Figure
1.
Arctic and its sub-regions analyzed in the study. Different colors indicate different Arctic sub-regions and land is in gray.
| Citation: | Menghan Wang, Zengan Deng. On the role of wave breaking in ocean dynamics under typhoon Matsa in the Bohai Sea, China[J]. Acta Oceanologica Sinica, 2022, 41(9): 1-18. doi: 10.1007/s13131-022-1995-3
|
The global temperature has continued to rise since the 1970s, which has a profound impact on the Arctic region, leading to a gradual warming of the Arctic (Cohen et al., 2014; Pachauri et al., 2014). Sea ice, as an integral component of the Arctic climate system (Aagaard and Carmack, 1989; Barry et al., 1993; Box et al., 2019; Wang et al., 2019), has undergone rapid changes in recent decades (Bi et al., 2016; Lemke et al., 2007; Stroeve and Notz, 2018). The melt onset dates (MOD) over Arctic sea ice can be used to depict the capability of the absorption of solar energy by the sea ice surface (Perovich et al., 2007), which significantly influences the Arctic surface radiation energy balance between sea ice and atmosphere (Belchansky et al., 2004a; Bliss et al., 2019; Perovich et al., 2007). During the melt season, the sea ice surface meltwater triggers a decrease in surface albedo, resulting in more solar energy absorbed by the sea ice surface and accelerating the reduction of sea ice volume (Curry et al., 1995; Stroeve et al., 2014). Consequently, an improved understanding of the spatio-temporal characteristics of the MOD variations at multiple time scales would contribute to assess the responses of Arctic sea ice to climate change in the future and predict the trend of Arctic sea ice MOD.
Due to the limited in situ observations, passive microwave remote sensing has become one of the most viable methods for monitoring MOD, which provides observation with all-time and all-weather coverage, especially in polar areas with the complex geographical environment and extreme climatic conditions (Markus et al., 2009; Smith, 1998). When liquid water appears in sea ice, the microwave emission measured by microwave sensors will change rapidly, and thus detects the earliest initiation of melt (Anderson, 1987). Although data from active microwave sensors can yield MOD with a reasonably high spatial resolution (100 m–5 km) (Kwok et al., 2003; Winebrenner et al., 1994), it is more common to estimate MOD from passive microwave brightness temperature (Belchansky et al., 2004b; Drobot and Anderson, 2001a; Smith, 1998), which can provide a consistent long-term observation record over the entire Arctic from 1979 onwards. Several MOD estimation algorithms have been developed based on passive microwave observations, including the advanced horizontal range algorithm (AHRA) (Drobot and Anderson, 2001b), the passive microwave method (Markus et al., 2009), and the combination of active and passive microwave data (Belchansky et al., 2004a), which have been successfully applied in the study of MOD over Arctic sea ice (Bliss et al., 2017). Some studies have reported different discoveries of MOD by using passive microwave products. Mortin et al. (2016) found the sea ice in the Arctic is experiencing an earlier melt onset over the past few decades. Moreover, the magnitude and trend of MOD vary with location and periods. Anderson and Drobot (2001) reported the MOD in the western central Arctic exhibited a significant earlier trend (8.9 d/decade, p<0.05), and the Beaufort Sea and Canadian Archipelago regions significantly advanced by 5.1 d/decade and 6.7 d/decade during 1979–1998, respectively. Stroeve et al. (2006) indicated the MOD displayed a significant earlier trend at a rate of 4.7 d/decade in the Beaufort Sea, 7.3 d/decade in the Kara, and 8.8 d/decade in the Barents from 1979 to 2005. Bliss and Anderson (2014a) identified the interannual mean of MOD from 1979 to 1987 was generally later than that during 1979–2012. Several studies pointed out that the interannual variations of MOD are likely related to many factors, such as atmospheric circulation (Drobot and Anderson, 2001b), surface downwelling longwave radiation (Barber and Thomas, 1998; Maksimovich and Vihma, 2012), and atmospheric moisture anomalies (Mortin et al., 2016). The abovementioned studies only focused on the interannual variations of MOD, while few studies have reported the results concerning the mutation monitoring of MOD variations over the Arctic so far. Additionally, some studies have preliminarily indicated the Arctic Oscillation (AO) has effects on MOD variation (Belchansky et al., 2004b; Drobot and Anderson, 2001a); however the correlation analysis between AO and MOD for different Arctic sub-regions is limited and is worthy of further investigation.
Therefore, in this study, more comprehensive statistical analysis, such as mutation detection and trend analysis of MOD variations, was conducted. The spatio-temporal variation characteristics of MOD for Arctic sea ice over a 39 years period (1979–2017) during the melt seasons were analyzed by using long-term passive microwave observations. The time-series approach (Sen’s slope estimator) was used to analyze the spatio-temporal variable patterns and trends for the Arctic and its 14 sub-regions from 1979 to 2017. A multi-time scales analysis of the trend variations of MOD was performed. Additionally, the Mann-Kendall (M-K) mutation test was adopted to investigate the transition point for different Arctic sub-regions. Finally, the correlation between monthly and seasonal AO and MOD was investigated. The paper is organized as follows. The MOD dataset and study region are briefly introduced in Section 2. Section 3 describes the analysis method used in the study, including the Sen’s slope estimator, the M-K method, and the correlation analysis. In Section 4, the spatial and temporal variation characteristics, different period variations, and transition point of MOD, and the correlation between MOD and AO for the Arctic and its sub-regions are analyzed. Finally, the conclusions are summarized in Section 5.
