a. Meteorological data

Daily mean sea level pressure (SLP), 2-m temperature, and 24-h accumulated precipitation data were acquired for three stations—Flesland near Bergen (60.29°N, 5.23°E, 48 m MSL), Tromsø (69.65°N, 18.94°E, 100 m MSL), and Oslo-Blindern (59.94°N, 10.72°E, 94 m MSL)—using the eKlima portal of the Norwegian Met Service (http://eklima.met.no). These stations were selected to cover the main meteorological features in southwestern, northern, and southeastern Norway, respectively. Six-hourly meteorological analysis data were retrieved for the period 15 November 2006–5 January 2007 from the European Centre for Medium-Range Weather Forecasts (ECMWF) at a spectral resolution of T799 and 91 vertical levels, and interpolated onto a 1.0° × 1.0° horizontal grid spacing.

b. CHRM model and water vapor tagging method

The High-Resolution Model (HRM) is a hydrostatic limited-area NWP model that had originally been part of the German Weather Service's operational weather forecasting model chain as the Europa Modell (Majewski 1991). Later on, a climate component was added, enabling the model to run seasonal and multiyear integrations (Schär et al. 1999). This so-called climate HRM (CHRM) has been used extensively for climate simulations and process studies in the midlatitudes and was, until recently, in operational use by several national meteorological services. CHRM uses a rotated coordinate system with an Arakawa C grid in the horizontal and hybrid sigma–pressure coordinates in the vertical direction. The bulk microphysics scheme follows Kessler (1969) and Lin et al. (1983), and for convection the Tiedtke (1989) mass-flux scheme is used. A Dickinson (1984) soil model with three layers is used at the surface.

Numerical water vapor tracers have been used by a number of authors in global models to diagnose the sources and transport pathways of water vapor (Joussaume et al. 1984; Koster et al. 1986; Numaguti 1999; Bosilovich and Schubert 2002). CHRM tagging is the first limited-area NWP model capable of advecting water vapor tracers through its own complete secondary water cycle (Sodemann et al. 2009).¹ The tracer water is released either through evaporation from predefined surface areas, or introduced into the domain by advection at the boundaries. The multidimensional positive-definite advection transport algorithm (MPDATA; Smolarkiewicz 2006) is used for advecting both the WV tracers and the total WV field. In this study a consistent advection scheme is used for the tracers and total WV field. This new modification leads to substantially better conservation of cloud water in comparison to the previous CHRM-tagging version, which applied a leapfrog scheme for total WV advection and MPDATA for the tracer advection. In all parameterized subgrid-scale processes that involve phase changes, the transfer rates of tagged moisture from one phase to another are considered proportional to the fraction of tagged water times the respective total transfer rate. Precipitation receives the tag from the vapor it condenses from. Further details on the tagging implementation are available in Sodemann (2006) and Sodemann et al. (2009). In this study a coding error has been corrected in the parameterization of the evapotranspiration of tagged WV from land surfaces, which leads to an increase in tagged land evapotranspiration compared to the previous version (Winschall et al. 2012). A further important new modification of the model is the deactivation of the calculation of modified turbulent transfer coefficients in the presence of clouds, which had led to a violation of mass conservation within clouds. With these modifications, the fraction of WV not assigned to any tracer (see section 2c), which is due to numerical diffusion and numerical inconsistencies, remains below 1%–2% for WV and cloud water throughout the simulation period for the whole domain, and below 5%–10% for local precipitation. For ease of interpretation, all tracer results in this paper have been scaled relative to the total known amounts of WV and precipitation.

c. Simulation setup and tracer experiments

We employ the CHRM WV tracers to follow the moisture evaporating from oceanic source regions to target domains in southern and northern Norway. The simulation domain includes most of the North Atlantic winter storm track region (Fig. 1). ECMWF analysis data have been supplied at the boundaries and for the nudging (see below) continuously by interpolating between the 6-hourly analyses. The CHRM grid spacing is 0.5° × 0.5° in the horizontal, with 40 vertical levels. During the 1.5-month integration starting at 0000 UTC 15 November 2006, a variable nudging of the wind fields has been applied, as described in section 2d.

View Full Size

Calculation domain of the CHRM model (solid black line) and interior model domain (dashed black line). Colors show moisture source tracers released by surface evaporation from every 10° latitude band of ocean area (T1–T6), land (TL, here plotted in white), and from advection through the southern (TS), western (TW), and northeastern (TNE) boundaries reaching from the bottom to top of the model domain. Initial atmospheric tracer (TA) is not displayed. Boxed areas over Scandinavia denote the target domains of northern Norway (green) and southern Norway (red).

