Abstract

This study examines the causes and predictability of the spring 2011 U.S. extreme weather using the Modern-Era Retrospective Analysis for Research and Applications (MERRA) and Goddard Earth Observing System Model, version 5, (GEOS-5) atmospheric general circulation model simulations. The focus is on assessing the impact on precipitation of sea surface temperature (SST) anomalies, land conditions, and large-scale atmospheric modes of variability. A key result is that the April record-breaking precipitation in the Ohio River valley was primarily the result of the unforced development of a positive North Atlantic Oscillation (NAO)-like mode of variability with unusually large amplitude, limiting the predictability of the precipitation in that region at 1-month leads. SST forcing (La Niña conditions) contributed to the broader continental-scale pattern of precipitation anomalies, producing drying in the southern plains and weak wet anomalies in the northeast, while the impact of realistic initial North American land conditions was to enhance precipitation in the upper Midwest and produce deficits in the Southeast. It was further found that 1) the 1 March atmospheric initial condition was the primary source of the ensemble mean precipitation response over the eastern United States in April (well beyond the limit of weather predictability), suggesting an influence on the initial state of the previous SST forcing and/or tropospheric–stratospheric coupling linked to an unusually persistent and cold polar vortex; and 2) stationary wave model experiments suggest that the SST-forced base state for April enhanced the amplitude of the NAO response compared to that of the climatological state, though the impact is modest and can be of either sign.

1. Introduction

According to Karl et al. (2011, p. 1), “the spring (March–May) of 2011, particularly April, brought extreme weather and climate events to many parts of the United States. Tornadoes, flooding, drought, and wildfires ravaged many parts of the country during the period, and each of these extremes broke long-standing records and have been compared to the ‘worst such cases’ in history.” Much of the Ohio River Valley in particular experienced historic rainfall amounts during April, with 2011 ranked as the wettest on record (NCDC 2011; http://www.ncdc.noaa.gov/monitoring-content/sotc/national/statewidepcpnrank/statewidepcpnrank-201104.gif).

As a first step in assessing the causes of such extremes, a preliminary NOAA assessment of the climate factors contributing to the 2011 tornado outbreak (ESRL 2011; http://www.esrl.noaa.gov/psd/csi/events/2011/tornadoes/) notes that the April 2011 period of very active severe weather occurred during a La Niña event and that, despite the fact that the La Niña event was then in its waning phase, it was still able to affect U.S. weather patterns.

Hoerling et al. (2013a) found that the record 2011 flooding of the Missouri River basin resulted from a combination of factors: antecedent wet conditions that preconditioned the upper basin toward greater runoff efficiency, a cold and wet 2010/11 winter with unusually high snowpack, and heavy rains in late spring (likely the most critical factor). They found a modest contribution of SST forcing to the heavy rains, concluding that they were a highly unlikely occurrence driven by random atmospheric variation.

The above-normal April precipitation across the northern half of the country was juxtaposed with near- to below-average precipitation in the southern plains and Southeast. In fact, by the end of April 2011, much of Texas was experiencing extreme-to-exceptional drought conditions (U.S. Drought Monitor 2011; http://droughtmonitor.unl.edu/MapsAndData/MapArchive.aspx, select 26 April 2011). While the drought is not the focus of this study, we note that several studies have shown an important link between this drought and the evolving La Niña conditions (e.g., Hoerling et al. 2013b; Seager et al. 2014; Wang et al. 2014). The Wang et al. (2014) study in particular highlighted the importance of the tropical Pacific SST during spring 2011 in producing a warming and drying tendency in the southern United States, though the amplitude of the response was weaker than observed.

In the present study we examine the causes and predictability of the springtime 2011 extreme weather activity with a focus on the record-breaking precipitation that occurred in the central and eastern United States during April using high-resolution simulations with the NASA Goddard Earth Observing System Model, version 5, (GEOS-5) atmospheric general circulation model (AGCM). We complement the modeling results with diagnostics derived from NASA’s Modern-Era Retrospective Analysis for Research and Applications (MERRA) data and a stationary wave model (SWM). Section 2 describes the AGCM, the simulations, the MERRA data, and the SWM. Results are presented in section 3, and section 4 provides a summary and conclusions.

2. The experiments and data

We base part of our analysis on MERRA (Rienecker et al. 2011). MERRA was produced with the NASA GMAO atmospheric data assimilation system (ADAS), which consists of the GEOS-5 AGCM (Rienecker et al. 2008; Molod et al. 2015) along with the Gridpoint Statistical Interpolation analysis system (GSI) (Wu et al. 2002; Kleist et al. 2009), which is jointly developed by the GMAO and the National Centers for Environmental Prediction (NCEP). MERRA was run at a resolution of ½° latitude × ⅔° longitude, with 72 vertical levels extending to 0.01 hPa. More information about MERRA can be found on the GMAO’s website (http://gmao.gsfc.nasa.gov/research/merra/). We also make use of the Climate Prediction Center (CPC) unified gauge-based 0.25° analysis of daily U.S. precipitation (Chen et al. 2008).

Our main results, however, are based on high-resolution runs with the GEOS-5 AGCM (Rienecker et al. 2008; Molod et al. 2015). The model employs the finite-volume dynamics of Lin (2004) and various physics packages (Bacmeister et al. 2006) integrated together using the Earth System Modeling Framework (Collins et al. 2005)—packages such as the Catchment Land Surface Model (Koster et al. 2000) and a modified form of the relaxed Arakawa–Schubert convection scheme (Moorthi and Suarez 1992). For our experiments, the model (with internal designation Fortuna 2.4) was run with 72 hybrid-sigma vertical levels extending to 0.01 hPa and with a ¼° horizontal resolution on a cubed sphere.

Table 1 lists the complete set of simulations. All simulations were initialized on 1 March and integrated through the end of May. The primary set of runs (A) consists of 44 hindcasts for spring 2011 forced with observed SST. All atmospheric and land initial conditions were taken from MERRA and thus are deemed reasonably realistic. To compute anomalies for 2011, we establish a climatology consisting of six ensemble members per year covering the period 2000–10, These runs are treated identically to those for 2011 in that they are forced by observed SST and initialized from MERRA (differing only in the number of ensemble members per year). For convenience we shall refer to that set of runs as Ac. The anomalies for 2011 were accordingly computed by subtracting out the mean of the Ac runs.

Table 1.

Here we list the complete suite of GEOS-5 AGCM hindcast experiments. Each set of runs consists of 22 ensemble members, except A, which has 44 members, and Ac, which has 66 members (6 for each year times 11 years). All model simulations were done at a horizontal resolution of ¼° latitude/longitude. Perturbations (only to the atmosphere) for the runs were produced as: ±(xy)/8, where x and y are two atmospheric states separated by 1 day, and where the states were obtained from a 1-month simulation starting 1 Nov. The impacts of the SST and initial conditions (ICs) are determined from the runs as follows: both SST and ICs = A − Ac; global SST = A − B; tropical Pacific (TPac) SST = C − B; atmosphere and land ICs = B − Bc; and land ICs over North America = A − D.