The MOD data used in the study are the “Snow Melt Onset Over Arctic Sea Ice from SMMR and SSM/I-SSMIS Brightness Temperatures, Version 4”, which are available from the National Snow and Ice Data Center (NSIDC) (https://nsidc.org/data/nsidc-0105) (Anderson et al., 2019). The data are provided in a polar stereographic grid with a spatial resolution of 25 km. In this study, the annual MOD in the period of 1979 to 2017 was used. The operational MOD algorithm for this data is the AHRA developed by Drobot and Anderson (2001b). No other change is made to the AHRA method since the publication by Bliss and Anderson (2014b) and thus, the V3 (Meier et al., 2017) and V4 data have almost no difference at all in most sea ice areas. The MOD in the new version is reprocessed by using a new version of the sea ice extent masks. In the dataset, the brightness temperatures from several passive microwave sensors are collected, including the scanning multichannel microwave radiometer (SMMR) during 1979 to 1987, the special sensor microwave imagers (SSM/I) and the special sensor microwave imager and sounder (SSMIS) during 1988 to 2017 to estimate the MOD over Arctic sea ice. To ensure consistent input data record, prior to determining the MOD over sea ice, the brightness temperature data from different platforms are calibrated using the DMSP F08 SSM/I as the baseline sensor through linear regression coefficients during overlap periods following previous studies (Abdalati et al., 1995; Jezek et al., 1991; Stroeve et al., 1998).
The earliest MOD can be identified by a change in the brightness temperature. Briefly, the theoretical basis of the AHRA method depends on the day-to-day variability in the increase of brightness temperature (Drobot and Anderson, 2001a). Furthermore, horizontal channels are more strongly related to snow conditions during melt than vertical channels. Moreover, different frequencies have different responses to melt onset. The change of brightness temperature at 19 GHz (18 GHz for SMMR data) is more dramatic than that at 37 GHz when the liquid water occurring in the snowpack. Therefore, the melt criterion of the AHRA method to detect MOD is the daily difference between horizontally polarized brightness temperature at 19 GHz (18 GHz for SMMR data) and 37 GHz channels (Drobot and Anderson, 2001a). For more details of the AHRA method, readers are referred to Drobot and Anderson (2001a). The MOD data set released by NSIDC has been used in many previous studies (Bliss and Anderson, 2014b, 2018; Singh et al., 2020). Some studies have confirmed and explained the accuracy of the MOD data (Bliss and Anderson, 2018; Drobot and Anderson, 2001a).
To better understand the temporal variability of the MOD in different regions, the Arctic region was divided into 14 sub-regions (i.e., Sea of Okhotsk, Bering Sea, Hudson Bay, St. Lawrence Gulf, Baffin Bay, Greenland Sea, Barents Sea, Kara Sea, Laptev Sea, East Siberian Sea, Chukchi Sea, Beaufort Sea, Canadian Archipelago, and central Arctic) following the criterion used in Bliss and Anderson (2018). The Arctic region defined here is relatively broad, spanning the Arctic Circle and the middle and low latitudes, which is identical to the region definitions used by Meier et al. (2017). Figure 1 shows the Arctic and its sub-regions analyzed in the study. The version of the Arctic section mask file used in the study also first appeared in Meier et al. (2017). The original section mask file is an array formatted to NSIDC’s 25 km Northern Hemisphere Polar Stereographic Grid, and we classified MOD into 14 sub-regions according to the mask file.
In this study, the Sen’s slope estimator (Sen, 1968) and the M-K trend test (Kendall, 1948) were used to investigate the temporal trends of MOD and its significance over the Arctic from 1979 to 2017 at the pixel level. The Sen’s slope estimator can not only reduce the interference of abnormal values in a time series but also diminish the influence of missing time series observations and non-normally distributed data (Liuzzo et al., 2016). Therefore, Sen’s slope estimator is widely used to analyze long-term sequence datasets. It is often used in the trend analysis of long-term series data to detect the magnitude of the trend:
| $$ {\rm{sen}} = {\rm{Median}}\left( {\frac{{{x_j} - {x_k}}}{{j - k}}} \right), \quad\; j=1,2,\cdots,n, $$ | (1) |
where xj and xk are the data values at times j and k (j>k), respectively. A positive sen value indicates an increasing trend, and a negative sen value indicates a decreasing trend.
The M-K trend test is a nonparametric test method, which has no requirement on data type and is often applied to determine the significance. The standard normal statistic of the M-K test (Z) can be calculated by
| $$ Z = \left\{ {\begin{array}{*{20}{c}} {\dfrac{{s - 1}}{{\sqrt {{\rm{Var}}\left( S \right)} }}}&{\rm{if}}\;{S > 0}\\ 0&{{\rm{if}}}\;{S = 0}\\ {\dfrac{{s + 1}}{{\sqrt {{\rm{Var}}\left( S \right)} }}}&{{\rm{if}}}\;{S < 0} \end{array}} \right., $$ | (2) |
where
| $$ S = \sum_{i = 1}^{n - 1} {\sum_{j = i + 1}^n {{\rm{sgn}} \left( {{x_j} - {x_i}} \right)} } , $$ | (3) |
| $$ {\rm{sgn}} \left( {{x_j} - {x_i}} \right) = \left\{ {\begin{array}{*{20}{c}} { + 1}&{{\rm{if}}}\;{{x_j} - {x_i} > 0}\\ 0&{{\rm{if}}}\;{{x_j} - {x_i} = 0}\\ { - 1}&{{\rm{if}}}\;{{x_j} - {x_i} < 0} \end{array}} \right., $$ | (4) |
| $$ {\rm{Var}}\left( S \right) = \frac{{n\left( {n - 1} \right)\left( {2n + 5} \right) - \displaystyle\sum_{i = 1}^m {{t_i}\left( {{t_i} - 1} \right)\left( {2{t_i} + 5} \right)} }}{{18}}, $$ | (5) |
where xi and xj indicate the pixels values at times i and j, respectively; sgn(xj−xi) represents the sign function; n and m are the number of data points and tied groups, respectively;
The M-K method was also employed to detect the mutation points of MOD over the Arctic region and its sub-regions. The M-K method is considered to be an effective method for detecting the transition point from a relatively stable state to another state (Da Silva et al., 2015; Fu et al., 2013).
