3.2. Forward Analysis: Relationships Between the Moisture Sink and the SIC
In the analysis of moisture sinks, for all seas with two moisture sources, the time series are added (for example, for the Ross Sea we added the IND and SAUS moisture sinks to obtain only one time series). Figure 5
shows the annual cycle of the moisture sink (moisture arriving in the target areas from source), obtained from forward trajectories ((E − P)i10 < 0), over the maximum annual area with sea ice concentration equal or greater than 0.1, which occurs in September (blue area in Figure 2
). To facilitate the interpretation of the results, the moisture sink values are multiplied by −1. The moisture sink annual cycles have different patterns from those of the moisture sources (Figure 3
), where the oceans have maxima from July to October. For the moisture sinks, the annual cycles are relatively flat over much of the year, presenting weak peaks as a function of the sea. The Weddell, King, and East Antarctica seas have a weak maximum from February to April (Figure 5
a–c), while in the Ross Sea (Figure 5
e) it is displaced to April–June, and in the Amundsen Sea to August (Figure 5
d). On the other hand, the minimum amount of the moisture sink occurs from September to December (Figure 5
). In terms of the average, the moisture sink is higher and lower, respectively, in the Amundsen and Weddell seas. These annual cycle shapes of the moisture sink, with the maximum at the end-summer/autumn and the minimum in spring-begin/summer, are similar to those in the Arctic [34
]. As moisture sinks over the target areas are mainly associated with precipitation, we can compare it with precipitation climatology over the Southern Ocean. In this context, the slight maximum of the moisture sink in the austral autumn in most of the austral seas (Figure 5
) agrees with the precipitation climatology southward of 67.5° S [57
]. Considering more regional comparisons, the relatively weak seasonality of the moisture sink in East Antarctica (Figure 5
c), with a slight peak in the austral autumn months, is in agreement with observations of rainfall in the Macquarie Island located in this sea [58
]. The maximum of the moisture sink between winter and spring in the Amundsen Sea (Figure 5
d) is also observed in the rainfall at the Rothera station located in the Antarctica Peninsula on the border of the Amundsen Sea [59
The moisture sink presents strong interannual variability, as measured by the standard deviation of the monthly values, in most of the seas (Figure 5
). Moreover, the interannual variability of the moisture sink is higher over all months of the year in the Weddell, King and East Antartic and weaker in the Amundsen and Ross seas.
Comparing the annual cycles of the moisture sink with those of the SIC, in general, there is an inverse pattern: the maximum SIC occurs when the moisture sink is weaker (Figure 5
). This behavior is reflected in the negative and high values (from −0.54 to −0.97) of correlation coefficients (top of Figure 5
) in most of the seas, except the Amundsen Sea (0.22). In other words, the moisture seems to make a negative contribution to the SIC over the blue area displayed in Figure 2
. This result is similar to [34
] for the Arctic region and the case study performed by [23
] for the austral seas. Figure 5
further shows that the SIC in the Weddell Sea has stronger interannual variability (as shown by standard deviation bars) from January-June, and equally large variability is observed in the Amundsen Sea throughout the year. For the Weddell Sea, the higher interannual variability of the SIC is observed in the period of the greater moisture sink, while in the Amundsen Sea it occurs in the periods of both higher and lower moisture sink. The King and East Antarctic seas register greater SIC interannual variability in the months of minimum moisture supply (July–December). A more complicated pattern is noted in the Ross and Amundsen seas where the SIC interannual variability is greater, respectively, from May-October and July–December.
For a better understanding the relationship between the SIC and the moisture sink, the Pearson correlation coefficient is computed considering different time lags (months), i.e., the monthly mean moisture sink time series are fixed in time, and the SIC are displaced from 1 until 6 months ahead (Table 1
). The correlations are calculated for the time series of the anomaly, i.e., the mean annual cycle is subtracted from the time series. The p
-value (α = 0.05) is also calculated as an indication of the statistical significance of the correlations. According to Table 1
, the higher negative and statistically significant correlations occur at a 2-month lag, except for the 4-month lag in the Ross Sea.
For the time series including all months of the year, Figure 6
presents the correlation maps (correlations between the anomalous moisture sinks and the SIC in each sea, which were merged to provide only one map) for the lags from 0 to 5 months. Only the statistically significant values are shown. For this reason, in general, there is no information near the Antarctic coastline. As expected from Table 1
, the negative correlations are dominant in almost all austral seas, and they are higher (in absolute value) in the northern sector of the sea ice edges. Considering the time lags, negative correlations increase (in absolute value) from lag-0 to lag-2, decreasing afterwards. At lag-2 the more intense negative correlations expand in the north-south direction covering a larger area compared to the other lags. Figure 6
also reveals a weaker but persistent positive correlation, from lag-0 to lag-2, between the moisture sink and the SIC over the northern limit of the Ross Sea.