Here we list the complete suite of GEOS-5 AGCM hindcast experiments. Each set of runs consists of 22 ensemble members, except A, which has 44 members, and Ac, which has 66 members (6 for each year times 11 years). All model simulations were done at a horizontal resolution of ¼° latitude/longitude. Perturbations (only to the atmosphere) for the runs were produced as: ±(x − y)/8, where x and y are two atmospheric states separated by 1 day, and where the states were obtained from a 1-month simulation starting 1 Nov. The impacts of the SST and initial conditions (ICs) are determined from the runs as follows: both SST and ICs = A − Ac; global SST = A − B; tropical Pacific (TPac) SST = C − B; atmosphere and land ICs = B − Bc; and land ICs over North America = A − D.
Here we list the complete suite of GEOS-5 AGCM hindcast experiments. Each set of runs consists of 22 ensemble members, except A, which has 44 members, and Ac, which has 66 members (6 for each year times 11 years). All model simulations were done at a horizontal resolution of ¼° latitude/longitude. Perturbations (only to the atmosphere) for the runs were produced as: ±(x − y)/8, where x and y are two atmospheric states separated by 1 day, and where the states were obtained from a 1-month simulation starting 1 Nov. The impacts of the SST and initial conditions (ICs) are determined from the runs as follows: both SST and ICs = A − Ac; global SST = A − B; tropical Pacific (TPac) SST = C − B; atmosphere and land ICs = B − Bc; and land ICs over North America = A − D.

Supplemental series of simulations utilized, for practical reasons, a reduced set of 22 ensemble members. The first supplemental set used climatological SSTs (B and Bc) but the same atmospheric and land initializations as in series A; thus, by comparing series A and B, we can assess the impact of the global SST field on the simulated weather. The second supplement set (C) utilized observed SSTs in the tropical Pacific (20°S–30°N, east of 180°) and climatological SSTs elsewhere, allowing us, by comparing series B and C, to isolate the impact of SST in that region. A final supplemental set (D) utilized observed 2011 SST as in A but used, over much of North America (35°–70°N, 130°–60°W), a set of initial conditions for land states (including soil moisture, soil temperature, and snow) that contain no useful information for the forecasts. In essence, the initial land states for D in this region were obtained from states occurring during the previous 11 years: of the 22 ensemble members, two used land states from 1 March 2000, two used land states from 1 March 2001, and so on. Comparing series A and D thus allows us to assess the impact of the 2011 spring land conditions on weather.

The various ensemble members were generated by perturbing the atmospheric state. The perturbations were computed as scaled differences [± (xy)/8] between two arbitrary November atmospheric states (x, y) taken 1 day apart. The somewhat arbitrary scaling factor of ⅛ was chosen simply in an attempt to produce perturbations that are of a size not inconsistent with initial errors typically found in numerical weather prediction. The November-based perturbations were imposed on 1 March atmospheric states; these perturbations were small enough to ensure that the 1 March atmospheric initialization remained climatologically sensible.1

We also make use of a stationary wave model (Ting and Yu 1998) to help diagnose the forcing of the NAO. (The specific SWM experiments we run are described in section 3d). The SWM is based on the three-dimensional primitive equations in σ coordinates and is nonlinear and time dependent. Model-generated transient disturbances are suppressed by strong damping. The model has rhomboidal wavenumber-30 truncation in the horizontal and 14 unevenly spaced σ levels in the vertical (R30L14). The SWM has been shown to be a valuable tool for diagnosing the maintenance of both climatological and anomalous atmospheric circulation patterns by evaluating the relative roles of stationary wave forcings over specific regions (e.g., Held et al. 2002; Schubert et al. 2011). More details regarding the SWM can be found in Ting and Yu (1998).

3. Results

We begin in section 3a with a look at the observations as represented by MERRA data. We focus here on April, the spring month most dominated by extremes. We follow this in sections 3b and 3c with an analysis of our AGCM experiments, focusing respectively on simulated ensemble means and intraensemble variances. Section 3d examines in more detail the role of the NAO.

a. Observations (MERRA)

Figure 1 (top-left panel) shows that the Northern Hemisphere upper-level (250 mb) height field during April 2011 was dominated by a positive NAO-like anomaly,2 with large negative anomalies spanning much of northern North America and the North Atlantic. Positive anomalies extend across the southern tier of the United States and across parts of Europe and Asia. This pattern is associated with positive zonal wind anomalies (Fig. 1, bottom-left panel) that extend from the North Pacific across the United States. The bottom-right panel of Fig. 1 shows the daily meridional wind variability at upper levels (a measure of storm intensity; e.g., Chang and Fu 2002). The variability is enhanced over much of the United States between 30° and 50°N, as well as upstream over the North Pacific (centered on 30°N), with reduced variability to the north and south of these regions. The precipitation anomalies (Fig. 1, top-right panel) highlight the aforementioned record values over the Ohio River basin, with dry conditions over the southern plains and parts of the Southeast. The juxtaposition of the positive precipitation anomalies over much of the Ohio River basin and the enhanced storminess is not surprising in view of the strong link between midlatitude cyclones and extreme precipitation events found in this region for the cold season (e.g., Pfahl and Wernli 2012).

Fig. 1.

April 2011 anomalies with respect to the climatology spanning the years 2000–10 of (top left) z250mb (m), (top right) precipitation (mm day−1; shaded) and 850-mb moisture flux (g kg−1 m s−1; vectors) anomalies, (bottom left) u250mb climatology (5 m s−1; contours) and anomalies (m s−1; shaded), and (bottom right) υ2 250-mb anomalies (m s−1)2. All fields are from MERRA, except the precipitation, which is from the CPC unified gauge-based analysis of daily precipitation over the continental United States (CONUS) at 0.25° latitude/longitude resolution.

Fig. 1.

April 2011 anomalies with respect to the climatology spanning the years 2000–10 of (top left) z250mb (m), (top right) precipitation (mm day−1; shaded) and 850-mb moisture flux (g kg−1 m s−1; vectors) anomalies, (bottom left) u250mb climatology (5 m s−1; contours) and anomalies (m s−1; shaded), and (bottom right) υ2 250-mb anomalies (m s−1)2. All fields are from MERRA, except the precipitation, which is from the CPC unified gauge-based analysis of daily precipitation over the continental United States (CONUS) at 0.25° latitude/longitude resolution.

To put these results in a historical perspective, we examine the leading modes of variability of the height field at 250 mb during boreal spring (March–May) over the last three decades (1979–2012) using a rotated empirical orthogonal function (REOF; Richman 1986) decomposition. Figure 2 (top-left panel) shows that the third leading REOF resembles the NAO pattern and has much in common with the April height anomalies shown in Fig. 1 (top-left panel). The first and second REOFs (not shown) consist of patterns resembling the northern annular mode (NAM) and the Pacific–North American teleconnection pattern (PNA), respectively. We will come back to the role of the NAM in the next section. The monthly time series [rotated principal component (RPC)] of REOF 3 further emphasizes the dominance of this mode during April of 2011, showing that during this month it attained its largest positive value over the 34-yr time span, consistent with the NOAA/CPC results mentioned above. Composites of the precipitation (Fig. 2, top-right panel) and daily 250-mb meridional wind variance (Fig. 2, bottom-right panel) based on the extremes of that RPC further reinforce the importance of this mode for producing anomalous precipitation and weather variability over the eastern United States (cf. Fig. 1).

