Limited Influence of Localized Tropical Sea‐Surface Temperatures on Moisture Transport into the Arctic

Arctic moisture transport is dominated by planetary‐scale waves in reanalysis. Planetary waves are influenced by localized Sea‐Surface Temperature (SST) features such as the tropical warm pool. Here, an aquaplanet model is used to clarify the link between tropical SST anomalies and Arctic moisture transport. In a zonally uniform setup with no climatological east‐west gradients, Arctic moisture transport is dominated by transient planetary waves, as in reanalysis. Warming tropical SSTs by heating the ocean strengthens Arctic moisture transport, mediated mostly by changes in water vapor rather than eddies. This strengthening occurs whether the tropical warming is zonally uniform or localized. Cooling tropical SSTs weakens Arctic moisture transport; however, unlike warming, the pattern matters, with localized cooling producing stronger transport changes owing to nonlinear feedbacks in the surface energy budget. Thus, the simulations show that localized tropical SST anomalies influence Arctic moisture transport differently than uniform anomalies, but only in cooling scenarios.

. Indeed, subseasonal to decadal variability in warm pool convection has been linked to corresponding variability in Arctic moisture transport and surface temperature in observations (Gong et al., 2017;Lee, 2012;Lee et al., 2011;Yoo et al., 2011).
Warming tropical SSTs could also drive Arctic moisture transport via processes that rely on changes in the zonal-mean state rather than in quasi-stationary planetary waves. For example, localized SST anomalies in the western Pacific produce a zonally uniform atmospheric temperature change in the tropics (Sobel et al., 2001;Yulaeva & Wallace, 1994) that can modify transient eddies and their associated energy transport in the extratropics (L'Heureux & Thompson, 2006;Robinson, 2002;Seager et al., 2003). Tropical atmospheric temperatures are also tightly coupled to water vapor (Trenberth et al., 2005;Wentz & Schabel, 2000), so changes in Arctic moisture transport might be expected to scale with the Clausius-Clapeyron relationship, in a manner similar to other aspects of the climate system under global warming (Held & Soden, 2006;Lorenz & DeWeaver, 2007;Seager et al., 2010).
In addition to how moisture transport changes driven by tropical SST anomalies are communicated to the Arctic, we must also consider that the effects of cooling (e.g., La Niña) may differ from those of warming (e.g., El Niño). For example, cooling produces a larger SST change per unit energy perturbation because, unlike heating, the resulting energy imbalance is not efficiently offset by evaporation (Shin et al., 2017). Cooling also induces changes in surface cloud radiative forcing that amplify the cooling (Shaw et al., 2015). Thus, cooling could more efficiently perturb tropical SSTs than warming, leading to a more pronounced change in the zonal-mean state and a greater change in Arctic moisture transport. Moreover, tropical convection depends sensitively on the absolute value of SST, so equal but opposite SST anomalies could excite different quasi-stationary planetary waves (Hoerling et al., 1997) that result in different Arctic moisture transport.
To elucidate how tropical SSTs affect moisture transport into the Arctic, and which structural features of SST anomaly patterns are most important, we carry out a series of perturbation experiments using a slabocean aquaplanet model. The model allows us to clarify the processes governing observed zonally integrated Arctic moisture transport (Section 3.1), as well as the role of localized tropical SST anomalies induced by warming or cooling in driving changes in moisture transport (Section 3.2). This, to our knowledge, provides a novel testbed for isolating the role of tropical SSTs on Arctic moisture transport mediated by planetary waves. The results suggest that an ongoing focus on transient, high-latitude, planetary waves is warranted, and show a concrete but limited influence of localized SST anomalies on zonally integrated Arctic moisture transport arising from tropical cooling (Section 4).

Model, Experiments, and Diagnostics
The National Center for Atmospheric Research Community Atmospheric Model version 5 (NCAR CAM5, Neale et al., 2010) is used. The atmosphere has a horizontal resolution of 1° latitude by 1.25° longitude, has 30 vertical hybrid-sigma levels and is coupled to a slab ocean. The control simulation is forced with zonally uniform boundary conditions following the Tropical Rain belts with an Annual Cycle and continent Model Intercomparison Project protocol (TRACMIP, Voigt et al., 2016). Specifically, the model is forced with present-day seasonal and diurnal cycles of insolation and concentrations of greenhouse gases, the slab ocean depth is set to 30 m, sea ice is turned off (SSTs can drop below freezing), zonal-mean ozone is prescribed following Blackburn and Hoskins (2013), and a time-independent ocean heat flux convergence (termed the "Q-flux") is prescribed to approximate the observed climatological zonal-mean (Equation 3 and Table 2 in Voigt et al. (2016)). The control simulation is spun up for 10 years and the next 40 years are used for analysis. The control and other simulations are compared to ERA-Interim data from 1979 to 2017 (Dee et al., 2011).
