1 INTRODUCTION

Sea surface temperature (SST) is one of the most important parameters to study ocean dynamics and marine meteorology. It represents a comprehensive result of the ocean's thermo and hydro-dynamics, as well as the interaction between sea surface and atmosphere. On the other hand, sea surface diurnal warming (i.e. day-night SST difference), which has drawn a lot of attenthion in recent years, is a combined result of solar radiation, turbulent mixing, sea surface heat conduction, etc. (Deschamps and Frouin, 1984; Flament et al., 1994).

Currently, studies on the sea surface diurnal warming mainly focused on two aspects. Some researchers studied on the strength of and factors that influence diurnal warming (Deschamps and Frouin, 1984; Stramma et al., 1986; Flament et al., 1994). While others focused on the influence of oceanic and atmospheric environment (Shinoda, 2005; Kawai and Wada, 2007; Li et al., 2013). Meanwhile, with the development of temporal and spatial resolution of satellite remote sensing technology and the extensive application of numerical models, an increasing number of in-depth studies have been conducted on sea surface diurnal warming, combining satellite data, measured data, and numerical models, regarding temporal, spatial features and the underlying mechanisms (Stramma et al., 1986; Kawai and Kawamura, 2002; Stuart-Menteth et al., 2003).

It has been discovered through the observation of satellite data that the surface diurnal warming of most regions is relatively small (about 0.5 K), but significant diurnal warming events (over 6 K) has also been found in some areas (Deschamps and Frouin, 1984; Stramma et al., 1986; Flament et al., 1994; Kawai and Kawamura, 2002; Notarstefano et al., 2006). According to other researchers' analysis, diurnal warming strength is closely related to the solar radiation and sea surface wind fields (Deschamps and Frouin, 1984; Flament et al., 1994; Tanahashi et al., 2003; Merchant et al., 2008; Hong et al., 2014), but there are regional differences. For instance in 2012, Filipiak et al., applied the wind data about sea surface heat flux and numerical weather prediction to build a statistical model for sea surface diurnal warming in which diurnal warming presents a linear dependence on the net surface heat flux integrated since (approximately) dawn and an inverse quadratic dependence on the maximum of the surface wind speed in the same period. Therefore, there are seasonal differences of the diurnal warming in the time scale (Kawai and Kawamura, 2005; Notarstefano et al., 2006; Eastwood et al., 2011; Lin et al., 2014). In addition, it is found that climate variations have certain impacts on diurnal warming (Stramma et al., 1986; Soloviev and Lukas, 1997; Kawai and Kawamura, 2005; Kawamura et al., 2008). Similarly, during the La Nina and El Nino periods, the sea surface diurnal warming is modulated in strength (Wu, 2013). According to research studies, the large diurnal warming of some areas is still influenced by other factors. For instance, Zhu et al. (2014) studied the large diurnal warming of coastal areas with in-situ temperature and weather datasets from the Caribbean and Great Barrier Reef (GBR), Australia. Results showed that, in contrast to the nearby deep areas, the shallower water areas have a higher maximum temperature and lower minimum temperature, and the diurnal heating is closely related to local environment, such as water depth, different types of rock, tidal fluctuation, etc.

In the present work, the East China Sea is selected for research, as shown in Fig. 1 (21.5°–34.5°N; 116.5°–130°E). With complicated topography and a long coastal line, it covers various islands, such as Taiwan Island, Jeju Island, the Ryukyu Islands, and the Zhoushan Islands. The marine topography becomes increasingly deep from the northwest to the southeast. On the northwest of the line between Taiwan Island and the Gotō Islands, there are continental shallow waters, and on the southeast, there is the continental shelf and trough. The continental shelf of the East China Sea is one of the world's widest continental shelves.

Since the inland-sea-air interaction is very complex in the shelf region, together with various influencial factors and complicated mechanisms, study of the diurnal warming process in the East China Sea is of great importance to the study and prediction of the land-sea climate variations and air-sea interactions in the area. Moreover, sea surface diurnal warming may further impact on the marine ecology and fishery activities.