Fig. 2.

(top left) REOF 3 spatial pattern. (bottom left) The monthly time series of RPC 3 for MAM of each year. (right) Composites based on the extremes (1.5 standard deviations) of RPC 3 (positive − negative composites): (top right) precipitation (mm day−1) and 850-mb q fluxes (g kg−1 m s−1); (bottom right) υ2 (m s−1)2 at 200 mb. The results are for MAM for the period 1979–2012. All fields are from MERRA, except the precipitation, which is from the CPC unified gauge-based analysis of daily precipitation over CONUS at 0.25° latitude/longitude resolution.

Fig. 2.

(top left) REOF 3 spatial pattern. (bottom left) The monthly time series of RPC 3 for MAM of each year. (right) Composites based on the extremes (1.5 standard deviations) of RPC 3 (positive − negative composites): (top right) precipitation (mm day−1) and 850-mb q fluxes (g kg−1 m s−1); (bottom right) υ2 (m s−1)2 at 200 mb. The results are for MAM for the period 1979–2012. All fields are from MERRA, except the precipitation, which is from the CPC unified gauge-based analysis of daily precipitation over CONUS at 0.25° latitude/longitude resolution.

b. The ensemble mean (forced) response

The spring 2011 SST anomalies computed with respect to a climatology spanning the years 2000–103 (Fig. 3) reflect decaying La Niña conditions with the cold eastern tropical Pacific anomalies weakening and the pattern resembling a negative Pacific decadal oscillation (PDO). During March and April, much of the eastern tropical Pacific is characterized by negative anomalies less than −½°C, with the largest negative values (south of the equator) exceeding 1.5°C in magnitude. By May, the eastern equatorial Pacific SST anomalies are nearly zero, while weak cold anomalies remain to the south and over the far eastern North Pacific, with positive anomalies occurring just to the west in the central North Pacific. Also seen are cold conditions over the North Atlantic, some positive anomalies over the tropical Atlantic, and negative anomalies over the Indian Ocean just south of the equator.

Fig. 3.

March–May 2011 monthly mean SST anomalies (°C) computed with respect to a climatology spanning the years 2000–10.

Fig. 3.

March–May 2011 monthly mean SST anomalies (°C) computed with respect to a climatology spanning the years 2000–10.

Figure 4 shows the March through May 2011 evolution of precipitation and 2-m air temperature (T2m) anomalies over North America from MERRA and from the ensemble mean of the model simulations (A − Ac, labeled “hindcasts”; see Table 1). We note that in these and the following comparisons between model results and observations we always use the same climatology (the period 2000–10) to compute the anomalies. The simulations successfully capture some of the main features of the observed (CPC) precipitation anomalies, including the positive anomalies over the eastern United States during March and April, though with reduced amplitude, especially during April—the month when the largest anomalies occurred. They also capture the dry conditions over the southern plains and the Southeast. The ensemble mean May precipitation anomalies are very diffuse but do show some resemblance to the observations. The spatial correlations over the United States (between 25° and 55°N) between the ensemble mean and observed precipitation anomalies are 0.72, 0.51, and 0.28 for March, April, and May, respectively. The simulated T2m anomalies also show a considerable resemblance to the observations, with cold anomalies over much of Canada and the northern tier of states and with warm anomalies to the south, especially over the southern plains. The successful simulation of the north/south temperature gradient over the United States during April is particularly noteworthy for our study, though the simulated (ensemble mean) gradient is weaker than observed. The month of May shows a transition in the observed (MERRA) T2m pattern, with the cold anomalies more confined to the west—a feature not well captured by the simulations. Nevertheless, the model does an overall better job of reproducing the observed T2m anomaly patterns (compared with those for precipitation). This is reflected in the spatial correlations over the United States (between 25° and 55°N) between the ensemble mean and observed T2m anomalies, which are 0.83, 0.78, and 0.53 for March, April, and May, respectively.

Fig. 4.

(left two columns) Precipitation anomalies (mm day−1) with respect to a climatology spanning the years 2000–2010 from the 0.25° CPC unified gauge-based analysis of daily precipitation, and the ensemble mean of the model simulations (A − Ac; see Table 1) for March–May 2011. (right two columns) As in (left two columns), but for T2m (°C), and the observed values are from MERRA.

Fig. 4.

(left two columns) Precipitation anomalies (mm day−1) with respect to a climatology spanning the years 2000–2010 from the 0.25° CPC unified gauge-based analysis of daily precipitation, and the ensemble mean of the model simulations (A − Ac; see Table 1) for March–May 2011. (right two columns) As in (left two columns), but for T2m (°C), and the observed values are from MERRA.

Figure 5 shows the ensemble mean 250-mb height anomalies (top panel) and the 250-mb daily meridional wind variance (υ2; bottom panel) for April. Clearly, some of the key features of the observations are reproduced in the simulations, including the negative height anomalies over northern North America and the positive anomalies across the southern tier of states, though again with reduced amplitude compared to the observations. These anomalies over North America should, however, not be interpreted as being part of the NAO pattern, which has the largest negative anomalies centered farther to the east (cf. Fig. 2). The pattern of anomalous storminess (as measured by υ2) across the North Pacific and North America is well reproduced, though also with weaker amplitude. The wave pattern in the height field across Eurasia (Fig. 1) is not reproduced in the simulations, suggesting that this feature is not forced by SST.

Fig. 5.

Ensemble mean results for April 2011 (A − Ac; see Table 1): (top) 250-mb height anomalies (m); (bottom) 250-mb υ2 anomalies (m s−1)2.

Fig. 5.

Ensemble mean results for April 2011 (A − Ac; see Table 1): (top) 250-mb height anomalies (m); (bottom) 250-mb υ2 anomalies (m s−1)2.

The ability of the model to reproduce some of the key features of the precipitation, T2m, 250-mb height, and υ2 fields associated with the unusual weather activity over the United States during spring 2011 gives us confidence that we can use the model to isolate the sources of the observed changes. Of course, as we have seen, the ensemble mean model response does not reproduce the observed anomalies exactly—something that we should not expect even for a perfect model. In particular, the model does not reproduce the NAO character of the observed anomalies (it places the negative height anomalies farther west, and they are overall weaker). This suggests that there is a significant unforced component to the observed anomalies. We will return to that issue in the next section.