We investigate the impact of tropical SSTs on moisture transport into the Arctic by imposing time-independent Q-flux perturbations in the tropics, which generate diabatic heating anomalies and extratropical circulation responses. Experiment L+ imposes a localized Q-flux perturbation centered on the equator with a spatial scale of ∼90° longitude and 30° latitude (Figure 1a). The exact pattern is taken from Equation 5 in Neale and Hoskins (2000) Figure 1. Annual-mean tropical Sea-Surface-Temperature (SST) anomalies defined relative to the control simulation for experiments L+, L−, U+, U−, 2L+, L k1 , and 2U+ (see Section 2 for description of the experiments). Purple contours denote Q-flux perturbations imposed in each experiment, with contour interval 30 W m −2 and solid lines denoting positive perturbations, and dashed lines denoting negative perturbations. Statistical significance is calculated at the 95% level using a two-tailed student's t-test and hatching indicates anomalies that are not statistically significant. Note the different color bar range for experiment L− and that the fields in all experiments are phase shifted westwards by 75° for ease of viewing. : , 0 otherwise where λ denotes longitude, ϕ denotes latitude, and the spatial extent of the perturbation is set by λ d = π/2 and ϕ d = π/6. The magnitude χ = 150 W m −2 is chosen to resemble the annual-mean east-west difference in ocean heat flux convergence across the equatorial Pacific in reanalysis. We have confirmed that the results are similar for different choices of the spatial scale λ d and that the resulting Arctic stationary waves have comparable amplitude to reanalysis (Figures S1 and S2). Experiment U+ imposes a zonally uniform Q-flux perturbation with the same total heat input as in experiment L+, but distributed equally over all longitudes ( Figure 1c). Thus, experiments U+ and L+ input identical amounts of globally integrated energy into the slab ocean. Additional experiments are performed where the sign of the Q-flux perturbation is reversed (L−, U−), the magnitude is doubled (2L+, 2U+), and a zonal wave number k = 1 pattern with amplitude 75 W m −2 is imposed (L k1 , Equation 6 in Neale and Hoskins (2000)). The last experiment inputs zero globally integrated energy into the slab ocean. The experiments are branched from the last day of the control simulation, spun up for 10 years and the next 40 years are used for analysis. Figure 1 shows annual-mean SST anomalies relative to the control simulation (shading) resulting from the Q-flux perturbations (purple contours). The SST warming (cooling) generally follows the pattern of heat input (output) by the Q-flux forcing. However, localized heat removal ( Figure 1b) produces stronger SST anomalies than localized heat input (Figure 1a), and dominates the response when both forcings are present ( Figure 1f). This asymmetry arises due to nonlinear feedbacks in the surface energy budget (Shaw et al., 2015;Shin et al., 2017) and becomes apparent when the Q-flux perturbations are larger (compare Figures 1a,1b and 1c,1d). This asymmetry will be useful for understanding the Arctic moisture transport responses discussed later. The tropical SST responses are communicated to the tropical atmosphere via changes in convection, reflected in changes in vertically integrated diabatic heating ( Figure S3). This, in turn, results in the generation or attenuation of quasi-stationary planetary waves (Sardeshmukh & Hoskins, 1988).
These experiments do not include climatological zonal asymmetries from the presence of land and orography and so the resulting changes in Arctic planetary waves and moisture transport should be interpreted with certain caveats in mind. For example, zonal asymmetries can modify the extratropical waveguide and poleward propagation of planetary waves (Simmons et al., 1983;Ting & Sardeshmukh, 1993). Moreover, a given SST anomaly can generate different planetary waves depending on its location relative to a zonally varying background state. This is because the planetary wave source depends on tropical convection, which is sensitive to the absolute value of SST (Hoerling et al., 1997) and the location of this convection relative to the climatological waveguide (Branstator, 2014). Nevertheless, these experiments represent a first step in understanding the link between tropical SST anomalies and Arctic moisture transport.
We diagnose daily vertically and zonally integrated eddy moisture transport as a function of zonal wave number following Graversen and Burtu (2016): where angled brackets denote the vertical integral, square brackets denote the zonal integral, and stars denote deviations from the zonal integral. The procedure decomposes specific humidity q and mass-flux

Comparing Arctic Moisture Transport in Reanalysis and the Control Simulation
Annual-mean zonally integrated moisture transport in the control simulation is qualitatively similar to reanalysis in the Arctic but differs at lower latitudes (Figures 2a and 2b). In the Arctic, the transport in the control simulation and reanalysis is dominated by transient planetary waves (k ≤ 3, see also transient vs. stationary moisture transport for reanalysis in Figures 1e and 1f of Lee et al. (2019)). At lower latitudes, moisture transport in the control simulation is dominated by synoptic waves (k > 3) whereas both synoptic and planetary waves are important in reanalysis. The differences at lower latitudes arise due to a lack of stationary planetary waves in the control simulation. Moisture transport is generally larger in the control simulation than reanalysis except in the tropics, likely due to more available moisture in the ocean-covered aquaplanet and/or more vigorous eddies.