Citation: Monthly Weather Review 141, 8; 10.1175/MWR-D-12-00256.1

• Download Figure
• Download figure as PowerPoint slide

Six oceanic source regions for WV tracer evaporation have been defined in the North Atlantic domain, by means of four 10°-wide latitude bands from 30° to 70°N, plus the regions south of 30°N and north of 70°N (Fig. 1, shaded areas). Tracers are numbered from south to north as T1–T6. Evapotranspiration from all land regions is tagged by one additional tracer (TL). Furthermore, at the boundaries, atmospheric inflow is tagged from the south (TS), west (TW), and north and east (TNE). All moisture present in the atmosphere at initialization time has unknown sources and is tracked as an additional tracer labeled TA. The sum of all tracers thus allows us to calculate the fraction of unassigned WV. Two target domains have been defined for analyzing moisture sources for precipitation over western Scandinavia. The northern Norway (NN) domain covers the region 63°–72°N and 8°–26°E (Fig. 1, green box), while southern Norway (SN) is defined here as the region 58°–63°N and 4°–14°E (Fig. 1, red box). This choice has been based on the very different weather conditions often prevailing in these regions during the study period (see section 3a).

d. Tagging spinup

During the spinup of the tagging simulation, the moisture of tracer TA residing in the domain initially with unknown evaporative source regions is continuously replaced by tracer moisture with known origin through evaporation from the surface and advection at the boundaries. We performed a two-stage simulation, consisting of a spinup and an analysis period. The period 15–30 November served as a spinup period to remove as much as possible of the moisture with unknown sources that is present in the domain initially (tracer TA), while the period 1–31 December has been used for further analysis. At the end of the spinup period less than 0.5% of the original TA tracer was left inside the model domain. During the spinup period model winds were nudged to the 6-hourly ECMWF winds interpolated in time with a fairly strong relaxation rate of γ = 10⁻⁴ s⁻¹. This tied the simulation closely to the ECMWF analyses while allowing spinup of the tracer distribution in the model domain. We did not nudge temperature and humidity fields to avoid a too direct interference with the model's hydrological cycle. For the analysis period, the relaxation rate of the wind fields was then lowered to γ = 10⁻⁵ s⁻¹ to allow the model to develop its own internal representation of the water cycle. At the same time, the nudging ensured that the model would not too strongly deviate from the analysis, allowing for a meaningful comparison to station data.

a. Meteorological evolution in Norway during December 2006

An almost continuous sequence of cyclones brought warm and moist air to all of western Scandinavia during December 2006. Precipitation was 100%–200% above average in many regions of Norway, temperatures were the warmest for a December since the beginning of recorded temperatures, and new temperature and precipitation records were set at many weather stations (Meteorologisk Institutt 2007). Similar weather continued into January and February 2007 (Luterbacher et al. 2007), with an extreme precipitation event taking place in central Norway at the end of January (Steensen et al. 2011). This anomalous weather pattern is reflected in a high value of the SLP-based winter season (December–March, DJFM) mean North Atlantic Oscillation (NAO) index of +1.83 (Osborn et al. 1999; updated as described online at http://www.cru.uea.ac.uk/~timo/datapages/naoi.htm). In the following, the regional model simulation is briefly validated by a comparison to the observations at three meteorological stations in Norway.

Flesland (Bergen Airport, 48 m MSL) is a weather station situated along the west coast of Norway, but at some distance from the steepest orographic rise. Given our fairly coarse model resolution, comparison with the simulation results is more meaningful for Flesland than for the better known nearby station Bergen-Florida, which has one of the highest climatological precipitation totals in Norway (2250 mm annual mean and 235 mm December mean, at Flesland these numbers are 1815 mm and 189 mm, respectively), partly due to precipitation enhancement along the steep orographic rise to 600-m altitude within a very short distance from the weather station. Due to the frequent landfall of cyclones at the end of the North Atlantic storm track, the climatological mean 2-m temperature T_2m during December of 1.8°C is relatively high for the station's latitude. During December 2006, however, T_2m was 4–7 K above normal on most days (Fig. 2a, thick line). A warm period until 17 December was accompanied by intense precipitation events, with 52.4-mm precipitation in 24 h on 11 December (Fig. 2g, open bars). SLP was fairly low during the first 2 weeks of December, with a minimum below 971 hPa on 4 December (Fig. 2d, thick line). While SLP increased during the second half of December 2006, precipitation amounts still remained substantial. The precipitation total at Flesland during December 2006 amounted to 323.5 mm, which is 70% more than the climatological mean. The thin lines in Fig. 2 show the results of the nudged regional NWP model simulation, taken from averaging all land grid points within 60 km from the station location. The simulated T_2m at Flesland shows a very large warm bias during a cold episode after 18 December, while the overall evolution is captured well (Fig. 2a). For SLP, correspondence between the model and observations is very good (Fig. 2d). While the model is not able to capture the precipitation amounts at Flesland at the 0.5° × 0.5° horizontal grid spacing employed here, precipitation timing has been simulated reasonably well (Fig. 2g). The exception is the period 20–25 December, when a significant precipitation event is simulated to be much weaker than observed. Detailed inspection of the simulation shows that the precipitation area, connected to a landfalling AR (see section 3b), was displaced slightly northeastward compared to the actual weather evolution.

View Full Size

Time series of simulated and observed meteorological variables at the stations (left) Flesland near Bergen, (middle) Oslo-Blindern, and (right) Tromsø during December 2006. (a)–(c) Observed (boldface line) and simulated (thin line) daily mean 2-m temperature (°C) compared to the climatology (thick gray line). (d)–(f) Observed (boldface line) and simulated (thin line) daily mean SLP (hPa). (g)–(i) Simulated (filled area) and observed (lines) 24-h accumulated precipitation (mm). Letters in (d)–(f) refer to labeled cyclones in Fig. 3.