We examine now, using the supplemental simulations listed in Table 1 (see section 2), the separate impacts of SST and land conditions on precipitation (Fig. 6). To help in the interpretation of the results, we present in Fig. 7 the corresponding changes in the 250-mb height field. The first column of Fig. 6 shows that the impact of the global SST forcing (A − B) is primarily to force dry conditions over the southern plains and parts of Mexico, consistent with previous studies (e.g., Seager et al. 2014; Wang et al. 2014). The corresponding height anomalies (first column in Fig. 7) show a wave response during March and April that is typical for La Niña conditions. The key feature during these months is the ridge over the southwest and south-central part of the United States that tends to block storms approaching from the west leading to the dry conditions in the southern plains, as seen in Fig. 6. The ridge is presumably also associated with subsidence, further contributing to the drying in that region. In May, the impact of the SST is different, producing a tendency for wet conditions over the northern plains and parts of southern Canada, consistent with the development of an anomalous trough covering much of the continental United States and southern Canada.

Fig. 6.

Precipitation anomalies (mm day−1). (left) Impact of global SST (A − B; see Table 1) for March–May 2011. (center left) Impact of tropical Pacific SST (C − B). (center right) Impact of the initial North American (35°–70°N, 130°–60°W) land conditions (A − D). (right) Impact of global land and atmospheric initial conditions (B − Bc). Contouring indicates values are significant at 90% based on a t test.

Fig. 6.

Precipitation anomalies (mm day−1). (left) Impact of global SST (A − B; see Table 1) for March–May 2011. (center left) Impact of tropical Pacific SST (C − B). (center right) Impact of the initial North American (35°–70°N, 130°–60°W) land conditions (A − D). (right) Impact of global land and atmospheric initial conditions (B − Bc). Contouring indicates values are significant at 90% based on a t test.

Fig. 7.

As in Fig. 6, but for the anomalies in the 250-mb height field (Northern Hemisphere) (m).

Fig. 7.

As in Fig. 6, but for the anomalies in the 250-mb height field (Northern Hemisphere) (m).

The second column of Fig. 6 shows the precipitation anomalies obtained when the observed SSTs are confined to the tropical Pacific (C − B). The anomalies are quite similar to those obtained with the global SST forcing. This is reflected in the spatial correlations over the United States (25°–55°N) between the precipitation responses of the global and tropical Pacific experiments, which are 0.70, 0.77, and 0.72 for March, April, and May, respectively. There is also considerable similarity between the height anomalies (second column of Fig. 7). This indicates that the tropical Pacific SST anomalies dominate the SST-related precipitation and height responses over North America.

The role of initial land conditions over North America (A − D) is shown in the third columns of Fig. 6 and Fig. 7 for precipitation and 250-mb height, respectively. The main impact of the realistic land initialization occurs during April, with a tendency for wet conditions to occur in a line extending from the central plains to the Northeast, and dry conditions in the Southeast. Given field significance considerations, we cannot rule out the possibility that the indicated precipitation anomalies are themselves insignificant. Nevertheless, the anomalies are consistent with an apparently significant ridge pattern that develops over the eastern United States (Fig. 7) during April. Using an extensive ensemble of AGCM simulations, Koster et al. (2014) show that dry (wet) conditions in the south-central United States tend to produce a large-scale planetary wave pattern that includes a southward (northward) anomaly in the 250-mb meridional wind field in the eastern half of the United States and that would be associated with an anomalous low (high) in the 250-mb height field along the East Coast. In our experiments here, soil moisture conditions on 1 March 2011 are characterized by a wet anomaly in the north-central United States that persists into April; extrapolating somewhat from the Koster et al. (2014) study (and as more directly supported by additional land-focused AGCM experiments, not shown), this could indeed induce a high along the East Coast as part of a broader planetary wave pattern. The pattern would not similarly appear in March given the presence of snow cover as well as a reduction in evaporation anomalies associated with reduced surface radiative forcing. The absence of the pattern in May may reflect the fact that, by this time, the initial soil moisture anomaly has weakened (not shown).

The last column in Fig. 6 shows the joint impact of realistic global atmospheric and land initialization (B − Bc). This column highlights the important contribution of the initial conditions (especially for the atmosphere) to the full precipitation response shown in Fig. 4. Indeed, some of the main features (especially the negative anomalies in the eastern United States) of the full precipitation response (second column of Fig. 4) can be attributed to the initial conditions. Figure 7 (last column) shows that the impact of the initial conditions is associated with large-scale height anomalies, with March showing the signature of a positive NAM and wavelike features extending across North America (with negative anomalies extending southwestward across the continent), the North Atlantic, and Europe. The NAM appears to persist into April along with the important negative height anomalies over eastern Canada and the northeastern United States and positive anomalies to the south. In May the anomalies are quite weak, indicating there is little impact of the initial conditions three months into the predictions.

The relatively large impact of the initial conditions extending well into April is rather surprising in light of the relatively weak impact of the North American land initial conditions (discussed above) and generally accepted estimates of weather predictability that suggest one should expect little if any skill from atmospheric initial conditions beyond a few weeks (e.g., Lorenz 1969; Chang et al. 2000). In particular, the NAO has been estimated to have an e-folding time of 9.5 days (Feldstein 2000). It is possible that the initial atmospheric conditions contain information from prior SST forcing, and this can impact the forecasts several weeks into the future (e.g., Barsugli et al. 1999; Branstator 2014), though the main height anomalies linked to the initial conditions also suggest a high-latitude connection (with a clear NAM component).

An alternative mechanism for extending predictability into the second month involves the dynamical coupling between the stratosphere and troposphere through downward propagation of zonal wind anomalies, through which NAM variability in the lower stratosphere can provide enhanced predictability of tropospheric NAM variability (e.g., Baldwin and Dunkerton 1999). The stratospheric Arctic vortex was unusually cold and persistent during the winter and spring of 2011. Hu and Xia (2013) found that beginning the middle of February the stratospheric NAM showed downward propagation that led to enhancement of the tropospheric NAM, reaching maximum values in late March and April. Karpechko et al. (2014) found in a modeling study of the impact of the 2011 spring Arctic ozone depletion that, while both SST and ozone depletion played a role in producing the downward propagation of the zonal wind anomalies to the troposphere following the strong polar vortex event in late winter and early spring, the magnitude of the event is only captured if preconditioning (likely associated with internal atmospheric dynamics) is reproduced both in the troposphere and lower stratosphere. To the extent our initialized runs capture the relevant tropospheric–stratospheric coupling, we can expect to have enhanced tropospheric predictability (extending into April), though the exact mechanisms by which that can produce skillful precipitation predictions over North America is unclear.

Finally, we summarize in Table 2 the similarity between the total precipitation response (second column in Fig. 4) and the separate responses shown in Fig. 6 in terms of the anomaly pattern correlation. The results clearly show the strong impact of the atmospheric conditions during March and the importance of the global SST during April and May. It is interesting that the tropical Pacific SST shows rather weak correlations despite the greater than 0.7 correlations between the global and Pacific SST responses (mentioned previously). This suggests that subtle phase shifts can have rather large impacts on such a metric and highlights the limitations of using pattern correlations to define similarity. The correlations associated with the North American land response are generally small (even negative during the first month) but increase with lead time, though the amplitude of the response is quite weak by May (see our previous discussion of the land impacts for why that might be the case).

Table 2.

