the raw and homogenized SST series appeared before 2000, indicating the break points caused by instrument changes may lead to biases in the estimated trend and probably underestimate the warming rate at the Chinese coast waters. Figure 4c also includes the SAT series during 1960–2015 for comparison. It is obvious that the homogenized SST series and homogenized SAT show much similar inter-annual variability as well as similar long-term trends ((0.22±0.06)°C per decade for the SAT series).

Figure  4.  Annual mean raw SST anomalies (a), homogenized SST anomalies (b) and homogenized SAT anomalies (c) averaged over the Chinese coast (1981–2010 base period). Gaussian low-pass filter is shown by the solid black line.

DownLoad: Full-Size Img PowerPoint

The trends of the raw and homogenized annual mean SST series of stations along the Chinese coast are shown in Fig. 5. Obviously, both the raw and homogenized SST series of the 26 stations show warming trends; but the warming rates are much stronger in the homogenized data than those in the raw data in general. Results show that the warming was not spatially uniform along the Chinese coast. There were two rapid warming areas, that is, the northern East China Sea and the other area along the Bohai Sea, exceeding 0.20°C per decade. However, the spatial characteristic of warming cannot be found in the raw SST series (Fig. 5). About 14 SST series show significant increasing trends, exceeding 0.18°C per decade and most of these stations located in the north of the PTN station. The large-scale contrast of warming between north and south may have been primarily caused by the weakening East Asian Winter Monsoon (Committee on China’s National Assessment Report on Climate Change, 2007; Ding, 2008), which generally exerted a larger influence in northern part of the coastal zone, and also partially caused by regional natural and human effects (Xie et al., 2002; Zhang et al., 2010; Park et al., 2015; Kako et al., 2016). However, further investigation is needed to better understand the spatial pattern of the SST increase.

Figure  5.  Linear trends of annual mean original SST series (a) and homogenized SST series (b) during 1960–2015 at stations from north to south. The legend represents the temperature variation per decade, and the unit is °C. See Table 1 and Fig. 1 for the full names and locations of the 26 coastal hydrological stations.

DownLoad: Full-Size Img PowerPoint

In order to test the validity of the homogenized SST series, a SST dataset which has been widely used in the regional and global climate change studies is employed here for comparison. The SST dataset with 0.25° by 0.25° spatial resolution, called OISST which spans from 1982 to present. Firstly, OISST is interpolated to the stations’ locations. Figure 6 includes the annual mean SST series in the coast of the China seas based on OISST during 1982–2015 for comparison with raw SST and homogenized SST respectively. It is notable that the annual mean homogeinzed SST series shows inter-annual variations very similar to that in the OISST (Fig. 6). The correlation coefficient calculated from Homogenized SST OISST is 0.87. However, there are discrepancies in the amplitudes between Raw SST and OISST (Fig. 6). The correlation coefficient calculated from Raw SST and OISST is 0.71 respectively which is lower than that from Homogenized SST. The result indicates that homogenized SST has a better consistency with the global gridded SST with high resolution than the raw SST in the coast area of China.

Figure  6.  Time series of the annual mean SST anomalies in the coast of the China seas from raw SST, homogenized SST, and OISST from 1982–2015 relative to the climate period of 1982–2010.

DownLoad: Full-Size Img PowerPoint

4.3 Seasonal mean SST trends and atmospheric circulation variability

The homogenized SST series is more reliable than the raw SST series and thus could be used to assess the seasonal mean SST variabilities in this section. Previous studies of marginal sea of China found that the most significant warming rate occurred in winter and the slowest warming trend occurred in summer (Yeh and Kim, 2010; Bao and Ren, 2014). Thus, seasonal mean homogenized SST series for winter (December, January, and February) and summer (June, July, and August) were calculated. Trends of the two seasons were examined for individual station shown in Fig. 1. Clearly, most stations show robust SST warming trends in winter and weak warming trends in summer (Fig. 7). There was an asymmetric seasonal temperature trends in the Chinese coastal water in the past decades. More interestingly, the winter warming rates of each SST series increased with increasing latitude. These indicate that the warming rates were stronger in winter than in summer and stronger in the relative cool area than in warm area during 1960–2015, which were consistent with previous reports of SAT changes in Chinese mainland (Chen et al., 2013; Li et al., 2015; Sun et al., 2018).