Citation: Monthly Weather Review 141, 8; 10.1175/MWR-D-12-00256.1

• Download Figure
• Download figure as PowerPoint slide

The station Oslo-Blindern (94 m MSL) is situated about 300 km east of Bergen, with less exposure to landfalling storms from the west. During the first half of December, T_2m was 6–8 K warmer than normal, which was accompanied by 22.6 mm day⁻¹ of precipitation on 8 December, and a drop in SLP to 981 hPa (Figs. 2b,e,h). The second half of December was fairly dry at that station, as a high pressure system influenced southeastern Norway, and temperatures were closer to the climatological mean. For T_2m the trend is captured well by the simulation, while the variability is too low (Fig. 2b). For precipitation, agreement between the model and observations is very reasonable, both in terms of timing and magnitude.

At Tromsø, several prolonged episodes with temperatures up to 9 K above the climatological mean occurred during December 2006, again accompanied by intense precipitation of up to 36.3 mm on 21 December (Figs. 2c,i). Precipitation episodes coincided with marked drops in SLP on 21 and 27 December, indicating the proximity of low pressure systems (Fig. 2f). Interestingly, most precipitation was recorded in Tromsø when Bergen-Florida and Oslo-Blindern received little precipitation. The simulated T_2m matches observations fairly well, again with a slight warm bias during cold episodes. For precipitation, the model underestimates high daily precipitation totals, while the timing and duration of the simulated precipitation events match very well with the observations.

b. Evolution of the synoptic situation in the study domain

The covariance between SLP and precipitation in Fig. 2 suggests that precipitation is largely driven by cyclone activity during the study period. Indeed, about six cyclones made landfall in Norway during the study period, out of a total of 29 cyclones identified in the simulation domain during December 2006. In the following, the evolution of the meteorological situation is described with an emphasis on moisture advection to the target domains.

Figure 3 displays the upper-level flow configuration at four selected times in terms of the isentropic potential vorticity (PV, shaded) and wind velocity at the 325-K potential temperature surface as contours in the panels on the left. On the right in Fig. 3, the meteorological situation in the lower troposphere is depicted as vertically integrated WV between 200 hPa and the surface (shaded, mm), SLP (contours, hPa), and winds at 850 hPa (arrows). Cyclones are labeled alphabetically by black letters following their sequence of occurrence in the simulation domain. When viewed jointly with the low-level winds, the direction and magnitude of the water vapor fluxes can be estimated from the IWV field. We use IWV as a proxy for water vapor fluxes throughout this manuscript.

View Full Size

Meteorological configuration (left) at the tropopause and (right) in the lower troposphere at 0600 UTC 10 Dec, 0000 UTC 14 Dec, 1200 UTC 19 Dec, and 0600 UTC 23 Dec 2006 as simulated by the CHRM model. (a),(c),(e),(g) Isentropic potential vorticity [PV, in potential vorticity units (PVU, shading, 1 PVU = 10⁻⁶ K m² kg⁻¹ s⁻¹)], wind direction (arrows), and wind speed (m s⁻¹, black contours at 30, 40, 50, and 60 m s⁻¹) on the 325-K isentrope. (b),(d),(f),(h) IWV (mm, shading), SLP (hPa, black contours, intervals of 5 hPa between 1040 and 950 hPa, with thick contours at 980, 960, and 940 hPa), and wind vectors at 850 hPa (arrows). Black letters refer to tracked cyclones (see text). Cross sections S1 and S2 indicated in (d) are shown in Fig. 7.

Citation: Monthly Weather Review 141, 8; 10.1175/MWR-D-12-00256.1

• Download Figure
• Download figure as PowerPoint slide

During 1–7 December 2006, a mostly zonally oriented jet at upper levels, combined with a pronounced stationary low pressure system centered north of the British Isles, dominated the atmospheric circulation pattern in the model domain (not shown; see the online supplemental material). At 0600 UTC 10 December 2006 the jet stream had attained a slight meridional orientation along a pronounced upper-level ridge (Fig. 3a), leading to the advection of a large air mass of high IWV toward the British Isles (Fig. 3b). South of Greenland, cyclone I intensified rapidly, reaching a minimum core pressure below 960 hPa. The area of high IWV apparently extended from poleward subtropical latitudes and fed into the cyclone's frontal system. Despite indications of evaporation in the cold sector, a fairly sharp gradient existed between the pre- and post-cold-frontal air (Fig. 3b). During the following 24 h, the frontal system of cyclone I passed over southern Norway, bringing 52.4 mm day⁻¹ of precipitation to Flesland (Fig. 2g). With its broad, mostly zonally oriented tropopause-level structure and the dominance of a single large cyclone in the area, this period resembles the LC2 wave breaking category.

At 0000 UTC 14 December, the upper-level jet had become more meridionally oriented in a subsequent Rossby wave breaking event (Fig. 3c). A fairly narrow, elongated band of high IWV was present at that time (Fig. 3d), reminiscent of an AR. The frontal area of cyclone K was broad near the occlusion and was wrapped partly around the cyclone center. As the cyclone moved into northern Norway during the following day, mean SLP remained low at 983 hPa in Tromsø (NN in Fig. 2f), while precipitation mostly fell in southern Norway.

