Along with the amplified warming and dramatic sea ice decline, the Arctic has experienced regionally and seasonally variable moistening of the atmosphere. Based on reanalysis data, this study demonstrates that the regional moistening patterns during the last four decades, 1979–2018, were predominantly shaped by the strong trends in horizontal moisture transport. Our results suggest that the trends in moisture transport were largely driven by changes in atmospheric circulation. Trends in evaporation in the Arctic had a smaller role in shaping the moistening patterns. Both horizontal moisture transport and local evaporation have been affected by the diminishing sea ice cover during the cold seasons from autumn to spring. Increases in evaporation have been restricted to the vicinity of the sea ice margin over a limited period during the local sea ice decline. For the first time we demonstrate that, after the sea ice has disappeared from a region, evaporation over the open sea has had negative trends due to the effect of horizontal moisture transport to suppress evaporation. Near the sea ice margin, the trends in moisture transport and evaporation and the cloud response to those have been circulation dependent. The future moisture and cloud distributions in the Arctic are expected to respond to changes in atmospheric pressure patterns; circulation and moisture transport will also control where and when efficient surface evaporation can occur.
During the recent decades, the Arctic has experienced drastic changes, the most recognized being the amplified warming and dramatic decline in sea ice concentration and thickness (IPCC 2019). The horizontal moisture transport and especially the associated increase in downward longwave radiation have been recognized among factors contributing to these major changes (Kapsch et al. 2013; Park et al. 2015; Graversen and Burtu 2016; Kapsch et al. 2016; Gong et al. 2017; Lee et al. 2017; Yang and Magnusdottir 2017; Dai et al. 2019; Hao et al. 2019). An example of impacts of moisture transport is the spring onset of surface melt on Arctic sea ice, which is predominantly driven by the increased downward longwave radiation associated with increased moisture transport to a region (Maksimovich and Vihma 2012; Persson 2012; Mortin et al. 2016; Cao et al. 2017). Furthermore, years with anomalously large poleward moisture transport in the spring have been followed by an anomalously low autumn sea ice concentration (Kapsch et al. 2013; Mortin et al. 2016). On daily time scales, moisture transport largely determines the spatial distributions of water vapor, cloud water, and surface downward longwave radiation in the Arctic (Devasthale et al. 2012; Nygård et al. 2019).
Climate model simulations have demonstrated that horizontal moisture transport responses to an increase in CO2 concentration (Hwang et al. 2011), but the response is strongly seasonal (Singh et al. 2017). In previous studies, which have addressed the Arctic as a whole, the estimates of the net moisture transport trend have ranged from a decreasing, although statistically insignificant, trend across 70°N in 1979–2013 (Dufour et al. 2016) to a statistically significant, positive trend across 60°N since 1959 (Villamil-Otero et al. 2018). In winter, there has been a significant increase in the number of intensive moisture intrusions (Woods and Caballero 2016), which contribute substantially to the moisture transport to the Arctic (Woods et al. 2013; Liu and Barnes 2015; Baggett et al. 2016).
In general, temporal changes in moisture transport can be attributed to some combination of changes in atmospheric circulation (Woods et al. 2013; Zhang et al. 2013; Vázquez et al. 2016; Gong and Luo 2017; Yang and Magnusdottir 2017; Kapsch et al. 2018; Zhong et al. 2018; Nygård et al. 2019), evaporation especially in the seasonally varying key source regions (Singh et al. 2017; Gimeno-Sotelo et al. 2018), and removal of water vapor by condensation. Atmospheric circulation patterns are responsible for determining whether the moist air masses from the source areas find their way to the Arctic, and for determining the regional variability of this transport. Midlatitude cyclone activity has a critical role in this (Villamil-Otero et al. 2018). Interactions between atmospheric circulation and moisture transport are in fact acting in two directions. For example, in the Barents and Kara Seas, recent studies have identified a positive feedback loop between Ural blocking and Arctic moisture transport; an increased frequency of the Ural blocking enhances poleward moisture transport and sea ice decline, while also regionally reducing the background meridional temperature gradient (Luo et al. 2016; Gong and Luo 2017; Zhong et al. 2018). This regional reduction in the meridional temperature gradient further enhances background conditions favorable for an increased frequency of the Ural blocking.
