The importance of coupled ocean-atmosphere prediction systems in providing seasonal predictability is well-known (Kim et al., 2012, and references therein). Sources of predictability in seasonal forecasting systems tend to be, by their very nature, coupled systems driven by teleconnections that are functions of climate modes, such as the North Atlantic Oscillation and the El Niño–Southern Oscillation that have geographically far-reaching consequences. However, as timescales shorten and the dominance of these coupled climate modes become less fundamental to predictability of the atmosphere-ocean system, it becomes less obvious whether the benefits of fully coupled systems justify the computational cost or the technical and scientific complexity required. The coupling between atmospheric and wind wave models was first introduced operationally in 1998 at ECMWF. The method based on the theoretical work of Janssen (1991) contributed to an improvement of both atmospheric and surface wave forecasts at the medium range on the global scale. The usual approach of forcing the ocean with atmospheric conditions (Takano et al., 1973), and referred to in this section as “forced”) using bulk parameterisations of Polar lows are of a decidedly less extreme nature than tropical cyclones, but they share the same dependence on sea surface temperature (Rasmussen and Turner, 2003). As winds blow off the sea ice, the air is rapidly warmed by the (relatively) warm ocean surface. Under the appropriate atmospheric conditions (Kolstad, 2015), this can lead to the formation of polar lows. These are small-scale, intense cyclones, typically with gale-force winds. If the cyclone is rather stationary, a shallow layer of warmer water can mix with cooler waters through Ekman pumping. As the ocean temperature is key to sustaining a cyclone, the water mixing can sometimes be enough to inhibit further growth of the polar low. Examples of instantaneous coupling between land, ocean, and atmosphere also include coastal inundation during landfall of tropical cyclones (Lee et al., 2019). In these cases, heavy precipitation leads to a swelling of local rivers, which is often coincidental with a large storm surge. The result is a rapid sea-level rise that may cause extensive damage to coastal infrastructure, especially when combined with large surface waves and strong winds. the fluxes (Large and Yeager, 2009) is computationally and structurally far easier and cheaper than coupling approaches. However, the key boundary layer processes (see Section 9.1 for details) are not taken into account and thus the feedback between the atmospheric boundary layer and the upper ocean is not represented. It is necessary to understand how important these processes might be, bearing in mind that coupled models can suffer from systematic errors as a result of positive feedback leading to drifts in the forecast (Hyder et al., 2018).
Ocean forecasting systems have become increasingly high-resolution, resolving coastlines, bathymetry, and eddy-scale processes. The effect of coupling on model predictions becomes more important with increasing grid resolution (Janssen et al., 2004), and so the question of the benefits of coupling to ocean forecasting is perhaps more relevant now than ever. A small but growing body of literature demonstrates the benefits to ocean prediction of coupling at shorter time-ranges (Brassington et al, 2015; Allard et al., 2010; Lewis et al., 2018 and 2019).
Understanding the advantages of coupled over uncoupled predictions in short-range ocean forecasting is in its infancy. Although the future of advanced systems is clearly coupling, as several processes are better represented, predictive modelling without coupling is however possible thanks to parameterizations and should never be discarded as an option. At a recent science meeting of OceanPredict (Vinayachandran et al., 2020), the need for a careful evaluation of how ocean and atmosphere components interact and impact each other was highlighted. At monthly or shorter timescales, the benefits of running coupled systems need to be evaluated, balancing scientific and service benefits against complexity and computing costs. Intermediate complexity coupling may also be an appropriate approach if full coupling is not viable and the service is not reliant on the atmosphere and ocean information. Lemarié et al. (2021) provided an example of an atmospheric boundary layer approach that gives some of the benefits of coupling whilst being significantly simpler and computationally cheaper.
The potential benefits of using a coupled framework is reinforced by the move towards a multi-hazard approach to predictions. Natural hazards from multiple sources may combine or occur concurrently (Lewis et al., 2015). Large waves, storm surges, high-wind speeds, and extreme precipitation are all hazards that are likely to co-occur, and influence each other through coupled feedbacks that can compound one another (for example through over-topping). Coupled systems that predict these coupled feedbacks may enable an improvement in the range and consistency of actionable information to be provided through hazard warnings and guidance.
When considering providing services in multi-hazards frameworks, the opportunities that coupling provides should be considered alongside the scientific benefits. A coupled system combining the full water-cycle – including consistent precipitation, river runoff, wave, currents, and surge forecasts - can give users mutually consistent products in a joint probability framework. This can be important in coastal flooding, where the impacts for coastal communities or industries can come from high river flows and local heavy precipitation events, alongside overtopping waves and extreme surges. From a service perspective, it is attractive to provide probabilistic frameworks in which the timings and intensities of events are consistently incorporated and interact appropriately; these services increasingly rely on probabilistic information for decision making. An area that has had limited attention but seems likely to prove significant is the impact of feedback among Earth-system components upon ensemble spread, and hence the quality of the probabilistic information.
