Australia: The Land Where Time Began
Atlantic Overturning Estimates at 25oN – Impacts of Atmospheric Reanalysis Uncertainty on it
According to Pillar et al. it is common for atmospheric reanalyses to be used to force numerical ocean models, though in spite of large discrepancies reported between different products, the impact of reanalysis uncertainty on the simulated ocean state is rarely assessed. In this study, the impact of uncertainty in surface fluxes of buoyancy and momentum of the modelled Atlantic meridional overturning at 25oN is quantified for the period January 1994-December 2011. The space and time origins of overturning uncertainty that result from air-sea flux uncertainty are fully explored. Uncertainty in overturning that is induced by previous air-sea flux can be greater than 4 Sv (Svedberg units), where 1 Sv ≡ 106 m3/s, within 15 years, at times exceeding the amplitude of the ensemble-mean overturning anomaly. On average a key result is that uncertainty in overturning at 25oN is dominated by zonal wind uncertainty at lags of up to 65 years and by the uncertainty in surface heat fluxes thereafter, with the uncertainty of the heat flux uncertainty over the Labrador Sea appearing to play a role that is of critical importance.
Ocean general circulation models (OGCMs) are an important tool used in climate research. Though Earth system models that are fully coupled, that have physical, biogeochemical, and ecosystem components, are now often used for prediction studies of the 21st century, OGCMs have remained extremely valuable for the exploration of the role of the ocean in shaping the climate of the Earth. The value of these models is in the reduced computational cost and a reduction of complexity of ocean-only model configurations. Thorough mechanistic investigations of processes that control the main state and variability of the global ocean circulation is facilitated by both of these factors.
Specification of the air-sea fluxes of buoyancy and momentum at the surface boundary is demanded by omission of an interactive atmosphere in OGCMs. These fluxes can be obtained directly from atmospheric reanalyses or computed in the OGCM via bulk formula (Large & Pond, 1981; Large & Yeager, 2004), where the latter requires estimates of near-surface atmospheric fields that have been reanalysed. Multidecadal reanalysed atmospheric datasets are now produced by several operational centres, which includes the European Centre for Medium-Range Weather Forecasts (Uppala et al., 2005; Dee et al., 2011), the National Centers for Environmental Prediction (Kalnay et al., 1996; Kanamitsu et al., 2002), and the Japan Meteorological Agency (Ebita et al., 2011; Kobayashi et al., 2015).
In this paper Pillar e al. present the results of their assessment of the way in which the estimates of the reanalysed air-sea flux that are provided by different agencies impacts the circulation in an OGCM. They focused on the impacts on variability in the monthly mean Atlantic Meridional Overturning Circulation (AMOC) at 25oN, which is referred to in this paper as AMOC25N. Their choice of metric was motivated by a vast body of work that linked variation in the AMOC to low frequency changes in the sea surface temperatures (SSTs) of the North Atlantic, e.g., Delworth & Mann, 2000; Knight et al., 2005; Zhang & Delworth, 2005; Klöwer et al., 2014), with impacts that are far-reaching on regional and global climate (see review by Buckley & Marshall, 2016, and references therein). It is important that there is growing support for potential predictability for the climate of the North Atlantic on decadal time scales (e.g., Robson et al., 2012a, b: Yeager et al., 2015; Årthun et al., 2017), that arises from the long memory and meridional connectivity of the AMOC. As there have been significant past and ongoing efforts to measure the basinwide circulation near latitudes 25oN (Cunningham et al., 2007; McCarthy et al., 2015), latitudes near 25oN are of particular interest. Also, previous studies support that much of the variability observed in AMOC25N results from external forcing (Roberts et al., 2013; Pillar at al., 2016) as opposed to internal variability, which suggests that uncertainty of the air-sea flux may also have an important role in driving uncertainty in AMOC25N.
Forcing the OGCM with an ensemble of atmospheric reanalyses and measuring the divergence in modelled transport (existing studies will be reviewed in the next section), can be used to explore the impact of air-sea flux uncertainty on the modelled ocean state. In this paper, Pillar et al. take a different approach by projecting an ensemble of reanalysed air-sea fluxes onto maps that are time-evolving of the linearised sensitivity of AMOC25N to air-sea exchange of buoyancy and momentum (Czeschel et al., 2010; Heimbach et al., 2011; Pillar et al., 2016). An OGCM and its adjoint are used to compute these sensitivity maps. By the use of the adjoint separate contributions from the uncertainty in surface fluxes of freshwater, heat, zonal momentum, and meridional momentum to the total uncertainty in the modelled AMOC25N, can be unambiguously quantified. Also, the exact time and space origins of the air-sea flux uncertainty dominating uncertainty in the modelled AMOC25N can be identified. This allows the examination of how AMOC25N uncertainty evolves with time and explore whether there is a difference between where and when the reanalysis air-sea flux uncertainty is largest and where and when this uncertainty has the largest impact on the modelled AMOC25N.
