SEARCH

SEARCH BY CITATION

Keywords:

  • Surface fluxes;
  • Denmark Strait;
  • Nordic Seas

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Summary
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] The impact of extreme Nordic Seas heat loss on Denmark Strait (DS) dense water transport is examined in (1) control runs of the Hadley Centre HadCM3 and HadGEM1 coupled climate models and (2) perturbation experiments with the fast coupled model FORTE that allow heat flux effects to be isolated from wind stress. All three models show an approximately linear increase in southward DS transport of cold dense water with increasing Nordic Seas winter heat loss in the range −80 to −250 Wm−2. The propagation of the cold anomaly from the Nordic Seas source along a trajectory through the DS and into the Irminger Basin is also examined. A common response time is found with the strongest decrease in DS temperature occurring within 8–12 months of the heat loss signal. Our results show that Nordic Seas heat loss must be considered in addition to other processes in understanding DS variability.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Summary
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] The formation of dense water in the Nordic Seas and its subsequent transport through the Denmark Strait (DS) is a key process within the global climate system. Warm, saline surface waters from the Atlantic are carried north over the Greenland-Scotland Ridge into the Nordic Seas where dense water forms as these surface waters undergo surface cooling and vertical mixing [Hansen and Østerhus, 2000]. The dense water that exits the Nordic Seas to the west of Iceland is termed Denmark Strait Overflow Water (DSOW). After passing through the DS, the DSOW becomes the densest part of North Atlantic Deep Water and as such is an important component of the Atlantic Meridional Overturning Circulation (MOC) [Macrander, 2004]. Despite the importance of theses processes, the potential impact of climate change on dense water formation and its transport through the DS, as well as the implications for the Atlantic MOC, are not well understood. Holland et al. [2006] recently examined projected changes in the Nordic Sea freshwater budget in the Community Climate System Model Version 3. They found that although convection in the Nordic Seas became less frequent by the end of the twenty first century, it was maintained to a greater extent than in the Labrador Sea. Earlier analyses of the Hadley Centre coupled ocean-atmosphere model HadCM3, have also indicated that reductions in the density of Nordic Seas overflows are important for weakening Labrador Sea convection which in turn leads to a slowdown of the MOC [Wood et al., 1999].

[3] Recent observations indicate that changes of up to 30% in the DS overflow transport can occur on interannual timescales [Macrander et al., 2005, 2007]. The task of understanding current variability and predicting future changes in DS transport is complicated by the fact that the strength of the overflow is determined by a number of factors operating on different time scales. In particular, there appears to be both a barotropic response to surface wind forcing and a baroclinic response to upstream changes in density [Biastoch et al., 2003; Macrander et al., 2005; Köhl et al., 2007]. In addition, some modes of atmospheric variability can jointly influence both of these processes. Here we examine the response of the DS overflow to anomalous heat flux forcing in the Nordic Seas. This is achieved using a range of models, (1) HadCM3, (2) HadGEM1, the latest Hadley Centre coupled model, included in the IPCC Fourth Assessment Report [IPCC, 2007] and (3) FORTE, a coarser resolution climate model [Sinha and Smith, 2002] which we use for a range of model experiments with varying heat flux forcing. The FORTE experiments enable us to isolate the spatial and temporal character of the ocean response to winter heat loss from other processes, in particular wind forcing. The research presented here is a short paper describing results obtained from an extension of an earlier analysis of HadCM3 alone [Grist et al., 2007]. In the earlier study, we found that the Denmark Strait transport was dependent on heat loss in the Nordic Seas but were unable to consider more than one model. Here we explore whether the same relationship holds in two further models, one of which enables us to isolate the effects of heat flux forcing from wind driven changes. In addition, we investigate in detail for the first time the propagation of the heat flux induced Nordic Seas cold anomaly toward and through the Denmark Strait into the Irminger Basin. We will show that the DS transport increases in response to stronger Nordic Seas winter heat loss in all 3 models and thus conclude that heat flux anomalies are a key process that must be considered in order to understand variations in the circulation and overflow characteristics.