At 1200 UTC 19 December, a pronounced Rossby wave with a partly meridional jet was present at upper levels (Fig. 3e), which had favored northward advection of moist subtropical air masses in a large area of the eastern North Atlantic (Fig. 3f). Cyclone P with a core pressure of 973 hPa had moved poleward along the east coast of Greenland, followed by a long trailing band of high IWV values (Fig. 3f). During the next 12 h, cyclone P traveled north to Svalbard, and became detached from the band of high IWV. The moist air mass caused precipitation first in southern Norway, then in northern Norway during the next 48 h. At Flesland, observed precipitation was 16.7 mm day⁻¹ on 20 December, which was missed in the simulation due to a northward displacement of the AR (Fig. 2g). At Tromsø the precipitation event was more pronounced, with more than 13 mm day⁻¹ precipitation on three consecutive days and a maximum of 36.4 mm day⁻¹ on 21 December (Fig. 2i). In the following 12 h, the remnants of cyclone M, which had developed in the subtropics, were incorporated into the IWV plume. At the same time cyclone Q had formed from a frontal wave instability in the IWV plume near 55°N (cf. Ralph et al. 2011), generating another filament of high IWV off Newfoundland from evaporation behind the cold front, which is incorporated into the large subtropical moisture excursion downstream. In the Thorncroft et al. (1993) cyclone life cycle classification, due to the meridionally oriented jet and the presence of many small cyclones, this flow configuration is similar to the LC1 type.

By 0600 UTC 23 December cyclone U had intensified and moved into the Greenland Sea west of Iceland, attaining a minimum core pressure of 977 hPa (Fig. 3h). A double frontal structure had developed from the remnants of the previous IWV filament adopted by cyclone U and the developing frontal system of cyclone V, which later on become indistinguishable from one another. As cyclone U moved north of Iceland during the next 12 h, steered by the northeast-oriented upper-level jet (Fig. 3g), its frontal system made landfall in southern Norway with direct onshore flow from the southwest, shedding 14.3 mm day⁻¹ precipitation at Flesland (Fig. 2h). Onshore flow remained stationary at about 64°N until 0000 UTC 25 December, as cyclone V split southeast of Greenland and moved north along the east coast, dragging a narrow IWV filament behind it. While further intensifying, cyclone V moved north of Svalbard before it slowly filled and drifted southeast, its trailing frontal system precipitating out over northern Norway after 26 December with above 12 m day⁻¹ of precipitation at Tromsø on four consecutive days (Fig. 2c).

At 1800 UTC 29 December, the upper-level jet attained a zonal orientation again, and an intense cyclone Y had developed in warm and moist air with a large area of high IWV ahead and south (see the online supplemental material). While further intensifying during the next 12 h, cyclone Y moved its center to Iceland and became the single dominant low pressure pattern over the North Atlantic. The band of high IWV was pushed southeast and crossed Norway during the following 24 h, with observed precipitation of over 8.0 mm day⁻¹ at Flesland and 11.6 mm day⁻¹ at Oslo-Blindern (Figs. 2g,h). Similar to the period before 10 December, this flow configuration can be classified as an LC2 case.

c. Temporal sequence of oceanic moisture tracer contributions in the target domains

From the previous discussion, it could be seen that precipitation in western Scandinavia coincided with poleward extensions of high IWV bands [for reasons of simplicity referred to as atmospheric rivers (AR) here] that appeared to contribute to the WCB airstream of cyclones. However, these ARs could not always be related to exactly one cyclone and apparently existed longer than individual cyclones. Instead, the precipitation generating WCBs of individual cyclones seemed to feed on the ARs but consumed only part of the available moisture. At the same time, narrow IWV filaments formed by evaporation in the cold sector were fused into the ARs at the upstream edge (e.g., Figs. 3f,h). A single AR thus could be supply the moisture for several small cyclones, which were spawned along and north of the AR. By means of the WV tracers, we now focus on the question of how the origin of the precipitation in western Scandinavia varied in the presence of ARs.

To identify periods when ARs contributed a significant share to total precipitation in one of the target regions, we defined AR periods when at least one grid box was above a threshold value of 14-mm IWV over land within the target regions. The 14-mm threshold value was chosen after careful testing and inspection of all individual cases of landfalling ARs in Norway during this period. Ralph et al. (2004) used a threshold value of 20 mm for the definition of atmospheric rivers at the eastern North Pacific at latitudes of ~30°–40°N, but due to the higher latitudes considered here, and since only grid points over land (and hence with underlying topography) are evaluated, it is reasonable to choose a lower threshold value for AR conditions. Figure 4 shows how the AR identification depends on the threshold value. Whenever the maximum IWV in a target region exceeds 14 mm (Figs. 4a,b, gray line), an AR period is identified (Figs. 4a,b, black dots). Fairly strong variability of IWV on a 3–5-day time scale is apparent in the two domains. Several of the ARs roughly precede or coincide with peaks in mean and maximum precipitation in the domains, for example, on 14–15 December in southern Norway and on 21–22 December in northern Norway (Figs. 4c,d). Figures 4e and 4f show the relative contribution of all moisture source tracers to the 6-hourly precipitation total in each domain. While the highest contributions come from the regions T4 (red) in SN and T5 (dark blue) in NN, during AR periods and peaks in maximum precipitation in both domains substantial contributions come from ocean sectors T1–T3 (pink, light blue, green, respectively). In these events the three southerly tracers can contribute up to 60% of the total precipitation. On the other hand, during some non-AR periods in NN (e.g., 15–19 December), local and northerly tracers (T5, T6) contribute most of the precipitation water.

View Full Size

Time series of AR conditions and simulated precipitation origin in the (left) SN and (right) NN domains. (a),(b) Maximum IWV (black line, mm) and oceanic sources latitude (blue line, ° latitude) in the respective domains at a 6-hourly interval. Thick gray line denotes the 14-mm threshold for AR conditions, and black dots denote AR periods. (c),(d) Simulated daily maximum (open bars) and mean (filled bars) accumulated precipitation in the respective domain (mm day⁻¹). (e),(f) Tracer fractions for the domain average precipitation at a 6-hourly interval (%).