Anomaly pattern correlations over the continental United States (25°–55°N) between the total ensemble mean precipitation response (A − Ac; second column in Fig. 4) and the responses to the individual contributions shown in Fig. 6: global SST = A − B, tropical Pacific SST = C − B, land ICs over North America = A − D, and atmosphere and land ICs = B − Bc.

Anomaly pattern correlations over the continental United States (25°–55°N) between the total ensemble mean precipitation response (A − Ac; second column in Fig. 4) and the responses to the individual contributions shown in Fig. 6: global SST = A − B, tropical Pacific SST = C − B, land ICs over North America = A − D, and atmosphere and land ICs = B − Bc.
Anomaly pattern correlations over the continental United States (25°–55°N) between the total ensemble mean precipitation response (A − Ac; second column in Fig. 4) and the responses to the individual contributions shown in Fig. 6: global SST = A − B, tropical Pacific SST = C − B, land ICs over North America = A − D, and atmosphere and land ICs = B − Bc.

c. The intraensemble (unforced) component

The differences we see between the ensemble mean results and the observations (e.g., in terms of U.S. precipitation or height anomalies in the North Atlantic) likely reflect a substantial noise (unforced) component to the anomalies, particularly given that the ensemble mean anomalies tend to be smaller. (We cannot, however, rule out the possibility that the differences reflect deficiencies in the model.) Evidence for large intraensemble variance is presented in Fig. 8, which shows the height anomalies for individual ensemble members. Of the 44 ensemble members displayed in the figure, only about four achieve the large negative anomalies over eastern North America and the North Atlantic found in the observations. Figure 9 compares one of those four ensemble members with the observed and the ensemble mean. Not only does this member have realistic height anomalies, it also includes the large precipitation anomalies over the Ohio River basin found in the observations. Figure 9 further supports the idea that the observed event has a large random (unforced by SST and land conditions) component.

Fig. 8.

April 2011 250-mb height anomalies (m; A − Ac; Table 1) for the 44 individual ensemble members and (bottom right) the observed (outlined by a red dashed line). Red solid lines outline cases that show a qualitative resemblance to the observed over North America and the North Atlantic.

Fig. 8.

April 2011 250-mb height anomalies (m; A − Ac; Table 1) for the 44 individual ensemble members and (bottom right) the observed (outlined by a red dashed line). Red solid lines outline cases that show a qualitative resemblance to the observed over North America and the North Atlantic.

Fig. 9.

(top) April 2011 250-mb height anomalies (m) with respect to the climatology, spanning the years 2000–10. (bottom) As in (top), but for precipitation anomalies (mm day−1). (left) For the ensemble mean (A − Ac). (middle) A single ensemble member chosen because of its resemblance to observations. (right) The results based on observations: the height anomalies are from MERRA, and the precipitation anomalies are from the CPC unified gauge-based 0.25° analysis of daily precipitation.

Fig. 9.

(top) April 2011 250-mb height anomalies (m) with respect to the climatology, spanning the years 2000–10. (bottom) As in (top), but for precipitation anomalies (mm day−1). (left) For the ensemble mean (A − Ac). (middle) A single ensemble member chosen because of its resemblance to observations. (right) The results based on observations: the height anomalies are from MERRA, and the precipitation anomalies are from the CPC unified gauge-based 0.25° analysis of daily precipitation.

To quantify the relative contributions of the SST-forced response and the unforced component to the April 2011 height anomalies, we follow the approach of Chang et al. (2012): we reconstruct the MERRA height anomaly field as a linear combination of the model’s ensemble mean anomaly and the leading noise components. We compute the noise components based on an REOF decomposition of the April intraensemble variance computed from the various model experiments.

The four leading noise REOFs consist of the NAM, a northern Eurasian wave mode, the NAO, and the North Pacific Oscillation (NPO). Using the leading RPCs as predictors, it turns out that the NAM (E1), the NAO (E3), and the eighth4 REOF (E8) provide a minimal set of predictors for the unforced part of the variance, allowing the reproduction (together with the forced response) of the observed height anomalies. In particular, we form the regression equation:

 
formula

where ε is a residual and α, β, and γ are the regression parameters. Also, is the April 2011 MERRA height anomaly, and is the model’s ensemble mean anomaly for that month.

The results of the reconstruction are presented in Fig. 10. The left panels show the observed anomalies (top panel), the forced response (ensemble mean; middle panel), and the reconstructed field (bottom panel), the latter consisting of the ensemble mean plus the weighted sum of the first, third, and eighth REOFs. The resemblance of the reconstructed field to the observed over the Western Hemisphere is remarkable.

Fig. 10.

(top left) The April 2011 250-mb height anomaly with respect to the climatology spanning the years 2000–10 from MERRA. (middle left) The ensemble mean anomaly of the AGCM simulations (A − Ac). (bottom left) The reconstructed field based on the ensemble mean and the leading REOFs of April z250mb model intraensemble variance [(right) REOFs 1, 3, and 8].

Fig. 10.

(top left) The April 2011 250-mb height anomaly with respect to the climatology spanning the years 2000–10 from MERRA. (middle left) The ensemble mean anomaly of the AGCM simulations (A − Ac). (bottom left) The reconstructed field based on the ensemble mean and the leading REOFs of April z250mb model intraensemble variance [(right) REOFs 1, 3, and 8].

These results (together with those shown in Fig. 2) indicate that the NAO plays a key role in producing the most extreme anomalies over the United States during April. We thus turn next to a more detailed analysis of that mode.

d. The NAO mode

A key question is whether there was a predilection for an extreme NAO-like mode to develop during April 2011: particularly, whether an SST-forced change in the base state made the development of the NAO more likely than average. Here we turn to experiments with the stationary wave model described in section 2. We begin with a look at observations-based experiments designed to establish the leading forcing terms and to assess the ability of the SWM to reproduce the NAO’s structure. This is followed by an examination of additional AGCM-based SWM experiments designed to isolate SST-forced changes in the base state during April 2011 that might have increased the likelihood of an extreme NAO event.

In the first set of experiments, the background state consists of the full three-dimensional March–May mean climatological flow computed from MERRA for the period 1980–2011. The stationary wave forcing includes anomalies of diabatic heating, vorticity and temperature transient flux convergences, and divergence terms [see Schubert et al. (2011) for details of the leading forcing terms]. To isolate the forcing associated with the NAO, we compute the forcing terms at each grid point and for each spring month based on the daily MERRA data for 1980–2011. The monthly forcing terms are then regressed against the normalized springtime monthly RPC 3 (i.e., the time series of the pattern most resembling the NAO from MERRA). Figure 11 provides an example of two of the leading forcing terms. The heating anomalies in the middle troposphere (top panel) consist of a tripole pattern in the North Atlantic, with heating to the south of Greenland juxtaposed with cooling to the south (located just north of 30°N) and heating over the Caribbean Sea and western tropical North Atlantic. There is also a heating anomaly centered over the central tropical Pacific. The inverse Laplacian5 of the transient vorticity forcing (bottom panel of Fig. 11) also shows a tripole pattern over the North Atlantic extending westward over eastern North America. The North Pacific is characterized by negative anomalies in middle latitudes with positive anomalies to the north extending into the Arctic.

