One of the worlds major

The influence of surface fluxes on Atlantic storms

Thanks to: Sebastian Scher, Eirini Tsartsali, Iris de Vries Rein Haarsma,Sybren Drijfhout

Image source:http://goes.gsfc.nasa.gov/

Hylke de Vries

Gulf Stream

l

Characterized by high gradients of Sea Surface Temperatures (SST)

l

One of the worlds major

birth regions of mid-latitude storm

l

Question: are these strong

SST-gradients important for

the development of winter

storms &, more generally the

storm track

There is impact (“obviously”)

convergence and divergence are strong (80u–40uW, 30u–48uN, red- dashed box in Fig. 1c). Furthermore, consistent with the MABL model¹⁶ where SST variations force pressure adjustments, the pattern of laplacian SST with sign reversed (2=²SST) exhibits some similarities to laplacian SLP and wind convergences (Fig. 1d). These results indicate that MABL pressure adjustments to SST gradients near the Gulf Stream are important for surface wind divergence.

Relatively high pressures on the colder flank and relatively low pressures on the warmer flank induce cross-frontal components of near-surface winds, leading to divergence and convergence (Supple- mentary Fig. 1).

Previous studies suggested that warmer SSTs induce stronger ver- tical momentum mixing, and the enhanced mixing is responsible for mesoscale features in the surface wind convergence field^9,10, consis- tent with a numerical model experiment focusing on near-surface adjustments¹⁷. Our observational result indicates the importance of the overlooked pressure adjustment mechanism, consistent with both a recent short (a few days) regional model experiment for the Gulf Stream¹⁸ and a numerical study of tropical instability waves¹⁹. Note that the observed surface wind convergence is roughly collo- cated with the axis of the Gulf Stream (Fig. 1e, Supplementary Fig. 1).

Satellite observations further reveal that the Gulf Stream anchors a narrow rain band roughly collocated with the surface wind conver- gence (Fig. 2a). Although there was evidence that the Gulf Stream affects precipitation²⁰, our high-resolution analysis reveals that the narrow rain band meanders with the Gulf Stream front and is con- fined to its warmer flank with SSTs greater than 16uC. This close co- variation in space is strongly indicative of an active role of the Gulf Stream. The precipitation pattern is well reproduced in the opera- tional analysis (Supplementary Fig. 2), with a bias of excessive rain rates compared to satellite observations.

The causality is further examined using an atmospheric general circulation model (AGCM)⁴. It successfully captures the rain band following the meandering Gulf Stream, although the rain rate near the coast is somewhat too weak compared with satellite observations (Fig. 2b). When the SST is smoothed (see Methods for details), however, the narrow precipitation band disappears in the AGCM (Fig. 2c). Compared to the smoothed SST run, rain-bearing low- pressure systems tend to develop along the Gulf Stream front in the control simulation (Supplementary Fig. 3). These results indicate that the narrow precipitation band in the western North

Atlantic results from the forcing by the sharp SST front of the Gulf Stream.

Similar to precipitation, surface evaporation also exhibits a narrow banded structure on the offshore side of the SST front (Supplementary Fig. 2). This evaporation band is consistent with a short-term field observation²¹. The amount of evaporation is slightly larger than that of precipitation, indicating that local evaporation supplies much of the water vapour for precipitation. The local enhancement of evaporation on the warmer flank of the Gulf Stream is due to enhanced wind speed and the large disequilibrium of air temperature from SST^9,13.

As precipitation off the US east coast is often associated with deep weather systems, the rainfall pattern described above suggests that the Gulf Stream’s influence may penetrate to the free atmosphere.

Indeed, the upward motion across the Gulf Stream displays a deep structure extending to the upper troposphere (Fig. 3a). The upward motion is anchored by wind convergence in the MABL (Fig. 3a). The latter peaks at the sea surface, and is strongly affected by SST (Fig. 1).

It is interesting to note that although surface convergence and diver- gence are similar in magnitude (Fig. 1), the upward motion over surface wind convergence is much stronger and deeper than the downward motion over the wind divergence (Fig. 3a). This is sug- gestive of the importance of condensational heating above the MABL in developing the asymmetry between the upward and downward motion.

The upward wind velocity is strongest just above the MABL between the 850 and 700 hPa levels (Fig. 3a). The horizontal distribution at these levels is quite similar to the distribution of the surface conver- gence. The structure trapped by the Gulf Stream is clearly visible at 500 hPa and remains discernible at the 300 hPa level (Supplementary Fig. 4). Remarkably, the divergence in the upper troposphere is also dominated by a meandering band following the Gulf Stream front (Fig. 3b)—such a pattern is required by mass conservation, with the tropopause acting virtually as a lid for the mean circulation.

Next we examine the occurrence of high clouds, and infer cloud- top temperature using three-hourly outgoing long-wave radiation (OLR) derived from satellite observations. Lower OLR levels indicate lower temperatures and higher altitudes of cloud tops. Figure 3c shows the occurrence rate of OLR lower than 160 W m²², which roughly corresponds to a cloud-top height of about 300 hPa. A nar- row band of high occurrence hugs the SST front of the Gulf Stream in

a Observed rain rate, satellite b Rain rate, AGCM c Rain rate, AGCM, smoothed

50° N 45° N 40° N 35° N 30° N 25° N

50° N 45° N 40° N 35° N

1 1.5 2 2.5 3 3.5

mm d^–1

4 4.5 5 5.5 6 30° N

25° N

50° N 45° N 40° N 35° N 30° N

25° N

80° W 70° W 60° W 50° W 40° W 80° W 70° W 60° W 50° W 40° W 80° W 70° W 60° W 50° W 40° W

Figure 2 | Annual climatology of rain rate.

a, Observed by satellites. ^b,^c, In the AGCM with observed (b) and smoothed (c) SSTs. Contours are for SST, as in Fig. 1.