Citation: Monthly Weather Review 141, 8; 10.1175/MWR-D-12-00256.1

• Download Figure
• Download figure as PowerPoint slide

A correlation analysis between the maximum IWV in the NN and SN domains and the respective time series of each tracer's relative precipitation contribution reveals tracer T2 (30°–40°N) as most influential in NN, while T3 (40°–50°N) has the strongest correlation in SN (Table 1). Weaker yet significant positive correlations are found for T1–T3 as well as the boundary tracers TW and TS in both domains. Local tracers are negatively correlated with IWV (T5 for SN, T5 and T6 for NN), emphasizing the role of long-distance advection for high IWV values over Norway. The oceanic source mean latitude, constructed from the precipitation-weighted center latitude of the oceanic source regions of T1–T6, is clearly anticorrelated with the maximum IWV in both domains, consistent with the correlation analysis (Figs. 4a,b, blue line). During AR conditions, the mean oceanic source region can very rapidly shift to locations 15°–20° farther south than during non-AR conditions.

Table 1.

Correlation of each water vapor tracer (Tx) contribution time series with mean IWV in the NN and SN domains during December 2006 at a 6-h time interval. Significant correlations at p < 0.001 are shown in boldface.

View Table

d. Total oceanic moisture tracer contributions to target domain precipitation

The most important moisture contributor to precipitation in SN during the whole study period is T4 (50°–60°N) with 36.3% (Table 2, red area in Fig. 4e). The more southerly ocean regions T1–T3 in combination contribute 21.7%. Further large contributors are land regions (TL) and tracer T5 (60°–70°N) with 19.6% and 15.2%, respectively. For NN, region T5 is the dominant contributor with 36.1%. Here, T1–T3 contribute 24.1% to the precipitation total. Further substantial contributions are from TL (13.6%) and T4 (13.1%). Essentially, this means that the majority of the precipitation in the target domains is contributed by moisture evaporating at local or nearby latitudes. However, since precipitation formation is a highly nonlinear process, the relevance of the remaining moisture from farther south may be larger than their relative contribution to total precipitation suggests. For example, enhanced moisture content above the freezing level will be important for the formation of ice particles, which would allow for more efficient precipitation formation in the atmospheric column below than when only warm-cloud microphysics are involved.

Table 2.

Contribution of all water vapor tracers (Tx) to the total accounted precipitation in the NN and SN domains during December 2006 (%), also separated into atmospheric river conditions (AR) and non-AR conditions (nAR). Contributions from tracer TA are below 0.1% for all cases.

View Table

The total source region contributions are further investigated by separating the precipitation origin for AR and non-AR conditions. During AR conditions, as has already been suggested by the oceanic source mean latitude time series (Figs. 4a,b), persistently higher contributions from ocean sources south of 60°N in SN and NN are identified (Figs. 5a,b). During AR conditions in SN, the southerly tracers T1–T3 increase their contributions by factors of 10–20 (Table 2). Tracer T5 shows the most pronounced decreases during AR conditions. The more distant boundary tracers TW and TS increase during AR conditions. For NN, tracer T5 also decreases most markedly during AR conditions, while the southerly tracers T1–T3 increase by a factor of 2–5. Moisture from outside the western and southern model boundaries (TW, TS) contributes more to AR precipitation events (Fig. 5b). The less clear-cut separation between AR and non-AR contributions for the NN domain could indicate that the 14-mm threshold value may be too high for these latitudes. Overall, this analysis quantitatively consolidates the previous picture that more long-range transport of moisture from southerly latitudes occurs during AR conditions.

View Full Size

Fraction of total precipitation contributed by the moisture source tracers (%) during AR and non-AR conditions for (a) SN and (b) NN. Histograms of categorized 6-hourly precipitation at all land grid cells for AR periods (black line) and non-AR periods (gray area) in (c) SN and (d) NN.

Citation: Monthly Weather Review 141, 8; 10.1175/MWR-D-12-00256.1

• Download Figure
• Download figure as PowerPoint slide

When considering histograms of the 6-hourly gridpoint precipitation rate for the AR and non-AR categories (Figs. 5c,d), it is apparent that the precipitation characteristics are markedly different for the AR versus non-AR conditions. First, there are fewer grid points with no precipitation [<0.1 mm (6 h)⁻¹] during AR conditions in both domains. Second, the strongest increase is for grid points with medium precipitation rates [0.5–2 mm (6 h)⁻¹], and less for higher precipitation rates. Precipitation is thus more widespread and more intense during AR conditions, but not necessarily extreme. Overall, in SN precipitation is 2.9 mm (6 h)⁻¹ higher during AR compared to non-AR conditions, yet with a large variability (Table 3). For NN, the precipitation increase during AR conditions is relatively small [from 3.4 to 3.7 mm (6 h)⁻¹]. The T_2m values are markedly different, with 1.8 K warmer conditions for AR conditions in SN and 1.4 K in NN (Table 3). AR and non-AR conditions thus mark periods when the study domains are under the influence of distinct air masses that have very different moisture source, temperature, and precipitation characteristics. It is now further investigated how the AR air masses obtain these distinct characteristics.

Table 3.