Here we address the regional and seasonal trends in the horizontal moisture transport, redistributing the atmospheric moisture in the Arctic during the last four decades, in 1979–2018. We demonstrate the dominant role of moisture transport for determining the regional moistening patterns and causing regionally uneven greenhouse effect. We also show that changes in local evaporation in the Arctic, although they previously received much attention (Screen and Simmonds 2010; Bengtsson et al. 2011; Bintanja and Selten 2014; Boisvert and Stroeve 2015; Boisvert et al. 2015; Morrison et al. 2018), have only had a minor role in shaping the moistening patterns. Compared to previous studies (Bintanja and Selten 2014; Dufour et al. 2016; Villamil-Otero et al. 2018), the advantage of our approach is that 1) the role of horizontal moisture transport is not only limited to transport to and from midlatitudes across a certain latitudinal belt (e.g., 70°N) and 2) evaporation in the Arctic is not only counted as a circumpolar mean. Accordingly, we specifically address the regionally varying moisture transport and evaporation within the Arctic.
2. Data and methods
a. Reanalysis data
The study is based on the most modern global reanalysis ERA5 (Copernicus Climate Change Service 2017), produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). In ERA5, a variety of atmospheric observations (e.g., from radiosoundings and satellites) have been assimilated into a numerical weather prediction model, applying a four-dimensional variational data assimilation method. The spectral model resolution of ERA5 is T639, and the horizontal grid of the data used in this study is 0.25° × 0.25°. The amount of vertical model levels is 137. Our analysis is based on the surface level and vertically integrated products of the reanalysis at 6-h intervals. Northward and eastward components of vertically integrated moisture transport are direct output variables of ERA5, calculated for the air column from the surface to the top of the atmosphere. ERA5 is the fifth generation of ECMWF atmospheric reanalyses, published in 2018. According to the producers of ERA5, strengths of ERA5 compared to its predecessor ERA-Interim (Dee et al. 2011) include its much higher spatial resolution, better global balance of precipitation and evaporation, and more consistent sea surface temperature and sea ice. ERA5 is generally warmer than ERA-Interim in summer and autumn over Arctic sea ice, although both of these reanalyses have a warm bias (Wang et al. 2019). ERA5 provides the best estimates of downward radiative fluxes in spring and summer, suggesting more realistic representation of Arctic cloud cover in it (Graham et al. 2019). Based on our own analyses, we found that the imbalance between moisture and cloud water convergence compared to the net precipitation (precipitation minus evaporation) in ERA5 is rather small in the region north of 50°N; the convergence is 6% larger than the net precipitation.
It has been previously reported that different reanalyses mostly agree on the sign of trend in precipitable water, but not in its magnitude (Rinke et al. 2019). In the online supplemental material, the trends in mean sea level pressure (MSLP) and moisture transport in ERA5 are compared to those in ERA-Interim (Dee et al. 2011), the Japanese Meteorological Agency (JMA) 55-Year Reanalysis (JRA-55) (Kobayashi et al. 2015), and radiosonde observations. The comparison indicates that the MSLP and moisture transport trends are consistent in these different datasets, which suggests that our conclusions on these trends are robust. However, uncertainty related to trends in evaporation and total cloud water in ERA5 is expected to be higher. The total cloud water is especially vulnerable to errors, due to possible misrepresentation of cloud-related processes in the reanalysis (Vihma et al. 2014). Nevertheless, the uncertainties of evaporation and total cloud water in ERA5 are assumed to be small enough to allow for an assessment of the directions of their long-term changes.