Wind divergence, 500–200 hPa (10⁻⁷ s⁻¹) Upward wind (10⁻² Pa s⁻¹)

Pressure (hPa)

200 300 400 500 600 700 800 900 1,000

32° N 34° N 36° N

–2 –1.5 –1 –0.5 0.5 1 1.5 2 2.5 3 –6 –5 –4 –3 –2 –1 1 2 3 4 5 6 3 4 5 6 7 8 9 10 11 38° N 40° N 42° N

a b c OLR<160 W m^{Occurrence–2} (%)

80° W 70° W 60° W 50° W 40° W 80° W 70° W 60° W 50° W 40° W 50° N

45° N 40° N 35° N 30° N 25° N

50° N 45° N 40° N 35° N 30° N 25° N

Figure 3 | Annual climatology of parameters connecting MABL and free atmosphere.

a, Vertical wind velocity (upward positive; colour), boundary layer height (black curve) and wind convergence (contours for 61, 2, 3310²⁶s²¹) averaged in the along-front direction in the green box inb, based on the ECMWF analysis.b, Upper- tropospheric wind divergence averaged between 200 and 500 hPa (colour).c, Occurrence frequency of daytime satellite-derived OLR levels lower than 160 W m²² (colour). Contours in ^b and^c are for SST, as in Fig. 1.

NATURE|^{Vol 452}|13 March 2008

LETTERS

207 Nature Publis hing Group

©2 0 0 8

Minobe et al. (2008)

Rain with and without GS-front

OBS CNTRL SMTH

Small et al, 2014 eddy meridional heat flux

There is impact (“obviously”)

Small (2014)

Storm track

Meridional eddy heat flux (At 950 hPa) Without SST-front:

• Closer to coast

• Weaker (much) With SST-front:

• Better location

• Too strong L

ERA-I SMTH

ATL

Number of cyclones

Figure S3. Annual mean climatology of cyclone number density (color) simulated by the AGCM using observed (a) and smoothed (b) SSTs (contours). A cyclone is detected as the sea-level pressure minimum within a circular range of 300 km radius on 6-hourly maps. The units of the color bar are number per year in a bin of 2.0°×2.0°, and contour interval is 2°C with the 10°C and 20°C contours dashed.

Minobe et al. (2008)

DIAGNOSING AIR-SEA

INTERACTION

Two approaches

1. Climatology & storm-track

•

Observations / reanalysis Imprint of the Gulf Stream in vertical structure of the atmosphere. [e.g. Minobe et al (2008)]

•

Experiments with global models. Difference in climatology with different SST-forcing. [e.g. Small et al (2014)]

2. Case studies- How are cyclones on an individual basis

influenced by GS surface fluxes / SST-gradients?

GS affects the “entire” troposphere

Figure S1. Summary of the climatic responses to the Gulf Stream. On the offshore flank of the SST front (black dashed curve) of the Gulf Stream (green long arrow), surface wind convergence associated with low pressure (positive SLP Laplacian), and enhanced rain occur (red shade). On the onshore flank of the front, surface wind divergence associated with high pressure (negative SLP Laplacian) occurs (blue shade). The distribution of the wind convergence and divergence is closely associated with surface winds across the SST front (short gray arrows).

Anchored by the wind convergence and enhanced precipitation, upward motion penetrates into the upper troposphere (yellow arrow). The upward wind velocity is associated with the upper-level horizontal divergence (blue oval) and frequent occurrence of high-level clouds.

Minobe et al. (2008)

Near surface: two mechanisms VMM: Vertical momentum

mixing; higher 10m-

windspeeds above warm sector

PAM: Pressure adjustment

warm SST => upward motion

=> wind convergence ((like a sea-breeze)

Diagnosing air-sea interaction from climatology

GS affects the “entire” troposphere

convergence and divergence are strong (80u–40uW, 30u–48uN, red- dashed box in Fig. 1c). Furthermore, consistent with the MABL model¹⁶ where SST variations force pressure adjustments, the pattern of laplacian SST with sign reversed (2=²SST) exhibits some similarities to laplacian SLP and wind convergences (Fig. 1d). These results indicate that MABL pressure adjustments to SST gradients near the Gulf Stream are important for surface wind divergence.

Relatively high pressures on the colder flank and relatively low pressures on the warmer flank induce cross-frontal components of near-surface winds, leading to divergence and convergence (Supple- mentary Fig. 1).

Previous studies suggested that warmer SSTs induce stronger ver- tical momentum mixing, and the enhanced mixing is responsible for mesoscale features in the surface wind convergence field^9,10, consis- tent with a numerical model experiment focusing on near-surface adjustments¹⁷. Our observational result indicates the importance of the overlooked pressure adjustment mechanism, consistent with both a recent short (a few days) regional model experiment for the Gulf Stream¹⁸ and a numerical study of tropical instability waves¹⁹. Note that the observed surface wind convergence is roughly collo- cated with the axis of the Gulf Stream (Fig. 1e, Supplementary Fig. 1).

Satellite observations further reveal that the Gulf Stream anchors a narrow rain band roughly collocated with the surface wind conver- gence (Fig. 2a). Although there was evidence that the Gulf Stream affects precipitation²⁰, our high-resolution analysis reveals that the narrow rain band meanders with the Gulf Stream front and is con- fined to its warmer flank with SSTs greater than 16uC. This close co- variation in space is strongly indicative of an active role of the Gulf Stream. The precipitation pattern is well reproduced in the opera- tional analysis (Supplementary Fig. 2), with a bias of excessive rain rates compared to satellite observations.

The causality is further examined using an atmospheric general circulation model (AGCM)⁴. It successfully captures the rain band following the meandering Gulf Stream, although the rain rate near the coast is somewhat too weak compared with satellite observations (Fig. 2b). When the SST is smoothed (see Methods for details), however, the narrow precipitation band disappears in the AGCM (Fig. 2c). Compared to the smoothed SST run, rain-bearing low- pressure systems tend to develop along the Gulf Stream front in the control simulation (Supplementary Fig. 3). These results indicate that the narrow precipitation band in the western North

Atlantic results from the forcing by the sharp SST front of the Gulf Stream.