Global atmospheric reanalyses are produced by the reprocessing of millions of archived atmospheric observations by the use of state-of-the art weather forecasting system to produce atmospheric reanalyses. This reprocessing involves deriving the best estimate of the state of the atmosphere over sequential short period of time, typically 1 day, analysis cycles that are nested within the full period of the reanalysis, which typically is multidecadal. A prior estimate of the atmospheric state is first provided from a model forecast that is initialised at the end of the previous cycle, for each analysis cycle. The misfit between the prior estimate and the observations, which are distributed irregularly over space and time, is then computed and then used to inform adjustments to the model state and/or parameters for the minimisation of this misfit. In order to provide an estimate that is corrected of the atmospheric state the forecast is then rerun. Multidecadal gridded records of both observed directly and derived variables is produced by repeating this procedure for consecutive analysis cycles. They are easy to handle and updated frequently and are extensively used to investigate variability of the atmosphere and to force OGCMs.
Discussion and conclusions
Atmospheric reanalyses are often used to provide air-sea fluxes of buoyancy and momentum to drive OGCMs. There are Notable discrepancies between the products, even though independent reanalyses assimilate many of the same observational datasets. These have been well documented in the scientific literature. Less attention has been paid to the impact of these discrepancies on the simulated ocean state. In this study, Pillar et al. explored the impact of reanalysis air-sea flux uncertainty, as applied over a period of 15 years, on the simulated ocean state. Their focus was on the variability in Atlantic Meridional Overturning Circulation about the mean seasonal cycle at 25oN (AMOC25N). According to Pillar et al. this study is a natural extension of their earlier study (Pillar et al., 2016), in which they presented model-based sensitivities of AMOC25N to air-sea fluxes at lead times up to 15 years and they discussed the adjustment mechanisms that were revealed in the temporal and spatial evolution of the patterns of sensitivity.
In this study they projected forcing anomalies from 5 reanalysis products onto the patterns of linear sensitivity in order to construct 5 time series of AMOC25N anomalies. They have focused throughout most of this paper on the period January 1994-December 2011, where an AMOC25N estimate that was constructed by using a 15 year forcing history was available for this investigation for all of the 5 reanalyses. They were able to make separate quantitative assessments of the uncertainty in AMOC25N that originated from uncertainty in reanalysis zonal wind stress, meridional wind stress, surface heat fluxes, and fluxes of surface freshwater. Also, their approach allowed them to explore where and when atmospheric forcing uncertainty is responsible for generating the largest uncertainty (ensemble spread) in AMOC25n.
It is shown by the ensemble of AMOC25N estimates that there is good qualitative agreement, notably in the timing of high amplitude AMOC25N anomalies that are driven by zonal wind and trends of apparently low frequency that are driven by surface buoyancy forcing. The ensemble spread in AMOC25N that is induced by prior air-sea flux uncertainty can, however, exceed 4 Sv within 15 years, and at times exceeding the amplitude of the of the ensemble mean-AMOC25N anomaly. This spread increases monotonically, however, with time; following a 15 year period, the AMOC25N spread that is induced by uncertainty in the zonal wind, meridional wind, surface heat flux, and surface freshwater flux is 0.8, 0.5, 2.2, and 0.8 Sv, respectively. There is a notable difference, however, in the evolution of the wind and ensemble spread that if induced by buoyancy. The latter is dominated by slow AMOC25N adjustment to the uncertainty of surface heat flux over the subpolar latitudes, becoming increasingly important with AMOC25N lag, whereas the former is dominated by a rapid adjustment of AMOC25N to the uncertainties of local wind at short lead, becoming less important with increasing AMOC25N lag. A key result is that, on average, AMOC25N uncertainty is dominated by the uncertainty of zonal wind lag of up to 6.5 years, and by the uncertainty of surface heat fluxes thereafter. Also, they are most potent in the vicinity of 25oN, though uncertainties of zonal wind stress are largest along the western boundary as well as throughout the North Atlantic subpolar gyre. Uncertainties in the heat flux in the surface water of the Atlantic are largest along the Gulf Stream and throughout the subpolar ocean, though they are most potent in the Labrador Sea.
This study by Pillar et al. was limited as they were only able to consider a response history using the adjoint of the AMOC25N for a period they was 15 years long. On time scales that are longer the air-sea flux uncertainties would continue to impact their AMOC25N estimate through advective teleconnections that are slower (e.g., Thomas et al., 2015) and there is a possibility that remote uncertainties, e.g., over the Antarctic Circumpolar Current, could become more important (e.g., Nikurashin & Vallis, 2012). Also, they neglected the impact of systematic offsets between the reanalysis products in their estimation of AMOC25N uncertainty that was externally forced. This omission is the result of the forcing anomaly for any product, i, is compounded about the climatological seasonal cycle in that same product (Eq. 1). A rigorous investigation of the uncertainty in AMOC25N that results is an important avenue for research in the future, since the product differences in seasonal climatology (e.g., Fsc/ERA-INT - Fsc/NECPII) are substantial.