Figure  7.  Linear trends of seasonal mean SST in winter (a) and summer (b) at stations from north to south along the coast of China. The legend represents the temperature variation per decade, and the unit is °C. See Table 1 and Fig. 1 for the full names and locations of the 26 coastal hydrological stations.

DownLoad: Full-Size Img PowerPoint

The general warming may have been mainly caused by anthropogenic increase in atmospheric CO₂ concentration and partly by the urbanization effect around the stations (Zhang et al., 2010). Considering the notable asymmetric seasonal trends and non-uniform spatial warming rate, however, the possible relationship between the largest warming rate occurring in winter and three large-scale atmospheric circulation modes were further analyzed. The AO index is often linked to winter climate variability in China and also correlated with the strength of the East Asian Winter Monsoon and Siberian high pressure system (Chen et al., 2013; You et al., 2013; Sun et al., 2018). It is notable that the winter AO index increases since the 1960s, accompanied by the weakening East Asian Winter Monsoon in general (Niu et al., 2010; He, 2013; Gong et al., 2018). The China seas are also generally influenced by East Asian Monsoon System and AO (Cai et al., 2017; Pei et al., 2017). Moreover, Yeh and Kim (2010) associated the increasing SST trends in the East China seas in winter with NPO-like sea level pressure change. They regarded the warming as driven by NPO mode through weakened northerly winds.

Figures 8a, c and e display the Pearson’s correlation coefficients between the winter SST series in each station from north to south of the China seas and the atmospheric circulation indices. The winter EAT index may have exerted a major influence, as indicated by the high positive significant (p<0.01) correlations ($ {r}_{1} $) of 0.49–0.70 for all series. Yet, the winter AO index is only significantly and positively correlated with winter SST series at the coast of the Bohai Sea and Yellow Sea. Similarly, the winter NPO index is significantly and positively correlated with winter SST series at coast of the Yellow Sea and the East China Sea. Figure 8 also shows the annual and decadal mean circulation indices of EAT (Fig. 8b), AO (Fig. 8d) and NPO (Fig. 8f) from 1960 to 2015. It is notable that the winter EAT index has been consecutively increasing since the 1960s and transferred from negative phase (strong EAT) to positive phase (weak EAT) at the 1980s. Meanwhile, winter AO index has increased since the 1960s generally. NPO index also increased since 1960s, but the inter-annual variation amplitude is weaker than that of the other two indices.

Figure  8.  Correlations between winter SST series at stations from north to south along the coast of China and the three atmospheric circulation indices: EAT index (a), AO index (c) and NPO index (e) (0.26 is the significant level at the 0.05 confidence level); and time series of winter EAT index (b), AO index (d) and NPO index (f) (black line). The smoother red line is the 9-year smoothing average for 1960–2015. See Table 1 and Fig. 1 for the full names and locations of the 26 coastal hydrological stations.

DownLoad: Full-Size Img PowerPoint

It is clear that all of the three indices have an increasing trend since 1960s in general. The weaker EAT and stronger AO reduce the cold air of the polar region from entering into the China seas, especially north of the China seas. Moreover, NPO mode in higher phase weakened northerly winds. The circulation modes benefited the less frequent bursts of the winter cold waves and the sharp increase of winter SST in northern China seas during 1960–2015. Winter SST series at the northern China seas are possibly under the synergistic effects of EAT, AO and NPO from the middle-high latitude. This may well explain the more rapid and significant warming in winter at the northern China seas. In contrast, these atmospheric circulation modes exerted a relatively weak influence in winter SST series at the southern China seas, especially the AO.

4.4 Some local effects

It is also possible that the SST observations themselves may have been affected by local tidal mixing, sea currents, fresh water discharge, urbanization and other kind of land use and land cover changes along the coastal zone, leading to locally heterogonous warming pattern. The more rapid warming along the Bohai Sea and the northern East China Sea as found in the Fig. 5b may have been related to the local factors in certain extents. The estuaries of the Changjiang River and the Huanghe River are located at the northern East China Sea and the Bohai Sea, respectively. Recently updated observations show that, from 1960 to 2015, the Changjiang River stably discharges fresh water of about 882 Gt each year (1960–2015 mean) to the East China Sea (Fig. 9a). Though, the Changjiang River discharge has shown no increasing trend, the stream water temperature in the Changjiang River Estuary increased by about 2.0°C since 1986 (Zhou et al., 2005), contributing to the extremely rapid warming observed in this sub-region.

