Projected Changes in Surface Air Temperature and Surface Wind in the Gulf of St. Lawrence

Abstract

The impacts of climate change on surface air temperature (SAT) and winds in the Gulf of St. Lawrence (GSL) are investigated by performing simulations from 1970 to 2099 with the Canadian Regional Climate Model (CRCM), driven by a five-member ensemble. Three members are from Canadian Global Climate Model (CGCM3) simulations following scenario A1B from the Intergovernmental Panel on Climate Change (IPCC); one member is from the Community Climate System Model, version 3 (CCSM3) simulation, also following the A1B scenario; and one member is from the CCSM4 (version 4) simulation following the Representative Concentration Pathway (RCP8.5) scenario. Compared with North America Regional Reanalysis (NARR) data, it is shown that CRCM can reproduce the observed SAT spatial patterns; for example, both CRCM simulations and NARR data show a warm SAT tongue along the eastern Gulf; CRCM simulations also capture the dominant northwesterly winds in January and the southwesterly winds in July. In terms of future climate scenarios, the spatial patterns of SAT show plausible seasonal variations. In January, the warming is 3°–3.5°C in the northern Gulf and 2.5°–3°C near Cabot Strait during 2040–2069, whereas the warming is more uniform during 2070–2099, with SAT increases of 4°–5°C. In summer, the warming gradually decreases from the western side of the GSL to the eastern side because of the different heat capacities between land and water. Moreover, the January winds increase by 0.2–0.4 m s⁻¹ during 2040–2069, related to weakening stability in the atmospheric planetary boundary layer. However, during 2070–2099, the winds decrease by 0.2–0.4 m s⁻¹ over the western Gulf, reflecting the northeastward shift in northwest Atlantic storm tracks. In July, enhanced baroclinicity along the east coast of North America dominates the wind changes, with increases of 0.2–0.4 m s⁻¹. On average, the variance for the SAT changes is about 10% of the SAT increase, and the variance for projected wind changes is the same magnitude as the projected changes, suggesting uncertainty in the latter.

Résumé

[Traduit par la rédaction] Nous examinons les impacts des changements climatiques sur la température de l'air et sur les vents en surface dans le golfe du Saint-Laurent, à l'aide de simulations allant de 1970 à 2099, réalisées avec le Modèle régional canadien du climat (MRCC), que pilote un ensemble de cinq membres. Trois des membres proviennent de simulations du Modèle couplé climatique global canadien (MCCG3), suivant le scénario A1B du Groupe d'experts intergouvernemental sur l’évolution du climat (GIEC). Un autre membre est issu de la troisième version du modèle communautaire du système climatique (CCSM3), suivant le même scénario. Le dernier membre vient de la quatrième version de ce même modèle (CCSM4), suivant le profil représentatif d’évolution des concentrations (RCP8.5). En comparant les données nord-américaines de réanalyse régionale (NARR) avec celles du MRCC, nous démontrons que le modèle peut reproduire les distributions spatiales des températures de l'air en surface. Par exemple, les simulations du MRCC et les données de NARR produisent une langue chaude de températures de l'air en surface dans l'est du golfe. Les simulations du MRCC reproduisent aussi les vents dominants du nord-ouest en janvier et du sud-ouest en juillet. En ce qui concerne les scénarios de climat futur, les caractéristiques spatiales de la température de l'air en surface reflètent des variations saisonnières plausibles. En janvier, entre 2040 et 2069, le réchauffement est de 3 à 3,5°C dans la portion nord du golfe et de 2,5 à 3°C près du détroit de Cabot, tandis que le réchauffement devient plus uniforme entre 2070 et 2099, avec une augmentation des températures de 4 à 5°C. En été, le réchauffement diminue graduellement, de la partie ouest du golfe du Saint-Laurent vers sa partie est, et ce, en raison des capacités calorifiques différentes entre la terre et l'eau. De plus, entre 2040 et 2069, les vents de janvier augmentent de 0.2 à 0.4 m s⁻¹ dû à l'atténuation de la stabilité dans la couche limite planétaire. Toutefois, entre 2070 et 2099, les vents diminuent de 0.2 à 0.4 m s⁻¹ au-dessus de la portion ouest du golfe. Ce changement suit le déplacement vers le nord-est des trajectoires des tempêtes de l'Atlantique Nord-Ouest. En juillet, une intensification de la baroclinicité le long de la côte est nord-américaine domine les changements de vents, entraînant des augmentations de 0.2 à 0.4 m s⁻¹. En moyenne, la variance des changements de la température de l'air en surface s’élève à environ 10% de la hausse de cette température. La variance des changements prévus de vents reste du même ordre que les changements prévus, laissant ainsi planer une incertitude relativement aux vents.