2. Method

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Summary
  7. Acknowledgments
  8. References
  9. Supporting Information

[4] We have analyzed 1000 years of the HadCM3 control run [Gordon et al., 2000] and 240 years of the HadGEM1 control run [Johns et al., 2006]. HadCM3 has 19 levels in the atmosphere with a horizontal resolution of 2.5° × 3.75° and 20 levels in the ocean with a horizontal resolution of 1.25° × 1.25°. HadGEM1 has an atmospheric resolution of 1.25° × 1.875° with 38 vertical layers. The ocean has 40 layers with 1° longitudinal resolution. The latitudinal resolution is 1° between the poles and 30°N and then increases smoothly to 1/3° at the equator. As well as increased resolution, HadGEM1 contains a number of other changes in formulation, including a new nonhydrostatic atmosphere and interactive aerosols. In addition, the sea-ice model within HadGEM1 now resolves subgrid-scale ice thickness with its temporal evolution being determined by thermodynamic growth/melt, advection, and redistribution by ridging. The control runs of both HadCM3 and HadGEM1 are run with preindustrial greenhouse gases and without flux adjustments.

[5] FORTE has a 2° × 2° horizontal resolution ocean and a T42 spectral atmosphere, roughly equivalent to 2.8° × 2.8° resolution. Both the atmosphere and the ocean have 15 vertical levels [Sinha and Smith, 2002]. It has been used in a number of previous climate studies [e.g., Blaker et al., 2006; Smith et al., 2006]. We have carried out 7 experimental runs of FORTE with different prescribed winter heat loss in a Nordic Sea box (67°N–79°N, 19°W–9°E). Each experiment is initiated from the model state after 160 years of a control run and carbon dioxide is held at 358 ppm. The prescribed heat loss values in the 7 runs are 100, 300, 400, 475, 550, 625 and 700 Wm−2 (referred to as P100, P300, P400, P475, P550, P625, P700, respectively). The heat loss is prescribed only for the first winter (December–March, 120 model days) with the model evolving freely thereafter over a total integration of 10 years in each experiment. As the temperature of some of the Nordic Seas box grid cells becomes cold enough for ice formation to occur and subsequently reduce the prescribed heat loss, the actual first winter mean heat loss in each case is less than the prescribed value (Table 1).

Table 1. Area Averaged DJFM Net Heat Flux for the Nordic Seas Box for the 7 FORTE Experiments Described in the Main Text and the Control Run
 ControlP100P300P400P475P550P625P700
Heat flux (Wm−2)−138−75−185−226−298−340−389−427

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Summary
  7. Acknowledgments
  8. References
  9. Supporting Information

3.1. Heat Flux Forcing and Mixed Layer Response

[6] We first consider the strength of the heat flux forcing in HadCM3 and HadGEM1. Time series of the Nordic Seas box winter (DJFM) net heat flux for these models over comparable 250 year periods are shown in Figure 1a. Both models exhibit substantial interannual variability and HadGEM1 has an overall mean heat loss, −197 Wm−2, which is about 20% stronger than the corresponding value, −165 Wm−2, for HadCM3. Much of this difference is due to a reduction in the fractional ice coverage (Figure 1b) in HadGEM1, mean (and standard deviation) 0.12 (±0.04), compared to HadCM3, mean 0.35 (±0.10). The HadGEM1 ice distribution reflects improvements to the ice model used by HadCM3 and is closer to present day observations [Connolley et al., 2006]. Specifically, using the HadISST1.1 Global sea-ice data set [Rayner et al., 2003] we calculated the mean DJFM ice cover for 1980–1999 to be 0.17 ± 0.04 for the Nordic Seas box. Possibly as a result of the aforementioned difference in net heat flux, the mean mixed layer depth for the Nordic Seas is less in HadCM3 than it is in HadGEM1. Figure 1c shows the time series of Nordic Seas area-averaged March mixed layer depth for the two models. In HadCM3 the area-averaged mixed layer depth is 179 (±64) m and for HadGEM1 it is 64 m greater at 243 (±52) m. Monterey and Levitus [1997] have calculated the Global Ocean mixed layer depths based on the World Ocean Atlas 1994 [Levitus and Boyer, 1994]. From this data we have calculated the Nordic Seas area-averaged March mixed layer depth to be 305 m. The HadGEM1 mixed layer depth is therefore closer to observations than HadCM3 in the study region.