Fig. 11.

The forcing anomalies obtained from MERRA by regressing the March–May monthly interannual anomalies over the period 1980–2011 against the NAO mode (RPC 3). (top) Midtropospheric heating. (bottom) Inverse Laplacian of the transient vorticity term. Values correspond to one standard deviation in RPC 3.

Fig. 11.

The forcing anomalies obtained from MERRA by regressing the March–May monthly interannual anomalies over the period 1980–2011 against the NAO mode (RPC 3). (top) Midtropospheric heating. (bottom) Inverse Laplacian of the transient vorticity term. Values correspond to one standard deviation in RPC 3.

We now examine the extent to which the MERRA-based NAO anomalies can be reproduced by the SWM when driven by these forcing terms. Figure 12 shows the responses of the streamfunction in the model upper troposphere (σ = 0.257) to the leading forcing terms (diabatic heating, temperature transients, and vorticity transients). The response to the transient vorticity flux convergences looks remarkably like the positive NAO streamfunction obtained via regression directly from MERRA (cf. top-left and bottom-left plots in Fig. 12). A key difference is that the negative anomalies over North America and the North Atlantic in MERRA are shifted farther east. The response to the temperature transients (bottom-right panel of Fig. 12) is largely out of phase with the response to the vorticity transients; it is also weaker, so that the combined response still resembles the regressed NAO field, but with reduced amplitude, especially over North America/Greenland. The response to the diabatic heating (middle-left panel) is generally weaker than that from the other terms, though it does make substantial contributions to the anomalies over the North Atlantic. Overall, we find that the response to the sum of the transients and the diabatic heating (top-right panel in Fig. 12) does resemble the NAO pattern, though the amplitude is too weak, especially in northern high latitudes (particularly over North America).

Fig. 12.

Eddy streamfunction patterns at σ = 0.257 based on various regressions against the MERRA-based RPC 3 (see Fig. 2) using the March–May monthly anomalies over the period 1980–2011. (top left) The regressed NAO pattern plotted for the positive phase. The remaining panels are the responses to the various forcing terms obtained by regressing the forcing against RPC 3. The forcing terms consist of (top right) the heating and total transients, (middle left) the heating, (middle right) the total transients, (bottom left) the vorticity transients, and (bottom right) the temperature transients. All values correspond to one standard deviation in RPC 3.

Fig. 12.

Eddy streamfunction patterns at σ = 0.257 based on various regressions against the MERRA-based RPC 3 (see Fig. 2) using the March–May monthly anomalies over the period 1980–2011. (top left) The regressed NAO pattern plotted for the positive phase. The remaining panels are the responses to the various forcing terms obtained by regressing the forcing against RPC 3. The forcing terms consist of (top right) the heating and total transients, (middle left) the heating, (middle right) the total transients, (bottom left) the vorticity transients, and (bottom right) the temperature transients. All values correspond to one standard deviation in RPC 3.

A key caveat of the above analysis concerns the nature of the transient eddy forcing. As Feldstein (2003) points out, the NAO is driven by nonlinear processes with the transient eddies playing a key role. While the high-frequency transients act in part as a feedback mechanism to prolong the life cycle of the NAO, the question of what processes organize the transients to drive the NAO in the first place is still not well understood. As such, one must be careful in interpreting the eddies as a forcing term (as we do above), when in fact the existence of the eddies depends at least in part on the existence of the NAO.

Despite the above limitation, we believe the SWM-based analysis of the forcing of the NAO is sufficiently realistic to allow the use of the SWM for a study of the impact on the NAO of changes in the base state. In particular, we examine whether the anomalous (SST-forced) three-dimensional April 2011 state was unusually conducive to the development of a large-amplitude NAO. We use the same approach as above, but now, instead of using MERRA fields, we compute the forcing terms for April 2011 based on the intraensemble variability of the model simulations. The monthly forcing terms are then regressed against the model’s normalized April RPC 3, which is the pattern most resembling the NAO from the model simulations. We examine two different April base states: a control consisting of the average of all the ensemble members for the years 2000–10, and the anomaly base state consisting of the mean of the 2011 ensemble members. The former is our estimate of the model’s climatological April base state, while the latter (when differencing it from the control state; A − Ac in Table 1) provides an estimate of how the SST impacted the base state during April 2011 (e.g., Fig. 5a).

The top-left panel of Fig. 13 shows the AGCM’s NAO streamfunction pattern (plotted for the positive phase) in the upper troposphere (σ = 0.257) computed by linearly regressing the intraensemble streamfunction anomalies against the model’s normalized RPC 3 values (the associated REOF is shown in the right, middle panel of Fig. 10). This is our target for what we expect to reproduce with the SWM. The results of the SWM are shown in the remaining panels. The base state is (as described above) the April 2011 three-dimensional ensemble mean of the AGCM runs. We see that the SWM-generated pattern for all of the forcing terms combined (top-right panel of Fig. 13) does resemble the AGCM NAO pattern, though it is again somewhat weaker in amplitude. The primary forcing comes from the vorticity transients (middle-left panel), with a secondary contribution from the temperature transients (middle-right panel; in quadrature with the response to the vorticity transients). Heating provides a negligible contribution (not shown).

Fig. 13.

SWM diagnosis of the NAO for April 2011 based on AGCM results: (top left) The AGCM’s NAO streamfunction pattern (plotted for the positive phase) at σ = 0.257, computed by linearly regressing the intraensemble streamfunction anomalies against the model’s normalized RPC 3 values. The remaining three upper panels show the results from an SWM at sigma level 0.257, in which the forcing associated with the NAO was obtained by linearly regressing the intraensemble forcing against the normalized RPC 3 values. The base state is the April 2011 three-dimensional ensemble mean of the AGCM runs. (top right) The results for all three forcing terms (vorticity and temperature transients and heating). (middle left) the results when the SWM was forced with just the vorticity transients. (middle right) The results for just the temperature transients. (bottom) The SWM results obtained with the April 2011 ensemble mean state [shown (top right)] minus the results obtained using a climatological April base state computed from the AGCM runs spanning 2000–10.

Fig. 13.

SWM diagnosis of the NAO for April 2011 based on AGCM results: (top left) The AGCM’s NAO streamfunction pattern (plotted for the positive phase) at σ = 0.257, computed by linearly regressing the intraensemble streamfunction anomalies against the model’s normalized RPC 3 values. The remaining three upper panels show the results from an SWM at sigma level 0.257, in which the forcing associated with the NAO was obtained by linearly regressing the intraensemble forcing against the normalized RPC 3 values. The base state is the April 2011 three-dimensional ensemble mean of the AGCM runs. (top right) The results for all three forcing terms (vorticity and temperature transients and heating). (middle left) the results when the SWM was forced with just the vorticity transients. (middle right) The results for just the temperature transients. (bottom) The SWM results obtained with the April 2011 ensemble mean state [shown (top right)] minus the results obtained using a climatological April base state computed from the AGCM runs spanning 2000–10.