2 METHOD AND DATA 2.1 MODIS-aqua, terra

The Earth Observation System (EOS), as a significant part of the ESE (Earth Science Enterprise) strategic plan of NASA (National Aeronautics and Space Administration), aims to strengthen the comprehensive scientific research of the earth's surface land, sea, air, and their mutual relations. Through long-term preparation and manufacturing, NASA launched Satellite Terra and Aqua on December 18, 1999 and May 4, 2002, respectively. Satellite Terra descends at 10:30 am and ascends at 10:30 pm local time, while satellite Aqua descends at 1:30 am and ascends at 1:30 pm. Consequently, Terra is also called the morning satellite while Aqua is called the afternoon satellite.

A Moderate-Resolution Imaging Spectroradiometer (MODIS) is one of the main sensors carried on Satellite Terra and Satellite Aqua. It has a spatial resolution of as high as 0.25–1 km. The two satellites cooperate with each other and are able to observe the entire earth surface repeatedly in one or two days, obtaining 36 wavebands of observational data. Standard data products can be divided into level 0 and level 1 according to their distinct content. After level 1B, data products are divided into levels 2, 3, and 4. In the present research, the level 4, mapped produce, is selected. It is mainly obtained through the analysis model and comprehensive analysis of data below level 3 with geographical and radiation corrections (Gentemann, 2014). In the Mapped data, there is a layer that stores the credibility of the data, which has five attributes: -1 means no value, 0 means good quality, 1 means edge data, 2 means snow coverage, 3 means cloud coverage. To ensure the credibility of the data and a high coverage rate, data points ranging between 0 and 1 were selected for statistics.

The accuracy of the MODIS level 4 product in the East China Sea has been validated by Zheng et al. (2006), which suggested the mean, standard and maximum errors to be 0.23 K, 0.29 K and 0.5 K, respectively.

2.2 Calculation of sea surface diurnal warming

SST Diurnal warming can be calculated by subtracting the night temperature (NSST) from the day temperature (DSST). Meanwhile, to increase the data coverage, the adjacent day temperature or night temperature will be used to replace the missing data. The calculation formula follows Stuart-Menteth et al. (2003):

Events with ΔT < 0 rarely occur with magnitude less than 0.5 K, which could be caused by algorithm error or some oceanic/atmospheric processes like storm surge. In this paper, we focused only on diurnal warming process, that is, data points with ΔT > 0 are selected for statistics analysis and those with ΔT < 0 are ignored.

3 RESULT 3.1 Comparison of MODIS-aqua and MODISterra diurnal warming

The Aqua and Terra data between January 1 and December 31, 2014 are used to calculate the diurnal warming of the East China Sea and for comparison.

It can be learned from Fig. 2 that the numbers of valid data vary similarly in ΔT-Aqua and ΔT-Terra. Both have 10 to 20 days of oscillation due to the impacts of scanning swath, clouds, water vapor, etc. During the rainy season in spring and summer, the data efficiency is relatively low; while in partly cloudly season in winter and autumn, the number is obviously higher than that in other periods.

Satellite Terra is descending at 10:30 am local time and ascending at 10:30 pm, while Aqua is descending at 1:30 am and ascending at 1:30 pm, respectively. According to the historical meteo data, the SST reaches its highest level at 14:00 due to the solar radiation, and the lowest temperature occurs at 6:00 due to the thermal longwave radiation at sea surface and the air-sea sensible heat exchange (Weng et al., 1993). Therefore, factors that may enhance the sea surface diurnal warming include the solar radiation during daytime, the thermal longwave radiation and sensible heat flux during night, while comparably stronger latent heat flux with higher SST and stronger wind during daytime may reduce the sea surface diurnal warming strength. Consequently, ΔT-Aqua is larger and has a wider threshold value than ΔT-Terra (Figs. 3 and 4). However, due to the huge ocean specific heat and air-sea interaction, seawater mixing may give rise to internal heat transmission or heatmomentum transformation, which may make the diurnal warming smaller (ranging between 0 and 5 K) than that on the land. As seen from the ΔT interval probability, for ΔT-Terra, the range between 0 and 0.5 K is the 40% of the total, while ΔT-Aqua ranges between 0 to 0.5 K or 0.5 to 1.0 K, accounting for about 30% to 40% of the total. It has less difference between ΔT-Terra and ΔT-Aqua in ranges above 1 K and their variations with month are similar. It is worth noting that the data above 1.5 K account for 5%–20% of the total in ΔT-Terra and ΔT-Aqua, which is obviously higher than other areas (Lin et al., 2014, in the South China Sea).