Similar to precipitation, surface evaporation also exhibits a narrow banded structure on the offshore side of the SST front (Supplementary Fig. 2). This evaporation band is consistent with a short-term field observation²¹. The amount of evaporation is slightly larger than that of precipitation, indicating that local evaporation supplies much of the water vapour for precipitation. The local enhancement of evaporation on the warmer flank of the Gulf Stream is due to enhanced wind speed and the large disequilibrium of air temperature from SST^9,13.

As precipitation off the US east coast is often associated with deep weather systems, the rainfall pattern described above suggests that the Gulf Stream’s influence may penetrate to the free atmosphere.

Indeed, the upward motion across the Gulf Stream displays a deep structure extending to the upper troposphere (Fig. 3a). The upward motion is anchored by wind convergence in the MABL (Fig. 3a). The latter peaks at the sea surface, and is strongly affected by SST (Fig. 1).

It is interesting to note that although surface convergence and diver- gence are similar in magnitude (Fig. 1), the upward motion over surface wind convergence is much stronger and deeper than the downward motion over the wind divergence (Fig. 3a). This is sug- gestive of the importance of condensational heating above the MABL in developing the asymmetry between the upward and downward motion.

The upward wind velocity is strongest just above the MABL between the 850 and 700 hPa levels (Fig. 3a). The horizontal distribution at these levels is quite similar to the distribution of the surface conver- gence. The structure trapped by the Gulf Stream is clearly visible at 500 hPa and remains discernible at the 300 hPa level (Supplementary Fig. 4). Remarkably, the divergence in the upper troposphere is also dominated by a meandering band following the Gulf Stream front (Fig. 3b)—such a pattern is required by mass conservation, with the tropopause acting virtually as a lid for the mean circulation.

Next we examine the occurrence of high clouds, and infer cloud- top temperature using three-hourly outgoing long-wave radiation (OLR) derived from satellite observations. Lower OLR levels indicate lower temperatures and higher altitudes of cloud tops. Figure 3c shows the occurrence rate of OLR lower than 160 W m²², which roughly corresponds to a cloud-top height of about 300 hPa. A nar- row band of high occurrence hugs the SST front of the Gulf Stream in

a Observed rain rate, satellite b Rain rate, AGCM c Rain rate, AGCM, smoothed

50° N 45° N 40° N 35° N 30° N 25° N

50° N 45° N 40° N 35° N

1 1.5 2 2.5 3 3.5

mm d^–1

4 4.5 5 5.5 6 30° N

25° N

50° N 45° N 40° N 35° N 30° N

25° N

80° W 70° W 60° W 50° W 40° W 80° W 70° W 60° W 50° W 40° W 80° W 70° W 60° W 50° W 40° W

Figure 2 | Annual climatology of rain rate.

a, Observed by satellites. ^b,^c, In the AGCM with observed (^b) and smoothed (^c) SSTs. Contours are for SST, as in Fig. 1.

Wind divergence, 500–200 hPa (10⁻⁷ s⁻¹) Upward wind (10⁻² Pa s⁻¹)

Pressure (hPa)

200 300 400 500 600 700 800 900 1,000

32° N 34° N 36° N

–2 –1.5 –1 –0.5 0.5 1 1.5 2 2.5 3 –6 –5 –4 –3 –2 –1 1 2 3 4 5 6 3 4 5 6 7 8 9 10 11 38° N 40° N 42° N

a b c OLR<160 W m^{Occurrence–2} (%)

80° W 70° W 60° W 50° W 40° W 80° W 70° W 60° W 50° W 40° W 50° N

45° N 40° N 35° N 30° N 25° N

50° N 45° N 40° N 35° N 30° N 25° N

Figure 3 | Annual climatology of parameters connecting MABL and free atmosphere.

a, Vertical wind velocity (upward positive; colour), boundary layer height (black curve) and wind convergence (contours for 61, 2, 3310²⁶s²¹) averaged in the along-front direction in the green box inb, based on the ECMWF analysis.b, Upper- tropospheric wind divergence averaged between 200 and 500 hPa (colour).c, Occurrence frequency of daytime satellite-derived OLR levels lower than 160 W m²² (colour). Contours in band care for SST, as in Fig. 1.

NATURE|Vol 452|13 March 2008 LETTERS

207 Nature Publis hing Group

©2 0 0 8

Minobe et al. (2008)

Upward wind Wind divergence

at upper levels High cloudtop Diagnosing air-sea interaction from climatology

Based on ECMWF analyses Satellite derived

Satellite observations, of 10m wind

convergence (top) and 10m wind curl (bottom).

Dec 2001 to Nov 2006 average.

(10^-6s^-1) (source: Takatama et al., 2012)

Wind convergence and curl

Diagnosing air-sea interaction from climatology

Vertical momentum mixing

Credits: Eirini Tsartsali

Cool SST ~ low wind speeds

Warm SST ~ higher wind speeds

Diagnosing air-sea interaction from climatology

Signs of VMM

4 Chapter . Introduction

reason: differences in MABL response for cross-front and along-front winds (e.g. Kilpatrick, Schneider, and Qiu, 2014; Chelton et al., 2001); differences in ocean current speed effects on the cross-isotherm and along-isotherm wind stress (e.g. Chelton et al., 2001); and small scale curl features are masked by the large scale wind field when spatial high-pass filters are not applied (e.g. Chelton et al., 2004).

F 1.5: Binned scatter plots of the wind stress divergence and curl as a function of the downwind and crosswind SST gradients respectively for four different regions: the Southern Ocean (60°S to 30°S, 0°to 360°E), the eastern tropical Pacific (5°S to 3°N, 150°W to 100°W ), the Kuroshio Extension (32°N to 47°N, 142°E to 170°W), and the Gulf Stream (35°N to 55°N, 60°W to 30°W). (source:

Chelton et al., 2004)

1.2.2 Pressure Adjustment Mechanism

After some years of the Sweet et al., 1981 explanation for the variations of near surface wind speed with the help of VMM, Hsu, 1984 introduced another mechanism. He proposed an analogue to the sea breeze circulation, with the warm ocean at GS having the role of the warm land.