Figure  9.  Time series of the measured annual water discharge at Datong station of Changjiang River (31.8°N,120.9°E) (a), and Lijin station (37.49°N, 118.25°E)^① of the Huanghe River (b) from 1960 to 2015.

DownLoad: Full-Size Img PowerPoint

The rapid warming in the Bohai Sea might have had a different origin. As shown in Fig. 9b, the water discharge from the Huanghe River to the Bohai Sea has notably decreased from 1960 to 2015, with the rate of −0.68 Gt/a. Considering that the Bohai Sea is the only inner sea of China which is surrounded by heavily populated and industrialized cities, the rapid warming in this sub-region might have resulted from the observed terrestrial warming affecting the adjacent coastal seas. Taken Dalian City (39°01′N, 121°44′E) as an example, the population grew from 1.28 million in 1953 to 6.69 million in 2010, at a 2.9% compound annual growth rate (the data source: reports of the National Census of China). Urbanization and the accompanying local warming may have contributed to the observed SST increase in some extents. The urbanization effect has been confirmed for annual and seasonal mean SAT trends in northern and eastern China (Ren et al., 2008, 2017; Yang et al., 2011), the observed SST trends along the Bohai Sea may have also been affected by urbanization. Therefore, the rapid warming along the Bohai Sea may have been partially caused by land-based anthropogenic activities, such as the local urbanization and the other kinds of land use and land cover change, in addition to the large-scale atmospheric circulation variability.

5. Dicussion and conclusions

In this study, monthly mean SST recorded at 26 stations along the coast of the China seas from 1960 to 2015 have been homogenized, using the PMTred algorithm together with metadata to detect break points, and using the QM adjustment method conducted with reference series to adjust these series to diminish inhomogeneities.

It is notable that the instrument change including station system transformation is the main cause for the identified inhomogeneities, followed by stations relocation, accounting for about 61.3% and 24.2% of the total change points respectively. These artificial shifts in monthly SST series were then adjusted using the results of statistical tests together with available metadata. The ratio of the negative shifts is much higher than that of the positive shifts, indicating that homogenized SST largely corrects the warm biases caused in the earlier artificial observing period. The warming rate of the whole SST series along the Chinese coast after adjustment is much larger than the raw data, up to 0.19°C per decade. Using the newly homogenized monthly mean SST data, seasonal mean SST trends in winter and summer were analyzed. Results show that the SST series located at the northern China seas have experienced robust warming trends in winter which are probably caused by the combined effect of multi-decadal variability of three large-scale atmospheric modes.

Due to the sparse of coastal hydrological stations in China, it is difficult to find an available SST series to make a reference sequence. In this case the SAT data as alternative data were used to construct the reference sequence. Based on the homogenized SST data, it is confirmed the rapid warming at the Chinese coast. The warming rate is comparable in magnitude to those reported in the eastern China based on SAT data (Ren et al., 2005, 2017) and in the China seas based on SST data (Bao and Ren, 2014) during the same period. The slightly lower warming trend of SST as compared to the increase of SAT along the coastal zone ((0.19±0.06)°C per decade vs. (0.22±0.06)°C per decade) may have been related to the fact that some of the SAT observations are made at urban stations, and the effect of urbanization on the estimated SAT trend may have been kept in the homogenized monthly mean SAT data series (Ren et al., 2008, 2015).

In this paper, the relationship between winter SST and several atmospheric circulation modes were analyzed. However, the dynamic linkage of the decadal and trend variation of the SST with oceanic oscillation modes and local oceanic forcing, such as Pacific Decadal Oscillation and El Niño-Southern Oscillation and Kuroshio Current, still needs to be investigated in future. In spite of the remaining issues mentioned above, this paper delivers the first results of SST data homogenization and the SST trend along the China seas based on the homogenized data of hydrological stations. The results are conductive to the understanding of regional climate change. The homogenized SST data could also be used as a benchmark for assessing the performance of analyses on globally gridded SST datasets.

Acknowledgements

We thank Hamburg University’s Cluster of Excellence CliSAP (Integrated Climate System Analysis and Prediction) for funding the visiting in Germany which benefits this study greatly.