The bottom panel of Fig. 13 shows the SWM results obtained with the April 2011 ensemble mean state (shown in the top-right panel) minus the results obtained using the control (climatological) April base state. It thus shows the impact of the base state changes in April 2011 induced by SST anomalies. A wavenumber-1 pattern appears at high latitudes in the difference plot, a pattern that acts to enhance the positive NAO (cf. top-left panel). Similar results (but with opposite sign) are obtained when the SWM is driven with forcing of the opposite sign, indicating that the results are quite linear. Overall, the SWM-based results indicate that the April 2011 SST anomalies did indeed have some modest influence on the NAO through changes in the base state, acting to produce a predilection for a more intense NAO, though the resulting NAO could have been of either sign.

We find some support for an enhanced intraensemble variance of RPC 3 (the NAO mode computed from experiment A) compared with the climatological value (computed from experiment Ac), though the results are not statistically significant.6 We note that Karpechko et al. (2014) found that the effect of SST and ozone forcing was to increase the probability of positive NAM events of large magnitude during the spring of 2011, a result that further supports the idea that the high latitudes had a predilection for increased variability. In any event, even if the April 2011 enhanced variance of RPC 3 is real, it would translate to a change in the probability of exceeding the observed value (assuming a normal distribution) from the already climatologically small probability of 0.26% to only 0.85%—still a very small probability.

4. Summary and conclusions

A series of AGCM simulations was carried out to examine the causes of the extreme weather activity and associated record-breaking precipitation that occurred throughout much of the Ohio River valley in the spring of 2011. Our analysis focused on April (the month with the most extreme precipitation anomalies), employing a moderately high (¼°) horizontal resolution version of the GEOS-5 AGCM run for the period 1 March through 31 May. Several different ensembles of runs were carried out in an attempt to isolate the role of SST forcing and initial (both atmospheric and land) conditions. The key result of the model simulations is that the record-breaking precipitation that occurred in April is primarily the result of the development of an unusually large-amplitude positive NAO-like (free) mode, while it was only weakly forced by SST and land conditions. The direct impact of the SST was primarily to produce drying and warming in the southern plains (with only weak wet anomalies in the northeast), while the impact of realistic initial North American land conditions was a tendency for wet conditions to occur in a line extending from the central plains to the Northeast and dry conditions in the Southeast.

A complicating factor in our assessment of the SST and land impacts is that the initial atmospheric condition7 was the primary contributor to the April ensemble mean precipitation response in the Ohio River basin. The source of that signal (which occurs well beyond the range of weather predictability) is unclear. One possibility is that it reflects in the initial conditions the impact of past SST forcing (e.g., Barsugli et al. 1999; Branstator 2014). Another possibility is that stratospheric–tropospheric coupling played a role. The polar vortex was unusually cold and persistent that winter and spring, and there is clear evidence for downward propagation from the stratosphere that led to an enhancement of the tropospheric NAM in March and April (Hu and Xia 2013). Karpechko et al. (2014) further showed that preconditioning (likely associated with internal atmospheric dynamics) played an important role in the downward propagation following the strong polar vortex event (though ozone depletion and SST forcing also played a role). As such, we cannot fully distinguish the contributions of the SST from other potential sources of the signal that might be reflected in the initial conditions.

The important role of the NAO has major implications for the predictability of the circulation and precipitation anomalies. Figure 14 shows the signal-to-noise ratios of the global upper-level height and U.S. precipitation fields for April 2011 based on hindcasts initialized at the beginning of March (experiment A). While the height field does show ratios larger than 2 over a substantial part of North America (and of course over much of the tropics), the ratios for precipitation over the United States are generally quite small (e.g., less than 0.5 over the key region of the Ohio River basin). As mentioned earlier, the NAO is known to have limited predictability at monthly and longer time scales, reflecting the intrinsic short time scales of this mode. This is the result of the dynamical processes that govern the growth, maintenance, and decay of the NAO—processes that are dominated by transient eddy fluxes (section 3d; see also Feldstein 2003). Our results are consistent with the study by Pegion and Webb (2014), which concluded that the meteorological factors leading to the 2011 flood are not accurately predicted at seasonal lead times by state-of-the-art operational and experimental forecast systems. There is some indication of larger ratios (exceeding 1) in the Southwest and along the southern tier of states that are likely linked to La Niña.

Fig. 14.

The square of the ensemble mean anomaly divided by the intraensemble variance for April 2011 (based on 44 ensemble members, experiment A); red contour indicates signal/noise = 1: (top) for 250-mb height; (bottom) for precipitation.

Fig. 14.

The square of the ensemble mean anomaly divided by the intraensemble variance for April 2011 (based on 44 ensemble members, experiment A); red contour indicates signal/noise = 1: (top) for 250-mb height; (bottom) for precipitation.

The question of whether there was some predilection for an extreme positive NAO during April 2011 was addressed here using an SWM. Our results indicate that the SST-forced (ensemble mean) base state for April 2011 did indeed enhance the amplitude of the NAO response compared to that of the climatological state, though the impact is modest and can be of either sign (contributing to either a positive or negative NAO, depending on the sign of the forcing). Our assessment of whether the April 2011 intraensemble variance of the NAO was elevated compared with its climatological value was inconclusive (results were not statistically significant), though even if significant, it would suggest at most only a slight enhancement (compared to the climatology) of the probability of obtaining a large-amplitude NAO during 2011 (still less than 1%).

Acknowledgments

Support for this work was provided by the NASA Modeling, Analysis, and Prediction (MAP) Program (NASA Research Announcement NNH12ZDA001N). We wish to thank the editor and two anonymous reviewers for their valuable comments that helped to substantially improve the paper.

