The impact of the AMV on Eurasian summer hydrological cycle

Impact studies of the Atlantic Multidecadal Variability (AMV) on the climate system are severely limited by the lack of sufficiently long observational records. Relying on a model-based approach is therefore mandatory to overcome this limitation. Here, a novel experimental setup, designed in the framework of the CMIP6-endorsed Decadal Climate Prediction Project, is applied to the CMCC climate model to analyse the remote climate impact of the AMV on the Northern Eurasian continent. Model results show that, during Boreal summer, an enhanced warming associated to a positive phase of the AMV, induces a hemispheric-scale wave-train response in the atmospheric circulation, affecting vast portions of Northern Eurasia. The overall AMV-induced response consists in an upper-tropospheric anomalous flows leading to a rainfall increase over Scandinavia and Siberia and to an intensified river runoff by the major Siberian rivers. A strengthening of Eurasian shelves’ stratification, broadly consistent with the anomalous river discharge, is found in the proximity of the river mouths during positive-AMV years. Considering that Siberian rivers (Ob’, Yenisei and Lena) account for almost half of the Arctic freshwater input provided by terrestrial sources, the implications of these findings for decadal variability and predictability of the Arctic environment are also discussed.


Winter response
During Boreal winter, the 2-m temperature response to fixed sea surface temperature (SST) anomalies resembles the AMV pattern over the North Atlantic even if the signal is less significant on the Western side of the basin, along the Gulf Stream. This means that the high heat transport from lower latitudes due to the fast current and the variability of the system is greater than the imposed SST anomalies. Nevertheless, the main features of the AMV are well reproduced over the Atlantic Ocean, including both the maxima over the sub-polar gyre (SPG) and Eastern Tropical Atlantic. Significant warming affects Central America, Tibetan plateau and the Saharan region, with maxima over the Middle East and Indian sub-continent (~0.4°C). The North Pacific Ocean is reminiscent of the IPO-like response, as estimated in DCPP protocol, indeed the structure of the negative signal is not interrupted by anomalies of the opposite sign [Sutton and Hodson 2006, Ting et al. 2011, Barcikowska et al. 2017.
Positive anomalies characterize also the Western Tropical Pacific (WTP), Maritime continent and the Indian Ocean, similar to Boreal summer response, reinforcing the idea that such feature is fed by the positive AMV phase, regardless of the season. The Arctic Ocean response reveals a significant increase of air temperature, consistently with existing literature [e.g. Miles et al. 2014], that results in thinner sea ice pack that is more prone to melt, leading to a smaller sea-ice cover at the end of the melting season. Significant variability emerges over the Mediterranean region with local increased 2-mter temperature, while European regions above 50°N are characterized by more uncertainties.
The sea level pressure (SLP) anomaly field can be considered a proxy of the atmospheric-flow changes and provides a measure of the alteration of the large-scale surface circulation induced by the polarities of the AMV. A broad-scale negative SLP anomaly features the North Atlantic Ocean, as a likely consequence of the Atlantic AMV-related warming ( Figure S.2). Nevertheless, non-significant areas covers the extra-tropical portion of the North Atlantic. This is partly linked to the efficiency of the restoring, especially over the SPG, representing a deep-water formation site. Seawater at the surface of the ocean is intensely cooled by the wind and surface air temperatures. Wind flowing over the water also produces a great deal of evaporation, resulting in an increase in the sea salinity and so in the water-mass density along with the temperature decrease. This process is obviously stronger during the colder season, leading to a deeper mixed layer in winter than in summer since the imposed SST is vertically distributed over a great depth. Thus, it is not so unexpected the model lack of significance due to a less effective SST restoring. The SLP response shows a meridional significant dipole, with a positive lobe over Greenland and a negative one over Central Europe extending over the Atlantic, associated with a statistically-significant precipitation increase ( Figure S.2). This pattern bears some resemblance with the negative phase of the North Atlantic Oscillation (NAO), in agreement also with 2-m temperature field, even though only a slight SLP weakening occurs at the Azores High.
In the North Pacific, the Aleutian low is weakened due to the positive SLP anomaly    In order to simulate the AMV impact, the model SST of the North Atlantic Ocean are nudged towards the spatial pattern. In the ocean component of the CMCC coupled model, NEMO uses flux formulation to globally restore SST, considering a negative feedback term to be added to the surface non-solar heat flux : where: • is the resulting surface non-solar heat flux. Q ns • is defined as the heat flux restoring term (hereafter hfcorr) which provides a measure of the restoring efficiency.
• is a negative feedback coefficient ( © T ) which indicates the restoring strength.
dT dQ According to the DCPP protocol, this term is fixed at -40 W m -2 K -1 equal to a relaxation time scale ( ⎮ R ) of 60 days for a 50m Mixed-Layer Depth (MLD).
• is the model SST.