image

Figure 1. Time series of 250 years of the control runs of HadCM3 (blue) and HadGEM1 (red) for DJFM of area-averaged Nordic Seas: (a) net heat flux (Wm−2), (b) fractional ice coverage, and (c) March mixed layer depth (m). The mean and standard deviation of the full 1000 years of the HadCM3 control run are shown with solid and dot-dashed black lines, respectively.

Download figure to PowerPoint

[7] Although the Hadley Centre models exhibit differences in terms of the average Nordic Seas net heat flux, ice-cover and mixed layer depth, the models exhibit a similar sensitivity of the region's mixed layer depth to heat flux variability. A composite figure showing the difference between HadGEM1 winter heat loss for the 10 strongest heat loss winters and the 10 weakest winters is shown in Figure 2. Differences greater than 200 Wm−2 are observed in the Nordic Seas. The corresponding difference in mixed layer depth is also shown. It is evident that in HadGEM1, extreme winter heat loss over the Nordic Seas is associated with a deepening of the mixed layer depth (up to 400 m); similar values have been previously found in HadCM3 [Grist et al., 2007]. The deepening of the Nordic Seas mixed layer is accompanied by a shoaling in the Labrador Sea. This is consistent with observational evidence that the convective activity at these two sites is in anti phase [Dickson et al., 1996].

image

Figure 2. Composites of the 10 strongest minus 10 weakest heat loss winters for the 240 years of the HadGEM1 control run: contours show net heat flux difference (solid contours are −50 Wm−2, −100 Wm−2, −150 Wm−2, and −200 Wm−2, dashed contours are 50 Wm−2 and 100 Wm−2) and colored field shows mixed layer depth difference (m). The black box delineates the area defined as the Nordic Seas (NS) box in the study. The diagonal black line denotes the location of the cross-section (Figure 7) from NS through the Denmark Strait (DS) to the Irminger Basin (IB).

Download figure to PowerPoint

3.2. Impact of Nordic Seas Heat Loss on Denmark Strait Transport

[8] Is the increased convection associated with stronger winter heat loss over the Nordic Seas in HadGEM1 sufficient to modify the properties and the quantity of the water subsequently transported south through the Denmark Strait? We have already found some evidence that this is the case within HadCM3 [Grist et al., 2007] but we now seek to strengthen this result using the more realistic HadGEM1 control run and the FORTE experiments to isolate the effects of heat flux forcing from wind anomalies. To this end we have grouped the 1000 (240) values of HadCM3 (HadGEM1) Nordic Seas box winter net heat flux into bins of equal heat flux interval. We then determined the mean and standard deviation of annual southward transport through DS within each bin for various density and temperature classes.

[9] The variation of dense water overflow with winter mean net heat flux is shown in Figure 3. For consistency with Grist et al. [2007], the overflow waters are characterized as having a density σθ > 27.5 kg m−3. Similar results are obtained using other reasonable choices for the water density threshold, for example σθ > 27.8 kg m−3. The transport values in Figure 3 have been normalized by the overall mean for each model to facilitate intermodel comparison. The figure reveals a similar response for both HadCM3 and HadGEM1 with an increase in the southward transport as the heat loss increases. The slope of this relationship is less steep for HadGEM1 than HadCM3 and this may reflect a reduction in the strength of the response given the more accurate ice model in HadGEM1. In terms of the dimensional change in DS transport water with σθ > 27.5 kg m−3, an increase in heat loss from −80 to −250 Wm−2 corresponds to an increases of 2.2 Sv/1.3 Sv for HadCM3/HadGEM1.

image

Figure 3. Variation of normalized Denmark Strait overflow with Nordic Seas winter net heat flux (Wm−2). The overflow is characterized as the annual mean southward transport of water with σθ > 27.5 kg m−3. The transport values of the Hadley models have been sorted into bins of equal winter heat flux intervals: HadCM3 (blue “+”) and HadGEM1 (red “x”). The error bars represent the standard deviation of transport within the bins. The green diamonds represent the transports from the FORTE experiments.