F 1.6: Schematic of the sea breeze circula- tion. In Hsu, 1984 sea-breeze-like circulation the warm water at GS had the role of land and the colder

slope waters that of the sea.

Warmer SSTs over the GS heat the atmo- sphere from below, resulting in less dense air which rises up and expands. In that way, a low pressure system is creating, and due to the relative higher pressure over the cooler waters, air is moving towards warmer waters.

Over the warmer waters, a high pressure sys- tem is developing at higher altitudes, forcing a return flow to the cooler waters and subsi- dence of air after cooling. This circulation causes an increase of the near surface wind speed over the GS. Moreover a convergence zone is created over the warmer waters due to convection, while divergence is observed Chelton (2004)

4 Chapter . Introduction

reason: differences in MABL response for cross-front and along-front winds (e.g. Kilpatrick, Schneider, and Qiu, 2014; Chelton et al., 2001); differences in ocean current speed effects on the cross-isotherm and along-isotherm wind stress (e.g. Chelton et al., 2001); and small scale curl features are masked by the large scale wind field when spatial high-pass filters are not applied (e.g. Chelton et al., 2004).

F 1.5: Binned scatter plots of the wind stress divergence and curl as a function of the downwind and crosswind SST gradients respectively for four different regions: the Southern Ocean (60°S to 30°S, 0°to 360°E), the eastern tropical Pacific (5°S to 3°N, 150°W to 100°W ), the Kuroshio Extension (32°N to 47°N, 142°E to 170°W), and the Gulf Stream (35°N to 55°N, 60°W to 30°W). (source:

Chelton et al., 2004)

1.2.2 Pressure Adjustment Mechanism

After some years of the Sweet et al., 1981 explanation for the variations of near surface wind speed with the help of VMM, Hsu, 1984 introduced another mechanism. He proposed an analogue to the sea breeze circulation, with the warm ocean at GS having the role of the warm land.

F 1.6: Schematic of the sea breeze circula- tion. In Hsu, 1984 sea-breeze-like circulation the warm water at GS had the role of land and the colder

slope waters that of the sea.

Warmer SSTs over the GS heat the atmo- sphere from below, resulting in less dense air which rises up and expands. In that way, a low pressure system is creating, and due to the relative higher pressure over the cooler waters, air is moving towards warmer waters.

Over the warmer waters, a high pressure sys- tem is developing at higher altitudes, forcing a return flow to the cooler waters and subsi- dence of air after cooling. This circulation causes an increase of the near surface wind speed over the GS. Moreover a convergence zone is created over the warmer waters due to convection, while divergence is observed Diagnosing air-sea interaction from climatology

Pressure adjustment (PAM)

• Over warm SSTs:

Ø Convection

Ø Low pressure system Ø Convergence

• Over cool SSTs:

Ø Subsidence

Ø High pressure system Ø Divergence

MSLP 1 / SST

Wind Divergence MSLP

Credits: Eirini Tsartsali

Diagnosing air-sea interaction from climatology

PAM - Wind conv. x SLP Lapl.

LETTERS

Influence of the Gulf Stream on the troposphere

Shoshiro Minobe¹, Akira Kuwano-Yoshida², Nobumasa Komori², Shang-Ping Xie^3,4 & Richard Justin Small³

The Gulf Stream transports large amounts of heat from the tropics to middle and high latitudes, and thereby affects weather phenom- ena such as cyclogenesis^1,2 and low cloud formation³. But its cli- matic influence, on monthly and longer timescales, remains poorly understood. In particular, it is unclear how the warm cur- rent affects the free atmosphere above the marine atmospheric boundary layer. Here we consider the Gulf Stream’s influence on the troposphere, using a combination of operational weather ana- lyses, satellite observations and an atmospheric general circula- tion model⁴. Our results reveal that the Gulf Stream affects the entire troposphere. In the marine boundary layer, atmospheric pressure adjustments to sharp sea surface temperature gradients lead to surface wind convergence, which anchors a narrow band of precipitation along the Gulf Stream. In this rain band, upward motion and cloud formation extend into the upper troposphere, as corroborated by the frequent occurrence of very low cloud-top temperatures. These mechanisms provide a pathway by which the Gulf Stream can affect the atmosphere locally, and possibly also in remote regions by forcing planetary waves^5,6. The iden- tification of this pathway may have implications for our under- standing of the processes involved in climate change, because the Gulf Stream is the upper limb of the Atlantic meridional over- turning circulation, which has varied in strength in the past⁷ and is predicted to weaken in response to human-induced global warming in the future⁸.

It is a challenging task to isolate the climatic influence of the Gulf Stream from energetic weather variability using conventional obser- vations, which are spatially and temporally sporadic. Recently, high- resolution satellite observations of surface winds made it possible to map the influence of the Gulf Stream^9,10 and other major sea surface temperature (SST) fronts^11–14 on the near-surface atmosphere. The Gulf Stream affects the 10-m wind climatology as observed by the QuikSCAT satellite¹⁵, with wind divergence and convergence on the cold and warm flanks, respectively, of the Gulf Stream front^9,10 (Fig. 1a). However, the mechanism by which the SST fronts influence surface winds is still under much debate^9,10

The identification of the mechanism responsible has been ham- pered by the need to know parameters not available from satellite observations, for which we turn to high-resolution atmospheric operational analyses from the European Centre for Medium-Range Weather Forecasts (ECMWF). The operational analysis successfully captures the observed pattern of wind divergence (Fig. 1b). Interestingly, the wind convergence closely resembles the pattern of the laplacian of sea-level pressure (=²SLP) (Fig. 1c).This correspondence is consistent with an immediate consequence of a marine atmospheric boundary layer (MABL) model¹⁶ (see Methods Summary). Note that it is virtually impossible to see the correspondence between the wind convergence and SLP itself without taking the laplacian. The laplacian operator acts as a high-pass filter, unveiling the SST frontal effect that is masked by large-scale atmospheric circulations.