b. Trend analyses
Linear trends in variables were calculated from the annual seasonal mean values of the variables at each grid point of the reanalyses. The traditional seasonal division into the 3-month periods [winter: December–February (DJF), spring: March–May (MAM), summer: June–August (JJA), and autumn: September–November (SON)] provided optimal groups for characterizing trends in MSLP and moisture transport. The trends were analyzed for three time periods: 1979–2018 (40 years), 1979–98 (20 years), and 1999–2018 (20 years). The 40-yr period indicates the direction of the long-term change, whereas the 20-yr periods give indications whether the trends have changed over the time. The annual average of Arctic sea ice extent was approximately 10% lower during the latter 20-yr period compared to the first 20-yr period. Statistical significance of linear trends was tested with the two-tailed Student’s t test, applying the 90% confidence level, which is commonly applied in studies related to atmospheric circulation and cyclones (Rudeva and Simmonds 2015; Zahn et al. 2018). As the horizontal moisture transport is a vector, with a magnitude and direction, the statistical significance was tested for the meridional and zonal components separately, and the trend was considered as statistically significant if at least one the components had a statistically significant trend. The trend in interannual variations of net horizontal moisture transport (see Fig. S4 in the online supplemental material) was calculated by first determining how much each value of the detrended time series of annual seasonal means deviated from the detrended 40-yr seasonal mean, and then calculating a linear trend based on the absolute values of the annual deviations. Linear correlations between moisture and cloud variables were calculated from their time series of annual seasonal means, without detrending. The correlations were considered statistically significant when the 90% confidence level was reached.
a. Changes in atmospheric circulation and moisture transport
Regional trends of moisture transport, and in particular trends in its direction, are strongly associated with the trends in mean sea level pressure (MSLP). This is because wind direction and speed in the lower troposphere, where most of the atmospheric moisture is located and being transported (Naakka et al. 2019), are largely regulated by MSLP patterns (Fig. 1). The distribution of MSLP in the Arctic has changed during the last four decades (Fig. 2). In Fig. 2, the trends in MSLP are shown with a color shading, and the trends in moisture transport are visualized with trend vectors, which contain information about the change in magnitude and direction of the transport. These trends (Fig. 2) need to be interpreted together with the mean moisture transport (Fig. 1) to perceive whether these trends have actually increased or weakened the moisture transport to a certain direction. Linear trends in MSLP for the 40-yr period (1979–2018) are on the order of ±1 hPa decade−1, and for the latter 20-yr period (1998–2018) on the order of ±3 hPa decade−1; seasonal and regional variability in MSLP trends is large. Regional trends in the magnitude of moisture transport have been drastic and coincident with the changes in the circulation, being at largest more than 10 kg m−1 s−1 decade−1.
In winter (DJF), the 40-yr trend of MSLP is characterized by increasing MSLP in Eurasia and decreasing MSLP in the North Atlantic (Fig. 2a). During the latter 20-yr period, these trends have been stronger and more confined to the western part of Russia and the Barents Sea, and in contrast to the 40-yr period, MSLP in the Pacific sector has increased by 3 hPa decade−1 (Fig. 2e). In the Atlantic sector, these trends have increased the zonal pressure gradient between the North Atlantic and Eurasia, and as a consequence, more moisture is transported over the North Atlantic and Scandinavia toward the pole. The regional trends in the magnitude of net moisture transport in the Arctic are substantial as the 40-yr trends (Fig. S3) correspond to a −6% to +13% change per decade compared to the mean magnitude, and the recent 20-yr trends to −23% to +36%. The trends in the latter 20-yr period have thus been 2–4 times larger than over the entire 40-yr period. Furthermore, interannual variations in the net moisture transport magnitude between the winters have significantly increased over the Kara Sea and near the Ural Mountains (Fig. S4).
In spring (MAM), the largest negative 40-yr trends in MSLP (−1 hPa decade−1) are located in the western part of Russia as well as over the Barents and Kara Seas, and positive (+1 hPa decade−1) in the Bering Sea (Fig. 2b). In the latter 20-yr period, the deepened low pressure over the Barents/Kara/Laptev Seas has strengthened the meridional gradients in MSLP and enhanced the eastward moisture transport along the Eurasian coast. On the eastern side of the negative MSLP trend region, the significantly increased meridional moisture transport has directed more moisture over the Laptev Sea and farther toward the Canadian Archipelago (Figs. 2b,f).
In summer (JJA), regional trends in MSLP are the weakest of all seasons; their magnitude is less than 1 hPa decade−1 (Figs. 2c,g). The 40-yr and latter 20-yr trends share many common features like the increase in MSLP in Greenland, causing a southward trend in moisture transport on its eastern side, and the decrease in MSLP in the Barents and Kara Seas, enhancing the zonal moisture transport along the Russian coast. The main difference is the increase in MSLP during the latter 20-yr period over central and eastern Russia, which has further strengthened the moisture transport along the coast and made the circulation over the Arctic Ocean more cyclonic. The 40-yr summer trends in moisture transport correspond to −4% to +8% change per decade compared the mean magnitude, and the recent 20-yr trends to −16% to +15%.