For a time series
| $$ {S_k} = \sum_{i = 1}^k {{r_i},\quad {k = 2,3,\cdots,n} }, $$ | (6) |
where
| $$ r_i = \left\{ \begin{array}{*{20}{c}} {1,}&{{x_i} > {x_j}}\\ {0,}&{{x_i} \leqslant {x_j}} \end{array} ,\right. {j = 1,2,\cdots,i} . $$ | (7) |
Assuming that the time series xi is random and independent, the statistic is defined as
| $$ {\rm{UF}}_{k} = \frac{{\left[ {S_k - E\left( {S_k} \right)} \right]}}{{\sqrt {{\rm{Var}}\left( {S_k} \right)} }}, \quad{k = 1,2,\cdots,n} , $$ | (8) |
where
| $$ E\left( {{S_k}} \right) = \frac{{n\left( {n - 1} \right)}}{4}, $$ | (9) |
| $$ {\rm{Var}}\left( {{S_k}} \right) = \frac{{n\left( {n - 1} \right)\left( {2n + 5} \right)}}{{72}}, $$ | (10) |
where Sk is the cumulative number of cases xi>xj (1≤j≤i), n is the number of samples, UFi follows a standard normal distribution, and UF1=0. If UFi value is greater than 0, it denotes an upward trend; otherwise, it denotes a downward trend. If UFi value exceeds the significance level, it indicates that there is a significant trend in the time series.
Moreover, this method also calculates the statistic UBk for the inverse series (xn, xn−1, ···, x1) following the same procedures as shown in Eqs (6)–(10), and UB1=0. The intersection of the UFk and UBk curves in confidence interval is determined as the mutation point.
The AO is a large scale mode of climate variability, also referred to as the Northern Hemisphere annular mode. It is not only an important parameter predicting El Niño (Chen et al., 2014), but also naturally has a great influence on the general Arctic climate (Moritz et al., 2002; Jevrejeva et al., 2003). Previous studies have explored the impact of AO on the MOD over the Arctic sea ice (Belchansky et al., 2004b; Drobot and Anderson, 2001a), sea ice motion (Rigor et al., 2002), and surface air temperature (Rigor et al., 2000). To further investigate the effect of AO on MOD changes in Arctic sub-regions, we calculated Pearson correlation coefficient between monthly and seasonal MOD and AO. The AO index is from NOAA National Centers for Environmental Information, which is obtained by projecting the AO loading pattern to the daily anomaly 1 000 hPa height field over 20°–90°N latitude. Besides, winter, spring, and summer here are defined as January–March, April–June, and July–September, respectively. The monthly mean AO index from 1979 to 2017 was used in this study and the seasonal AO index was calculated from the monthly mean AO index. More information of the AO index data can be seen at http://ncdc.noaa.gov/teleconnections/ao/.
The spatial patterns of MOD from 1979 to 2017 are shown in Fig. 2, which is described by mean values, standard deviation, trend, and significance. The processing method of data is the same as Bliss and Anderson (2014b). All statistics were calculated on a per-pixel basis where the MOD exists in all 39 years of the data record. A pole hole area (>84.5°N) is observed in Fig. 2 since data were not collected by the SMMR sensor within this region, and those pixels have been discarded from the subsequent statistical analysis. During the past 39 years, the mean MOD on Arctic sea ice has noticeable regional differences. The MOD occurs earlier at low latitude (<70°N), and the MOD gradually increases with increasing latitude spatially. In addition, the MOD mostly ranges from 90 to 170 day of year (DOY), accounting for 87.9% of the study area. The earlier MOD is found in the peripheral sea ice (Sea of Okhotsk, Bering Sea, Hudson Bay and St. Lawrence Gulf) and the southernmost regions, with occurrence before 100 DOY (April 9). However, a small area in the Canadian Archipelago and northwestern coast of Greenland shows later MOD around late May. By contrast, the regions with later MOD, occurring in above 150 DOY (May 30), are mainly located in the central Arctic and a part of the Canadian Archipelago.
Overall, most regions (81.7% of the study area) have considerable inter-annual variability as reflected by the high standard deviation (more than 15 days). Specifically, the regions with the largest standard deviation (more than 40 days) are located in the ice edge of the Laptev Sea and Kara Sea, indicating that the MOD in these regions fluctuates more significantly during the past 39 years. However, a part of study area (about 2.3%), mainly distributed in the western parts of the central Arctic, the Canadian Archipelago, the edge of Sea of Okhotsk and parts of Laptev Sea, exhibits a smaller variability within 10 days. The spatial pattern of MOD is consistent with that in Bliss and Anderson (2014b), though there is a slight difference in study period.
The spatial pattern of the change trends of MOD is displayed in Fig. 2c. Over the past 39 years (from 1979 to 2017), there are noticeably larger areas experiencing advanced MOD than those with delayed trends. Specifically, 85.9% of the study area shows an earlier trend (58.2% are statistically significant), while 14.1% of those exhibits a delaying trend (1.2% are statistically significant). Concurrently, 60% of study areas with advanced MOD (>8 d/decade) at the statistically significant level of p<0.05 mainly occur in higher latitude (above 70°N) and the northern part of Hudson Bay, whereas more than 1.2% of delayed MOD (>8 d/decade) at the significance level of p<0.05 occur in the northeast of Hudson Bay and eastern coast of Greenland Sea.
The temporal variations in the MOD were analyzed with Sen’s slope estimator based on the mean annual MOD trends, and the trend significance was determined by using the p-value from the two-tailed student’s t-test after testing for normal distribution and auto-correlation. Additionally, the coefficient of determination (R2) between MOD and year was estimated for each region.