KEYWORDS:

• climate change
• surface air temperature and winds
• Gulf of St. Lawrence
• Canadian Regional Climate Model (CRCM)
• IPCC scenario A1B

1 Introduction

The Gulf of St. Lawrence (GSL) is almost an inland sea opening to the Atlantic Ocean through Cabot Strait and the Strait of Belle Isle. It is the habitat of many fishery stocks, which are susceptible to possible changes in water temperature and salinity. Because the GSL is a relatively shallow area surrounded by land, the accuracy of the atmospheric surface driving fields is critical for simulations of its ocean circulation (Long et al., 2015; Saucier et al., 2003).

To represent the detailed features of surface fields over the GSL, the Canadian Regional Climate Model (CRCM) is employed to downscale global climate model (GCM) simulations to provide high-resolution surface fields, which can be used to drive a coastal ocean model (Guo et al., 2013; Long et al., 2015) or a wave model. Previous studies suggest that CRCM can be an alternative to GCMs for climate change studies at relatively small scales (Laprise et al., 2003; Pan et al., 2001; Plummer et al., 2006; Long et al. 2009, 2015). The advantages of the CRCM are that it can be run for comparatively long periods of time at relatively high resolutions, at an affordable computational cost, implemented over a limited regional area, driven with atmospheric data simulated by a GCM. Moreover, the CRCM is shown to have skill in downscaling large-scale information to regional scales and in successfully reproducing the means and variations of a number of surface fields (Denis et al., 2002; Guo et al., 2013; Long et al., 2015).

Driven with outputs from the CRCM simulations, the Canadian Océan Parallélisé (OPA) model, denoted CANOPA, has been shown to successfully reproduce sea ice, water temperature, and salinity in the GSL (Long et al., 2015). Compared with observations, CANOPA is able to simulate the sea surface temperature well and to capture the observed vertical structure of water temperature and salinity in the central GSL. In addition, the simulated sea-ice concentration and volume are shown to have similar patterns to those seen in observations. In terms of the possible future climate, CANOPA simulations suggest that by the late 2060s, the GSL will be largely ice free in January, with the ice volume in March steadily decreasing from about 80 km³ in the 1980s to near zero. On average, the water in the GSL will become warmer and fresher in a warmer climate. Moreover, compared with the present climate, the cold intermediate layer (CIL) is expected to become significantly weaker in 2040–2069 than it was in 1980–2009 (Long et al., 2015).

However, the results in Long et al. (2015) are based on a simulation with a one-member ensemble, and uncertainty could present a key challenge for future climate projections. Recent studies show that the dominant source of uncertainty in the simulated climate response at middle and high latitudes is internal atmospheric variability, which is estimated to account for at least half the inter-model spread in projected climate trends (Deser et al., 2012). Here, the CRCM is used to downscale five GCM simulations in order to understand the impacts of climate changes on surface air temperature (SAT) and wind fields in the GSL. The results presented here are based on a five-member ensemble. Section 2 describes the models and the experimental design. Section 3 shows the current climate. The impacts of climate change are given in Section 4. The effects of SAT adjustment are discussed in Section 5, and Section 6 presents the conclusions.

