Uncertainties in projecting future changes in atmospheric rivers and their. impacts on heavy precipitation over Europe

Transcription

1 1 2 Uncertainties in projecting future changes in atmospheric rivers and their impacts on heavy precipitation over Europe 3 Yang Gao, Jian Lu and L. Ruby Leung 4 5 Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, Richland, Washington, USA 6 7 Correspondence to: Dr. Yang Gao (Yang.Gao@pnnl.gov) Dr. L. Ruby Leung (Ruby.Leung@pnnl.gov)

2 12 Abstract This study investigates the North Atlantic atmospheric rivers (ARs) making landfall over western Europe in the present and future climate from the multi-model ensemble of the Coupled Model Intercomparison Project Phase 5 (CMIP5). Overall, CMIP5 captures the seasonal and spatial variations of historical landfalling AR days, with the large inter-model variability strongly correlated with the inter-model spread of historical jet position. Under RCP 8.5, AR frequency is projected to increase a few times by the end of this century. While thermodynamics plays a dominate role in the future increase of ARs, wind changes associated with the midlatitude jet shifts also significantly contribute to AR changes, resulting in dipole change patterns in all seasons. In the North Atlantic, the model projected jet shifts are strongly correlated with the simulated historical jet position. As models exhibit predominantly equatorward biases in the historical jet position, the large poleward jet shifts reduce AR days south of the historical mean jet position through the dynamical connections between the jet positions and AR days. Using the observed historical jet position, which is more poleward than simulated, as an emergent constraint, dynamical effects further increase AR days in the future above the large increases due to thermodynamical effects. In the future, both total and extreme precipitation induced by AR contribute more to the seasonal mean and extreme precipitation compared to present primarily because of the increase in AR frequency. While AR precipitation intensity generally increases more relative to the increase in integrated vapor transport except in western Europe, AR extreme precipitation intensity increases much less relative to the increase in extreme integrated vapor transport, or even decreases, most notably in mountain regions. Future improvements in simulating the midlatitude jet and orographic clouds and precipitation may potentially yield further insights on changes in ARs and their impacts on precipitation in a warmer climate. 2

3 35 Keywords: Atmospheric rivers, CMIP5, jet position, RCP 8.5, extreme precipitation 36 3

4 Introduction Atmospheric rivers (ARs) are narrow corridors of water vapor, usually with a length of 2000 km or more, that account for over 90% of the meridional moisture transport associated with storm tracks in the extratropical atmosphere [Zhu and Newell, 1998]. During winter, AR induced precipitation could account for 15-30% of the total precipitation in Europe and western US [Lavers and Villarini, 2015]. In northwestern US, ARs contribute to 25-55% of extreme precipitation [Rutz et al., 2014]. Because of their importance to floods and water resources, ARs have been extensively studied in the last decade [Gao et al., 2015; Hagos et al., 2015; Lavers and Villarini, 2015; Leung and Qian, 2009; Neiman et al., 2011; Ralph and Dettinger, 2012; Ralph et al., 2006] As prominent features over the Pacific Ocean, ARs that make landfall in the west coast of North America have been investigated both in the context of historical climatology and extreme events [Neiman et al., 2011; Ralph and Dettinger, 2012; Ralph et al., 2006] and future changes under climate warming [Gao et al., 2015; Payne and Magnusdottir, 2014]. Neiman [2008] and Gao et al. [2015] found higher number of ARs in the cool (warm) season in the southern (northern) coast. Changes in AR frequency in the future are closely related to changes in water vapor as well as atmospheric circulation, both strongly modulated by global warming [Barnes and Polvani, 2013]. In particular, changes in the winds associated with AR moisture transport were found to predominantly counter the effects of increasing water vapor that substantially increases the frequency of landfalling ARs in western North America [Gao et al., 2015]. However, with a poleward shift of the storm tracks, wind changes could also increase AR days in the high latitudes such as coastal Alaska in spring time. 4

5 In the North Atlantic, Lavers et al. [2013] investigated the dynamical and thermodynamical modulations of future landfalling ARs in the United Kingdom (50-60 N) and found that the increase of future winter AR is mainly a result of thermodynamical effect whereas dynamical effect plays very little role. However, similar to AR frequency in western North America, the number of AR days in western Europe could vary dramatically by seasons and locations. Thus, to fully understand the changes of North Atlantic ARs that make landfall in Europe and the driven mechanism, this study investigates the seasonal changes in AR days and extreme precipitation across the entire coastal area in western Europe. Using a multi-model ensemble of climate projections, uncertainty in projecting the changes in AR days is investigated, with the goal of exploring emergent constraints for AR changes for more robust projections of future changes In what follows, we first investigate the seasonality of landfalling ARs over Europe and examine the sources of the inter-model spread of AR days in a multi-model ensemble. Using outputs from the same set of models, the projected changes of the number of AR days under climate warming and the thermodynamical and dynamical modulations of the AR changes are evaluated. Lastly, the total and extreme precipitation associated with ARs and the future changes are discussed Data and method In this study, the same 24 CMIP5 models used by Gao et al. [2015] and listed in Table S1 of the supplementary material are analyzed for the historical period of and future period of under the RCP 8.5 scenario [Moss et al., 2010; Vuuren et al., 2011]. Details regarding the CMIP5 data and the four reanalysis data used for model evaluation are discussed in Text S1 in the supplementary material of Gao et al. [2015]. To summarize briefly, outputs from one 5