10.1029/2020GL091540 5 of 10 Poleward moisture transport into the Arctic in the control simulation causes near surface warming similar to reanalysis (Figures 2c and 2d). Lagged correlations between anomalous moisture transport at 70°N and zonal-mean temperature at 850 hPa show that poleward transport events are preceded by weak cold anomalies followed by stronger warm anomalies resulting in net warming over the polar cap. The Arctic warming is primarily associated with moisture transport by planetary waves ( Figure S4) and consistent with enhanced downward longwave radiation associated with a stronger water vapor greenhouse effect and stronger cloud radiative forcing ( Figure S5).
Overall, we find that zonally integrated Arctic moisture transport and its associated warming impacts are dominated by transient planetary waves in an aquaplanet model driven by zonally uniform boundary conditions. The results reveal the key role of transient, high-latitude, planetary waves in determining Arctic climate in a highly idealized setup, as in reality (Baggett & Lee, 2015;Graversen & Burtu, 2016;Heiskanen et al., 2020;Lee et al., 2019;Papritz & Dunn-Sigouin, 2020). Thus, we consider the aquaplanet model a useful framework for investigating how tropical SST perturbations affect zonally integrated Arctic moisture transport mediated by planetary waves.

Clarifying How Tropical SSTs Impact Arctic Moisture Transport
Warming tropical SSTs by heating the tropical ocean gives rise to strengthened moisture transport into the Arctic. But notably, in our simulations, zonally uniform and zonally localized heating result in a similar strengthening. Figures 3a and 3b show the percent change (shading) from the control simulation (gray contours) of annual-averaged moisture transport in experiments L+ (localized warming) and U+ (uniform warming). Moisture transport is generally stronger at all latitudes, with larger increases in the tropics. Assuming a purely thermodynamic response following the Clausius-Clapeyron relationship, larger tropical changes can arise either from a uniform temperature change, yielding greater water vapor increases for warmer background temperatures, or from a change in the gradient of temperature, yielding greater water vapor increases where the heating is applied. For the same total energy input, both experiments show a 15-20% increase in Arctic moisture transport. The increase is primarily due to stronger transport by transient eddies (compare black and gray circles in Figure 3c), even in experiment L+ where the heating creates locally warm SSTs reminiscent of the tropical warm pool (Figure 1a). Similar but larger increases in Arctic moisture transport occur when doubling the magnitude of the warming (40-50%, experiments 2L+ and 2U+ in Figure 3c).
The stronger eddy transport is consistent with changes in water vapor rather than changes in the eddies themselves ( Figure 3c). Blue circles show the percent change in lower-tropospheric water vapor while red circles show changes arising from eddy strength, quantified using the zonal-mean root-mean-square of meridional wind at 850 hPa. The increase in Arctic moisture transport is consistent with an increase in lower-tropospheric water vapor (compare black and blue circles for all experiments), in agreement with Clausius-Clapeyron scaling (green circles) obtained by multiplying the temperature change at 70°N and 850 hPa in each experiment with 7.3%/K calculated from the control simulation (Equations 1 and 2 in Held and Soden (2006)). Increases in total (stationary plus transient) eddy strength are small in comparison (red circles), as are increases in transient eddy strength (pink circles). These results point to thermodynamic rather than dynamic processes dominating the response of zonally integrated Arctic moisture transport to tropical SST perturbations, whether the perturbations are zonally localized or not.
Cooling tropical SSTs weakens Arctic moisture transport; however, unlike in the warming case, it matters whether the perturbation is zonally localized or zonally uniform (Figure 3c). For the same zonally integrated energy output, the Q-flux perturbation in experiment U− (uniform cooling) reduces Arctic moisture transport by 15-20% while that in experiment L− (localized cooling) reduces the transport by 40-45% (black circles). The asymmetry is consistent with nonlinearity in the surface energy budget response, which is larger for the strong localized cooling in experiment L− (Figures 1b and 1d). Indeed, although not shown here, the simulations exhibit different responses in surface evaporation and cloud radiative forcing, leading to different SST responses. Thus, reductions in Arctic moisture transport due to tropical cooling are most likely sensitive to differences in the water vapor response to different cooling patterns.