2 Model description and experimental design

The CRCM is based on the dynamical formulation of the Canadian Mesoscale Compressible Community (MC2) model and solves the fully elastic non-hydrostatic Euler equations using a semi-implicit, semi-Lagrangian numerical scheme (Caya and Laprise, 1999). The physical parameterization package of the second-generation Canadian Global Climate Model (CGCM2) follows McFarlane et al. (1992) and is implemented to solve the subgrid-scale processes. The Kain-Fritsch scheme is used for deep convection, and large-scale condensation is simulated using the CGCM2 physics formulation (Kain and Fritsch, 1990; Paquin and Caya, 2000). In this study, CRCM is implemented over a domain focused on eastern Canada and the northwest Atlantic (Fig. 1). The simulations are performed at a horizontal resolution of 25 km, with 29 levels in the vertical. A 15-minute time step is employed. A detailed description of CRCM is given by Caya and Laprise (1999), Laprise et al. (2003), and Caya and Biner (2004).

Fig. 1 Model domain for CRCM.

Five GCM simulations (1970–2099) were downscaled using CRCM, which include three simulations for the Intergovernmental Panel on Climate Change (IPCC) scenario A1B from CGCM3, one simulation for scenario A1B from the Community Climate System Model, version 3 (CCSM3), and one simulation for the Representative Concentration Pathway 8.5 (RCP8.5) scenario from CCSM, version 4 (CCSM4). The A1B scenario is one of three A1 scenario groups adopted by the IPCC for the fourth Assessment Report (AR4). These scenarios include fossil fuel energy intensive (A1FI), non-fossil fuel energy intensive (A1T), and balanced energy intensive across all energy sources (A1B). The A1B scenario predicts increasing carbon dioxide emissions until around 2050 and decreasing thereafter. By comparison, RCPs are four greenhouse gas concentration trajectories adopted by the IPCC for its fifth Assessment Report (AR5). The four RCPs, specifically RCP2.6, RCP4.5, RCP6, and RCP8.5, are named for the possible range of radiative forcing values estimated for the year 2100 relative to pre-industrial values (+2.6, +4.5, +6.0, and +8.5 W m⁻², respectively). The GCM simulations provide CRCM with initial and lateral boundary conditions (wind, temperature, surface pressure, and specific humidity). Because CRMC is not coupled to an ocean model, sea surface temperature (SST) and sea-ice fields have to be specified by the GCM outputs. However, the CGCM3 simulations tend to overestimate SST in the GSL, which leads to significant bias in surface wind speeds in CRCM simulations (Guo et al., 2013). Therefore, following Guo et al. (2013), the SST fields from the GCMs were corrected using the North American Regional Reanalysis (NARR) six-hourly SST climatology data to reduce the bias in SST fields from the GCMs (Mesinger et al., 2006). This is achieved by application of the relation(1)

where the and are six-hourly SST averages of outputs from the GCM simulations and NARR data over the 1979–2008 period. The horizontal resolution of NARR is 32 km and includes assimilation of precipitation along with other variables. The variable in Eq. (1) is used to drive CRCM, which effectively removes the bias in the GCM SST present climatology but keeps the GCM SST variability and long-term trends. In this study, surface wind and SAT for the present climate, represented as 1970–1999, are validated with NARR data over the 1979–2008 period, and the resulting analyses are based on the five-member ensemble for 1970 to 2099.

3 Present climate

a Surface Air Temperature

The SAT is used to estimate longwave radiation and surface heat flux and constitutes an important variable for driving a coastal ocean model. In January, the SAT over land is relatively colder than that over the GSL, and its spatial variability is dominated by land–sea contrast. However, because of the presence of sea ice along the shallow north and west coasts of the GSL (Long et al., 2015), there is a warm SAT tongue along the eastern GSL, showing the effects of air–sea interactions. Although CRCM can reasonably reproduce the observed SAT spatial patterns in NARR, it slightly overestimates SAT over water and underestimates SAT over land (Fig. 2), suggesting an overestimated land–sea contrast.

Fig. 2 SAT (°C) in (upper) January and (lower) July, averaged for the 1970–1999 period.