In autumn (SON), MSLP over Greenland has had a negative trend, which has become stronger during the latter 20-yr period (Figs. 2d,h). The positive trend in MSLP over Scandinavia is seen during both of these periods, whereas increase in MSLP over North America and central Siberia has only occurred during the latter 20-yr period. These changes have significantly strengthened the meridional moisture transport from the North Atlantic all the way to the central Arctic Ocean. At the same time, interannual variations of moisture transport magnitude have become larger in the North Atlantic (Fig. S4). Over Russia, the magnitude of dominantly zonal moisture transport has decreased, but mostly statistically insignificantly.
The trends in moisture transport have thus distinct characteristics in all four seasons, but all these trends are largely driven by changes in atmospheric circulation. The seasonal trends can be briefly summarized as follows: Winter has been characterized by enhanced northward moisture transport in the Atlantic and Pacific sectors, and spring by the increased moisture transport across the Arctic Ocean. In summer, the zonal moisture transport has been intensified along the continental coasts, whereas in autumn the role of the moisture transport from the North Atlantic has increased.
b. Interactions with trends in total column water vapor and local evaporation
The role of atmospheric circulation in distributing the moisture in the Arctic is evident: Positive trends in total column water vapor, which is defined as the total amount of water vapor in an atmospheric column, accompany the wind direction trends along the main moisture transport pathways, especially during the cold seasons: winter, spring, and autumn (Fig. 3). For example, in autumn, the trend toward more northward wind and moisture transport from the North Atlantic, together with increased northward transport from the Pacific sector have been associated with an increase in total column water vapor over most of the Arctic sea areas (Figs. 2d,h and 3d,h). During all the seasons, except in summer, annual means of total column water vapor are strongly correlated (correlation coefficient r > 0.8) with the magnitude of the net moisture transport. Even if the close interaction between these variables is apparent, it is difficult to determine the relative importance of circulation changes that enhance moisture transport versus an increase in moisture available for transport. Such a distinction is outside of the scope of this paper as the focus here is on comparison of the roles of transport driven or enhanced by the large-scale circulation, and local evaporation.
Changes in evaporation have also contributed to the trends in total column water vapor. Comparison of the earlier 20-yr period with the latter one clearly indicates that the area with a strongly positive evaporation trend (Fig. 4) is mostly limited to the marginal ice zone (Fig. 5) during the cold seasons (winter, spring, and autumn). In winter, the positive trend in evaporation has moved from the central Barents Sea (Fig. 4e) to the northern Barents Sea and Kara Sea (Fig. 4i), following the retreat of sea ice (Figs. 5a,e). In autumn, the positive trends in evaporation have occurred on the coastal seas off Russia and North America (Figs. 4d,h,l).
Over the sea areas, the marginal ice zone is practically the only region where the annual means of evaporation and total column water vapor in winter, spring, and autumn correlate positively (r > 0.6) (Fig. 6b; see also Figs. S5b,g); however, the mean evaporation there is small. This positive correlation may be causally related to evaporation, or it may also be a consequence of interactions between horizontal moisture transport and sea ice retreat. Analyses support the latter one, as the correlation between evaporation and moisture transport magnitude is positive in the marginal ice zone, but mostly negative or small in the other sea areas (Fig. 6c and Figs. S5c,h). Furthermore, correlation between moisture transport magnitude and sea ice fraction in the marginal ice zone is negative (not shown). This correlation clearly indicates that in the marginal ice zone, winters, springs, and autumns with strong moisture transport have been associated with a larger sea ice retreat and thus with a warmer surface. However, it remains as an open question whether 1) the sea ice retreat has actually been caused by the increased moisture transport [by decreasing the sea ice growth (Persson et al. 2017) and by the associated southerly winds pushing the ice edge northward (Stroeve and Notz 2018)] or 2) the increased horizontal moisture transport is a consequence of the warmer surface allowing for larger moisture transport due to reduced cooling and condensation. In any case, the sea ice retreat has also allowed for an increase in surface evaporation, which has provided additional moisture available for transport. Hence, both increased transport and evaporation have probably increased total column water vapor in the marginal ice zone.