Figure 3 presents the mean annual MOD for each Arctic sub-region. The average MOD statistics for the 14 Arctic sub-regions in the period of 1979 to 2017 are summarized in each subgraph. Over the past 39 years, the MOD exhibits an earlier trend in most Arctic sub-regions. The largest trend is in the East Siberian Sea, with a significant earlier trend of 9.45 d/decade. Anderson and Drobot (2001) also found the East Siberian Sea displayed the earlier MOD from 1979 to 1998 and indicated that it may be caused by different climate regimes due to land or ice-locked area. The regions with earlier trends exceed 7 d/decade, are mainly located in the Barents Sea, Kara Sea, Laptev Sea, Chukchi Sea, and Beaufort Sea, which may due to the large-scale weather patterns (such as a cyclonic activity) (Anderson and Drobot, 2001). In addition, the MOD has a significant trend of −6.97 d/decade in the central Arctic located in the highest latitudes, with a relatively high R2 value of 0.7. These MOD trends are generally consistent with the findings of Bliss et al. (2019). For the peripheral sea ice regions including the Sea of Okhotsk (Fig. 3a), Hudson Bay (Fig. 3c) and Baffin Bay (Fig. 3e), St. Lawrence Gulf (Fig. 3d), Bering Sea (Fig. 3b) and Greenland Sea (Fig. 3f), the trends are generally smaller, and the R2 values are relatively low compared to central Arctic (Fig. 3n). Statistically significant trends mainly occur in the Baffin Bay (Fig. 3e) at the significant level of p<0.05 with a low R2 value of 0.14. In addition, it is interesting to find that the MOD presents delayed trends in the Bering Sea (Fig. 3b) and St. Lawrence Gulf (Fig. 3d) regions typically being statistically different from other regions in the Arctic, with the weaker positive trend of 1.62 d/decade and 0.46 d/decade (in red mark in panels of Fig. 3), respectively. The anomalous results of the Bering Sea are similar to the previous study (Bliss and Anderson, 2014b) and the reasons need to be further investigated.
Compared with the results of Bliss and Anderson , it is found the variation trends of MOD are generally in agreement for the majority of Arctic regions, though the magnitude of change is slightly different. This may be due to the difference in the study period, i.e., 39 years in this study and 34 years in Bliss and Anderson (2014b). The different regional variability is most likely due to the variations of geographical environment and atmospheric conditions for different regions of the Arctic during melting (Bliss and Anderson, 2014a). Figure 4 displays the Pearson correlations between MOD for different Arctic sub-regions, which implies that peripheral seas are nearly independent of each other, whereas the regions in the middle and high latitudes (Barents Sea, Kara Sea, Laptev Sea, East Siberian Sea, Chukchi Sea, Beaufort Sea, Canadian Archipelago, and central Arctic) dominate the variation of MOD in the Arctic.
From the above analysis results, the MOD variations not only show pronounced spatial distribution differences but also present distinct temporal heterogeneity. Decadal trends may conceal the strong short-term fluctuations in the MOD. In this section, we analyzed MOD trends with various window sizes (i.e., the length of the study period) ranging from 5 years to 35 years (from 1979 to 2013), across all start-year and window-size combinations, averaged over the Arctic sea ice and its sub-regions, shown in Fig. 5.
The mean annual MOD universally displays earlier trends (i.e., earlier sea ice melt date) for the whole Arctic sea ice, whereas some short-term (5–10 years) changes in MOD demonstrate later trends since the mid-1990s. Meanwhile, the whole Arctic sea ice exhibits larger short-term earlier trends during two periods (1982–1987 and 1999–2001) from Fig. 5o. For Arctic sea ice regions, the change trends of MOD show complex variation characteristics in different study periods, and the trend variations of MOD display noticeable differences in different regions. The trend variations of MOD in peripheral sea ice regions, e.g., Sea of Okhotsk (Fig. 5a), Bering Sea (Fig. 5b), and St. Lawrence Gulf (Fig. 5d), are insignificant for different study periods, while the trend of MOD for other sub-regions are significantly positive at time frames of 15–39 years when the study period is before 1985. In addition, we can find the short-term (9–15 years) trend of MOD is significantly negative with a larger change rate (>2 d/a) for Kara Sea (1980–1985) (Fig. 5h), Laptev Sea (1982–1985) (Fig. 5i), East Siberian Sea (1979–1985 and 1997–2000) (Fig. 5j), Chukchi Sea (1983–1986) (Fig. 5k), and Beaufort Sea (1983–1986) (Fig. 5l).
From the multi-time scale analysis of MOD trend, it indicates that the fluctuation of the MOD trend varies with time periods. This also explains the slight difference in the MOD trends in Section 4.2 with those of (Bliss and Anderson, 2014b). As can be seen in Fig. 5, temporal variability and statistical significance of MOD trend exhibit large interannual differences with different time windows for most regions in the Arctic. Thus, when analyzing the long-term changes of MOD, different time scales may get distinctly different results.
The M-K mutation analysis method was adopted to investigate the abrupt change of MOD. Figure 6 presents the curves of the UF and UB obtained from the M-K mutation test method for the Arctic and its sub-regions from 1979 to 2017, and the 0.05 confidence level was used. If the intersection of the UF and UB curves is between the confidence interval, the time of the intersection corresponds to the time of mutation. It should be noted that the mutation points detected by the M-K method do not mean the minimum or maximum value of a time series data. They represent the trends of the time series data before and after the mutation point are significantly different (Mann, 1945; Kendall, 1948).