Download figure to PowerPoint

[10] HadCM3 and HadGEM1 produce a similar response of the overflow to heat flux forcing despite differences in the mean water characteristics between the models in the DS. These differences are evident in Figure 4, which shows cross-sections of temperature, salinity and potential density for each model. We note that in each case the DS cross section is the average of two latitude bands on the tracer grid. As a consequence the Strait is calculated as being the narrower of the two bands. The mean temperature in the deepest part of the DS (below 500 m) is 4.3°C in HadCM3 and 1.8°C in HadGEM1, whereas salinity is 35.4 psu in HadCM3 and 35.0 psu in HadGEM1. These differences partially compensate in density terms, such that in both models overflow waters have a potential density of about 27.8 kg m−3 and greater. A key difference between the two models is that HadCM3 has a significant proportion of overflow at very high densities, i.e., greater than 28.2 kg m−3 [Johns et al., 2006] The FORTE model is warmer (7.2°C) and more saline (35.8 psu) in the deepest part of the Denmark Strait (Figure 5). This corresponds to a density of around 28.1 kg m−3. Although this is greater than the mean for the overflow in the Hadley models, it is within the limit of the observable range [Macrander, 2004]. As the DS values for T and S in the different models span the range of observational uncertainty we are able to assess the robustness of the DS response to heat flux for various possible water mass characteristics.

image

Figure 4. Cross sections of Denmark Strait mean potential temperature (°C), salinity (psu) and potential density (kg m−3) for the control runs of (a, c, and e) HadCM3 and (b, d, and f) HadGEM1.

Download figure to PowerPoint

image

Figure 5. Cross sections of Denmark Strait mean (a) potential temperature (°C), (b) salinity (psu) and (c) potential density (kg m−3) for the control run of FORTE. The color scale is different to that used in Figure 4.

Download figure to PowerPoint

[11] Results from the Hadley Centre models indicate a strong influence of the heat flux on DS transport. However, in order to isolate the impact of the heat exchange from other processes, we have undertaken the heat flux forcing only experiments with FORTE, described in section 2. Results from these experiments are also included in Figure 3 and these again show increasing DS transport with stronger Nordic Seas heat loss, with the relationship being in good agreement with HadGEM1 over the heat flux range common to both models (−70 to −250 Wm−2). Thus our result that DS southward dense water transport increases with Nordic Seas heat loss appears robust across both sophisticated general circulation models and a relatively coarse resolution fast coupled climate model. In FORTE, the response in DS transport is only associated with prescribed anomalous heat forcing. In the Hadley control runs the response in the DS may be partially due to other factors, such as wind forcing. However, the fact that the transport-heat flux gradients are similar to that in FORTE suggests that heat flux effects may also be dominant in the Hadley models.

[12] FORTE also enables us to investigate the ocean response under more extreme heat loss conditions (stronger than −300 Wm−2) than encountered in the Hadley Centre model runs and in the much shorter observational record. A fuller discussion of the observed North Atlantic surface fluxes is included in Grist et al. [2007]. We have calculated the mean, standard deviation, maximum and minimum observed net heat flux from 1948–2005 from NCEP reanalysis [Kalnay et al., 1996] over the Nordic Seas Box (see Table 2). Comparison of Tables 1 and 2 shows that in the FORTE experiments we are exploring the effect of heat loss, across the observable range and beyond. Using heat flux anomalies greater than ordinarily observed enables us to more easily identify the effects of heat loss. In addition, the stronger heat loss conditions may become more frequent in the future as the amount of winter ice insulating the Nordic Seas reduces. The extreme heat loss FORTE experiments show that the dense transport continues to increase as the winter mean heat loss approaches 400 Wm−2, although there is some indication that this increase levels off at the strongest heat loss values (Figure 3). Further work will be required to clarify if this leveling off feature is robust and what mechanisms are responsible for it.