The trends of the monthly moisture sink in each austral sea and the SIC are also evaluated. For this reason, the slope of the time series is computed using Sen’s formulation [60
], while its 95% statistical confidence level is provided by the nonparametric test from Man Kendall [61
]. The trends are listed in Table 2
, where statistically significant values are highlighted in bold. The monthly moisture sink over the Weddell and King seas presents a positive trend, which is statistically significant only in the Weddell Sea. On the other seas, the trend is negative, and they are not statistically significant. On the other hand, there is a positive and statistically significant trend of the SIC in most austral seas, except in the Amundsen Sea, which presents a negative trend in agreement with [13
]. The positive trend in the SIC is also documented in several studies [64
3.3. Extremes in the Moisture Sink for the Weddell Sea and the SIC
For the Weddell Sea, Figure 7
a presents the area occupied for SIC values exceeding 0.4 for four different months: March, April, June, and September since, according to Figure 5
a, the SIC has its minimum and maximum, respectively, in March and September. Higher SIC values occupy larger and smaller areas, respectively, in September and March (Figure 7
According to [29
], much of the SIC and its variability in the Weddell Sea are modulated by changes in the depth and location of the ASL, which is the deepest of the three pressure centers observed around Antarctica (Figure 7
b). The ASL is located in the Pacific sector of the Southern Ocean, which includes the Ross, Amundsen, and Bellingshausen seas, over the latitude band 60°–70° S. Over this region, Figure 7
c indicates that ASL and northerly winds are stronger in September compared to March, while in the Weddell Sea the presence of anomalous low pressure limited to the north by anomalous high pressure is also registered.
Earlier studies focusing on Weddell Sea [20
] indicated that a higher moisture sink (precipitation that reaches the surface) leads to stronger ocean stratification, reducing the heat transport from the deep ocean to the surface with a consequent increase in the sea ice. On the other hand, [12
] have shown that the moisture sink may be associated with northerly winds transporting warm air from lower latitudes to Antarctica, and, consequently, reducing the sea ice concentration and coverage. Therefore, the knowledge of the forcings associated with changes in the SIC over the Weddell Sea helps us to understand the sea ice feedback on the climate, with consequent impacts on oceanic water and circulation. In this context, for the Weddell Sea, we investigated the atmospheric patterns associated with low (P20) and high (P80) values of the moisture sink (or precipitation) affecting the SIC.
For the Weddell Sea, Figure 8
presents the time series of the anomaly of the moisture sink together with the annual frequency and intensity of the extreme moisture sink events (P80 and P20). For a moisture sink with an annual mean value of the ~1.4 mm/day (Figure 5
a), the time series of the monthly anomaly (from −0.22 to +0.35 mm/day) indicates strong variability (Figure 8
a). In absolute values, the positive anomalies are greater than the negative ones. The frequency of events higher than P80 is increasing with time, while it is decreasing for events lower than P20 (Figure 8
b,d). Only the negative linear trend has statistical significance (α = 0.05). In terms of intensity (Figure 8
c,e), events higher than P80 are becoming more intense, and the trend is statistically significant (Figure 8
c), while there is no statistical significance for the weakening of the events lower than P20.
We selected the events (months) identified with the percentile method to perform composites of the SIC and atmospheric variables in order to analyze the effect of the moisture sink on the SIC. We also separated the events into two periods: from June to November (maximum SIC), and from December to May (minimum SIC) and considering the lags 0, 1 and 2 months (Figure 9
). Events exceeding P80 contribute to the SIC decrease in both periods of the year, while the events lower than P20 are associated with an increase in the SIC (Figure 9
Here we present the possible relationships between the SAM and the anomalous moisture sink by calculating the frequency of months with P20 and P80 occurring in each phase (Table 3
). P20 months usually occur during the negative SAM, with 65% (78%) of events in the period from December–May (June–November). P80 moisture sink events usually occur in the SAM positive phase, reaching 79% (64%) from December-May (June–November).
For a more detailed analysis of the influence of the moisture sink on the SIC, the month of the year that presented the highest frequency of P80 and P20 is selected. September and April, with 10 events each, represent the months with the highest occurrences, respectively, of P80 and P20. Figure 10
confirms the results of Figure 6
and Figure 9
, i.e., the increase (decrease) in the moisture sink is associated with the SIC decrease (increase). In addition, the reduction of the SIC is considerable and occupies a larger area two months after the month of a high moisture sink (P80) (Figure 10
). A quasi opposite change in the SIC occurs during a low moisture sink (P20).
indicated that in most of the austral seas the periods with a higher moisture sink coincide with a minimum in the SIC. For the Weddell Sea, larger interannual variability of SIC, as measured by the standard deviation (Figure 6
a), occurs when the moisture sink is greater. In order to understand this question, the composites of the atmospheric variables and their anomalies (differences between composites and the climatology for the period 1980–2015) are calculated.