First, we analyzed the regional changes in MOD for Arctic sub-regions, see Figs 6a–n. For the Arctic sub-regions, the regions with more than one intersection between UF and UB curves are mainly located in the Bering Sea, Baffin Bay and St. Lawrence Gulf, implying these regions are more prone to abnormal change. Drobot and Anderson (2001a) also found that abnormally late (early) snowmelt onset occurs near the Baffin Bay using principal component analysis. As shown in Fig. 6a, the MOD in the Sea of Okhotsk exhibits a short period of increasing trend (during 1979–1980) and a relatively long period of decreasing trend (after 1981). An abrupt change of MOD in the Sea of Okhotsk is detected in 1982 (significant at the 0.05 level). The MOD in the UF and UB statistic curves has an intersection in 1982 within the confidence interval. The MOD over the Sea of Okhotsk can be separated into two periods. The UF curve of MOD shows an increase before 1982, and then a decrease with fluctuations during 1982–2017, while these trends do not exceed the 95% confidence level. For other peripheral sea ice regions, there are a few intersections of UF and UB curves within the two confidence lines, which indicates that extreme MOD occurs more frequently. The abrupt changes in the Bering Sea occur in 1983, 1986 and 2015 (Fig. 6b). The mutations of the Hudson Bay occur in 1997 and 2013 (Fig. 6c), and those of the Baffin Bay take place in 1988 and 1993, see Fig. 6e. Based on the M-K mutation analysis and change trend analysis, it can be drawn that there is no significant mean annual MOD variation in the St. Lawrence Gulf and Greenland Sea, see Figs 6d and f, respectively. The more mutation points that occur in the St. Lawrence Gulf may be related to the air temperature. The air temperatures are generally warmer in these southerly locations than in other regions (Bliss and Anderson, 2014a). The onset of melting is a sensitive indicator of the near-surface atmospheric conditions, particularly the air temperature.
In contrast to the peripheral sea ice regions, there is only one intersection of UF and UB in 2005 for the Barents Sea (Fig. 6g), 1994 for Chukchi Sea (Fig. 6k), 1993 for Beaufort Sea (Fig. 6l), and 1984 for Canadian Archipelago (Fig. 6m). Additionally, there is a crossing point between the UF and UB curves in 2002 for Kara Sea (Fig. 6h), and 1992 for central Arctic (Fig. 6n). However, both Kara Sea and central Arctic do not have abrupt changes during the study period, because these crossing points do not pass the reliability test of significant level α=0.05.
Overall, as seen in the UF curve, the value of UF statistic is less than 0 for all the study periods except in 1984 (see Fig. 6o), indicating the MOD increases slightly from 1983 to 1984. From the mid 1980s, the MOD displays an earlier trend, and a significant earlier tendency starts at the end of the 1980s. The results of M-K tests show the value of UF exceeds the critical value of −1.96 since 1989, implying that the MOD experiences a significant earlier trend. Further analysis reveals that there is an intersection of UF and UB curves in 1993 in the whole Arctic. However, since the crossing point does not pass the reliability test of significant level α=0.05, it indicates that there are no abrupt changes for MOD over the Arctic sea ice.
As discussed above, the MOD at different Arctic sub-regions displays different mutation points. An extreme MOD variation year in a region may not appear to be an abnormal year of other regions. For example, the mutations of the Sea of the Okhotsk, Bering Sea, Hudson Bay, Baffin Bay, Barents Sea, Chukchi Sea, and Canadian Archipelago do not affect the mutation of the whole Arctic. Therefore, the regional mutation analysis of MOD may be indicative of the close association between MOD variation and the local atmospheric activity (such as cyclone) during the Arctic spring, rather than interaction between regions (Bi et al., 2019; Markus et al., 2009). Bliss and Anderson (2014a) analyzed the effect of cyclonic activity on the MOD and indicated that these small-scale events need to be further investigated to determine how it contributes to ongoing changes in sea ice. Besides, for the entire Arctic, there are no mutation points of MOD occur in the years of obvious sea ice extent minimum such as 2007 and 2012, and thus, the MOD may have no direct contribution toward the retreat of sea ice, which is also mentioned in Markus et al. (2009).
The Pearson correlations between the MOD of Arctic sub-regions and monthly AO can be seen in Fig. 7, which reveals the influence of monthly AO on the MOD of Arctic sub-regions is not significant and shows a weak correlation. The weak correlations between monthly AO and MOD suggest that the MOD over Arctic sea ice is not affected by the short-term effects of AO. Table 1 illustrates the correlations of seasonal AO and MOD, which indicates that the winter AO has a more significant impact on MOD in portions of Arctic sub-regions. However, snowmelt usually occurs from March to June in the Arctic, implying the lag effect of AO. Baffin Bay and Greenland Sea show positive correlations, whereas other regions (e.g., Sea of Okhotsk, Laptev Sea, East Siberian Sea, and Chukchi Sea) exhibit negative correlations indicating the stronger AO and the earlier MOD. Besides, Drobot and Anderson (2001a) found Baffin Bay and East Siberian Sea have the opposite relationship with AO, which is in good agreement with our results. Some studies pointed out positive (negative) phases of the AO are associated with abnormally warm (cool) surface air temperatures along the Siberian coast (Thompson and Wallace, 1998), which may be the reasons for the positive correlation between AO and MOD. The warmer air temperature contributes to the earlier snowmelt. However, Thompson and Wallace (1998) also indicated abnormally cool (warm) surface air temperatures near Baffin Bay. This may be the reason why Baffin and Greenland exhibit positive correlations, and the cooler air temperature lead to the later MOD over the Arctic sea ice. The weak correlation indicates MOD is not only affected by AO, but may be also associated with other factors, such as air temperature, ocean heat flux, wind fields and clouds.