Ocean phenomena are usefully classified depending on their nature, which determines the timescale for oceanic predictive skill and whether a coupled ocean-atmosphere model would be advantageous. Some phenomena have strong dependence, and a rapid response, to the atmosphere forcing and can be thought of as forced-dissipative systems. This category includes, surface waves, responses to surface heating and wind in the ocean boundary layer, and storm surges. These systems largely depend upon skill in the atmosphere model, and so the benefits of coupling to the atmosphere can be a leading-order driver of the ocean system skill. The advantage of coupling and its impact upon predictability often focus on the benefits to the atmosphere (Brunet et al., 2010; Belcher et al., 2015). The impact of ocean coupling on tropical meteorology is well documented with tropical cyclones (Bender and Ginis 2000; Mogensen et al., 2017; Smith et al., 2018; Neetu et al., 2019), monsoons (Fu, 2007), and the Madden–Julian Oscillation (Bernie et al., 2008; Shelly et al., 2014; Seo et al., 2014), which predictability improved in coupled systems. There is also an increasing body of evidence that the oceans have a significant local (important for short-range forecasts) and non-local (increasingly significant at longer lead-times) influence on the extra tropics (Minobe et al., 2008).
In the literature, there is limited quantification of the impact of the coupled improvement in atmospheric parameters on ocean services but it is an increasing area of study. Guiavarc’h et al. (2019) explored the impact of a coupled (atmosphere-ocean) system on short-range ocean forecast skill and showed that there are benefits in SST predictability at the short-range, but with mixed results for other parameters. Given that the research system they used is at a relatively early stage in development, and the resolution of the atmosphere is significantly lower than in comparable forced systems, these results are encouraging.
Although the importance of coupling the wave-ocean interface for improving forecasts of surge and waves is well documented (Wolf, 2008; Lewis et al., 2018), most storm surge and wave prediction systems remain largely independent. As well as the atmospheric forcing, ocean currents have a significant role in modifying ocean wave properties. The presence of eddies, fronts, and filaments with length scales of tens to hundreds km and ubiquitous in the world’s oceans, can be the main source of variability in significant wave heights at these scales. Ardhuin et al. (2017) made a compelling case for the importance of coupling the ocean surface currents to a wave model allowing adequate representation of wave height variability in the world’s open oceans. Wave predictions in shelf seas environments are shown to be improved as a result of coupling to an ocean model (Allard et al., 2012; Wahle et al., 2017; Lewis et al., 2018). as well as the predictions of ocean current and other ocean parameters, including upwelling due to stokes drift effects, were enhanced (Wu et al, 2019). Fan et al. (2009) showed that time and spatial variations in the surface wave field, as a result of coupling to winds, are particularly strong in hurricanes, with significant additional feedback from ocean currents and near-surface temperatures.
The ocean eddy kinetic energy is damped when taking into account the feedbacks between ocean surface current and winds (Oerder et al., 2018; Jullien et al., 2020). As ocean models increasingly resolve the mesoscale explicitly, they are likely to have the tendency to over-predict the eddy activity. In uncoupled systems, there is an option to calculate the wind stress using relative wind speeds (taking into account the eddies and other ocean current interactions). However, in these systems there is no imprint of ocean eddies on the atmospheric wind stress curl (due to the lack of ocean eddies in the uncoupled atmospheric modelling system), and so the feedback onto the wind stress results in over-damping of the eddies. A fully coupled system will correctly allocate the feedback between the winds and currents, allowing the eddy and wind fields to co-evolve correctly. This coupling between the winds and currents can also lead to upscaling to the large scale, e.g. Renault et al. (2016) showed that current/wind feedback, through its eddy killing effect, resolves long-lasting biases in Gulf Stream path.
Marine heatwaves have recently been recognised for their importance (Holbrook et al., 2019). They are high impact events that can be induced by anomalous heating at the ocean surface; their predictability is dependent upon air-sea coupled phenomena (Jacox, 2019). At the other end of the temperature scale, Pellerin et al. (2004) showed that coupling can also have strong impacts in ice-infested seas even down to sub-daily time scales, due to rapid changes in coastal sea ice cover (i.e. the formation of coastal polynyas). The sea ice acts as a barrier between a relatively warm–wet ocean and cold–dry atmosphere, and changes in the sea ice cover can have dramatic effects on heat and moisture fluxes. The importance of coupling has also been recognized in polar regions (Jung and Vitart, 2006).
Coastal regions are particularly impacted by coupled processes, both between the ocean and atmosphere and coupling with river and estuaries. The impact of freshwater discharges on the ocean circulation is highlighted by Røed and Albretsen (2007) and, more broadly on the coastal marine environment, by Dzwonkowski et al., 2017. The inputs from the land surface, mediated through estuaries and lagoons, are generally poorly represented in ocean forecasting systems due to their scale (time and space) and their complexity. It is extremely difficult to accurately model nutrient inputs, which are mediated strongly by land use and societal factors, and the associated plankton response is therefore compounded. Although this problem is not fundamentally a coupling problem, there is still scope for improving the inputs to the coastal environment through specifying better the river-estuary-ocean interface.
Previous chapter 9
Follow us