6 member for each CMIP5 model are interpolated to a 1.25 o latitude by o longitude grid. To evaluate how well the CMIP5 models capture the seasonal variations of ARs, four reanalysis datasets are used in this study: NCEP Climate Forecast System Reanalysis (CFSR) [Saha et al., 2010], ECMWF Interim Reanalysis Data (ERA-Interim) [Dee et al., 2011], MERRA [Rienecker et al., 2011] and National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) Reanalysis 1 (NCEP1) [Kalnay et al., 1996]. The common period of these four datasets ( ) is used for evaluating the historical ARs. For consistency, all reanalysis data are also interpolated to the 1.25 o by o grid. Variables used in this study mainly include daily mean temperature, specific humidity, zonal and meridional winds from 1000 hpa to 500 hpa and daily total precipitation from CMIP5 and the reanalysis data The vertically integrated vapor transport (IVT) is estimated by integrating the moisture transport between the 1000 hpa and 500 hpa pressure levels as 93 IVT =!!!""!""" qu dp! +!!!""!""" qv dp!, where g represents the gravitational acceleration, q represents the layer mean specific humidity, u represents zonal wind and v represents meridional wind Following [Gao et al., 2015; Lavers and Villarini, 2013; Lavers et al., 2012], we first identify ARs that make landfall in the west coast of Europe between 30 N to 70 N. The coastal grids, indicated by the colored grid cells in Figure 1, are grouped into 8 bins shown by the different colors separated by the gray dashed lines for each 5-degree latitudinal band. Daily IVT was computed first and on each day, the grid cell with the maximum IVT along the west coast of Europe is recorded, and the 85 th percentile of IVT in each bin was determined as the threshold 6

7 for that bin. If the IVT in the recorded grid cell exceeds the threshold of its corresponding bin, a backward (northwest/west/southwest/south) and forward (north/northeast/east/southeast) search was performed. The search continues only if one of the adjacent grids exceeds the IVT threshold. If the total trajectory (including both backward and forward) spans longer than 2000 km [Neiman et al., 2008; Ralph et al., 2004], and the mean vertically integrated water vapor (IWV) over the path is greater than 2 cm, all the grids along the path is defined to have an AR day Results Historical number of AR days in CMIP To gain confidence in how well the CMIP5 models are able to simulate North Atlantic ARs, the seasonal total number of AR days simulated across the eight bins over the coastal area in Europe is compared with four reanalysis datasets, shown in Figure 2. Similar to ARs in eastern Pacific, ARs occur more frequently in fall and winter along the European coast, with a maximum of about three to four AR days between 45 o 50 o N, while spring has the lowest likelihood for AR occurrence. Overall, the latitudinal variations in the CMIP5 multi-model ensemble mean (MME) correspond relatively well with the reanalysis data across the four seasons. However, there is significant inter-model spread and the CMIP5 models overestimate and underestimate the number of ARs in a few locations and seasons, with the most prominent overestimation occurring over the 35 o -40 o bin during the winter

8 As a prominent atmospheric circulation feature, uncertainties and biases in simulating AR frequency may be related to those of the model simulated large-scale environment. Since ARs are associated with large IWV and IVT, we first analyze the relationships between temperature and winds, which influence water vapor and its transport, with AR frequency in the models and global reanalyses. In all seasons, the relationships between the inter-model spread of temperature and water vapor with the inter-model spread of historical number of ARs are weak, especially in the areas with larger model biases. On the contrary, there are significant correlations between the historical AR numbers and the jet stream. Jet speed and position are defined as the maximum of the seasonal mean 850 hpa zonal wind averaged between 30 W and 10 E and the corresponding latitude, respectively Focusing on the winter season because of the larger biases and inter-model spreads (Figure 3), we find that the mean winter jet position from the four reanalyses is around 53 N, but the CMIP5 jet position ranges from 41 N to 54 N with a mean position at 48 N, showing an evident equatorward bias. There are distinct relationships between AR days and jet position on both sides of the CMIP5 mean jet position. South of the CMIP5 mean jet position (40-45 N, Fig. 3a), models that have larger equatorward biases in the jet position simulate more ARs. Conversely, north of the CMIP5 mean jet position (i.e., N, Fig. 3c), models with more poleward jet position simulate more ARs. The identical but opposite slopes from linear regression of AR days on jet position in the two regions suggest an asymmetric response of AR days to the jet position in the models. For the latitude bin (45-50 N, Fig. 3b) that coincides with the mean CMIP5 jet position, AR frequency shows a weak relationship with jet position, but correlates more strongly with the jet speed. Hence the overestimation (underestimation) of AR days between N (50-55 N) in Figure 2a is largely attributable to the equatorward biases in the CMIP5 jet 8