Table 2. Area Averaged DJFM Net Heat Flux (Wm−2, Maximum, Mean and Minimum) for the Nordic Seas Box as Calculated From the NCEP Reanalysis 1948–2005
 MaximumMean (± Std. Dev.)Minimum
Nordic Seas Box−137−185 ± 25−264
Nordic Seas Box (ice free part)−153−215 ± 31−294

3.3. Anomaly Propagation

[13] In this section, we investigate the propagation of the heat flux induced cold water anomalies toward the Denmark Strait. Before doing so, we examine the variation of temperature and potential density profiles in the Nordic Seas in response to heat flux forcing.

[14] The Nordic Seas March mean potential temperature profile for the FORTE control run is shown along with the profile from the March of experiment P550 in Figure 6a. In P550, as a consequence of the intense cooling of the surface waters there is mixing down to a depth of around 800 m. The mean control profile is cold (of the order 0°C) at the surface, with a subsurface maximum at 400 m, thus the vertical mixing leads to a surface warming and cooling centered around 400 m, consistent with the results by Gamiz-Fortis and Sutton [2007]. A similar ocean response occurs following strong heat loss events within the control runs of HadCM3 and HadGEM1. This is seen in Figures 6c and 6e, which show the potential temperature profiles from the composites of the 10 strongest and 10 weakest Nordic Seas heat loss winters from the respective control runs. The weak heat loss profiles are cooler than the FORTE control run and the subsurface temperature maximum is found at a shallower depth, (approximately 200 m). The profiles in the Hadley models strong heat loss years show the characteristic surface warming, and subsurface cooling associated with vertical mixing; however, because they are composites of different years, the mixing profiles are not as clearly defined as for P550. The temperature profiles indicate mixing in the Hadley models penetrates to about 400 m during the strong heat loss years compared to about 700 m in P550.

image

Figure 6. Mean vertical profiles of potential temperature and potential density in March for grid cells in the Nordic Seas for the different models. (a) and (b) FORTE (73.00°N, 7.00°W), where the solid black line is the mean of the control run, and the dashed gray line is for the first winter of the P550 experiment. (c) and (d) HadCM3 (73.13°N, 7.50°W), where the solid black line is the composite of the 10 weakest heat loss years (WL) over the Nordic Seas, and the dashed gray line is the composite of 10 strongest heat loss years (SL) over the Nordic Seas. (e) and (f) same as (c) and (d) but for HadGEM1 (75.00°N, 8.00°W).

Download figure to PowerPoint

[15] The corresponding potential density profiles are shown in Figures 6b, 6d, and 6f. The surface heat loss and the resultant vertical mixing causes the density increase in the top 600 m in FORTE, the top 200 m in HadCM3 and the top 150 m in HadGEM1. The effect of heat loss on potential density above 600 m is of interest because this is the approximate height of the Denmark Strait sill. Under conditions of hydraulic control the transport of water denser than a given density, σθcrit, through the Denmark Strait will be proportional to the upstream height of the σθcrit isopycnal above the sill [Whitehead, 1998]. Köhl et al. [2007] have recently argued that this relationship holds for transport variability of timescales greater than a week. Although not sufficient to confirm the hydraulic control of DSOW transport [Girton et al., 2006], the results in Figure 6 are consistent with this mechanism. That is, to say in years of strong heat loss (and increased DSOW transport), the density of the water in the near surface layers increases which implies an upward displacement of specific isopycnals relative to the depth of the sill.