SST M ODEL
• is the target SST which is, in this study, the positive (or negative) 1 SST T ARGET standard deviation of the AMV anomalies superimposed to the model climatology.
Temperature restoring introduces density anomalies over the SPG: for instance, during a simulated positive phase, warm anomalies tends to stabilize the ocean column reducing the deep ocean convection and generating a weakening of the subpolar gyre (SPG) density. These negative density anomalies may induce a decline in the AMOC and SPG strength, also affecting the poleward ocean heat transport. In the opposite case, imposing cold anomalies To reduce the non-linear uncertainty related to the internal variability of the climate system, an ensemble approach is used. The DCPP protocol recommends to generate the ensemble using at least 25 "macro-perturbations" [Hawkins et al. 2016], i.e. considering 25 different ocean states as initial condition of each realization, in addition to only different atmospheric states ("micro-perturbation"). This ensures a better spread among the ensemble members. In this study, 32 members are considered taken from a multicentury pre-industrial run. To counterbalance the additional flux term introduced by SST restoring during the simulations, sea surface salinity (SSS) is also restored to the preindustrial-run climatology. The SSS relaxation is used to avoid the progressive alteration of the mean ocean circulation and thermodynamical balance, reducing at least as possible the density anomaly. SSS restoring can be considered as a flux correction on freshwater fluxes which has no physical meaning.
As for the SST, a flux formulation is used to restore SSS adding a feedback term in the freshwater budget to the freshwater flux EMP: where: • EMP is the resulting freshwater flux (evaporation minus precipitation).
• is the climatological SSS for both the AMV phases.

SSS T ARGET
• is the model SSS.

SSS M ODEL
• i s the negative SSS restoring term which indicates the magnitude of the damping γ S in the salinity. Unit is mm/d.
Given the model-dependence nature of , a tuning of salinity damping term has been γ S necessary to assess the best value which contributes to perturb the least possible the climate-system equilibrium. According to hfcorr diagnostics, setting equal to -432.00 γ S mm/d (T R ≅ 4 months for a 50-m MLD) determines a better restoring in terms of SST convergence towards the observed AMV phase, without excessively altering the ocean circulation and preserving the characteristics of the water masses.
The SST/SSS restoring is applied over the North Atlantic (from 10°N to 65°N) based on a mask provided by DCPP protocol on a 2-degree horizontal grid. Outside this target region, the model is allowed to freely evolve. Following the protocol [Boer et al. 2016], the original AMV mask is extrapolated over land to minimize the interpolation errors and, using a bilinear method, is regridded onto the ORCA1 tripolar grid of NEMO ocean model ( Figure   S.4). An 8-degree wide buffer zone is opportunely designed at the edge of the nudging area to minimize shocks and to avoid instabilities in the no-restoring region.
To reduce the non-linear uncertainty related to the internal variability of the climate system, an ensemble approach is used. The DCPP protocol recommends to generate the ensemble using at least 25 "macro-perturbations" [Hawkins et al. 2016], i.e. considering 25 different ocean states as initial condition of each realization, in addition to only different atmospheric states ("micro-perturbation"). This ensures a better spread among the ensemble members. In this study, 32 members are considered taken from a multicentury pre-industrial run.
Two sets of idealized experiments have been performed where the perturbation via restoring is applied: • AMV+ experiments: North Atlantic (10°N -65°N) SSTs are restored to positive time-independent AMV anomaly (i.e. +1 standard deviation of the AMV index) superimposed on 12-month model climatology. No restoring is performed where the sea ice fraction is greater than 15%, which is the Sea-Ice Extent definition.
• AMV-experiments: they are analogous to AMV+ experiments, but it is considered the negative AMV anomaly (i.e. -1 standard deviation of the AMV index).
Each simulation is integrated over a 10-year period, allowing to catch the climate response to the AMV input, while longer simulations may lead to the aforementioned drift, introduced by the experimental setup [Ruprich-Robert et al. 2017].