9 position, while the overestimation of jet speed contributes to the overestimation of AR days between N. Similar relationships are also found for the fall (Figure S3) and spring (Figure S1) seasons. In summer (Figure S2), the mean CMIP5 jet position (52 N) is very close to that of the four reanalyses, so the AR days are more tightly correlated with jet speed instead of jet position. Overall, uncertainties in model simulated jet position and speed contribute importantly to uncertainties in the simulated AR days Thermodynamical and dynamical contributions to changes in AR days in the future To investigate the impact of climate change on landfalling ARs, Figure 4 shows the numbers of AR days at present ( ; black) and future under RCP 8.5 ( ; red), with the percentage change ([RCP 8.5 Present]/Present) indicated by the numbers in the top row. A majority of coastal areas show significant increases in the number of AR days by a few times under warming. The AR frequency peaks between 45 N to 55 N for all seasons in both current and future climate. This region corresponds to the mean CMIP5 jet position, where the peak AR days increase between 127% and 275% To investigate the thermodynamical and dynamical contributions to the increase of AR days, a scaling method is used to separate the effects of changes in water vapor and winds that influence the IVT. As described in detail in Gao et al. [2015] and briefly summarized here, we rescaled the water vapor simulated for the present climate by a factor of!!!!!!, where q!! and q!! are the thirty-year mean IWV averaged over the eastern North Atlantic basin (20 N to 60 N, 60 W to 15 W) for the present and future, respectively. ARs were detected using the rescaled IVT that combines the rescaled water vapor with the present day winds, referred to as V! Q!, for comparison with the AR days simulated for the present (black line). Contribution to changes in 9

10 AR days from the increase of water vapor in the future is estimated as the difference in AR days between the rescaled and present-day IVT and shown as percentage increases, quantified by the numbers in the second row in Figure The thermodynamical contribution can also be evaluated by rescaling the present-day water vapor based on the Clausius-Clapeyron (C-C) relationship with warming as in Lavers et al. [2013] for winter AR changes over United Kingdom (50-60 N). The C-C scaling approximated with a 7% increase of water vapor per degree K warming is shown by the dashed line in Figure 4. As discussed by Gao et al. [2015], the C-C scaling may underestimate the water vapor changes in ARs, which carry high percentile water vapor content. As shown in Figure 4, the underestimation using the C-C scaling (comparing the dashed line and the blue line) is strong particularly in spring, summer and fall, although the agreement with the rescaling of IVT is better during winter in the latitudinal range of the United Kingdom (50-60N) (see also Lavers et al. [2013] ) To examine the dynamical modulations of AR events, the effects of wind changes on the ARs can be inferred by rescaling the future IVT by a factor of!!!!!!, referred to as V! Q! (orange line in Figure 4) and comparing the AR days with that of V! Q! (black line in Figure 4). The resulting difference in AR days due to dynamical effects is shown in Figure 5 together with the corresponding changes in zonal wind speed. It is clear that the AR days increase and decrease following the changes in zonal wind speed averaged over each 10-degree latitudinal bin and 20 degrees west of the coastal area. A dipole pattern of positive and negative changes on each side of the peak AR frequency for the respective season (see the black curves in Figure 5) indicates a poleward dynamical shift, in concert with the shift of the zonal wind. South of the peak area, the 10

11 number of ARs is reduced dynamically due to the decreases in zonal wind speed, while north of the peak area, the increases of zonal wind speed largely drive the increase of ARs. The change in AR days due to dynamical effects can also be estimated by comparing the AR days simulated for the present and future, each detected using its respective 85 th percentile threshold for IVT, thus eliminating the large influence of enhanced water vapor with warming that drives a significant increase in IVT. As discussed in Gao et al. [2015], this method circumvents errors in the rescaling method related to the covariance between water vapor and winds. Overall, we found consistent changes in AR days due to dynamical effects using the rescaling method and the respective percentile IVT thresholds for present and future climate Emergent constraint on AR changes in a warmer climate Analysis of the CMIP5 model biases and inter-model spreads in the historical AR days indicates an important control of the jet stream on ARs. As ARs are associated with extratropical storms, and storm tracks are steered by the jet stream, the equatorward bias of historical jet position in CMIP5 models displaces the AR frequency equatorward compared to the global reanalyses (Figure 3). The dipole changes in AR days due to dynamical effects that increase (decrease) the AR days equatorward (poleward) of the historical mean jet position (Figure 5) further hints at a role for jet stream changes in future ARs Kidston and Gerber [2010] found a statistically significant correlation between the modelprojected changes in jet position and their climatological jet position in CMIP3 models. Similar correlations are shown in Figure 6 for the CMIP5 models for the Atlantic jet during winter and spring. Models with a larger equatorward bias in the historical jet position project a larger 11

12 poleward jet shift in the future. This relationship is stronger in winter than spring. Unlike the robust poleward shift of the southern hemispheric jet stream [Kidston and Gerber, 2010], both poleward and equatorward shifts are projected for the North Atlantic jet, with an ensemble mean shift of jet position close to zero in winter and about 1.5 degrees poleward in spring (gray horizontal dashed line in Figure 6). Were the reanalysis jet positions used as emergent constraints, the mean shift of jet position would be calibrated to be equatorward in winter and ~1 equatorward in spring To further link the historical jet position with the change of AR days, Figure 7 shows the relationships between the changes in AR days and the historical jet position for winter. The changes in AR days are those associated with the dynamical effects (i.e., excluding the large changes due to changes in water vapor in a warmer climate) and shown for two regions between N, as suggested by the dipole changes displayed in Figure 5a. In this analysis, the dynamical effects are estimated by comparing the AR days based on the respective 85% IVT thresholds for the current and future climate, which yielded higher correlations between the historical jet position and the change of AR days than if the latter were estimated based on the rescaling method, possibly because the rescaling method does not account for covariance between moisture and wind changes [Gao et al., 2015]. The CMIP5 ensemble mean change of AR days due to dynamics in N (Portugal and Spain) is close to zero (which is also shown in Figure S4a in the supporting information). Using the historical mean jet position from the reanalysis as an emergent constraint, that is, adjusting the ensemble mean projection of AR days along the gray regression slope in Figure 7, AR days are projected to increase by extra 0.4 days above the CMIP5 ensemble mean change over N and 0.7 days over N in winter, as the more realistic jet (should be more poleward in its mean position) would shift less 12