| Region | Jan.–Mar. | Apr.–Jun. | Jul.–Sept. | |||
| AO & MOD | AO & MOD | AO & MOD | ||||
| Sea of Okhotsk | −0.43* | −0.08 | −0.06 | |||
| Bering Sea | −0.11 | −0.01 | 0.01 | |||
| Hudson Bay | 0.11 | 0.03 | 0.12 | |||
| St. Lawrence Gulf | −0.14 | −0.14 | −0.38* | |||
| Baffin Bay | 0.44* | 0.14 | −0.01 | |||
| Greenland Sea | 0.45* | 0.17 | 0.00 | |||
| Barents Sea | −0.28 | −0.13 | −0.01 | |||
| Kara Sea | −0.31 | −0.06 | 0.03 | |||
| Laptev Sea | −0.34* | −0.25 | 0.17 | |||
| East Siberian Sea | −0.39* | −0.21 | 0.09 | |||
| Chukchi Sea | −0.36* | −0.05 | −0.03 | |||
| Beaufort Sea | −0.15 | 0.04 | −0.02 | |||
| Canadian Archipelago | −0.13 | 0.06 | −0.07 | |||
| Central Arctic | −0.29 | −0.17 | 0.01 | |||
| Arctic | −0.24 | −0.10 | 0.05 | |||
| Note: * indicates statistically significant correlation at a 95% confidence level. | ||||||
DownLoad:
CSV
A good knowledge of the spatio-temporal variation of MOD over Arctic sea ice is of keen interest and significance for Arctic climate research. This study systematically investigated the spatio-temporal variation characteristics of annual MOD over the Arctic sea ice from 1979 to 2017, by using a set of mathematical and statistical methods (i.e., Sen’s slope estimator, M-K mutation test, and correlation analysis). The main conclusions are summarized as follows:
(1) The spatial patterns of mean annual MOD have noticeable regional differences over the Arctic, as well as interannual trend variations. The mean annual MOD over Arctic sea ice is related to the latitude, and earlier MOD is generally located in the peripheral and southern regions of sea ice. Moreover, the MOD is trending universally earlier for the majority of Arctic sea ice from 1979 to 2017, while the magnitude of trend varies from region to region. In addition, the regions in the middle and high latitudes (Barents Sea, Kara Sea, Laptev Sea, East Siberian Sea, Chukchi Sea, Beaufort Sea, Canadian Archipelago, and central Arctic) dominate the variation of MOD in the Arctic.
(2) The mutations in temporal trends of annual MOD exhibit a complex and dynamic behavior during the past 39 years. Although some significant mutations have occurred in different sub-regions, there is no direct connection between them. Therefore, it is inferred that the regional extreme variation of MOD is likely related to their local atmospheric conditions and activities. Besides, the mutation in sub-regions does not have a significant impact on the overall Arctic mutation.
(3) The response of MOD to AO varies greatly in different regions. The seasonal AO has more influence on MOD than monthly AO. The MOD has a negative correlation with winter AO in most Arctic sub-regions except for Baffin Bay and Greenland Sea. The MOD may be affected by many factors (e.g., air temperature, and ocean heat flux) leading to the weak correlation between MOD and AO, which is worthy of further investigation in the future.
Our results show that the effects of the mutations and variation on the temporal trends of MOD over Arctic sea ice are complex and dynamic during the past 39 years. In summary, while these preliminary results look promising, this study does not fully explore the physical mechanism that affects MOD variation from the aspect of large-scale atmospheric circulation. Further analysis of the MOD patterns in each region and their relationship to the atmospheric conditions is necessary to thoroughly explain the causes of the variability in MOD at a large scale. In the future, the effect of atmospheric circulation changes on the MOD and the length of the melt season will be further investigated (Ballinger et al., 2019; Huang et al., 2019; Stroeve et al., 2014), which will help us to get a better and more complete understanding of the physical processes controlling the evolution of sea ice melt and break-up patterns, and in turn improve the accuracy of global climate model predictions.
We thank NSIDC and NOAA for making the MOD and AO index data publicly available respectively. The MOD data used in the study are obtained from NSIDC (https://nsidc.org/data/nsidc-0105) and the AO index data are from NOAA National Centers for Environmental Information (http://ncdc.noaa.gov/teleconnections/ao/).
| [1] |
Agrawal Y C, Terray E A, Donelan M A, et al. 1992. Enhanced dissipation of kinetic energy beneath surface waves. Nature, 359(6392): 219–220. doi: 10.1038/359219a0
|
| [2] |
Blumberg A F, Mellor G L. 1987. A description of a three-dimensional coastal ocean circulation model. In: Heaps N S, ed. Three-Dimensional Coastal Ocean Models. Washington: American Geophysical Union, 1–16. doi: 10.1029/CO004p0001
|
| [3] |
Booij N, Ris R C, Holthuijsen L H. 1999. A third-generation wave model for coastal regions: 1. Model description and validation. Journal of Geophysical Research, 104(C4): 7649–7666. doi: 10.1029/98JC02622
|
| [4] |
Bye J A T. 1988. The coupling of wave drift and wind velocity profiles. Journal of Marine Research, 46(3): 457–472. doi: 10.1357/002224088785113559
|
| [5] |
Cao Yu, Deng Zengan, Wang Chenxu. 2019. Impacts of surface gravity waves on summer ocean dynamics in Bohai Sea. Estuarine, Coastal and Shelf Science, 230: 106443,
|
| [6] |