Mean 2-m temperature T_2m and precipitation P during atmospheric river (AR) and non-AR conditions (nAR) in the NN and SN domains calculated from 6-h model simulation data.

View Table

e. Structures of moisture transport in atmospheric rivers

Here, we introduce a new visualization of the coherence and overlap of the WV from the various source areas in a compact fashion. To display the dominant evaporation sources in a two-dimensional plot, the vertically integrated water vapor was calculated separately for each tracer and, then, expressed in percent relative to the total IWV. To show the coherence and overlap of the tracers in different regions of the model domain, the IWV contributions for each tracer were composited with transparent shading. Figure 6 shows the percent contribution of the various tracers to IWV below 500 hPa with transparent color shading. Each panel thereby presents an instantaneous snapshot of the water vapor tracer composition of the atmosphere. More intense shading of the same color denotes that a higher fraction of the same tracer contributes to the total IWV. When viewed jointly with plots of IWV (Fig. 3, right column), it can be identified what share the tracers contribute to the total IWV at each location. For reference to the IWV maps, the 14-mm IWV contour is shown as a thick black line on the composite maps. When viewed as a time lapse sequence, the various transparent shadings allow us to follow the evolution of the tracers. These maps are available for the entire simulation period and in different subsets of tracers in the online supplemental material.

View Full Size

Tracer composition of total IWV below 500 hPa (shaded) for the same times as in Fig. 3. More opaque shading indicates more than 10%, 30%, and 50% IWV contributions. SLP (gray contours, contour interval of 5 hPa) and cyclone symbols as in Fig. 3. Thick black contour denotes IWV of 14 mm. White solid and dashed circles around cyclone U and the X indicate a 500-km analysis radius (see text for details).

Citation: Monthly Weather Review 141, 8; 10.1175/MWR-D-12-00256.1

• Download Figure
• Download figure as PowerPoint slide

At 0600 UTC 10 December, the single large cyclone I dominated the SLP pattern in the domain. A broad AR was located southeast, with a strong northeasterly flow behind the cold front, as well as a zonally oriented upper-level jet axis (Figs. 3a,b). The tracer composition for this instance in time shows a mostly zonal distribution of moisture sources near the center of cyclone I, coincident with the respective tracer sources in the North Atlantic (Fig. 6a, cf. also Fig. 1). Substantial amounts of lower-latitude moisture advection interrupt the zonal moisture distribution in the AR area, over the British Isles from tracer T3, and near the Azores from tracers T1 and T2.

At 0000 UTC 14 December, a long, narrow AR was present in the eastern North Atlantic, oriented in the southwest–northeast direction, and underneath a strong upper-level jet (Figs. 3a,b). The tracer composition shows a prominent excursion of southerly tracers in the AR area, with blue shades from tracer T2 reaching all the way to southern Norway (Fig. 6b). In south-to-north progression, contributions from tracers TS (gray), T1 (pink), T2 (light blue), and T3 (light green) dominate the composition of the AR. The excursion covers about 10°–20° latitude for the successive tracers, pointing toward the loss of the WV due to precipitation, as the moist air is lifted during northward advection. In the areas west of the AR (the cold sector of the cyclone) where the low-level flow is weaker and has a northwesterly orientation, the moisture source composition is dominated by the respective local tracers.

A vertical cross section of specific humidity along the AR's axis at 0000 UTC 14 December (coordinates 25°N, 39°W and 63°N, 30°E; line S1 in Fig. 3d) shows that most of the WV resides at levels below 750 hPa with q > ~5 g kg⁻¹ (Fig. 7a, shading). This corresponds well with the finding by Ralph et al. (2005) based on 19 soundings of AR events that 75% of the water vapor transport in ARs takes place below ~800 hPa. North of 58°N the humidity values drop quickly, marking the northern edge of the occluded front. Substantial amounts of humidity persist throughout the troposphere, and with few exceptions the air mass is cloudy (RH > 80%, thick blue contour). High meridional wind velocities reach all the way from the upper troposphere to the surface, in particular toward the north end of the cross section with ~15–20 m s⁻¹ between 45° and 57°N (Fig. 7a, solid black contours), leading to strong meridional moisture advection in the lower to middle troposphere (Fig. 7c). The corresponding tracer composition for the oceanic tracers (Fig. 7c, shading, cf. Fig. 6b) shows that at levels below ~850 hPa, the local tracers T1–T4 dominate in south-to-north sequence (colored boxes at the bottom indicate local source tracers), while the tracer contribution slopes northward vertically. This creates at most locations a vertical gradient of the tracer composition: toward higher levels progressively more remote tracers with southerly origin dominate (e.g., at 46°N, T3 up to 900 hPa, T2 up to 650 hPa, and T1 above 650 hPa). The tracer air masses thereby seem to follow or slightly exceed the sloping of the isentropes toward the north (e.g., T3 between 39° and 59°N). The initial rapid ascent of tracers (e.g., for tracer T2 in Fig. 7c) suggests that water vapor tracer is ventilated from the boundary layer into the WCB inflow and could be regarded as a WCB footprint region (Sinclair et al. 2008).