13 poleward (i.e., equatorward by as depicted in Figure 6a) with warming and the associated reduction of the wind at the equatorward flank is smaller Changes of landfalling AR induced precipitation in the future By virtue of the enhanced water vapor transport, heavy precipitation often accompanies landfalling ARs as they encounter mountainous terrains that provide a lifting mechanism for cloud formation. We define AR induced precipitation as the precipitation over land that occurs on the same day as the AR and within 250 km of the AR trajectory defined in section 2. The AR induced precipitation is considered extreme if the daily precipitation amount exceeds the 95 th percentile of all daily precipitation in each season. The fractional contributions of AR induced precipitation to the seasonal total precipitation are shown in Figure 8, with the ERA-Interim in the top row and present and future CMIP5 multi-model ensemble (MME) mean in the middle and bottom rows. Similar fractional contributions but for AR induced total seasonal extreme precipitation are shown in Figure 9. Note that our estimates of AR contributions are slightly lower than that of Lavers and Villarini [2015] because a smaller IVT threshold of 50 th percentile (ranges from 218 kg m -1 s -1 to 295 kg m -1 s -1 with an average of 244 kg m -1 s -1 ) was used in their study compared to the 85 th percentile (ranges from 311 kg m -1 s -1 to 435 kg m -1 s -1 with an average of 376 kg m -1 s -1 ) used here for AR detection. The ERA-Interim global reanalysis shows the largest contributions of ARs to seasonal total precipitation of up to 15% in Portugal and Spain during fall, followed by up to 10% in the same areas as well as the coastal regions of France and United Kingdom in winter. Similar seasonal and spatial variability is noted for the AR contributions to seasonal extreme precipitation, except that the latter more than doubles the 13

14 AR contributions to seasonal total precipitation, because ARs are sporadic events with the potential to generate intense precipitation Overall, the CMIP5 MME captures the spatial variability of the AR contributions to seasonal total and extreme precipitation (Figure 8, 9 top and middle rows), albeit a slight overestimation from 35 N to 45 N in fall and winter. The overestimation is primarily contributed by the positive bias in the number of ARs in CMIP5 as discussed in section Under climate warming, the contributions of AR induced precipitation to both the seasonal total and extreme precipitation increase substantially. For instance, the contributions of ARs induced precipitation to the total precipitation are projected to increase by 10-20% (bottom row in Figure 8) from 35 N to 45 N in Portugal, Spain, Ireland and United Kingdom, with more modest increase projected for other areas. Similar increases (15-25%) in the contributions of ARs to extreme precipitation are found in the west coast of Europe from 35 N to 60 N, as a result ARs are projected to contribute to 25-70% of seasonal extreme precipitation in the future The analysis presented above shows that ARs contribute more to seasonal total and extreme precipitation compared to other precipitation systems in a warmer climate. To understand the changes in AR precipitation, Figure 10 shows the percentage changes in AR total and extreme precipitation and the corresponding changes in IVT for winter. Lavers et al. [2014] showed that in Europe, AR induced precipitation is highly correlated with the IVT, particularly in mountainous regions where moisture flux convergence from orographic uplift is quasi-stationary. In particular, IVT may be a useful proxy for moisture flux convergence upwind of mountains. Hence a comparison between the changes in precipitation and IVT may provide some insights on potential changes in AR related precipitation processes in a warmer climate. Most areas in 14

15 Europe are marked by large percentage increases in total precipitation of above 100% (Fig. 10a). The changes are particularly significant in central Europe near 50 o N. Changes in IVT (Fig. 10c) generally follow a similar spatial pattern, except they are more modest especially in central Europe suggesting possible changes in precipitation processes. Compared to the changes in AR total precipitation (Fig. 10a), the increases in AR extreme precipitation (Fig. 10b) are generally smaller, and more comparable to the changes in AR extreme IVT (Fig. 10d) To delineate the contributions to changes in AR precipitation from changes in AR precipitation frequency versus precipitation intensity, Figure 11 shows the same changes corresponding to Figure 10, but with the precipitation and IVT normalized by the AR days, hence representing the changes in intensity rather than the total amount. For all the quantities shown in the Figure 11, the percentage changes in intensity are much smaller than that of the total amount, demonstrating that the significant increases shown in Figure 10 as well as the increased contributions of ARs to total and extreme precipitation shown in Figures 8 and 9 are largely associated with the increases in AR days in the future. Comparing the intensity changes in AR precipitation (Fig. 11a) and IVT (Fig. 11c), most areas especially in central Europe still exhibit amplifications of the changes from IVT to precipitation, potentially related to changes in precipitation processes in a warmer climate. However, decreases in AR precipitation intensity (Fig. 11a) are notable in Portugal, Spain, and Morocco, despite increases in IVT intensity corresponding to warming in the future. Similar results are also obtained for the fall season when ARs also occur frequently Comparing the percentage changes in AR total (Fig. 11a) and extreme precipitation (Fig. 11b) intensity, changes in the latter are generally much smaller. An even more striking difference between the two is that while AR total precipitation intensity (Fig. 11a) is mostly amplified compared to the IVT (Fig. 11c) changes, the increases in AR extreme precipitation (Fig. 11b) 15