South of the marginal ice zone, in the open water areas where the mean evaporation is relatively high in the cold seasons, the trend in evaporation is mostly negative, although not everywhere statistically significant (Fig. 4). Here, evaporation correlates negatively with total column water vapor (Fig. 6b and Figs. S5b,g), suggesting that winters, springs, and autumns with enhanced evaporation have occurred when the air has been relatively dry allowing for efficient evaporation. A negative correlation between evaporation and total column water vapor is also seen in summer (Fig. 6g), although the mean evaporation over the sea is then small. Increases in total column water vapor in the open water areas are thus likely associated with circulation and moisture transport changes rather than local evaporation. A clear example of this relationship is seen in the area south of Greenland in winter, where evaporation has a trend of more than +0.3 mm day−1 decade−1 (Fig. 4i), negative correlation between evaporation and total column water vapor is particularly strong (r = −0.8) (Fig. 6b), and also evaporation and moisture transport are anticorrelated (Fig. 6c).
In contrast to the cold seasons, most of summertime evaporation is taking place on land (Vihma et al. 2016). Over land, summers with high evaporation have typically been associated with a high amount of total column water vapor (Fig. 6g) but a low amount of total cloud water (Fig. 6j), suggesting that evaporation is most efficient when the reduction of clouds allows more surface radiative heating. However, regional trends of evaporation in summer (Figs. 4c,k) have not directly been translated into increased total column water vapor (Figs. 3c,g).
We summarize the interactions between moisture transport, surface evaporation, and total column water vapor as follows: Probably both increased moisture transport and local evaporation have increased total column water vapor at the marginal ice zone, whereas evaporation is of limited importance elsewhere. Over the open sea, years with a high amount of total column water vapor have been linked to large moisture transport but relatively low evaporation. In general, large evaporation strengthens the horizontal transport of moisture, whereas large horizontal moisture transport tends to suppress local evaporation by decreasing the humidity difference between the surface and the air above.
c. Trends in relative humidity and cloud water
Long-term changes in atmospheric moisture in the Arctic have also included changes in relative humidity (RH) in the whole air column from the surface to 300 hPa. During the latter 20-yr period, RH at 850 hPa has had positive trends in the regions, where the mean sea level pressure has had a negative trend (Fig. 7), suggesting that increased cyclonic activity and/or weaker high pressure patterns with reduced subsidence have been associated with the increased RH. Averaged over the whole Arctic (from 60°N northward), the 40-yr trend in RH has been slightly negative at all pressure levels, whereas the 20-yr trend has been negative below 950 hPa and above 500 hPa and positive at 950–500 hPa. The magnitude of 40- and 20-yr trends for the whole Arctic has been relatively small (on the order of ±0.5% decade−1), but the largest regional and seasonal trends of RH are on the order of 5% decade−1.
The largest regional trends in total cloud water (Fig. 8) are collocated with the largest trends in total column water vapor (Fig. 3). In the trends of total cloud water (Fig. 8), there is a sharp division between the marginal ice zone and open water area south of it (Fig. 5). In the marginal ice zone, winter, spring, and autumn means of total cloud water correlate positively with moisture transport magnitude and evaporation (Figs. 6d,e and Figs. S5d,e,i,j), suggesting that the positive trend in cloud water (Fig. 8) is in most regions linked to the displacement of the sea ice zone. This is the zone where northward-transported moisture meets the cold sea ice surface, leading to lifting due to the upward-tilting isentropes (Komatsu et al. 2018), cooling (Vihma et al. 2003), and condensation; the lifting does not strictly occur along the isentropes, due to diabatic processes. In the same zone, southward-moving cold and dry air masses from the sea ice trigger high evaporation and convective cloud forming when meeting the open water. On the open water side in the Barents Sea, the total cloud water has had a negative trend during the cold seasons. This is the only area, except regions with orographically induced cloud decay, where annual winter means of net moisture transport magnitude only weakly correlate with total cloud water (Fig. 6d and Figs. S5d,i). The weak correlation is presumably due to flow dependency, because the on-ice and off-ice flows may have compensating effects on cloud water (see section 4).