In contrast to the free atmosphere where wind velocities are nearly non-divergent, substantial divergence occurs in the MABL in the presence of strong friction and is proportional to the SLP laplacian in the MABL model described in the Methods Summary.

Such a linear relation approximately holds in observations (Fig. 1f), with a correlation coefficient as high as 0.70 for a region where wind

1Department of Natural History Sciences, Graduate School of Science, Hokkaido University, Sapporo 060-0810, Japan.²Earth Simulator Center, Japan Agency for Marine-Earth Science and Technology, Yokohama 236-0001, Japan.³International Pacific Research Center,⁴Department of Meteorology, University of Hawaii at Manoa, Honolulu, Hawaii 96822, USA.

a Wind convergence,

satellite (10⁻⁶ s⁻¹) Wind convergence, ECMWF (10⁻⁶ s⁻¹)

SLP laplacian (10⁻⁹ Pa m^–2)

SLP laplacian (10⁻⁹ Pa m^–2) –SST laplacian (10⁻¹⁰ K m^–2)

50° N

45° N

40° N

35° N

30° N

25° N

80° W –8

–2

10 20 30 40 50 60

–3 –2 –1 1 2 3

–1.2 –0.4 0.4 1.2 2

–5 –4 –3 –2 –1 1 2 3 4 5

–6 –4 –2 2 4 6 8

70° W 60° W 50° W 40° W 80° W 70° W 60° W 50° W 40° W 50° N

45° N

40° N

35° N

30° N

25° N

50° N

45° N

40° N

35° N

30° N

25° N

80° W 70° W 60° W 50° W 40° W

10 5 0 –5 –10

50° N

45° N

40° N

35° N

30° N

25° N

80° W 70° W 60° W 50° W 40° W

80° W 70° W 60° W 50° W 40° W 50° N

45° N

40° N

35° N

30° N

25° N

b

c d

e Surface current speed

(cm s⁻¹) f

−4 −3 −2 −1 0 1 2 3 4 Wind convergence (10–6 s–1)

Figure 1 | Annual climatology of surface parameters. a, b, 10-m wind convergence (colour) in QuikSCAT satellite observations (a) and in the ECMWF analysis (^b).^c,^d, SLP laplacian (^c) and sign-reversed SST laplacian (d) in the ECMWF analysis.e, Surface geostrophic current speed. Ina–e, SST contours (2uC interval and dashed contours for 10uC and 20uC) are shown.

f, Relationship between the SLP laplacian and wind convergence based on monthly climatology in the red-dashed box inc; the regression line is shown red. Error bars,61 s.d. of wind convergence for each bin of SLP.

Vol 452|13 March 2008|doi:10.1038/nature06690

206

Nature Publis hing Group

©2 0 0 8

LETTERS

Influence of the Gulf Stream on the troposphere

Shoshiro Minobe

, Akira Kuwano-Yoshida

, Nobumasa Komori

, Shang-Ping Xie

& Richard Justin Small

The Gulf Stream transports large amounts of heat from the tropics to middle and high latitudes, and thereby affects weather phenom- ena such as cyclogenesis

and low cloud formation

. But its cli- matic influence, on monthly and longer timescales, remains poorly understood. In particular, it is unclear how the warm cur- rent affects the free atmosphere above the marine atmospheric boundary layer. Here we consider the Gulf Stream’s influence on the troposphere, using a combination of operational weather ana- lyses, satellite observations and an atmospheric general circula- tion model

. Our results reveal that the Gulf Stream affects the entire troposphere. In the marine boundary layer, atmospheric pressure adjustments to sharp sea surface temperature gradients lead to surface wind convergence, which anchors a narrow band of precipitation along the Gulf Stream. In this rain band, upward motion and cloud formation extend into the upper troposphere, as corroborated by the frequent occurrence of very low cloud-top temperatures. These mechanisms provide a pathway by which the Gulf Stream can affect the atmosphere locally, and possibly also in remote regions by forcing planetary waves

. The iden- tification of this pathway may have implications for our under- standing of the processes involved in climate change, because the Gulf Stream is the upper limb of the Atlantic meridional over- turning circulation, which has varied in strength in the past

and is predicted to weaken in response to human-induced global warming in the future

.

It is a challenging task to isolate the climatic influence of the Gulf Stream from energetic weather variability using conventional obser- vations, which are spatially and temporally sporadic. Recently, high- resolution satellite observations of surface winds made it possible to map the influence of the Gulf Stream

and other major sea surface temperature (SST) fronts

on the near-surface atmosphere. The Gulf Stream affects the 10-m wind climatology as observed by the QuikSCAT satellite

, with wind divergence and convergence on the cold and warm flanks, respectively, of the Gulf Stream front

(Fig. 1a). However, the mechanism by which the SST fronts influence surface winds is still under much debate

The identification of the mechanism responsible has been ham- pered by the need to know parameters not available from satellite observations, for which we turn to high-resolution atmospheric operational analyses from the European Centre for Medium-Range Weather Forecasts (ECMWF). The operational analysis successfully captures the observed pattern of wind divergence (Fig. 1b). Interestingly, the wind convergence closely resembles the pattern of the laplacian of sea-level pressure (=

SLP) (Fig. 1c).This correspondence is consistent with an immediate consequence of a marine atmospheric boundary layer (MABL) model

(see Methods Summary). Note that it is virtually impossible to see the correspondence between the wind convergence and SLP itself without taking the laplacian. The laplacian operator acts as a high-pass filter, unveiling the SST frontal effect that is masked by large-scale atmospheric circulations.

In contrast to the free atmosphere where wind velocities are nearly non-divergent, substantial divergence occurs in the MABL in the presence of strong friction and is proportional to the SLP laplacian in the MABL model described in the Methods Summary.