16 intensity are overall subdued compared to the extreme IVT (Fig. 11d) changes. In Portugal, Spain, and France, AR extreme precipitation intensity (Fig. 11b) increases by less than 5% compared to increases of up to 20% in the extreme IVT (Fig. 11d). In the Scandinavian mountains and the Alps, AR extreme precipitation intensity (Fig. 11b) decreases by a few percent compared to increases in AR extreme IVT by up to 50%. This suggests potentially different processes governing changes in AR total and extreme precipitation in a warmer climate Discussions and Conclusions This study investigates the landfalling atmospheric rivers over western Europe in the present and future climate. The CMIP5 models reasonably capture the seasonal and spatial distributions of AR days. Although the multi-model mean AR days are generally comparable to those determined from four global reanalysis products, there are significant inter-model spreads, indicating large uncertainties in simulating AR days by state-of-the-art global climate models. Analysis of the atmospheric circulation demonstrates statistically significant correlations of model biases and uncertainties in simulating AR days with those of the jet position and strength simulated by the models. As most CMIP5 models show an equatorward jet bias, more AR events are simulated south of the mean jet position (45-55 N) by the CMIP5 models than the reanalyses and vice versa for the poleward side of the jet A poleward shift of the annual mean jet position in a warmer climate has been identified in several generations of coupled climate simulations [Barnes and Polvani, 2013; Kidston and Gerber, 2010; Miller et al., 2006; Swart and Fyfe, 2012; Yin, 2005]. Despite the rather robust tendency for the poleward shift, models disagree on the extent of the poleward shift and the 16

17 eastward extension of the jet that results in large uncertainty in projecting regional precipitation changes in the extratropics [Langenbrunner et al., 2015; Neelin et al., 2013]. The dependence of ARs on the jet stream has been demonstrated for ARs in aquaplanet simulations [Hagos et al., 2015] and ARs making landfall in western North America [Gao et al., 2015; Hagos et al., 2016]. Analysis in this study for ARs making landfall in western Europe provides further evidence of the relationships between AR frequency and jet position and speed. Thus, uncertainty in projecting the changes in jet stream may also project onto uncertainty in projecting changes in ARs By the end of this century, the number of AR events in western Europe was projected to increase by a few times compared to the historical level. Through a rescaling method, we found that thermodynamical changes play a dominate role in the future increase of ARs, but dynamical effect due to changes in wind also plays a significant role. The changes in AR days due to dynamical changes show a dipole feature with an overall negative (positive) effect south (north) of MME mean jet position. This dipole feature aligns well with the changes in zonal wind speed and consistent with the poleward jet shifts projected by the CMIP5 models Previous studies have identified significant correlations between the model-projected shifts in the midlatitude jet positions with the simulated climatological jet position in the southern hemisphere [Grise and Polvani, 2014; Kidston and Gerber, 2010]. In the North Atlantic, we found similar correlations for winter and spring. With the overall equatorward bias in the jet position in CMIP5, the projected future jet positions are predominantly poleward shifted. Recognizing the relationships between the jet positions and AR frequency, we tested the use of the historical jet positions as emergent constraints on the projection of AR days in the future and found statistically significant correlations between the two. Accounting for the equatorward jet 17

18 bias in CMIP5 climatology compared with the global reanalyses, the projected increase of winter AR days at the equatorward side of the mean jet by the CMIP5 MME should be adjusted upward, in addition to the larger associated with water vapor increases in the future For an emergent constraint to be effective, the empirical relationship between inter-model variations of an observable quantity and the inter-model variations in a future climate prediction must have a physical explanation [Klein and Hall, 2015]. The dynamical basis for the historical jet position as an emergent constraint for AR changes consists of two relationships linking the historical jet position with the projected jet shift, and the projected jet shift with the projected AR frequency changes. Barnes and Hartmann [2010a] found that when the Atlantic jet is more equatorward during the negative phase of the North Atlantic Oscillation (NAO), positive eddy feedback due to the enhanced baroclinicity and anomalous northward eddy propagation away from the jet helps maintain the jet in an equatorward position and a persistent negative NAO. Consistent with the above finding and more generally, Barnes and Hartmann [2010b] and Barnes and Polvani [2013] found that jet variability decreases as the mean jet is located more poleward. Hence the eddy feedback mechanism may also explain the over-persistence of the equatorward biased jet simulated by the CMIP5 models [Gerber et al., 2008], and hence their possible exaggeration of the poleward shift in response to external forcing [Barnes and Polvani, 2013; Kidston and Gerber, 2010]. As ARs are closely coupled to the extratropical cyclones, a larger poleward jet shift reduces the likelihood for extratropical storm tracks to tap tropical moisture or reduces the wind speeds for transporting significant moisture from the lower latitudes. Given the relationships between ARs and jet stream, improvements in simulating the jet position and strength may lead to improvements in simulating the number of ARs and reduce uncertainties in projecting AR changes in the future. 18