The use of smoothed forcing is an additional caveat of this study both
1) The climatological forcing that is used in the spinup integration and
2) The forcing anomalies that are projected onto the linear sensitivities are monthly values.
It is thought that high frequency atmospheric variability, that is believed to play a critical role in ocean convective variability, was not included in the model of Pillar et al. (Pickart et al., 2008; Våge et al., 2009). In this study Pillar et al. identified heat flux uncertainty in the Labrador Sea as an important driver of AMOC25N uncertainty, though should validate this result in future work, by the use of a model that has better representation of synoptic air-sea flux variations (e.g., Jones et al., 2018; Smith & Heimbach, 2018, manuscript submitted to J. climate).
According to Pillar et al. their results are also subject to any deficiencies in the OGCM, which lead to a misrepresentation of the adjustment of AMOC25N to the air-sea fluxes. It is important that the strongest response of AMOC25N uncertainties over the subpolar gyre in the North Atlantic is set by the regions of principal connectivity and dominant time scale of internal variability in the model (Delworth & Zeng, 2016). These vary significantly from model to model (Zhang & Wang, 2013; MacMartin et al., 2013; Branstator & Teng, 2012), as a result of biases in the mean state (e.g., Drews & Greatbatch, 2016), insufficient resolution by the model (Marshall & Johnson, 2013; Thomas & Zhai, 2013; Shaffrey et al., 2017), uncertain model parametrisation (Farneti & Vallis, 2011; Yeager & Danabasoglu et al., 2012; Danabasoglu et al., 2012), and the restoration of temperature and salinity (Latif et al., 2006; Behrens et al., 2013; Zhai et al., 2014). A challenge remains to assess the reliability of a model in the in simulating low frequency AMOC variability, though lacking observational time series that are sufficiently long and consensus on the mechanisms responsible (Buckley & Marshall, 2016), which highlights the importance of sustaining continuous observation of North Atlantic overturning and the formation of deep water (e.g., at the RAPID and Overturning in the Subpolar North Atlantic Program (OSNAP) arrays; see McCarthy et al., 2015; Lozier et al., 2017), respectively. Restoration of SST and salinity may also impact the seasonality of the sensitivity (Y. Kostov, 2018, pers. Comm.) and alter seasonal potency of the reanalysis uncertainty. Finally it was noted that the results presented in this paper may be sensitive to the method that is chosen for interpreting the atmospheric reanalysis onto the ocean model grid, as a result of the strong sensitivity of AMOC25N variations in density at the east and west boundaries.
In spite of these limitations, the time scales and regions of AMOC25N uncertainty that is forced externally appear plausible in that they bear many similarities to time scales and regions of AMOC variability that is forced externally in the scientific literature. The dominance of subpolar heat fluxes in driving decadal North Atlantic overturning variability, in particular, agrees with studies that have been published exploring the response of AMOC to the North Atlantic Oscillation, by the use of idealised (Zhai et al., 2014), geographically realistic (Böning, 2006; Biastoch et al., 2008; Yeager & Danabasoglu, 2014), and coupled fully (Delworth et al., 2106) nonlinear models. The critical importance of local heat flux forcing over the regions of deep convection is consistent with AMOC estimates that were presented by (Yeager & Danabasoglu, 2014), showing that variations in air-sea buoyancy exchange over the interior of the Labrador Sea alone drive almost all decadal variability in the AMOC25N that is modelled from 1958-2007.
Pillar et al. found that uncertainty in surface fluxes and buoyancy fluxes that are reanalysed can produce net uncertainties as large as ~5 Sv in the AMOC25N that is modelled. For period of the time that is considered, the time-averaged AMOC25N uncertainty is ~2 Sv. Recalling again that the buoyancy-driven AMOC25N estimates do not converge as the time window over which the convolution is performed, Pillar et al. stressing that the amplitude of AMOC25N uncertainty that is reported in this paper may continue to change, and may even decrease, when accounting for the response of the AMOC25N over a longer time window.
According to Pillar et al. their estimates are consistent with atmosphere-forced AMOC uncertainty (Brodeau et al., 2010; He et al., 2016). Though of most importance the amplitude of the AMOC25N uncertainty that is presented in this paper is comparable to the AMOC uncertainty in the CORv2-forced hindcasts (Danabasoglu et al., 2016), where the ensemble of AMOC estimates exhibits a time–mean standard deviation of ~3 Sv about the ensemble mean. As a common forcing history is shared by these hindcasts, AMOC divergence across the ensemble is due to model differences alone. It has recently been demonstrated (Huber & Zanna, 2017) that air-sea flux uncertainty is at least as important as parameter uncertainty in driving divergence in the steady-state AMOC in the Coupled Model Intercomparison Project phase 5. The results of this study further support the claim of Pillar et al. that atmospheric forcing uncertainty is a key source of ocean uncertainty in climate models and therefore a key area to target now for improving skill in climate prediction in the future.
|Author: M.H.Monroe Email: email@example.com Sources & Further reading|