Charnock H. 1955. Wind stress on a water surface. Quarterly Journal of the Royal Meteorological Society, 81(350): 639–640. doi: 10.1002/qj.49708135027
|
| [7] |
Craig P D, Banner M L. 1994. Modeling wave-enhanced turbulence in the ocean surface layer. Journal of Physical Oceanography, 24(12): 2546–2559. doi: 10.1175/1520-0485(1994)024<2546:MWETIT>2.0.CO;2
|
| [8] |
Deng Zengan, Zhao Yifei. 2020. Impact of tidal mixing on water mass properties and circulation in the Bohai Sea: a typhoon case. Journal of Marine Systems, 206: 103338. doi: 10.1016/j.jmarsys.2020.103338
|
| [9] |
Denman K L, Miyake M. 1973. Upper layer modification at ocean station Papa: observations and simulation. Journal of Physical Oceanography, 3(2): 185–196. doi: 10.1175/1520-0485(1973)003<0185:ULMAOS>2.0.CO;2
|
| [10] |
Donelan M A. 1990. Air-sea interaction. In: Le Mehaute B, Hanes D M, eds. The Sea: Ocean Engineering Science. New York: Wiley-Interscience Press, 239–292
|
| [11] |
Drennan W M, Kahma K K, Terray E A, et al. 1992. Observations of the enhancement of kinetic energy dissipation beneath breaking wind waves. In: Banner M L, Grimshaw R H J, eds. Breaking Waves: IUTAM Symposium Sydney, Australia 1991. Berlin, Heidelberg: Springer, 95–101
|
| [12] |
Drennan W M, Taylor P K, Yelland M J. 2005. Parameterizing the sea surface roughness. Journal of Physical Oceanography, 35(5): 835–848. doi: 10.1175/JPO2704.1
|
| [13] |
Feddersen F. 2012. Observations of the surf-zone turbulent dissipation rate. Journal of Physical Oceanography, 42(3): 386–399. doi: 10.1175/JPO-D-11-082.1
|
| [14] |
Gao Shuanzhu, Meng Zhiyong, Yang Guiming. 2009. Study on the predictability of the recurvature of Typhoon Matsa (0509) in Bohai. Meteorological Monthly, 35(2): 8–14
|
| [15] |
Gaspar P. 1988. Modeling the seasonal cycle of the upper ocean. Journal of Physical Oceanography, 18(2): 161–180. doi: 10.1175/1520-0485(1988)018<0161:MTSCOT>2.0.CO;2
|
| [16] |
Gemmrich J R, Farmer D M. 1999. Near-surface turbulence and thermal structure in a wind-driven sea. Journal of Physical Oceanography, 29(9): 480–499. doi: 10.1175/1520-0485(1999)029<0480:NSTATS>2.0.CO;2
|
| [17] |
Gerbi G P, Trowbridge J H, Terray E A, et al. 2009. Observations of turbulence in the ocean surface boundary layer: energetics and transport. Journal of Physical Oceanography, 39(5): 1077–1096. doi: 10.1175/2008JPO4044.1
|
| [18] |
Ginis I. 2002. Tropical cyclone-ocean interactions. In: Perrie W, ed. Atmosphere-Ocean Interactions. Southampton: WIT Press
|
| [19] |
Han Guijun, Li Wei, He Zhongjie, et al. 2006. Assimilated tidal results of tide gauge and TOPEX/POSEIDON data over the China seas using a variational adjoint approach with an onlinear numerical model. Advances in Atmospheric Sciences, 23(3): 449–460
|
| [20] |
Janssen P A E M. 2001. Reply. Journal of Physical Oceanography, 31(8): 2537–2544. doi: 10.1175/1520-0485(2001)031<2537:R>2.0.CO;2
|
| [21] |
Jones N L, Monismith S G. 2008. Modeling the influence of wave-enhanced turbulence in a shallow tide- and wind-driven water column. Journal of Geophysical Research, 113(C3): C03009. doi: 10.1029/2007JC004246
|
| [22] |
Kraus E B, Turner J S. 1967. A one-dimensional model of the seasonal thermocline II. The general theory and its consequences. Tellus, 19(1): 98–106. doi: 10.3402/tellusa.v19i1.9753
|
| [23] |
Liu Jiajun, Zhang Lifeng, Guan Jiping, et al. 2010. Application of ensemble prediction products to the forecast of Typhoon Masta. Meteorological Monthly, 36(5): 21–31
|
| [24] |
Mellor G L. 2001. One-dimensional, ocean surface layer modeling: a problem and a solution. Journal of Physical Oceanography, 31(3): 790–809. doi: 10.1175/1520-0485(2001)031<0790:ODOSLM>2.0.CO;2
|
| [25] |
Mellor G, Blumberg A. 2004. Wave breaking and ocean surface layer thermal response. Journal of Physical Oceanography, 34(3): 693–698. doi: 10.1175/2517.1
|
| [26] |
Mellor G L, Durbin P A. 1975. The structure and dynamics of the ocean surface mixed layer. Journal of Physical Oceanography, 5(4): 718–728. doi: 10.1175/1520-0485(1975)005<0718:TSADOT>2.0.CO;2
|
| [27] |
Mellor G L, Yamada T. 1974. A hierarchy of turbulence closure models for planetary boundary layers. Journal of the Atmospheric Sciences, 31(7): 1791–1806. doi: 10.1175/1520-0469(1974)031<1791:AHOTCM>2.0.CO;2
|
| [28] |
Mellor G L, Yamada T. 1982. Development of a turbulence closure model for geophysical fluid problems. Reviews of Geophysics, 20(4): 851–875. doi: 10.1029/RG020i004p00851
|
| [29] |
Meng Qingjun, Li Peiliang, Zhai Fangguo, et al. 2020. The vertical mixing induced by winds and tides over the Yellow Sea in summer: a numerical study in 2012. Ocean Dynamics, 70(7): 847–861. doi: 10.1007/s10236-020-01368-2
|
| [30] |
Phillips O M. 1977. Dynamtics of the Upper Ocean. Cambridgeshire: Cambridge University Press, 366
|
| [31] |
Polton J A, Lewis D M, Belcher S E. 2003. The Role of Wave-Induced Coriolis-Stokes Forcing on the Wind-Driven Mixed Layer. Journal of Physical Oceanography, 35(4): 444–457
|
| [32] |
Ren Huiru, Li Guosheng, Cui Linlin, et al. 2016. Wave climate changes in Bohai Sea related to the East Asian circulation oscillations over the last sixty years. Climatic and Environmental Research, 21(4): 490–502. doi: 10.3878/j.issn.1006-9585.2016.15281