In September climatology, the southerly winds flow from the Weddell Sea to the midlatitudes (Figure 7
b). However, for periods of high moisture sink (P80) the anomaly presents an inverse pattern in the atmospheric circulation, i.e., the air flows from the extratropics to the Weddell Sea (Figure 11
). This anomalous circulation acts by transporting warm air from midlatitudes to the Weddell Sea, which negatively affects the SIC, even in the presence of extreme precipitation. As a result of the anomalous circulation, a warm (cold) air bubble develops over the Weddell Sea (to the west), being more intense in September (lag-0). In the midlatitudes, the anomalous sea level pressure field shows strong low (ASL intensified) and high pressures centers, respectively, westward and eastward of the Weddell Sea. These changes in the atmospheric circulation are more intense in September (lag-0 month), weakening in the following two months (Figure 11
) when the SIC is decreasing (Figure 10
a–c). In addition, in Figure 11
d–f an anomalous convergence at low levels (1000 hPa) over the Weddell Sea is noted, indicating that the changes in the atmospheric circulation can reduce the movement (drift) of the sea ice, as mentioned by [10
], and/or drive ice toward the south with the warm air advection [12
]. This hinders the occurrence of the polynyas (free areas of ice), which are important to generate new sea ice and impact the SIC (later in this section more details about polynyas will be provided).
indicated a preference for high moisture sink events (P80) in the SAM positive phase. However, this phase is observed in only 50% of the months used in the composites in Figure 11
. This explains, at least in part, why the anomalous sea level pressure pattern (Figure 11
d–f) does not exactly resemble the SAM positive phase (i.e., negative anomalies of pressure around Antarctica). However, Figure 11
d highlights a strong negative anomaly of pressure from the Ross Sea to the Antarctica Peninsula, which indicates the ASL intensification. The combination of stronger ASL westward and of the high-pressure anomaly over the Weddell Sea intensifies the anomalous northerly winds that transport warm air from the extratropics to the higher latitudes, with a consequent SIC decrease. A similar circulation pattern was identified by [17
] in September 2016, the first of the 3 months of the highest extreme sea ice retreat since the observational period (1979–present) started, and this was a month with an extremely positive SAM phase (2.33). According to [12
], the decrease in sea ice extension/concentration in the positive SAM phase may be a result of the strengthened circum-Antarctic westerlies, which move warm subsurface water upward in the water column due to Ekman suction. As the sub-superficial water is warmer compared with that on the surface, it contributes to a decrease in sea ice.
For the low moisture sink events (P20, Figure 12
), there is almost a reversal pattern in the anomalies compared with P80 (Figure 11
). In these events, strong negative (positive) temperature anomalies predominate in the Weddell Sea (westward), while a weaker ASL is observed concomitantly with an anomalous low-pressure northward of the Weddell Sea (Figure 12
). The anomalous high pressure over the western Weddell Sea and the low pressure northward generate a horizontal pressure gradient favoring the air to flow from south to north. This anomalous pattern advects cold polar air from the interior of Antarctica to the Weddell Sea. However, this circulation pattern may associate with the katabatic winds and thus explain the low level (at 1000 hPa) wind divergence in the Weddell Sea during the P20 events (Figure 12
b). Katabatic winds force sea ice away from the coast of the Weddell Sea, which causes the formation of coastal polynyas [67
]. On the other hand, polynyas are areas of intense heat loss to the atmosphere [68
], which promote high ice production rates, but this sea ice is advected away, and new polynyas are created (feedback effect) [69
]. Therefore, this process is important to increase the area and concentration of sea ice.
Another interesting aspect is that during P20 events the spatial pattern of the anomalous sea level pressure resembles the SAM negative phase, i.e., there are anomalous high and low pressures, respectively, around Antarctica and over midlatitudes. Indeed, 7 out of the 10 months used in the composites for P20 of Figure 12
occurred during the SAM negative phase. As discussed in the introduction (and also shown by [70
]), the effect of the atmospheric variability modes on sea ice presents regional characteristics, which prevents us from extending the result for the Weddell Sea to all austral seas. In addition, Turner et al. [24
] mention that in some places the SAM positive phase helps to increase the sea ice extension/concentration, while in other places it has the opposite effect.