In July, the land–sea contrast is relatively weak, and the SAT shows strong meridional gradients (Fig. 2). Moreover, land warms faster than water because of the smaller heat capacity of the former, and the SAT in the western GSL is relatively high compared with that in the eastern GSL. In addition, the Earth's surface receives more solar radiation at low latitudes than at high latitudes, and accordingly, the SAT is warmer at low latitudes than at high latitudes, showing a strong south–north gradient. In contrast to SAT in January, there is a minimum over the Labrador Sea, showing the effects of the cold Labrador Current. Although CRCM reproduces the observed land–sea contrast and meridional gradients in NARR data, on average, it slightly underestimates the SAT (Fig. 2).

b Surface Wind

Surface winds play a fundamental role in air–sea exchanges. Winds not only affect the transfer of momentum from the atmosphere to the ocean surface but also influence heat and moisture fluxes. In the boundary layer, surface wind is a diagnostic variable, depending on the wind at the lowest level of the simulation model, the surface drag coefficient, and static stability (Edson, 2008; McFarlane et al., 1992). On large scales, because of the smaller drag coefficients of the ocean surface, the surface winds are stronger over ocean than over land (Fig. 3). For example, the average wind speed is about 1–3 m s⁻¹ over land and 4–7 m s⁻¹ over water in January. Moreover, atmospheric static stability has an important role in surface wind speed. In winter, the ocean surface is warmer than the atmosphere, and the static stability over the ocean is relatively weak, which enhances the energy transport from the upper atmosphere to the surface (Long and Perrie, 2012). In summer, the ocean surface is colder than the atmosphere, and the atmospheric planetary boundary layer is relatively stable. Therefore, the surface winds along the east coast of Canada are weaker in summer than in winter (Fig. 3). In addition, the wind over the GSL tends to be northwesterly in winter but southwesterly in summer.

Fig. 3 Wind at 10 m (m s⁻¹) in (upper) January and (lower) July, averaged for the 1970–1999 period.

Compared with NARR data, the CRCM simulates the surface winds well. Both CRCM simulations and NARR data show stronger winds over the ocean than over land. In addition, the CRCM simulation can capture the maximum wind speed in the southeast part of the model domain because of the strong SST gradients in the region. In terms of the surface winds in the GSL, although CRCM reproduces the northwesterly winds in winter and the southwesterly winds in summer, the CRCM simulations are more northwesterly in winter and more westerly in summer compared with NARR data. On average, surface winds in the GSL are stronger in CRCM simulations than in NARR data (Fig. 3).

4 Effects of climate change

a Surface Air Temperature

Although land warms faster than the ocean, under climate warming the warming is not uniform over land, particularly in winter, because of the positive feedback of snow. At high latitudes, most of the solar radiation reaching the Earth's surface is reflected into space as a result of the presence of snow and sea ice. However, the loss of snow and sea ice in a warmer climate can significantly increase the shortwave radiation received at the Earth's surface, which will further enhance surface warming. Therefore, Figs 4 and 5 suggest that the maximum warming in January is located in northern Quebec, with a maximum of about 5°C in 2040–2069 and 8°C in 2070–2099. In the GSL, the warming in 2040–2069 is about 3°–3.5°C in the northern GSL and 2.5°–3°C near Cabot Strait. However, in 2070–2099, the SAT increase is more uniformly distributed, with an average SAT increase of about 5°C.

Fig. 4 Differences in SAT (°C) in (upper) January and (lower) July between 2040–2069 and 1970–1999.

Fig. 5 Differences in SAT (°C) in (upper) January and (lower) July between 2070–2099 and 1970–1999.

Moreover, in summer, the SAT changes are dominated by the land effect because of its smaller heat capacity compared with that of water. In the domain of our study, the maximum warming is located over the land areas, between 40°N and 50°N, with a maximum of about 3°C in 2040–2069 and 5°C in 2070–2069. In addition, the magnitude of the warming over the GSL gradually decreases, from the western portion of the GSL to the eastern portion. Thus, in 2040–2069, the maximum warming along the west coast is about 2.4°–2.8°C compared with 2.0°–2.4°C in the eastern portion, whereas in 2070–2099, the warming along the west coast is 4.2°–4.8°C compared with 3.6°–4.2°C over most of the GSL, including the eastern portion.