|
| [33] |
Scully M E, Trowbridge J H, Fisher A W. 2016. Observations of the transfer of energy and momentum to the oceanic surface boundary layer beneath breaking waves. Journal of Physical Oceanography, 46(6): 1823–1837. doi: 10.1175/JPO-D-15-0165.1
|
| [34] |
Shao Jinchao, Zhao Hui, Shen Chunyan, et al. 2015. Influence of typhoon Matsa on phytoplankton Chlorophyll-a in the Northwest Pacific Ocean offshore and alongshore. Journal of Guangdong Ocean University, 35(4): 67–74
|
| [35] |
Smith S D, Anderson R J, Oost W A, et al. 1992. Sea surface wind stress and drag coefficients: the hexos results. Boundary-Layer Meteorology, 60(1−2): 109–142. doi: 10.1007/BF00122064
|
| [36] |
Song Wenpeng. 2009. The analysis of the structure of T-S and the current characteristics in Baohai Sea during winter and summer (in Chinese)[dissertation]. Qingdao: Ocean University of China
|
| [37] |
Song Yaoling. 2012. Effect of activity characteristics of Typhoon Matsa on aviation meteorological observation and support. Meteorological, Hydrological and Marine Instruments (in Chinese), 29(1): 99–104,
|
| [38] |
Stacey M W. 1999. Simulation of the wind-forced near-surface circulation in Knight Inlet: a parameterization of the roughness length. Journal of Physical Oceanography, 29(6): 1363–1367. doi: 10.1175/1520-0485(1999)029<1363:SOTWFN>2.0.CO;2
|
| [39] |
Sun Qun, Guan Changlong, Song Jinbao. 2006. Effect of wave breaking on turbulent energy budgets in ocean surface mixed layer. Oceanologia et Limnologia Sinica, 37(1): 69–74. doi: 10.3321/j.issn:0029-814X.2006.01.011
|
| [40] |
Terray E A, Donelan M A, Agrawal Y C, et al. 1996. Estimates of kinetic energy dissipation under breaking waves. Journal of Physical Oceanography, 26(5): 792–807. doi: 10.1175/1520-0485(1996)026<0792:EOKEDU>2.0.CO;2
|
| [41] |
Terray E A, Donelan M A, Agrawal Y C, et al. 1997. Reply. Journal of Physical Oceanography, 27(10): 2308–2309. doi: 10.1175/1520-0485(1997)027<2308:R>2.0.CO;2
|
| [42] |
Terray E A, Drennan W M, Donelan M A. 1999. The vertical structure of shear and dissipation in the ocean surface layer. In: Proceedings of Symposium on the Wind-driven Air-Sea Interface—Electromagnetic and Acoustic Sensing, Wave Dynamics, and Turbulent Fluxes. Sydney: University of New South Wales, 239–245
|
| [43] |
Toba Y, Smith S D, Ebuchi N. 2001. Historical drag expressions. In: Jones I S F, Toba Y, eds. Wind Stress over the Ocean. New York: Cambridge University Press
|
| [44] |
Wunsch C, Ferrari R. 2004. Vertical mixing, energy, and the general circulation of the oceans. Annual Review of Fluid Mechanics, 36: 281–314. doi: 10.1146/annurev.fluid.36.050802.122121
|
| [45] |
Yablonsky R M, Ginis I. 2008. Improving the ocean initialization of coupled hurricane–ocean models using feature-based data assimilation. Monthly Weather Review, 136(7): 2592–2607. doi: 10.1175/2007MWR2166.1
|
| [46] |
Yang Yongzeng, Qiao Fangli, Xia Changyong, et al. 2003. Effect of ocean wave momenfum and mixing on upper ocean. Advances in Marine Science, 21(4): 363–368. doi: 10.3969/j.issn.1671-6647.2003.04.001
|
| [47] |
Zhang Xuefeng, Han Guijun, Wu Xinrong, et al. 2011. Effect of surface wave breaking on upper-ocean structure revealed by assimilating sea temperature data. Journal of Tropical Oceanography, 30(5): 48–54. doi: 10.11978/j.issn.1009-5470.2011.05.048
|
| [48] |
Zhao Yifei, Deng Zengan, Yu Ting, et al. 2019. Numerical study on tidal mixing in the Bohai Sea. Marine Geodesy, 42(1): 46–63. doi: 10.1080/01490419.2018.1539055
|
| [49] |
Zhao Xin, Sun Qun. 2013. Influence of reclamation on hydrodynamic environment in Bohai Bay. Advanced Materials Research, 726–731: 3262–3265,
|
| [50] |
Zippel S F, Thomson J, Farquharson G. 2018. Turbulence from breaking surface waves at a river mouth. Journal of Physical Oceanography, 48(2): 435–453. doi: 10.1175/JPO-D-17-0122.1
|
| 1. | Ting Yu, Zengan Deng, Chi Zhang, et al. Detecting the role of Stokes drift under typhoon condition by a fully coupled wave-current model. Frontiers in Marine Science, 2024, 11 doi:10.3389/fmars.2024.1364960 |
Figures(17) / Tables(2)
Supported by:
Beijing Renhe Information Technology Co. Ltd
Menghan Wang, Zengan Deng. On the role of wave breaking in ocean dynamics under typhoon Matsa in the Bohai Sea, China[J]. Acta Oceanologica Sinica, 2022, 41(9): 1-18. doi: 10.1007/s13131-022-1995-3
| Region | Jan.–Mar. | Apr.–Jun. | Jul.–Sept. | |||
| AO & MOD | AO & MOD | AO & MOD | ||||
| Sea of Okhotsk | −0.43* | −0.08 | −0.06 | |||
| Bering Sea | −0.11 | −0.01 | 0.01 | |||
| Hudson Bay | 0.11 | 0.03 | 0.12 | |||
| St. Lawrence Gulf | −0.14 | −0.14 | −0.38* | |||
| Baffin Bay | 0.44* | 0.14 | −0.01 | |||
| Greenland Sea | 0.45* | 0.17 | 0.00 | |||
| Barents Sea | −0.28 | −0.13 | −0.01 | |||
| Kara Sea | −0.31 | −0.06 | 0.03 | |||
| Laptev Sea | −0.34* | −0.25 | 0.17 | |||
| East Siberian Sea | −0.39* | −0.21 | 0.09 | |||
| Chukchi Sea | −0.36* | −0.05 | −0.03 | |||
| Beaufort Sea | −0.15 | 0.04 | −0.02 | |||
| Canadian Archipelago | −0.13 | 0.06 | −0.07 | |||
| Central Arctic | −0.29 | −0.17 | 0.01 | |||
| Arctic | −0.24 | −0.10 | 0.05 | |||
| Note: * indicates statistically significant correlation at a 95% confidence level. | ||||||
DownLoad:
CSV
DownLoad:
