Sensitivity of modelled atmospheric circulation to the representation of stable boundary layer processes



[1] We examine the sensitivity of the modelled climate of the third generation Hadley Centre climate model to changes in the parametrization of surface and boundary-layer fluxes under stable stratification. Replacing the model's standard parametrization with one in which fluxes decrease more rapidly with increasing stability generates statistically-significant changes in modelled 500 hPa height. The largest changes are seen across the North Atlantic and North Pacific and occur during Northern Hemisphere summer, when persistent stable atmospheric boundary layers form over the western sides of these oceans. The atmospheric response in a coupled version of the model is stronger than in the atmosphere-only version, suggesting that feedbacks involving sea surface temperatures, surface fluxes and atmospheric circulation are important in determining the response.

1. Introduction

[2] Stable boundary layers (SBLs) are a widespread atmospheric phenomenon. In mid- and low latitudes, SBLs are usually transient features associated with nocturnal cooling. Long-lived SBLs are generally confined to the polar regions and the boreal continents in winter but may also occur wherever there is climatological advection of warm air over a cold ocean surface.

[3] The development of adequate parametrizations of fluxes of heat, momentum and moisture within SBLs for use in large-scale models has proved to be quite challenging. Classical surface- and boundary-layer similarity theory has provided some guidance to the development of practical parametrizations [e.g., Louis, 1979]. However, all parametrizations ultimately rely on the results of field experiments. These reveal that SBLs are complex, characterized by low levels of turbulence, intermittency and the presence of large-scale motions that contribute significantly to the measured fluxes. As a result of this complexity there are considerable uncertainties associated with SBL parametrizations and it is important to assess the contribution of these uncertainties to overall model performance.

[4] Studies of the sensitivity of models to changes in SBL parametrization have mostly focussed on the impact of such changes on near-surface conditions. Viterbo et al. [1999] demonstrated that changes to SBL parametrization could improve forecasts of surface temperature, while King et al. [2001] studied the response of surface winds and temperatures over the Antarctic in a climate model to similar changes. While changes to surface and boundary-layer fluxes will clearly have a direct impact on surface climate, the changes thus caused may themselves influence atmospheric circulation. In this paper, we study the sensitivity of global atmospheric circulation to changes in the SBL parametrization used in a climate model.

2. Methods

[5] We carried out experiments with the atmosphere-only (HadAM3) and coupled atmosphere-ocean (HadCM3) configurations of the third generation Hadley Centre climate model. The atmospheric component of the model [Pope et al., 2000] runs on a 2.5° latitude × 3.75° longitude grid, with 19 levels in the vertical. The lowest two model levels are at heights of approximately 30 and 300 m, giving a reasonable representation for surface and boundary-layer processes. The ocean component of the coupled model [Gordon et al., 2000] runs on a 1.25° latitude × 1.25° longitude grid, with 20 vertical levels. The ocean model includes a dynamic-thermodynamic sea ice model in which ice is assumed to drift with the surface ocean current.

[6] Turbulent fluxes of momentum, heat, and other scalars are parametrized through a simple first-order closure scheme. This approach is widely used in large-scale models but does have limitations [see, e.g., Etling and Brown, 1993]. As the purpose of the present study was to investigate model sensitivities, rather than to develop a radically-improved turbulence parametrization, we retain the framework of the first order closure scheme but make small changes to its implementation under stably stratified conditions. The parametrization of surface and boundary-layer fluxes in the stably-stratified atmosphere in this model follows the framework proposed by Louis [1979] and is described in detail by King et al. [2001]. Briefly, fluxes are first calculated using a first-order closure model assuming neutral stratification. The calculated fluxes are then multiplied by a stability function which is a decreasing function of Richardson number, Ri. This simulates the reduction of fluxes associated with the suppression of turbulence as stability increases. In the standard version of the model, the stability function is prescribed as

equation image

With this functional form, fluxes decrease relatively slowly with increasing stability. King and Connolley [1997] argue that observations support a more rapid decrease of f(Ri) with increasing stability. Following the methodology of King et al. [2001], we have investigated the sensitivity of modelled atmospheric circulation to changing f(Ri) from the form given by (1) to the SHARP parametrization [Derbyshire, 1999]

equation image

[7] The SHARP parametrization has the desirable properties that f(Ri) decreases more rapidly with increasing stability than with the standard parametrization but, as with (1), f(Ri) remains finite for all values of Ri, thus preventing decoupling of the model atmosphere from the surface under very stable conditions.

[8] We have carried out four model experiments:

[9] (a) AM3CTL. A 43-year control run of the atmosphere-only model with f(Ri) given by (1). Annually-repeating surface boundary conditions for sea surface temperature and sea ice extent were specified from 1961–1990 climatology and atmospheric composition was set appropriately for this period.

[10] (b) AM3SHARP. A 23-year run of the atmosphere-only model with f(Ri) given by (2) but otherwise as AM3CTL.

[11] (c) CM3CTL A 94-year run of the coupled model with pre-industrial atmospheric composition, starting from a fully spun-up state, with f(Ri) given by (1).

[12] (d) CM3SHARP. A 50-year run of the coupled model, as for CM3CTL, but with f(Ri) given by (2).

[13] We used the full length of the AM3CTL, AM3SHARP and CM3CTL runs as they are free of long-term trends. However, only the last 30 years of the CM3SHARP run were analysed in order to avoid the transient response associated with the change of parametrization.

[14] The impact of changing the SBL parametrization was investigated by examining changes in the modelled 500 hPa geopotential height (h500) field. h500 is an appropriate variable for studying large-scale tropospheric circulation at mid- and high latitudes. Examination of changes at other levels, from the surface to 250 hPa, indicates that the modelled response to changed parametrization is essentially equivalent barotropic.

3. Results

[15] Figure 1 shows the changes in modelled 500 hPa height (h500) during the June–August (JJA) season associated with implementing the SHARP parametrization into the atmosphere-only model. The largest changes are seen around Antarctica, where implementation of SHARP results in strong surface cooling and an increase in the strength of the katabatic winds over the continent [King et al., 2001]. In the northern hemisphere small (but significant at the 1% level using a t-test) falls in h500, are seen across North America and the North Pacific. A further area of reduced h500 over northern Europe is not statistically significant. During the December-February, (DJF) season (not shown) the changes around Antarctica are smaller than in JJA, reflecting the weaker stability of the boundary layer over the continent in summer. In the northern hemisphere, changes in h500 are generally small during DJF.

Figure 1.

JJA 500 hPa height (gpm) for run AM3SHARP minus that for the corresponding control run, AM3CTL.

[16] Implementing SHARP in the coupled model causes a somewhat different response. The globally and annually averaged top-of-atmosphere (TOA) net radiative balance changes from −0.15 W m−2 in the coupled control run, CM3CTL, to −1 W m−2 in the first year of CM3SHARP. This imbalance reduces to −0.5 W m−2 as the model climate cools, resulting in a general reduction of surface temperature and lowering of 500 hPa heights. The effect of this global cooling needs to be borne in mind when comparing CM3SHARP and CM3CTL. In JJA (Figure 2), changes in h500 around Antarctica are smaller and more zonally-symmetric than those seen in the atmosphere-only experiments although the impact of SHARP on Antarctic surface temperatures and near-surface winds (not shown) is similar. Some of the difference in the h500 response may be a result of differences between the modelled atmospheric circulation at high southern latitudes in AM3CTL and CM3CTL [Turner et al., 2006]. The largest differences between CM3SHARP and CM3CTL during JJA occur around 45°N, with maximum response to the changes (significant at the 1% level) seen over the NE Atlantic and NW Pacific. During the DJF season (not shown), the pattern of change associated with implementing SHARP is mostly similar to that seen with the atmosphere-only model, particularly once the global reduction in h500 in CM3SHARP, associated with the global cooling in that run, has been taken into account. Though not as large as in JJA, there is a band of reduced h500 across the North Pacific and, in contrast to the atmosphere-only experiment, there is an enhanced reduction in h500 over the Southern Ocean at around 45°S.

Figure 2.