b Surface Wind Speed

In a warmer climate, changes in surface wind speed are dominated by enhanced energy transport from the upper troposphere and the northeastward shift in the storm tracks in the northwest Atlantic in January. Results are shown in Figs 6 and 7. The increases over land are about 0.2–0.4 m s⁻¹, and the maximum increases occur over the Labrador Sea, with a magnitude of 0.6–0.8 m s⁻¹ because of the reduced sea ice in the region in a warmer climate. In addition, wind speed is projected to increase in the region to the southeast of Newfoundland and decrease to the south of Nova Scotia in 2070–2099, again reflecting the effects of the northeastward shift of storm tracks in the northwest Atlantic, as a result of reduced meridional baroclinicity (Long et al., 2009). Although the changes in wind speed in the GSL are mainly associated with weakening stability in the atmospheric planetary boundary layer in 2040–2069, the shift of the storm tracks also has an effect, particularly in 2070–2099. As shown in Figs 6 and 7, although the January wind speed over the GSL increases by about 0.2–0.4 m s⁻¹ in 2040–2069, it will decrease by 0.2–0.4 m s⁻¹ over the western GSL in 2070–2099.

Fig. 6 Differences in 10 m wind speed (m s⁻¹) in (upper) January and (lower) July between 2040–2069 and 1970–1999.

Fig. 7 Differences in 10 m wind speed (m s⁻¹) in (upper) January and (lower) July between 2070–2069 and 1970–1999.

In July, the increased baroclinicity along coastal areas plays an important role in the changes in winds. In summer, the land surface is warmer than the ocean surface; moreover, the land surface warms faster than the ocean surface in a warmer climate (Figs 4 and 5). Therefore, there is a significant increase in baroclinicity along coastal areas. Because of the increased baroclinicity, surface wind speed will increase by about 0.2–0.4 m s⁻¹ in the GSL, but maximum increases are located along the east coast of the United States (e.g., south of Cape Cod) and northern Quebec (Figs 6 and 7).

c Uncertainties

Uncertainty presents a key challenge for estimated projections of future climate. Overall uncertainty is related to forcing uncertainty, inter-model variability, and atmospheric internal variability. Internal variability is the natural variability of the climate system and arises from the non-linear dynamics processes intrinsic to the atmosphere. Recent studies suggest internal variability is estimated to account for at least half the inter-model spread in projected climate trends (Deser et al., 2012). Figures 8 and 9 show the variance of projected changes in SAT. The variance is estimated as

Fig. 8 Variance associated with the differences in SAT (°C) in (upper) January and (lower) July between 2040–2069 and 1970–1999.

Fig. 9 Variance associated with the differences in SAT (°C) in (upper) January and (lower) July between 2070–2099 and 1970–1999.

where n is the number of members; represents the changes in the individual member; and represents the mean change, of the n members. On average, the variance in SAT changes is about 0.2°C, representing about 10% of the increase in SAT in the region. However, the variance for projected changes in surface wind speed is about 0.2 m s⁻¹, which has the same magnitude as the projected changes, suggesting significant uncertainty in the projected changes in the surface wind speed (Figs 10 and 11). Therefore, the uncertainties for surface wind speeds are generally larger than those for SAT, as also shown in GCM simulations (Deser et al., 2012).

Fig. 10 Variance associated with the differences in 10 m wind (m s⁻¹) in (upper) January and (lower) July between 2040–2069 and 1970–1999.

Fig. 11 Variance associated with the differences in 10 m wind (m s⁻¹) (upper) January and (lower) July between 2070–2099 and 1970–1999.

5 Discussion

Atmospheric responses to the changes in SSTs are non-linear processes. For example, the changes in SSTs can affect moisture and latent heat fluxes in the upper atmosphere, which can have significant effects on atmospheric circulation. Moreover, the surface fields are computed from atmospheric values at the lowest model level and the static stability at the planetary boundary layer. Therefore, changes in SSTs can affect upper atmospheric circulation as well as atmospheric surface fields. When driven by the unadjusted SSTs, CRCM significantly underestimates SAT (Fig. 12) and surface winds in the GSL in July (Guo et al., 2013). To correct the biases in the SSTs from GCMs, the SSTs are adjusted, based on NARR climatology (Guo et al., 2013). Compared with the simulations driven by the unadjusted SSTs, the simulations driven by the adjusted SSTs simulate SAT (Fig. 12) and surface winds reasonably well compared with the NARR climatology (Guo et al., 2013). Moreover, the adjustment significantly affects SAT responses to the effects of climate change in the GSL (Fig. 13). Although the SAT responses in the simulations driven by the unadjusted SSTs show patterns similar to those driven by the adjusted SSTs, in fact, the warming in the GSL in the simulations driven by the unadjusted SSTs is about 0.4°C colder (Fig. 13). These results suggest the presence of non-linear processes in the atmospheric responses to the SSTs.