19 Due primarily to the more abundant moisture but also the poleward jet shift that increases the wind speeds in the higher latitudes, ARs contribute more importantly to both total and extreme precipitation in western Europe in the future. However, changes in AR total and extreme precipitation intensity reveal smaller increases compared to the amount, suggesting that the increased contributions from AR are mainly a result of increased AR frequency in a warmer climate. Generally the increase in AR precipitation intensity is larger compared to the increase in IVT, except in Portugal, Spain, and Morocco where AR precipitation intensity decreases despite an increase in IVT in the future AR extreme precipitation intensity increases much more modestly compared to the AR extreme IVT intensity or even decreases in some regions. Previous studies found that extreme precipitation in the extratropics scales approximately with thermodynamics following the CC relationship as the dynamical effects from changes in vertical velocity are small [Emori and Brown, 2005; O'Gorman and Schneider, 2009] except potentially for regions influenced by the poleward jet shift [Lu et al., 2014; O'Gorman and Schneider, 2009]. For non-convective events in the extratropics, O Gorman [2015] further showed weak dependence of extreme precipitation on changes in static stability. With dynamical influence potentially limited, possible reasons for reduction in AR extreme precipitation include changes in microphysical processes related to freezing level and hydrometeor fall speed [Singh and O'Gorman, 2014] and changes in orographic precipitation, with precipitation shifted downwind, reducing the total precipitation compared to the CC scaling [Siler and Roe, 2014]. The latter may be particularly relevant for the negative changes in AR extreme precipitation in the Scandinavian mountains and the Alps. However the simulated changes reported here may be hampered by the relatively coarse spatial resolution and uncertainties in cloud parameterizations in the CMIP5 models, so further 19

20 investigations are warranted to understand the dynamical, thermodynamical, and microphysical factors that modulate the AR extreme precipitation response to warming. As model resolution increases with advances in computational resources, potential improvements in simulating the jet stream [Lu et al., 2015] and orographic effects may improve understanding and lead to more reliable projections of future changes in AR days and extreme precipitation Acknowledgments This study was supported by the U.S. Department of Energy Office of Science Biological and Environmental Research (BER) as part of the Regional and Global Climate Modeling program. PNNL is operated for DOE by Battelle Memorial Institute under contract DE-AC05-76RL We acknowledge the World Climate Research Programme's Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modeling groups for producing and making available their model output References Barnes, E. A., and D. L. Hartmann (2010a), Dynamical Feedbacks and the Persistence of the NAO, Journal of the Atmospheric Sciences, 67(3), Barnes, E. A., and D. L. Hartmann (2010b), Influence of eddy-driven jet latitude on North Atlantic jet persistence and blocking frequency in CMIP3 integrations, Geophysical Research Letters, 37(23), L Barnes, E. A., and L. Polvani (2013), Response of the midlatitude jets, and of their variability, to increased greenhouse gases in the CMIP5 models, Journal of Climate, 26(18), Dee, D. P., et al. (2011), The ERA-Interim reanalysis: configuration and performance of the data assimilation system, Quarterly Journal of the Royal Meteorological Society, 137(656),

21 Emori, S., and S. J. Brown (2005), Dynamic and thermodynamic changes in mean and extreme precipitation under changed climate, Geophysical Research Letters, 32(17), L Gao, Y., J. Lu, L. R. Leung, Q. Yang, S. Hagos, and Y. Qian (2015), Dynamical and thermodynamical modulations on future changes of landfalling atmospheric rivers over western North America, Geophysical Research Letters, 42(17), 2015GL Gerber, E. P., L. M. Polvani, and D. Ancukiewicz (2008), Annular mode time scales in the Intergovernmental Panel on Climate Change Fourth Assessment Report models, Geophysical Research Letters, 35(22), L Grise, K. M., and L. M. Polvani (2014), Is climate sensitivity related to dynamical sensitivity? A Southern Hemisphere perspective, Geophysical Research Letters, 41(2), 2013GL Hagos, S., L. R. Leung, Q. Yang, C. Zhao, and J. Lu (2015), Resolution and dynamical core dependence of atmospheric river frequency in global model simulations, Journal of Climate, 28(7), Hagos, S. M., L. R. Leung, J.-H. Yoon, J. Lu, and Y. Gao (2016), A Projection of Changes in Landfalling Atmospheric River Frequency and Extreme Precipitation over Western North America from the Large Ensemble CESM Simulations, Geophysical Research Letters, 2015GL Kalnay, E., et al. (1996), The NCEP/NCAR 40-Year reanalysis project, Bulletin of the American Meteorological Society, 77(3), Kidston, J., and E. P. Gerber (2010), Intermodel variability of the poleward shift of the austral jet stream in the CMIP3 integrations linked to biases in 20th century climatology, Geophysical Research Letters, 37(9), L Klein, S. A., and A. Hall (2015), Emergent Constraints for Cloud Feedbacks, Current Climate Change Reports, 1(4), Langenbrunner, B., J. D. Neelin, B. R. Lintner, and B. T. Anderson (2015), Patterns of Precipitation Change and Climatological Uncertainty among CMIP5 Models, with a Focus on the Midlatitude Pacific Storm Track, Journal of Climate, 28(19), Lavers, D. A., and G. Villarini (2013), The nexus between atmospheric rivers and extreme precipitation across Europe, Geophysical Research Letters, 40(12), Lavers, D. A., and G. Villarini (2015), The contribution of atmospheric rivers to precipitation in Europe and the United States, Journal of Hydrology, 522(0),