JJA 500 hPa height (gpm) for the last 30 years of run CM3SHARP minus that for the corresponding control run, CM3CTL.

4. Discussion

[17] Persistent SBLs are usually thought of as being associated with high-latitude continental landmasses during winter. However, examination of the modelled surface sensible heat flux, Hs, during JJA (Figure 3) reveals regions of climatological negative Hs (i.e., persistent SBLs) over both the NW Atlantic and NW Pacific oceans. We henceforth refer to these regions as western ocean stable boundary layers (WOSBLs). In the WOSBLs, which are also present in climatologies of observed fluxes [e.g., Josey et al., 1998], warm continental air is being advected over a relatively cool ocean, leading to a downward sensible heat flux and stable stratification. Changing to the SHARP parametrization will reduce the magnitude of air-sea fluxes in the WOSBLs and will thus act to change atmospheric temperatures and humidities. In the coupled model runs, changes in ocean-atmosphere heat fluxes will also affect modelled sea surface temperatures (SSTs). We suggest that the changed atmospheric structure associated with modifications to the surface fluxes will affect the growth rates of synoptic-scale weather systems originating over the northwestern oceans and, as these systems propagate eastwards, they will generate the patterns of h500 changes seen in Figures 1 and 2.

Figure 3.

Contours of mean surface temperature for JJA in run CM3SHARP minus that for the corresponding control run, CM3CTL Contour interval is 1 K. Shading indicates regions where the JJA mean sensible heat flux is negative (i.e., persistent stable boundary layers) in CM3CTL.

[18] The larger changes in h500 in the coupled model experiment relative to those in the atmosphere-only experiment suggest that feedbacks associated with SST changes are important in determining the enhanced response of the coupled model to changes in SBL parametrization. Surface temperature changes caused by introducing the SHARP parametrization into the coupled model are shown in Figure 3. As discussed above, surface temperatures in CM3SHARP are everywhere lower than in CM3CTL. However, there are regions of enhanced reduction in SSTs associated with the WOSBLs. Figure 4 shows the change in the net ocean-atmosphere turbulent heat flux, HT (the sum of Hs and the latent heat flux, HL) associated with SHARP. In the WOSBLs, Hs is (by definition) directed downwards while HL, which can be of either sign, is generally of smaller magnitude than Hs. HT is therefore directed from the atmosphere towards the ocean and the impact of SHARP is to reduce this flux, imposing a cooling tendency on surface temperatures. Comparing Figures 3 and 4 confirms this correspondence between changes in SST and HT in the WOSBLs. However, outside of the WOSBLs there is rather poor correspondence between changes in HT and those in SST. In such regions, changes in other factors controlling SSTs, such as Ekman transport, mixing and upwelling must dominate over the changes in HT. A full investigation of such processes is outside the scope of this paper and remains a topic for further investigation.

Figure 4.

JJA mean total heat flux (Hs + HL, W m−2) in run CM3SHARP minus that in the corresponding control run, CM3CTL. Negative contours are shown dashed. Where the difference is positive (negative), implementing SHARP is acting to cool (warm) the surface.

[19] Our experiments have demonstrated that making relatively small changes to the way in which SBL fluxes are parametrized in a GCM can have a significant impact on modelled atmospheric circulation. We also note that changes in SSTs in the coupled experiments are of comparable magnitude to the regional SST biases in this model reported by Gordon et al. [2000]. There are still considerable uncertainties associated with SBL parametrization and our results suggest these may contribute significantly to uncertainties in climate simulation and prediction. The model response to changed SBL parametrization is largest in coupled model experiments, where changes in surface fluxes can modify SSTs. The largest circulation changes that we observe in response to changed SBL parametrization are not associated with the persistent SBLs that form in the polar regions and over the continental interiors during winter. Rather, they are associated with persistent SBLs that form over the oceans during summer as a result of warm air advection While the existence of persistent SBLs over the oceans has been recognised for some time, we believe that this is the first time that their influence on large-scale atmospheric circulation has been demonstrated.


[20] We thank L. Shaffrey for helpful discussions. A. Jrrar was supported by the UK HiGEM project, funded by the UK Natural Environment Research Council.