Fig. 12 SAT (°C) (Guo et al., 2013; reproduced by permission of the Canadian Meteorological and Oceanographic Society) in July (upper) for NARR data, and simulations (middle) without and (lower) with SST adjustment, driven by one member of the CGCM3 simulations.

Fig. 13 Differences in SAT (°C) in July between 2040–2069 and 1970–1999, (upper) without and (lower) with SST adjustment, driven by one member of the CGCM3 simulations.

6 Conclusions

The focus of this study is to understand the impacts of climate change on SAT and surface wind over the east coast of Canada. The CRCM is used to downscale five GCM simulations following IPCC climate change scenarios: three CGCM3 simulations following the A1B scenario; one CCSM3 simulation, also following the A1B scenario; and one CCSM4 simulation following the RCP8.5 scenario. The CRCM integrations are conducted for the 1970–2099 period. For the present climate, CRCM can reproduce the spatial patterns of SAT and surface wind shown in the NARR reanalysis data. For example, both the CRCM simulation and NARR data show a warm SAT tongue along the eastern GSL. In addition, CRCM captures the northwesterly wind in winter and southwesterly wind in summer. However, in terms of SAT, the land–sea contrast is overestimated in winter and underestimated in summer by CRCM. In addition, the simulated surface winds are more northerly in winter and less southerly in summer compared with NARR data.

Under the climate change scenarios, the SAT responses show a significant seasonal variation. In winter, the reduction in snow and sea ice plays an important role at high latitudes, and the maximum warming in winter is located over northern Quebec. In the GSL, the warming in 2040–2069 is about 3°–3.5°C in the northern GSL and 2.5°–3°C near Cabot Strait. In 2070–2099, the warming is more uniformly distributed in the GSL, with a SAT increase of about 4°–5°C. In summer, surface warming is dominated by the difference in heat capacities between land and water, and the effect is notably located over land between 40° and 50°N in the study domain. In the GSL, the warming gradually decreases from the western portion of the GSL to the eastern potion.

The changes in surface wind are dominated by weakening stability in the atmospheric planetary boundary layer in winter and enhanced baroclinicity along the east coast of North America in summer. In January, wind increases over land are about 0.2–0.4 m s⁻¹, and there is a maximum increase in wind over the Labrador Sea, with a magnitude of 0.6–0.8 m s⁻¹ as a result of reduced sea ice in the Labrador Sea. Although wind speed over the GSL increases by about 0.2–0.4 m s⁻¹ in 2040–2069, it is projected to decrease by 0.2–0.4 m s⁻¹ over the western GSL in 2070–2099 because of the northeastward shift of storm tracks in the northwest Atlantic (Long et al., 2009). In July, the land surface warms faster than the ocean surface, which enhances baroclinicity along the east coast. The surface wind speed will increase by about 0.2–0.4 m s⁻¹ in the GSL, and maximum increases are located along parts of the east coast of the United States and northern Quebec.

On average, the uncertainties are generally larger for surface wind speeds than for SATs. The variance of SAT changes is about 0.2°C, which is about 10% of the SAT increase. However, the variance for projected changes in surface wind speed is about 0.2 m s⁻¹, which is the same magnitude as the projected changes, suggesting significant uncertainty in the projected changes in surface wind speed (Figs 10 and 11).

Acknowledgements

The authors thank Michel Giguere of Ouranos for his support in setting up CRCM3.7 on the Bedford Institute of Oceanography Linux cluster.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

We thank the Department of Fisheries and Oceans Climate Change Science Initiative and the Aquatic Climate Change Adaptation Services Program (ACCASP) for support of this work.