View Full Size

Vertical cross sections of specific humidity (q, g kg⁻¹, shading), zonal–meridional wind speed (m s⁻¹, black contours), and relative humidity (80% and 90%, blue contours) at 0000 UTC 14 Dec 2006 along (a) 25°N, 39°W and 63°N, 30°E and (b) 61°N, 60°W and 43°N, 28°E, as indicated in Figs. 3d and 5b. (c),(d) Oceanic tracer fractions (shading, 10% contribution) and meridional moisture fluxes (g m⁻² s⁻¹, boldface lines), and isentropes (K, thin lines) along the same transects as in (a),(b). For clarity only five oceanic tracers at one opacity level are shown in (c) and (d). Colored bars below the topography indicate the tracers evaporating at the surface.

Citation: Monthly Weather Review 141, 8; 10.1175/MWR-D-12-00256.1

• Download Figure
• Download figure as PowerPoint slide

A second vertical cross section through the cyclone center and the head of the AR (coordinates 61°N, 60°W and 43°N, 28°E; line S2 in Fig. 3d) shows the difference between cold and warm sector air masses (Fig. 7b). The AR begins east of ~10°W, where a strong meridionally oriented jet at tropopause levels almost reaches to the surface, again with velocities of 10–20 m s⁻¹ (black contours; cf. also Fig. 3c). The maximum poleward water vapor flux is located at the intersection between the moist air masses and high wind velocities (Fig. 7d, thick contours). To the west of this jet structure, specific humidities are lower at all levels, and the cloud cover becomes broken with RH < 80% (thick blue contour). The cyclone center is also located just west of the jet structure (62°N, 15°W; see Fig. 3d). The tracer composition for this cross section (Fig. 7d) shows that WV of local origin (T5) dominates west of the cyclone center, while the cyclone core itself contains a mixture of tracer T5 and T4 (Fig. 6b). Directly east of the upper-level jet, a very pronounced change in moisture tracer composition occurs, including substantial fractions of the southerly tracers T1–T3. The tracers form a vertical stack, starting with tracer T3 from 970 to 730 hPa, T2 from 950 to 260 hPa, and T1 from 820 to the top of the model atmosphere. The stack is more clearly visible if tracer contributions larger than 10% are inspected at the same cross section (not shown). Considering the high isentrope levels that these vapor contributions are found at compared to the potential temperatures near the surface for the evaporation region of the tracers, further processes must have lifted the WV to higher altitudes. While we expect that diabatic heating due to large-scale ascent is responsible for most of the vertical lifting at that time of the year, it is possible that at southerly latitudes convective mixing contributed to the spread of the tracer across a wider range of potential temperatures, as has been observed by Purvis et al. (2003). Note also that the tropopause has been lifted clearly in the region above the moisture flux maximum at 300 hPa in a wide area. In combination with the proximity to the jet structure, this suggests that the cyclone's WCB has been intersected at this location (Eckhardt et al. 2004; Roiger et al. 2011). At this instance in time, tracers T1 and T2 contribute to the WCB outflow near the tropopause, while tracer T3 mostly contributes to the meridional transport at midlevels. It is interesting to note the fairly clear vertical separation between the air masses, which may have been enhanced during northward advection by the more intense diabatic heating in the more warm and moist southerly air masses.

At 0600 UTC 19 December, a major, meridionally oriented AR extended far into the northern latitudes (Fig. 3f). The tracer composition for this situation shows a substantial disruption of the zonal tracer distribution by the AR (Fig. 6c), with a meridional progression of tracer distributions similar to what has been identified in Fig. 7. However, T2 is dominant even up to Iceland. Tropical cyclone M, which had been incorporated into this AR near 50°N, 35°W, at 1200 UTC 19 December, further enhanced the fraction of southerly tracers (Fig. 3f). Cyclone P, which was located southwest of Svalbard near 77°N at that time, had incorporated substantial amounts of moisture from 60° to 70°N (dark blue shading extending toward NE Greenland), yet tracer T2 also contributed more than 10% to the moisture in the occlusion.

At 1800 UTC 23 December, cyclones U and V were located south of Greenland, with their frontal bands joining to form one large AR (Fig. 3h). The eastern AR, mostly associated with cyclone U, shows again a meridional sequence of the different tracer contributions from south to north (Fig. 6d). The western AR associated with cyclone V is in its initial formation phase. As the joint analysis of the IWV map and the tracer composition reveals, in some regions substantial amounts of moisture added by this frontal band originate from local evaporation due to airmass transformation in the cold sector. It is important to note here that the ARs are not exclusively formed by meridional advection, but also by zonal advection and evaporation occurring en route. Near the centers of cyclones U and V, the local tracer corresponding to the current latitude of the cyclone dominates the moisture composition.

As indicated by the 14-mm IWV contours in Fig. 6, there is typically a substantial separation, on the order of 1000 km or more, between the cyclone center and the location where the poleward moisture transport is strongest (see also Fig. 7b). This finding is also consistent with results from an AR climatology (Neiman et al. 2008) and cyclone-centered composites of IWV in the North Atlantic (Rudeva and Gulev 2011). In the following section, it is examined how airmass properties in the cyclone center and at a location within the AR region east of the center change during a cyclone's life cycle.

f. Moisture composition changes in the vicinity of a cyclone

Based on SLP and absolute vorticity at 850 hPa, cyclone centers were manually tracked at 6-h intervals for the simulation period. In total, 29 individual cyclones were present and tracked in the model domain during December 2006. We focus our analysis here on cyclone U, which was associated with a well-defined frontal system and AR (Figs. 3f, 6d). The track of cyclone U starts farther south than for most other cyclones during this month (not shown). The minimum sea level pressure, vertically integrated latent and sensible heat energy, and moisture tracer composition were extracted in a 500-km radius around the cyclone center along the track at each 6-h time interval (solid white circle in Fig. 6d). Latent heat (LH) and sensible heat (SH) energy content were calculated following Oort (1971) by vertically and horizontally integrating LH = c_p × T and SH = L × q within the given radius, where c_p is the specific heat at constant pressure, L is the heat of condensation, and q is specific humidity.