Such a linear relation approximately holds in observations (Fig. 1f), with a correlation coefficient as high as 0.70 for a region where wind

1Department of Natural History Sciences, Graduate School of Science, Hokkaido University, Sapporo 060-0810, Japan.²Earth Simulator Center, Japan Agency for Marine-Earth Science and Technology, Yokohama 236-0001, Japan.³International Pacific Research Center,⁴Department of Meteorology, University of Hawaii at Manoa, Honolulu, Hawaii 96822, USA.

a Wind convergence,

satellite (10⁻⁶ s⁻¹) Wind convergence, ECMWF (10⁻⁶ s⁻¹)

SLP laplacian (10⁻⁹ Pa m^–2)

SLP laplacian (10⁻⁹ Pa m^–2) –SST laplacian (10⁻¹⁰ K m^–2)

50° N

45° N

40° N

35° N

30° N

25° N

80° W –8

–2

10 20 30 40 50 60

–3 –2 –1 1 2 3

–1.2 –0.4 0.4 1.2 2

–5 –4 –3 –2 –1 1 2 3 4 5

–6 –4 –2 2 4 6 8

70° W 60° W 50° W 40° W 80° W 70° W 60° W 50° W 40° W 50° N

45° N

40° N

35° N

30° N

25° N

50° N

45° N

40° N

35° N

30° N

25° N

80° W 70° W 60° W 50° W 40° W

10 5 0 –5 –10

50° N

45° N

40° N

35° N

30° N

25° N

80° W 70° W 60° W 50° W 40° W

80° W 70° W 60° W 50° W 40° W 50° N

45° N

40° N

35° N

30° N

25° N

b

c d

e Surface current speed

(cm s⁻¹) f

−4 −3 −2 −1 0 1 2 3 4

Wind convergence (10–6 s–1 )

Figure 1 | Annual climatology of surface parameters. a, b, 10-m wind convergence (colour) in QuikSCAT satellite observations (a) and in the ECMWF analysis (^b). ^c, ^d, SLP laplacian (^c) and sign-reversed SST laplacian (d) in the ECMWF analysis.e, Surface geostrophic current speed. Ina–e, SST contours (2uC interval and dashed contours for 10uC and 20uC) are shown.

f, Relationship between the SLP laplacian and wind convergence based on monthly climatology in the red-dashed box in^c; the regression line is shown red. Error bars, 61 s.d. of wind convergence for each bin of SLP.

Vol 452|13 March 2008|doi:10.1038/nature06690

206

Nature Publis hing Group

©2 0 0 8

Minobe (2008)

. . Air-Sea Interaction 5

over the slope waters due to the sinking of

air. The effect of secondary circulations in the atmosphere on the near surface wind field, driven by pressure gradients due to the differential heating of the atmosphere in each side of the front, is the second mechanism called PAM. This suggestion was supported by many studies which used numerical and boundary layer models to understand these processes (e.g.

Lindzen and Nigam, 1987; Wai and Stage, 1989; Warner et al., 1990).

Minobe et al.,2008, noticed that the wind divergence pattern co-locates with the sea-level pressure Laplacian (r²SLP), and that both of them have many similarities with the pattern of SST Laplacian (r²SST) (see fig1.7). A linear relationship between the convergence and r²SLP was also found, which was in agreement with the MABL model of Lindzen and Nigam, 1987. Moreover, a linear relationship between r²SST and i)r²SLP, ii) convergence was expected with the latter being weaker than the former.

F 1.7: 10m wind con- vergence (top right), r²SLP (bottom left) and r²SST re- versed sign (bottom right), at the region of GS in the ECMWF analysis by Mi- nobe et al., 2008. At the top left, the regression analy- sis between r²SLP and con- vergence based on the red box in the bottom left fig- ure. The error bars are the ± 1 std (standard deviation) of wind convergence for each bin of SLP. (source:Minobe

et al., 2008)

Therefore, the strength of the PAM can be quantified by the slope of the regression analysis between these three fields as we can see below (e.g. Minobe et al., 2008; Putrasahan, Miller, and Seo, 2013; Shimada and Minobe, 2011).

⇢ r ·UÆ / r²SLP

r²SLP / r²SST (1.2)

We have to mention that these relationships between the three fields would not be visible without making use of the Laplacians. This is because the mesoscale features are masked by the large scale circulations and they are revealed by the Laplacian as it acts like a high-pass filter.

1.2.3 The Complex Truth

These two mechanisms and their role on the response of near surface wind to the sharp SSTs have concerned the scientists for many years. However, there are so many processes Diagnosing air-sea interaction from climatology

2. Individual cases

l

Controlled Regional

Experiment, Regional climate model Harmonie driven with ERA-Interim reanalysis data

l

24 single winter storm cases

l

For every storm case, one

simulation with observed SSTs (REF), and one with smoothed SSTs (SMTH)

How does the removal of the SST-front (by smoothing)

influence the development of individual storms?

15

Credits: Sebastian Scher

16

Measures

l

“Along-track”:

- Compute the track of the storm

- At each time-step, average the desired variable in a box

around the minimum

- → time-line of the variable

- average over time → one single value

- Compute SMTH-REF → one value to quantify the impact of smoothing

Credits: Sebastian Scher

H. d. Vries et al.

1 3

using a randomly varying field. The random perturbations are different for each 6 h time-step. For the single-storm ensemble we analyse the storm-strength parameters in the same way as the other storms.

3 Results

3.1 Storm-strength changes

Figure 3 shows a histogram of the relative difference in mean along-track maximum 10-m windspeed (denoted W10X), which is our measure of storm strength. For this statistic we search in each 3-h time interval for the maximum 10-m windspeed within a box of 100 × 100 km around the track.