3.2 Diurnal warming

Aqua data from January 1, 2005 to December 31, 2014 are adopted for evaluating the diurnal warming of the East China Sea by region. Figure 5 shows the monthly averaged surface diurnal warming of the East China Sea, from which it can be seen that the trends of the diurnal warming and sea surface temperature variation are different, and sea surface diurnal warming is characterized by strong variation in summer and autumn and weak variation in spring and winter. The main reason may be that key factors controlling sea surface diurnal warming are wind speed and solar radiation. The East China Sea spans the temperate zone and subtropical zone with relatively weak solar radiation in winter. Meanwhile, due to the impact of the Asian Continental High Pressure, the East China Sea is dominated by the northerly strong wind, whose average wind speed reaches 9–10 m/s. In summer, the solar radiation is relatively strong, and it is mainly dominated by the southerly wind, with a low wind speed.

The East China Sea has been divided into three regions for comparison, i.e. the Coastal Region (25.5°–33°N; 120°–123°E), the Taiwan Strait (21.5°–25.5°N; 117°–122°E), and the Eastern Rigion (25.5°–33°N; 123°–129°E), shown with dotted frames in Fig. 1.

The monthly averaged diurnal warming in the Coastal Region is relatively larger than that in the other two regions all through the year (Fig. 6). This is closely related to the greater daily temperature variability over the land and the river runoffs. In most years, the largest diurnal warming in the Coastal Region occurs in spring when the solar radiation increases and the surface wind is weaker during monsoon switching. However, another peak always occurs in winter, when the solar radiation is the weakest and the strong wind dominates. This might be explained by the much larger day-night air temperature difference under clear-sky situation in the Coastal Region, compared to the other seasons. On the other hand, the southward intrusion of the Changjiang (Yangtze) River plume may also count, but it should be verified by a further research work with numerical models.

In winter and spring, the surface diurnal warming is larger in the Eastern Region than in the Taiwan Strait, but the situation reverses in summer and autumn. This might because the strongest monsoon wind always dominates during winter and spring in the Taiwan Strait, while in summer and autumn, enhanced intrusion of the Kuroshio and the Taiwan Warm Current to the Eastern Region will bring stronger evaporation and reduces the surface diurnal warming. The highest peak always occurs in summer in the Taiwan Strait, under the situation of stronger solar radiation and weaker monsoon wind. But the annual cycle in the Eastern Region is more complicate and shows obvious interannual variability. Further numerical modeling works are needed to reveal the underlying mechanisms.

3.3 Large diurnal warming events in the East China Sea

ΔT≥5 is taken as the criterion of the large diurnal warming phenomenon in the East China Sea. Figure 7 shows the average ΔT of the East China Sea region and the spatial and temporal trend of points of diurnal warming, from which it can be seen that the large diurnal warming mainly occurs in areas and periods with high average ΔT, suggesting that the large diurnal warming is not an isolated phenomenon, but is closely tied to the overall diurnal warming of the sea area. Generally, in spring and autumn, the East China Sea is featured by a frequent occurrence of large diurnal warming, and according to the comparison by region, the large diurnal warming mainly occurs in nearby coastal areas, especially near the Hangzhou and Changjiang River estuary, which is greatly related to land-sea distribution, runoff, and water depth. Comparatively, large diurnal warming events are less frequently happening in the Taiwan Strait than in the Coastal Region because there are more days with strong wind and huge waves, which may mix the sea water significantly and then reduce the diurnal warming.