Cyclone U was initially tracked at 1200 UTC 20 December at 37°N (Fig. 8a). During the following 4 days, it fairly continually progressed up to 70°N, reaching a core pressure minimum of 977 hPa at 0600 UTC 23 December at ~64°N (Fig. 6d, solid white circle; see also Fig. 3f). The vertically integrated sensible and latent heat energy content in the core of the cyclone continuously decreased as the cyclone was intensifying and moving poleward (Fig. 8c, solid lines). During the strongest intensification phase on 22 December, the latent heat content decreased more rapidly than the sensible heat content, indicating heat supply by condensation processes. The minimum energy content was reached 12 h after the minimal core pressure at 1200 UTC 23 December, before the cyclone started to decay and disappeared from the SLP chart at 0600 UTC 24 December. Note that due to the fixed 500-km radius, Fig. 8c does not depict the energy content of the complete cyclone.

View Full Size

Time series of quantities analyzed within a 500-km radius about the cyclone center defined as the minimum SLP for cyclone U. (a) SLP minimum (hPa, black line) and center latitude (in ° latitude, blue line). (b) Fractional tracer contribution to total IWV in the analysis area (%, solid lines). Colored bar indicates the tracer evaporating from the ocean underneath the cyclone center at a given time. (c) Area-integrated latent heat [petajoules (PJ), blue lines] and sensible heat (PJ, red lines) for the cyclone center region (solid lines) and with a 1000-km offset toward the east (dashed lines). (d) As in (b), but for the analysis region with a 1000-km offset toward the east.

Citation: Monthly Weather Review 141, 8; 10.1175/MWR-D-12-00256.1

• Download Figure
• Download figure as PowerPoint slide

The contributions by the WV tracers during the life cycle of cyclone U show a rapid transformation over time (Fig. 8b). During the first 24 h, most of the moisture (up to 62%) in the cyclone core originated from 30° to 40°N (T2), while contributions from the southern boundary tracer TS decreased rapidly (from ~38% to 10%) and those from T3 started to increase (from ~10% to 20%). As the cyclone continued moving north, incorporation of moisture from tracers T3, T4, and T5 in sequence replaced the more southerly tracers and dominated the moisture composition in the cyclone. As indicated by the colored bars at the top of Fig. 8b, the contribution of each of the more northerly tracers reached a maximum within 24 h of moving over the respective evaporation source. This highlights the rapid turnover of latent heat energy in the cyclone center, in particular during the strongest intensification phase on 22 December (Fig. 8c).

To investigate how the tracers progressed in the AR associated with the cyclone, the analysis was repeated for a location offset 1000 km to the east of the cyclone center, near the maximum of the meridional transport (Fig. 6d, dashed circle; cf. also Fig. 3h). The tracer composition is markedly different compared to the cyclone center. Tracer T2 dominates the AR until the end of the intensification phase, together with smaller contributions from TS and T3 (Fig. 8d). At this point (0000 UTC 24 December), the air mass contains equal contributions from tracers T2 to T5, and thereafter is dominated by T5 and T6 in sequence. Thus, the tracer composition in the AR area is clearly dominated by the continuous long-range advection of tracers T2 and TS. Rather than becoming dominant, local tracers contribute at most 20% to the WV in the air mass, which may be partly from evaporation and partly from zonal advection (see section 3e).

The total latent heat content above the offset area correspondingly decreases slower than in the cyclone center and, in fact, remains nearly constant until the phase of strongest cyclone intensification (Fig. 8c, dashed blue line). The sensible heat content remains higher than in the cyclone center, reflecting the warmer temperatures in the AR region (Fig. 8c, dashed red line). Similar transformations of the air mass in the cyclone core and the associated AR region can be identified for many of the poleward-progressing cyclones tracked during December 2006 (not shown).

The mean meridional moisture transport from the oceanic source areas throughout the analysis period of December 2006 is investigated by the zonal mean fractional contribution of each tracer to total precipitation (Fig. 9). Each oceanic tracer (T1–T6) dominates locally with 60%–35%. It is interesting to note the skewed shape of the contributions, which are similar for each tracer, marked by a steep decline toward more southerly latitudes, barely extending farther than 10° latitude, and having a long tail toward the north. For tracers T2 and T3, the tail extends more than 30° latitude poleward at a contribution of ~10%. This shape is most likely a result of moisture being preferentially transported northward from the southern midlatitudes and subtropics where evaporation is strongest (Trenberth and Guillemot 1998). Together with the reduced size of the tracer area toward the north (Fig. 1), the long tails contribute to the decreased tracer concentration maxima with increasing latitude. During December 2006, much of this mean northward moisture transport occurred in the form of distinct and pronounced ARs.

View Full Size

Zonal mean of the precipitation fraction contributed by each water tracer during December 2006 (%, solid lines) as a function of latitude.

Citation: Monthly Weather Review 141, 8; 10.1175/MWR-D-12-00256.1

• Download Figure
• Download figure as PowerPoint slide