22 Lavers, D. A., F. Pappenberger, and E. Zsoter (2014), Extending medium-range predictability of extreme hydrological events in Europe, Nat Commun, 5. Lavers, D. A., G. Villarini, R. P. Allan, E. F. Wood, and A. J. Wade (2012), The detection of atmospheric rivers in atmospheric reanalyses and their links to British winter floods and the large-scale climatic circulation, Journal of Geophysical Research: Atmospheres, 117(D20), D Lavers, D. A., R. P. Allan, G. Villarini, B. Lloyd-Hughes, D. J. Brayshaw, and A. J. Wade (2013), Future changes in atmospheric rivers and their implications for winter flooding in Britain, Environmental Research Letters, 8(3), Leung, L. R., and Y. Qian (2009), Atmospheric rivers induced heavy precipitation and flooding in the western U.S. simulated by the WRF regional climate model, Geophysical Research Letters, 36(3), L Lu, J., G. Chen, L. R. Leung, D. A. Burrows, Q. Yang, K. Sakaguchi, and S. Hagos (2015), Toward the Dynamical Convergence on the Jet Stream in Aquaplanet AGCMs, Journal of Climate, 28(17), Lu, J., L. Ruby Leung, Q. Yang, G. Chen, W. D. Collins, F. Li, Z. Jason Hou, and X. Feng (2014), The robust dynamical contribution to precipitation extremes in idealized warming simulations across model resolutions, Geophysical Research Letters, 41(8), 2014GL Miller, R. L., G. A. Schmidt, and D. T. Shindell (2006), Forced annular variations in the 20th century Intergovernmental Panel on Climate Change Fourth Assessment Report models, Journal of Geophysical Research: Atmospheres, 111(D18), D Moss, R. H., et al. (2010), The next generation of scenarios for climate change research and assessment, Nature, 463, Neelin, J. D., B. Langenbrunner, J. E. Meyerson, A. Hall, and N. Berg (2013), California Winter Precipitation Change under Global Warming in the Coupled Model Intercomparison Project Phase 5 Ensemble, Journal of Climate, 26(17), Neiman, P. J., F. M. Ralph, G. A. Wick, J. D. Lundquist, and M. D. Dettinger (2008), Meteorological characteristics and overland precipitation impacts of atmospheric rivers affecting the west coast of North America based on eight years of SSM/I satellite observations, Journal of Hydrometeorology, 9(1),

23 Neiman, P. J., L. J. Schick, F. M. Ralph, M. Hughes, and G. A. Wick (2011), Flooding in western washington: the connection to atmospheric rivers, Journal of Hydrometeorology, 12(6), O'Gorman, P. A., and T. Schneider (2009), The physical basis for increases in precipitation extremes in simulations of 21st-century climate change, Proceedings of the National Academy of Sciences, 106(35), O Gorman, P. A. (2015), Precipitation Extremes Under Climate Change, Current Climate Change Reports, 1(2), Payne, A. E., and G. Magnusdottir (2014), Dynamics of landfalling atmospheric rivers over the north Pacific in 30 years of MERRA reanalysis, Journal of Climate, 27(18), Ralph, F. M., and M. D. Dettinger (2012), Historical and national perspectives on extreme west coast precipitation associated with atmospheric rivers during December 2010, Bulletin of the American Meteorological Society, 93(6), Ralph, F. M., P. J. Neiman, and G. A. Wick (2004), Satellite and CALJET aircraft observations of atmospheric rivers over the eastern north Pacific Ocean during the winter of 1997/98, Monthly Weather Review, 132(7), Ralph, F. M., P. J. Neiman, G. A. Wick, S. I. Gutman, M. D. Dettinger, D. R. Cayan, and A. B. White (2006), Flooding on California's Russian River: role of atmospheric rivers, Geophysical Research Letters, 33(13), L Rienecker, M. M., et al. (2011), MERRA: NASA s Modern-Era Retrospective Analysis for research and applications, Journal of Climate, 24(14), Rutz, J. J., W. J. Steenburgh, and F. M. Ralph (2014), Climatological Characteristics of Atmospheric Rivers and Their Inland Penetration over the Western United States, Monthly Weather Review, 142(2), Saha, S., et al. (2010), The NCEP climate forecast system reanalysis, Bulletin of the American Meteorological Society, 91(8), Siler, N., and G. Roe (2014), How will orographic precipitation respond to surface warming? An idealized thermodynamic perspective, Geophysical Research Letters, 41(7), Singh, M. S., and P. A. O'Gorman (2014), Influence of microphysics on the scaling of precipitation extremes with temperature, Geophysical Research Letters, 41(16),

24 Swart, N. C., and J. C. Fyfe (2012), Observed and simulated changes in the Southern Hemisphere surface westerly wind-stress, Geophysical Research Letters, 39(16), L Vuuren, D., et al. (2011), The representative concentration pathways: an overview, Climatic Change, 109(1-2), Yin, J. H. (2005), A consistent poleward shift of the storm tracks in simulations of 21st century climate, Geophysical Research Letters, 32(18), L Zhu, Y., and R. E. Newell (1998), A proposed algorithm for moisture fluxes from atmospheric rivers, Monthly Weather Review, 126(3), List of Figure captions Figure 1. The grids used to detect atmospheric rivers. The color-coded squares are the grid cells used to detect ARs, with different colors indicating different latitudinal bins, which are also separated by the dashed gray lines Figure 2. Box and whisker plots of the number of atmospheric river days in each season from 30 to 70 N for each of the eight bins from CMIP5 models from The horizontal line within the box indicates the CMIP5 median, the boundaries of the box indicate the 25th and 75th percentile, and the whiskers indicate the highest and lowest values of the results from CMIP5. The marked inside the box indicates the CMIP5 MME mean. The number of atmospheric rivers from the four reanalysis data sets is marked with triangles, i.e., CFSR (blue), ERA- INTERIM (red), MERRA (green), and NCEP1 (gold) during Figure 3. Historical number of AR days versus jet position (top row) and speed (bottom row) in winter for three latitudinal bins from 40 to 55 degrees north. Each black dot corresponds to one CMIP5 model whereas the triangles of different colors denote the reanalysis datasets. An asterisk indicates statistically significant correlations shown in red. 24