[16] The detailed time evolution of the ocean changes occurring in response to strong Nordic Seas heat loss are now investigated using the FORTE model runs. Here we present results for the P550 experiment. The variation of temperature anomaly with depth along a path (see Figure 2) running from the Nordic Seas through DS and into the Irminger Basin, at various times with respect to the start of the imposed anomaly, is shown in Figure 7. During the second month in which the heat flux anomaly is applied there is a reduction in the surface temperature of 2.5°C but no major change with depth (Figure 7a). By month 4, vertical mixing of the cold surface water causes a 2.5°C temperature decrease at around 700 m. In addition, because the Nordic Seas have a subsurface temperature maximum at around 400 m, the mixing results in a warming of the surface layer [Gamiz-Fortis and Sutton, 2007]. The cold anomaly begins to spread along the track to ward the DS (Figure 7b). By month 8 this anomaly has propagated over the DS sill and is confined to the column adjacent to the sill, reaching a depth of 2000 m (Figure 7c). Nearly a year after the heat flux signal was first applied the cold water anomaly can be seen as a plume spreading out into the Irminger basin while a strong residual feature of up to 1.5°C remains at depths of 400–800 m in the Nordic Seas (month 12, Figure 7d). A feature of z-layer models is that the dense water tends to mix too efficiently on the down stream side of a sill. A well-documented consequence of this is that the return flow of the meridional overturning circulation is too shallow [e.g., Winton et al., 1998; Tang and Roberts, 2005]. Observations suggest that on exiting the DS, cold overflow waters are conserved to a greater extent than indicated in Figure 7d [Tang and Roberts, 2005]. This feature of z-layer models does not influence the manner of response of DSOW to Nordic Seas heat flux anomalies, which is our main result.

image

Figure 7. Cross section showing the variation of temperature anomaly (°C) with depth (m) in the P550 FORTE experiment along the path (shown in Figure 2), running approximately NE to SW from the Nordic Seas (NS) through Denmark Strait (DS) and into the Irminger Basin (IB). The anomalies are with respect to the control run and the fields are shown for various lags with respect to the start (December) of the imposed heat flux anomaly: (a) 2 months, (b) 4 months, (c) 8 months, and (d) 12 months. The letters denote grid cells for which time series are plotted in Figure 7.

Download figure to PowerPoint

[17] The temporal evolution of the temperature anomaly at 4 key grid cells, labeled in Figure 7a, is shown in Figure 8. The region beneath the heat flux signal (point A) experiences a rapid drop in temperature, with the strongest cooling of 2.6°C reached 4 months after the start of the perturbation. This minimum is maintained for 2 months, subsequently the temperature returns over the course of the next 4–5 years to the value found in the control run. Points B and C, located north and south of the DS respectively, and point D, 1000 m below the southern edge of the sill, respond with increasing time and decreasing magnitude with distance from the heat flux anomaly. At B, the peak response is −1.7°C and occurs after 9 months; with values below −1.5°C being maintained for nearly a year. The potential density anomaly at point B is also shown in Figure 8 (right hand axis). The increase in the density of the DS waters associated with the temperature anomaly is clearly evident. South of the sill, the minimum values are −0.8°C at C and −0.1°C at D after 14 months. The response at point D is less pronounced than at C and is less than the standard deviation of the temperature anomaly time series. Analysis of the temperature at the southern side of the DS (point C) in the other integrations revealed the timescale of the response to be relatively insensitive to the magnitude of the heat flux.

image

Figure 8. (a) Time series of the temperature anomalies (FORTE P550 minus control run) for grid cells indicated in Figure 7a). A is shown by the black line. B is shown by the gray line. C is shown by the dashed black line. D is shown by the dashed gray line. The thick gray line (right-hand axis) is the corresponding potential density anomaly at point B. (b) Time series of temperature anomaly for point C (Forte P550 minus control run), green line and the corresponding temperature anomalies at point C in HadCM3, blue line and HadGEM1, red line, averaged over the five strongest heat loss years.