While thermodynamic changes in Arctic moisture transport are most important, they can be opposed by dynamic changes. Applying a Q-flux perturbation with heat input over half the tropical band and heat removal over the other half (experiment L k1 , where k1 indicates a zonal wave number-1 pattern) results in mostly colder SSTs (experiment L k1 , Figure 1f), and correspondingly, reduced water vapor and a thermodynamic weakening of Arctic moisture transport by transient eddies (blue and gray circles in Figure 3c). However, the weakening of total eddy moisture transport is relatively smaller (compare black and gray circles in Figure 3c), implying a dynamic strengthening in stationary eddy moisture transport, likely arising from planetary waves generated in the tropics. Strengthened stationary eddies also partly compensate weakened transient eddies, resulting in a near zero change in total eddy strength (compare red and pink circles in Figure 3c).
In summary, these aquaplanet simulations show that localized tropical SST anomalies can influence Arctic moisture transport more than zonally uniform SST anomalies, but only in the case of cooling relative to a DUNN-SIGOUIN ET AL.
10.1029/2020GL091540 7 of 10 Annual-mean vertically and zonally integrated eddy moisture transport as a function of zonal wave number and latitude for the control simulation (contours) and percentage change from control for the L+ (local warming) and U+ (uniform warming) experiments (shading). Control contour interval is 0.04 PW starting at 0.01 PW and the percent change is shown where the climatology is greater than 0.01 PW. Statistical significance is calculated at the 95% level using a two-tailed student's t-test and percent changes that are not statistically significant are hatched. (c) Percentage change from control of the following annual-mean variables at 70°N for all warming (+) and cooling (−) experiments: (black) vertically and zonally integrated eddy moisture transport, (gray) vertically and zonally integrated transient eddy moisture transport, (blue) zonal-mean specific humidity at 850 hPa, (red) zonal-mean root-mean-square eddy meridional wind at 850 hPa, and (pink) zonal-mean root-mean square transient eddy meridional wind at 850 hPa. Green circles show the moisture transport change estimated from Clausius-Clapeyron scaling: zonal-mean temperature change in each experiment multiplied by 7.3%/K calculated from the control simulation (Equations 1 and 2 in Held and Soden (2006) zonally uniform background state. The transport responses are consistent with thermodynamic changes in water vapor resulting from zonal-mean atmospheric temperature changes. Dynamic changes in stationary eddy transport, likely induced by planetary waves excited in the tropics, can partly enhance or oppose these thermodynamic changes, depending on the exact nature of the tropical perturbation. Similar results are also found for the extended November to March winter season ( Figure S6).

Concluding Remarks
This study clarifies how tropical SST anomalies affect zonally integrated Arctic moisture transport in an aquaplanet model. We find that warming tropical SSTs strengthens Arctic moisture transport, and the response is similar whether the warming pattern is zonally uniform or localized (Figures 3a and 3b). However, localized cooling weakens the transport more than zonally uniform cooling (Figure 3c). The differences arise due to nonlinearity in the tropical surface energy budget response to warming vs. cooling (Figures 1a  and 1b). Thus, the simulations show that localized tropical SST anomalies influence Arctic moisture transport differently than uniform SST anomalies, but only in cooling scenarios.
The results help clarify the processes by which the tropics influence zonally integrated moisture transport into the Arctic. Thermodynamic processes involving changes in zonal-mean atmospheric temperature and water vapor are most important (Figure 3c), as for responses to climate change (Held & Soden, 2006;Lorenz & DeWeaver, 2007;Seager et al., 2010). Dynamic processes involving eddy changes (Baggett & Lee, 2015;Lee et al., 2011;Robinson, 2002;Seager et al., 2003) play a secondary role, acting to enhance or oppose thermodynamic processes, depending on the exact nature of the tropical perturbation. The relative importance of these processes in the real world is difficult to address directly using our results, however, because of the simplified model configuration (in particular, the lack of stationary waves in our control simulation). Moreover, dynamic processes could be more important for local rather than zonally integrated transport similar to climate change (Seager et al., 2014;Simpson et al., 2016;Wills et al., 2016), and strong nonlinearity in the surface energy budget (Shaw et al., 2015;Shin et al., 2017), of the kind shown in response to SST cooling (Figures 1b and 1d), could be model dependent. Further work is needed to investigate the processes linking tropical SST anomalies and Arctic moisture transport in more complex and varied model configurations.
The results reveal the key role of transient, high-latitude, planetary waves in mediating zonally integrated Arctic moisture transport in an idealized model (Figure 2), as in reality (Baggett & Lee, 2015;Graversen & Burtu, 2016;Heiskanen et al., 2020;Lee et al., 2019;Papritz & Dunn-Sigouin, 2020). Observations further suggest that planetary-scale blocking patterns help steer moisture-laden synoptic weather systems into the Arctic (Papritz & Dunn-Sigouin, 2020;Ruggieri et al., 2020;Woods et al., 2013). Using our idealized model to relate weather systems with Arctic moisture transport could clarify the role of interactions between synoptic and planetary scales.