These wind speeds are averaged along the track and com- pared. There is no sign of a systematic change. Instead, there are large variations, up to ± 25% in the storm-strength response, as measured by this parameter. So both weakening and strengthening storms are found. Because the underlying SST-change is the only difference in boundary conditions, and because these differ substantially, depending on whether storms pass north or south of the original SST-front, we use colouring in the histogram to denote the mean absolute dif- ference (SMTH–REF) in along-track SST. It appears that storms that meet on average cooler SST in SMTH (i.e., more southerly located tracks), tend to produce lower wind speeds, while those that experience on average warmer conditions show a more mixed picture. Thus storm-strength change and SST-change are correlated (Spearman rank correlation between relative windspeed difference and absolute SST difference is + 0.70). Similar results are found if we use

sea-only points, increase the search-box to 500 × 500, 1000

× 1000 km (not shown), or if we use as averaging period the time from the start of the simulation to the time of mini- mum in 𝜙925 . Other storm-strength measures such as 850 hPa relative vorticity (zeta850), or the minimum in 𝜙925 show a similar broad spread in the response (Fig. 9), with some storms getting weaker and some stronger. This mixed response is consistent with Roebber (1989a) who remarked that there is a tendency towards a Gaussian response if mul- tiple mechanisms are acting simultaneously. From the large spread and the quasi-gaussian shape of the histogram, one could argue that from a statistical point of view there is sim- ply no systematic response. However, we do not want to make that conclusion yet. Instead we will try to understand the reason underlying the mixed response.

3.2 Single-storm ensemble

To examine whether the response at the level of a single storm is significant, or at least systematic, we created an ensemble for a single storm (Storm 16 from the list). This storm travelled quite close to the SST-front and through the middle of the simulation domain. Figure 4 shows the time- series of a number of storm indicators for the ensemble, with color coding indicating the various groups within the ensemble. First of all, the tracks (top-left panel) are scat- tered closely around the tracks of REF (black) and SMTH (thick-red), except for two members (labelled 14 and 19) where the tracking failed. These two members are discarded from subsequent analysis. The SMTH-based members (thin- red) appear to propagate slightly further northward than the REF-based members, but we did not examine this aspect in detail. As expected, along-track average SST changes quite systematically as one cycles through the “linear-combina- tion” members 1–10 (for the definition of the members, see Sect. 2.3) and is scattered around REF for members 11–20 and around SMTH for members 21–30. For 925 hPa-geo- potential 𝜙925 we had only 6-h output, and duplicated the points to 3 h (bottom-left). It shows gradual changes for members 1–10 and a clear separation of the REF-based and SMTH-based. A similar result holds for the variations in the wind-speed (bottom-right). These gradual, but system- atic changes seen in especially members 1–10, and the clear separation of the two other groups (REF-based and SMTH- based), increase our confidence that the responses we see in other parameters are indeed systematic, and not influenced too strongly by the chaotic nature of the flow.

3.3 Environmental storm parameters

We now turn from the single-storm ensemble to the pair- simulations of all storms. Removing the SST front results in predictable changes in two environmental parameters that

Fig. 3 Histogram of relative difference (SMTH–REF)/(REF) of the mean along-track 10-m wind speed maximum W10X. The maxi- mum is computed using a window of 100 × 100 km around the track.

The colouring denotes mean along-track absolute difference of SST (SMTH–REF)

Mixed response

strengthening weakening

l No clear response

l About 1/3 of the storms change only very little, the other 2/3 either strengthen or weaken

How Gulf-Stream SST-fronts influence Atlantic winter storms

1 3

along the Gulf Stream front. The SST of SMTH is rescaled to have the same spatially average SST as REF, but still contains the large-scale meridional gradient of SST enforced by the boundary conditions.

Figure 2 shows the smoothed SST-field of all SMTH storms, and the absolute difference SMTH–REF, aver- aged over all cases. The difference pattern is a dipole of considerable amplitude, with increased SST north of the SST-front and decreased values south of it. The grey box outlines the HCLIM domain. It encompasses the region with the SST-front, but is still small enough to constrain the upper-level flow and large-scale baroclinicity by the boundaries. Indeed, the upper-level (above 500 hPa) flow and temperature structure are found to hardly differ between the REF and SMTH simulations.

2.2 Along-track statistics

An along-track measure is developed to summarise the response of a chosen variable (e.g. wind speed) to the smoothing of the SSTs. First the center of the storm is identified as the local minimum of the geopotential at 925 hPa ( 𝜙925 ) for every 6-hourly time step. This results in the track of the storm. The tracking is done for REF and SMTH separately, to account for deviations in the tracks. Generally these track-differences are small (Fig. 1), implying that the tracks are only weakly influenced by fine-scale details of the underlying SST pattern and more constrained by the lateral boundary conditions. Then the variable of interest is aver- aged over a square box around the center of the storm. This results in a single timeline for each variable. The size of the box is chosen to be rather small (30 × 30 km for 𝜙925 , 100 × 100 km for all other variables) to focus on near-center storm response. Finally the timelines are time-averaged, yielding a single value for each simulation. Analyses are repeated for larger boxes (500 × 500 and 1000 × 1000 km) and systematic differences across scales will be discussed.

2.3 Single-storm ensemble

The tracks shown in Fig. 1 indicate a large variability in the simulated storms. Because each storm is simulated “only”

twice, it is impossible to judge on a storm-by-storm basis whether the differences are significant. To alleviate this, we selected one storm (Storm 16) and constructed a 30-member ensemble for it. Member 0 is the REF storm. Members 1 to 10 are determined by using a linear combination of REF and SMTH, i.e., SST_i= [(10−i)SST_ref+ (i)SST_smth]∕10 voor i∈[0, 10] . Members 11–20 are obtained using SST of REF augmented with a spatially non-correlated random perturba- tion field with an amplitude between ± 0.5 K. Finally, Mem- bers 21–30 are obtained by perturbing the SST of SMTH Fig. 1 Mean SST of all REF-storms (shading, units: K). Lines denote

the tracks; REF (full) and SMTH (dashed). The simulation domain is outlined in grey

Fig. 2 As in Fig. 1 but for SST of SMTH (left) and the absolute difference SMTH–REF (right)

How Gulf-Stream SST-fronts influence Atlantic winter storms

1 3

are known to be relevant extratropical storm development:

(1) surface latent heat fluxes (LHF) and (2) low-level baro- clinicity (B). LHF will influence the surface signature of storms by modifying boundary layer processes (e.g. vertical mixing) and may also influence cyclogenesis via the asso- ciated changes in diabatic heating. Low-level baroclinicity associated with strong horizontal temperature gradients and vertical stratification also plays a role in cyclogenesis. In addition to these two parameters the storms will be influ- enced and steered by the mid- to uppertropospheric flow configuration. The average vertical wind-shear of the tropo- sphere determines to first order the maximal incipient cyclo- genesis rate via the classic Eady and Charney mechanisms for baroclinic instability. However, by construction this part of the flow is largely kept constant for each storm-pair in our simulations.