It is thus evident that long-term changes in atmospheric circulation and moisture transport have also been reflected to the trends in RH and total cloud water, and further to trends in downward longwave radiation at the surface, which are discussed in the online supplemental material.
Our results show that the regional moistening patterns in the Arctic (Rinke et al. 2019) have predominantly been shaped by trends in moisture transport rather than trends in local evaporation. Trends in moisture transport are associated with changes in the atmospheric circulation and have most likely arisen from a combination of inherent variability of atmospheric dynamics (Ding et al. 2017; Gong et al. 2017) and response to the increased amount of anthropogenic greenhouse gases (Hwang et al. 2011; Singh et al. 2017). We have demonstrated that not only local evaporation but also the magnitude of horizontal moisture transport is affected by the diminishing sea ice cover during the cold seasons, from autumn to spring. Retreat of the sea ice from a region causes an evident stepwise increase in evaporation, due to removal of the insulating layer. This increase has been visible in previous studies in which evaporation trend from a single period (Boisvert et al. 2015) has been addressed, or mean evaporation amounts of two periods (Bintanja and Selten 2014; Singh et al. 2017) have been compared. This study is the first to show a negative trend in open water evaporation after the sea ice has disappeared from a region. Hence, the positive trends in evaporation during a local sea ice decay are restricted to the vicinity of sea ice margin over a limited period. Negative trends in wintertime evaporation, due to reduced surface-air specific humidity difference, have previously been reported by Boisvert et al. (2013) in the Kara/Barents Seas, East Greenland Sea, and Baffin Bay regions where there is open water year round. We suggest that the negative trend in evaporation in open water areas, including the newly opened areas, is tightly linked to the influence of moisture transport to suppress the local evaporation (Park et al. 2015; Nygård et al. 2019).
In the vicinity of the marginal ice zone, the trends and impacts of moisture transport and evaporation are dependent on the flow direction. In a schematic figure (Fig. 9), we summarize the current understanding and the new results of flow dependency of the trends in moisture transport and evaporation, and the cloud response to those. When the flow is from the open ocean, the retreat of sea ice (Figs. 9a,c) is not associated with major changes in the open water zone (zone 1 in Fig. 9c), whereas the zone from where the sea ice has been removed (zone 2 in Fig. 9c) experiences higher evaporation due to the warmer surface. Also the moisture transport is larger, as the evaporation and lack of air mass cooling and condensation allow for more water vapor in the air. Low clouds and fog, forming due to cooling downstream of the ice edge, are displaced farther north to the retreated sea ice zone (zone 3 in Fig. 9c). The moisture transport over this retreated sea ice zone (zone 3) (Fig. 9c) is larger than previously in this zone (Fig. 5a), because the air mass has had shorter time and fetch to gradually cool and dry.
Conversely, when the flow is from the sea ice zone (Figs. 9b,d), the retreat of sea ice induces much higher evaporation in the newly opened sea zone, leading to convective cloud formation (Kay and Gettelman 2009) and increased southward moisture transport (Fig. 9d). Originally cold and dry air mass rapidly moistens over the warmer sea surface. Consequently, farther on the open sea, evaporation is not as efficient as near the sea ice edge, but moisture transport is larger (zone 1 in Fig. 9d).
The increased amount of clouds in the vicinity of the marginal ice zone has been in many previous studies attributed solely to increased local evaporation and strong air–sea coupling (Kay and Gettelman 2009; Boisvert et al. 2015; Morrison et al. 2018; Taylor et al. 2018; Morrison et al. 2019), but we emphasize the fundamental role of moisture transport and the flow dependency of the cloud response. An example of the flow dependency is that the moisture transport magnitude presumably increases in the newly opened sea (zone 2 in Figs. 9c,d) both in southward and northward flows, but these flows are most likely linked to opposing effects on cloud water in this zone. Indeed, the weak correlation between cloud water and moisture transport magnitude (Fig. 6d and Fig. S5) suggests that these processes related to cloud formation partly compensate each other.