Download figure to PowerPoint

[18] Figure 9 shows the variation of the temperature anomaly at point C for the experiments P300, P400, P550 and P700. A common feature of the experiments is that the strongest temperature anomaly typically occurs between 8 and 14 months after the onset of the prescribed heat flux. However, as one might expect the magnitude of the change in DS temperature increases with the size of the prescribed heat flux anomaly. For example, the maximum anomaly between 8 and 14 months was just −0.2°C for P300 but it was −1.0°C for P700. Out of the 7 experiments conducted, the 5 with the strongest heat flux anomalies yielded a lagged response in temperature at point C (the southern side of the DS) that was at least twice the standard deviation of the control temperature at that point.

image

Figure 9. Time series of the temperature anomalies (°C) at the Denmark Strait (Point C in Figure 7a) following a winter of strong heat loss: P300 minus control (dashed black line), P400 minus control (dashed gray line), P550 minus control (solid black line) and P700 minus control (solid gray line).

Download figure to PowerPoint

[19] The time series of the temperature anomaly from the winters of the 5 strongest Nordic Seas heat loss for grid cells comparable to point C in HadGEM1 and HadCM3 are shown in Figure 8b. In HadCM3, the response appears slightly stronger than in FORTE, whilst in HadGEM1 it is slightly weaker. There are also some differences in the time-lag associated with the DS response. Despite these differences, all 3 models show a clear decrease in DS temperature at point C within a year of the onset of strong Nordic Seas winter heat loss.

4. Summary

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Summary
  7. Acknowledgments
  8. References
  9. Supporting Information

[20] The relationship between Nordic Seas net heat flux and Denmark Strait (DS) dense water transport has been explored using three different coupled climate models: HadCM3, HadGEM1 and FORTE. Our analysis has built on earlier work with HadCM3 alone [Grist et al., 2007] and has allowed us to establish a common response of the DS transport to Nordic Seas heat loss. Furthermore, we have now also examined for the first time the propagation of the cold water anomaly, generated by extreme winter surface cooling, from the Nordic Seas source region to the Denmark Strait.

[21] Each of the three models shows an approximately linear increase in southward DS transport of dense water with increases in heat loss over the Nordic Seas. In addition, the FORTE experiments have enabled us to isolate the spatial and temporal character of the ocean response to winter heat loss, in the absence of other processes, in particular wind forcing. In these experiments, strong winter heat loss results in cooling of the water down to 700 m. The associated temperature anomaly propagates south to the Denmark Strait and subsequently descends to a depth of 2000 m on the southern side of the sill. In all three models there is a response in the DS temperature within 8–12 months of the start of Nordic Seas winter heat loss.

[22] The consistent results obtained with HadCM3, HadGEM1 and FORTE demonstrate that extreme heat loss in the Nordic Seas is a key factor in determining the magnitude of the Denmark Strait overflow which must be considered in addition to other effects such as wind forcing. This underlines the importance of improved heat flux estimates for the Nordic Seas, particularly in light of the increasing role of deep water formation in this region in projections of twenty first century climate [Holland et al., 2006].

[23] Recent observations have revealed dramatic reductions in Arctic ice cover over the past two decades [Serreze et al., 2007; Stroeve et al., 2007]. Such reductions in ice cover are likely to lead to a substantial modification of the air-sea heat exchange with the potential for major increases in heat loss in regions newly exposed to the atmosphere. Our study has revealed that variations in the Nordic Seas air-sea exchanges have a significant impact on dense water formation and subsequently the transport through the Denmark Strait. More generally, as the ice cover in the Arctic continues to retreat in response to anthropogenic climate change, variations in the surface heat exchange are likely to play a key role in modifying the circulation characteristics of the Arctic ocean and its marginal regions, including the Nordic Seas.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Summary
  7. Acknowledgments
  8. References
  9. Supporting Information

[24] The Hadley Centre model output was obtained from the British Atmospheric Data Centre (BADC). This research has been funded by the Grant NER/T/S/2002/00427 of the UK Natural Environment Research Council Rapid Climate Change program.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Summary
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Method
  5. 3. Results
  6. 4. Summary
  7. Acknowledgments
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
jgrc10944-sup-0001-t01.txtplain text document0KTab-delimited Table 1.
jgrc10944-sup-0002-t02.txtplain text document0KTab-delimited Table 2.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.