Over the Gulf Stream, especially over the warm-tongue extension, the atmospheric air-temperature in the winter is generally lower than the SST underneath. This temperature contrast gives rise to considerable LHF from the sea to the air. The LHF peaks over the warm tongue, and strongly decreases northward due to the much colder SST (top-left panel, Fig. 5). LHF also decreases southward of the warm

tongue due to higher air temperatures. If the SST-front is absent or replaced by a longitudinal average (as approxi- mately is the case in SMTH) LHF south of the front is much lower and north of it much higher. In the LHF-anomaly field, this becomes manifest as a dipole (bottom-left panel).

The second important difference between SMTH and REF is the change of low-level baroclinicity (B). Baroclinic- ity is crucial for storm development as it renders the basic flow inherently unstable to small disturbances. As discussed before, the upper-level flow is strongly constrained by the lateral boundary conditions, and will thus be quite similar for both REF and SMTH storms (i.e., they will still differ from storm to storm, but the SMTH–REF difference for a single storm case is small). Therefore it makes more sense to study the lower troposphere. We approximate the low- level baroclinicity by the absolute value of the horizontal temperature gradient at 850 hPa

B is proportional to the vertical shear 𝛬 of the wind and to the Eady (1949) growth rate. Averaged over all storms, B is positive everywhere in the domain for all levels up to the mid-troposphere, consistent with a northward decreasing (1) B= |∇T|.

Fig. 4 Timeseries of various parameters for the mini-ensemble of storm 16. The thick black line denotes CTRL, the thick red line SMTH (the other members are explained in the text). Top row: tracks (left) and SST (right). Bottom row shows 𝜙925 (left) and wspdmax (right)

Single-storm ensemble (#16)

track SST

phi925 wspdmax

Example (storm 19)

Latent Heat flx wspd

REF SMTH

20

Latent heat flux & baroclinicity

latent heat flux REF

H. d. Vries et al.

1 3

temperature and positive vertical wind shear. The top- right panel in Fig. 5 shows the average B (850 hPa) for all REF storms. The pattern is not uniform, with the largest amplitudes in the vicinity of the SST-front. The imprint of the underlying SST-front on B is much stronger at 925 hPa (Fig. 10 in Appendix 3), but also remains visible up to altitudes of 700 hPa (not shown). In the SMTH runs, the peak in amplitude of B is completely gone; low-level baro- clinicity over the SST-front is therefore much less and we expect storms traveling over this region to be influenced by it. Meanwhile low-level baroclinicity in the adjacent side- bands is enhanced. The resulting anomaly field resembles a tripole (bottom-right panel). Again, at 925 hPa this is even more clear (Fig. 10).

Given these robust patterns of change in the storm param- eters, the question is whether these can explain the differ- ences in the storm-strength. Prior to answering this question let us first assess whether the two proposed parameters (so not their changes) do actually matter for the current set of REF-storms. Figure 6 shows a 2d-binned statistics plot of instantaneous W10X (scale: 100 km) of all REF-storms,

Fig. 5 Storm parameters. averaged over all simulations. Top-row: REF values for (left) surface latent heat flux and (right) absolute value of the horizontal temperature gradient at 850 hPa. Bottom row: absolute changes (SMTH–REF)

Fig. 6 2d-binned statistics of 3 h W10X using all REF-storms and all times (scale: 100 km). The storm parameters B (850 hPa) and LHF are used as coordinates. Each hexagon is coloured with the average W10X of all points within

H. d. Vries et al.

1 3

temperature and positive vertical wind shear. The top- right panel in Fig. 5 shows the average B (850 hPa) for all REF storms. The pattern is not uniform, with the largest amplitudes in the vicinity of the SST-front. The imprint of the underlying SST-front on B is much stronger at 925 hPa (Fig. 10 in Appendix 3), but also remains visible up to altitudes of 700 hPa (not shown). In the SMTH runs, the peak in amplitude of B is completely gone; low-level baro- clinicity over the SST-front is therefore much less and we expect storms traveling over this region to be influenced by it. Meanwhile low-level baroclinicity in the adjacent side- bands is enhanced. The resulting anomaly field resembles a tripole (bottom-right panel). Again, at 925 hPa this is even more clear (Fig. 10).

Given these robust patterns of change in the storm param- eters, the question is whether these can explain the differ- ences in the storm-strength. Prior to answering this question let us first assess whether the two proposed parameters (so not their changes) do actually matter for the current set of REF-storms. Figure 6 shows a 2d-binned statistics plot of instantaneous W10X (scale: 100 km) of all REF-storms,

Fig. 5 Storm parameters. averaged over all simulations. Top-row: REF values for (left) surface latent heat flux and (right) absolute value of the horizontal temperature gradient at 850 hPa. Bottom row: absolute changes (SMTH–REF)

Fig. 6 2d-binned statistics of 3 h W10X using all REF-storms and all times (scale: 100 km). The storm parameters B (850 hPa) and LHF are used as coordinates. Each hexagon is coloured with the average W10X of all points within

850hPa temp.gradient ~ baroclinicity

21

Changes SMTH-REF

difference in latent heat flux SMTH-REF difference in low level baroclinicity SMTH-REF Stronger at 925hPa, less above

Sensible heat fluxes near surface Latent heat fluxes higher up