Generally, an increase of total column water vapor with the increasing temperature in the Arctic has been expected, as a consequence of the Clausius–Clapeyron relation. However, the increase in the Arctic moisture is not only a direct consequence of the temperature increase. On one hand, locally originating increase of surface temperature (e.g., due to decreased sea ice concentration) allows for increases in local evaporation, sensible heat flux, and air temperature, enabling an increase in moisture transport at the location. On the other hand, temperatures in many Arctic subregions are driven by midlatitude circulation (McGraw and Barnes 2020). Possibilities for poleward transport of an air mass that is both warm and dry are very limited in the cold seasons and effective moisture transport from the midlatitudes has to be associated with transport of warm air, as cold air cannot hold and carry moisture. As a consequence, temperature and moisture in the Arctic are highly correlated. However, we want to emphasize that moisture content is not a passive component responding to the temperature changes in the Arctic; a large body of evidence has demonstrated the two-way feedback between atmospheric moisture and temperature (Kapsch et al. 2013; Park et al. 2015; Graversen and Burtu 2016; Kapsch et al. 2016; Gong et al. 2017; Lee et al. 2017; Yang and Magnusdottir 2017; Dai et al. 2019; Hao et al. 2019). If transport of moist and inevitably warm air increases to a region, it is seen as an increase in temperature not only due to the direct effect of transport of dry static energy but also because of the release of the transported latent heat. In addition, temperatures along the moisture transport path are typically further affected by radiative effects of water vapor and clouds (Nygård et al. 2019). Hence, individual impacts of temperature and moisture are difficult to isolate, and the increase in the Arctic moisture is not only a direct consequence of the temperature increase. Furthermore, we have shown, based on our analyses (section 3c), that the often-made assumption of constant relative humidity in the changing climate is not valid at regional and local scales, especially when MSLP and occurrence of certain circulation patterns in a region have changed.
In this paper, we have addressed the regional and seasonal trends in the horizontal moisture transport in the Arctic during the last four decades (1979–2018). We have also compared the roles of moisture transport and local evaporation for the moistening of the Arctic atmosphere. Our main conclusions are the following:
Regional and seasonal trends of moisture transport are strongly associated with the changes in atmospheric circulation.
Long-term changes have most likely arisen from a combination of inherent variability of atmospheric dynamics and response to the increased amount of anthropogenic greenhouse gases. In winter, the changes are characterized by enhanced northward moisture transport in the Atlantic and Pacific sectors, in spring by the increased moisture transport over the Laptev Sea and farther toward the Canadian Archipelago, in summer by the intensified zonal moisture transport along the Eurasian coast, and in autumn by the increased moisture transport from the North Atlantic.
Moisture transport has had a dominant role for determining the regional changes in total column water vapor, whereas the role of local evaporation has been minor.
Increases in evaporation have been restricted to the vicinity of the sea ice margin over a limited period during the local sea ice decline. After the sea ice has disappeared from a region, evaporation over the open sea has had negative trends due to the effect of horizontal moisture transport to suppress evaporation.
At the marginal ice zone, both increased moisture transport and local evaporation have probably increased total column water vapor, whereas evaporation is of limited importance elsewhere. Near the sea ice margin, the trends in moisture transport and evaporation and the cloud response to those have been circulation dependent.
Over the open sea, years with a high amount of total column water vapor have been linked to large moisture transport but relatively low evaporation.
The trends in total cloud water are in many regions collocated with the trends in total column water vapor and moisture transport. Relative humidity, in turn, has increased in the regions, where the mean sea level pressure has had a negative trend.
In the future, there will be more open water area in the Arctic (IPCC 2019) to potentially increase evaporation. However, the sea area where cold, dry air masses can be formed is simultaneously diminishing (Pithan et al. 2018), which may limit the evaporation efficiency. As the moisture transport from the lower latitudes is projected to increase (Bengtsson et al. 2013; Skific and Francis 2013), it is likely that also in the future the atmospheric circulation and horizontal moisture transport patterns continue to distribute the moisture and control where and when efficient evaporation can take place. The regional moistening and cloud trends in the Arctic will probably be very sensitive to future changes in circulation patterns.
The study was supported by the Academy of Finland via projects TODAy (308441) and AFEC (317999). We acknowledge Copernicus and ECMWF for providing the ERA5 and ERA-Interim data, and the Japan Meteorological Agency and NCAR for providing JRA-55 data.