25 Figure 4 The CMIP5 MME seasonal total numbers of AR days over 8 latitudinal bins (30-70N) for present ( , black) and future climate conditions in RCP8.5 ( , red). Also shown are the numbers of AR days from rescaling of the future IVT by the present IWV (V! Q!, orange) and present IVT by future IWV (V! Q!, blue) and using the C-C scaling (dashed black). The shaded areas represent one standard deviation of the CMIP5 inter-model spread. The numbers on the top row in each panel indicate the percentage changes of AR, calculated based on (V! Q! V! Q! )/V! Q! 100% whereas the bottom row shows the thermodynamical effect calculated through ( V! Q! V! Q! )/V! Q! 100%, with the red numbers indicating statistical significance at 95% level Figure 5 Changes in the number of AR days due to dynamics (black curve), calculated using ARs detected with V! Q! (orange line in Figure 4) minus that detected with V! Q! (the black line in Figure 4) as a function of latitude. Also shown are the changes of zonal wind speed (red curves), calculated by averaging the zonal wind speed for each 10-degree latitudinal bin extending from the coast to 20 degrees west. The black and red dots indicate statistically significant changes for AR days and zonal winds, respectively, at 95% level. Shading corresponds to one standard deviation above and below the multi-model mean Figure 6. Correlation between the CMIP5 simulated historical jet position and projected changes of jet position in the future under RCP 8.5 for winter and spring. The four colored circles indicate the jet position from the four reanalysis datasets. The multi-model mean historical jet positions and changes of jet positions are indicated by the grey and red dashed lines, respectively Figure 7. Similar to Figure 6 but for correlation between the historical jet position and the changes of AR days in winter at two latitudinal bins due to dynamical effects. 25

26 Figure 8. The fractional contribution of AR induced precipitation to the total precipitation in each season from ERA-Interim ( ; top row), CMIP5 MME at present ( ; middle row), and the difference between CMIP5 MME in RCP 8.5 and present ( minus ; bottom row). 557 Figure 9. The same as Figure 8 but for extreme precipitation Figure 10. Percentage change in winter AR total precipitation (a), extreme precipitation (b), total IVT (c), and extreme IVT (d) from the CMIP5 MME comparing the present ( ) with the future ( ) Figure 11. Same as Figure 10, but for intensity of AR total precipitation (a), extreme precipitation (b), IVT (c), and extreme IVT (d)

27 Figure 1. The grids used to detect atmospheric rivers. The color-coded squares are the grid cells used to detect ARs, with different colors indicating different latitudinal bins, which are also separated by the dashed gray lines

28 Figure 2. Box and whisker plots of the number of atmospheric river days in each season from 30 to 70 N for each of the eight bins from CMIP5 models from The horizontal line within the box indicates the CMIP5 median, the boundaries of the box indicate the 25th and 75th percentile, and the whiskers indicate the highest and lowest values of the results from CMIP5. The marked inside the box indicates the CMIP5 MME mean. The number of atmospheric rivers from the four reanalysis data sets is marked with triangles, i.e., CFSR (blue), ERA- INTERIM (red), MERRA (green), and NCEP1 (gold) during

29 Figure 3. Historical number of AR days versus jet position (top row) and speed (bottom row) in winter for three latitudinal bins from 40 to 55 degrees north. Each black dot corresponds to one CMIP5 model whereas the triangles of different colors denote the reanalysis datasets. An asterisk indicates statistically significant correlations shown in red

30 Figure 4 The CMIP5 MME seasonal total numbers of AR days over 8 latitudinal bins (30-70N) for present ( , black) and future climate conditions in RCP8.5 ( , red). Also shown are the numbers of AR days from rescaling of the future IVT by the present IWV (V! Q!, orange) and present IVT by future IWV (V! Q!, blue) and using the C-C scaling (dashed black). The shaded areas represent one standard deviation of the CMIP5 inter-model spread. The 30

31 numbers on the top row in each panel indicate the percentage changes of AR, calculated based on (V! Q! V! Q! )/V! Q! 100% whereas the bottom row shows the thermodynamical effect calculated through ( V! Q! V! Q! )/V! Q! 100%, with the red numbers indicating statistical significance at 95% level

32 Figure 5 Changes in the number of AR days due to dynamics (black curve), calculated using ARs detected with V! Q! (orange line in Figure 4) minus that detected with V! Q! (the black line in Figure 4) as a function of latitude. Also shown are the changes of zonal wind speed (red curves), calculated by averaging the zonal wind speed for each 10-degree latitudinal bin extending from the coast to 20 degrees west. The black and red dots indicate statistically significant changes for AR days and zonal winds, respectively, at 95% level. Shading corresponds to one standard deviation above and below the multi-model mean