4 DISCUSSION 4.1 Diurnal warming and wind speed over the sea surface

AMSR-E (The Advanced Microwave Scanning Radiometer—Earth Observing System) wind data (Wentz and Meissner, 2007) and ΔT-Aqua data of 2014 are used to examine the relation between the surface diurnal warming of East China Sea and wind speed. Both are from the same satellite, with the same transit time, but due to the different of but due to the difference in spatial size of the data (AMSR-E is 25 km, and Aqua is 4 km), the ΔT-Aqua data close to the AMSR-E data points are selected. Meanwhile, to eliminate the impact of other factors, the data with a high valid rate are selected as representative for observation. Figure 8a is a scatter plot of the diurnal warming versus, from which it can be learned that the diurnal warming is closely related to the wind speed, and diurnal warming greater than 3℃ mainly occurs in the interval in which the wind speed is smaller than 4 m/s. When the wind speed is higher than 4 m/s and lower than 12 m/s, diurnal warming increases with the wind speed while the threshold decreases gradually. When the wind speed is greater than 12 m/s, the diurnal warming fails between 0.5 and 1, with relatively slight changes in distribution. Similarly, it can be seen from Fig. 8b that when the wind speed ranges between 0 and 4 m/s, diurnal warmings greater than 1.0 is account for more than 60% of the total. When the wind speed is greater than 12 m/s, diurnal temperatures between 0.5 and 1 accounts for more than 80% of the total. The probability of diurnal warming which is greater than 2.5 K is nearly 0. The mixing effect of wind speed on sea water is mainly considered. When the wind speed is small, the sea water mixing is relatively slight, but the sea surface temperature varies greatly with the solar radiation, and as a result, it is relatively large. With the increase of the wind speed, the ocean mixing is enhanced, but in the case of an obvious temperature decrease, diurnal warming is suppressed. When the wind speed is greater than a definite value, the sea water will be fully mixed at vertical and horizontal direction, and the diurnal warming will be weak.

5 CONCLUSION

In the present work, we firstly compared the surface diurnal warming in the East China Sea calculated by the MODIS-Aqua and MODIS-Terra data in 2014. Features are similar and significant events can be detected in both. However, due to diffetent transit time of the two satellites, the strength calculated by MODIS-Aqua data is relatively larger. The following conclusions were reached:

1) Surface diurnal warming in the East China Sea presents seasonal variations. In general, it's stronger in summer and autumn but weaker in winter and spring, just in opposite to the surface wind speed. But the seasonality performs differently in different regions. In the Coastal Region, the largest surface diurnal warming occurs in spring, while a second peak always exists in winter, which might be induced by the comparably larger day-night air temperature difference and the southward intrution of the Changjiang River plume. In the Taiwan Strait, the highest peak shows in summer, under the situation of stronger solar radiation and weaker monsoon wind. But the annual cycle in the Eastern Region is more complicate and shows obvious interannual variability.

2) Large surface diurnal warming events in the East China Sea mainly occurs in coastal region, and there is a high-incidence period from April to July. Temporal and spatial distributions of the large events show consistency with the averaged surface diurnal warming strength.

3) A very close relationship between the surface diurnal warming strength and the surface wind speed was revealed. Strong diurnal warming mainly gathers in areas with low wind speed. The possibility of surface diurnal warming strength less than 1 K will be over 80% and no point with diurnal warming strength larger than 2.5 K is detected, when the wind speed reaches to over 12 m/s.

6 DATA AVAILABILITY STATEMENT

Sea surface temperature data that support the findings of this study have been deposited in the NASA's OceanColor Web:

https://oceandata.sci.gsfc.nasa.gov/MODIS-Aqua/Mapped/Daily/4km/, and

https://oceandata.sci.gsfc.nasa.gov/MODIS-Terra/Mapped/Daily/4km/.

Sea surface wind speed data that support the findings of this study have been deposited in the Remote Sensing Systems (RSS) with the primary accession code:

ftp://ftp.remss.com/.

All those databases are public and free.

7 ACKNOWLEDGEMENT

The authors would like to thank the editor and the reviewers for their remarkable advices, and Mr. FAN Jianlei for proofreading.