REFERENCES

REFERENCES
Bacmeister
,
J. T.
,
M. J.
Suarez
, and
F. R.
Robertson
,
2006
:
Rain reevaporation, boundary layer–convection interactions, and Pacific rainfall patterns in an AGCM
.
J. Atmos. Sci.
,
63
,
3383
3403
, doi:.
Baldwin
,
M. P.
, and
T. J.
Dunkerton
,
1999
:
Propagation of the Arctic Oscillation from the stratosphere to the troposphere
.
J. Geophys. Res.
,
104
,
30 937
30 946
, doi:.
Barnston
,
A. G.
, and
R. E.
Livezey
,
1987
:
Classification, seasonality, and persistence of low-frequency atmospheric circulation patterns
.
Mon. Wea. Rev.
,
115
,
1083
1126
, doi:.
Barsugli
,
J. J.
,
J. S.
Whitaker
,
A. F.
Loughe
,
P. D.
Sardeshmukh
, and
Z.
Toth
,
1999
:
The effect of the 1997/98 El Niño on individual large-scale weather events
.
Bull. Amer. Meteor. Soc.
,
80
,
1399
1411
, doi:.
Branstator
,
G.
,
2014
:
Long-lived response of the midlatitude circulation and storm tracks to pulses of tropical heating
.
J. Climate
,
27
,
8809
8826
, doi:.
Chang
,
E. K. M.
, and
Y.
Fu
,
2002
:
Interdecadal variations in Northern Hemisphere winter storm track intensity
.
J. Climate
,
15
,
642
658
, doi:.
Chang
,
Y.
,
S. D.
Schubert
, and
M. J.
Suarez
,
2000
:
Boreal winter predictions with the GEOS-2 GCM: The role of boundary forcing and initial conditions
.
Quart. J. Roy. Meteor. Soc.
,
126
,
2293
2321
, doi:.
Chang
,
Y.
,
S. D.
Schubert
, and
M. J.
Suarez
,
2012
:
Attribution of the extreme U.S. East Coast snowstorm activity of 2010
.
J. Climate
,
25
,
3771
3791
, doi:.
Chen
,
M.
, and
P.
Xie
,
2008
: CPC unified gauge-based analysis of global daily precipitation. Eos, Trans. Amer. Geophys. Union, 89 (Western Pacific Geophysics Meeting Suppl.), Abstract A24A-05.
Collins
,
N.
, and Coauthors
,
2005
:
Design and implementation of components in the Earth System Modeling Framework
.
Int. J. High Perform. Comput. Appl.
,
19
,
341
350
, doi:.
CPC
,
2016
: NOAA Climate Prediction Center Monthly mean NAO index since January 1950. Accessed 5 September 2016. [Available online at http://www.cpc.ncep.noaa.gov/products/precip/CWlink/pna/norm.nao.monthly.b5001.current.ascii.table.]
ESRL
,
2011
: Preliminary assessment of climate factors contributing to the extreme 2011 tornadoes. Accessed 5 September 2016. [Available online at http://www.esrl.noaa.gov/psd/csi/events/2011/tornadoes/.]
Feldstein
,
S. B.
,
2000
:
The timescale, power spectra, and climate noise properties of teleconnection patterns
.
J. Climate
,
13
,
4430
4440
, doi:.
Feldstein
,
S. B.
,
2003
:
The dynamics of NAO teleconnection pattern growth and decay
.
Quart. J. Roy. Meteor. Soc.
,
129
,
901
924
, doi:.
Held
,
I. M.
,
M.
Ting
, and
H.
Wang
,
2002
:
Northern winter stationary waves: Theory and modeling
.
J. Climate
,
15
,
2125
2144
, doi:.
Hoerling
,
M.
,
J.
Eischeid
, and
R.
Webb
,
2013a
: Understanding and explaining climate extremes in the Missouri River basin associated with the 2011 flooding. NOAA Climate Assessment Rep., 34 pp.
Hoerling
,
M.
, and Coauthors
,
2013b
:
Anatomy of an extreme event
.
J. Climate
,
26
,
2811
2832
, doi:.
Hu
,
Y. Y.
, and
Y.
Xia
,
2013
:
Extremely cold and persistent stratospheric Arctic vortex in the winter of 2010–2011
.
Chin. Sci. Bull.
,
58
,
3155
3160
, doi:.
Karl
,
T.
,
E.
O’Lenic
, and
H.
Brooks
,
2011
: Spring 2011 U.S. climate extremes. NOAA’s National Climatic Data Center, Asheville, NC, Tech. Rep., 3 pp.
Karpechko
,
A. Yu.
,
J.
Perlwitz
, and
E.
Manzini
,
2014
:
A model study of tropospheric impacts of the Arctic ozone depletion 2011
.
J. Geophys. Res.
,
119
,
7999
8014
, doi:.
Kleist
,
D. T.
,
D. F.
Parrish
,
J. C.
Derber
,
R.
Treadon
,
W.-S.
Wu
, and
S.
Lord
,
2009
:
Introduction of the GSI into the NCEPs Global Data Assimilation System
.
Wea. Forecasting
,
24
,
1691
1705
, doi:.
Koster
,
R. D.
,
M. J.
Suárez
,
A.
Ducharne
,
M.
Stieglitz
, and
P.
Kumar
,
2000
:
A catchment-based approach to modeling land surface processes in a general circulation model: 1. Model structure
.
J. Geophys. Res.
,
105
,
24 809
24 822
, doi:.
Koster
,
R. D.
,
Y.
Chang
, and
S. D.
Schubert
,
2014
:
A mechanism for land–atmosphere feedback involving planetary wave structures
.
J. Climate
,
27
,
9290
9301
, doi:.
Lin
,
S.-J.
,
2004
:
A vertically Lagrangian finite-volume dynamical core for global models
.
Mon. Wea. Rev.
,
132
,
2293
2307
, doi:.
Lorenz
,
E. N.
,
1969
:
The predictability of a flow which possesses many scales of motion
.
Tellus
,
21A
,
289
307
, doi:.
Molod
,
A.
,
L.
Takacs
,
M.
Suarez
, and
J.
Bacmeister
,
2015
:
Development of the GEOS-5 atmospheric general circulation model: Evolution from MERRA to MERRA2
.
Geosci. Model Dev.
,
8
,
1339
1356
, doi:.
Moorthi
,
S.
, and
M. J.
Suarez
,
1992
:
Relaxed Arakawa–Schubert. A parameterization of moist convection for general circulation models
.
Mon. Wea. Rev.
,
120
,
978
1002
, doi:.
NCDC
,
2011
: April 2011 statewide ranks. Accessed 31 August 2016. [Available online at http://www.ncdc.noaa.gov/monitoring-content/sotc/national/statewidepcpnrank/statewidepcpnrank-201104.gif.]
Pegion
,
K.
, and
R.
Webb
,
2014
: Seasonal precipitation forecasts over the Missouri River Basin: An assessment of operational and experimental forecast system skill and reliability. NOAA Seasonal Precipitation Forecast Assessment Rep., 26 pp.
Pfahl
,
S.
, and
H.
Wernli
,
2012
:
Quantifying the relevance of cyclones for precipitation extremes
.
J. Climate
,
25
,
6770
6780
, doi:.
Richman
,
M. B.
,
1986
:
Rotation of principal components
.
J. Climatol.
,
6
,
293
335
, doi:.
Rienecker
,
M. M.
, and Coauthors
,
2008
: The GEOS-5 Data Assimilation System—Documentation of versions 5.0.1, 5.1.0, and 5.2.0. NASA Tech. Rep. Series on Global Modeling and Data Assimilation, NASA/TM-2007-104606, 95 pp.
Rienecker
,
M. M.
, and Coauthors
,
2011
:
MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications
.
J. Climate
,
24
,
3624
3648
, doi:.
Schubert
,
S.
,
H.
Wang
, and
M.
Suarez
,
2011
:
Warm season subseasonal variability and climate extremes in the Northern Hemisphere: The role of stationary Rossby waves
.
J. Climate
,
24
,
4773
4792
, doi:.
Seager
,
R.
,
L.
Goddard
,
J.
Nakamura
,
N.
Henderson
, and
D. E.
Lee
,
2014
:
Dynamical causes of the 2010/11 Texas–northern Mexico drought
.
J. Hydrometeor.
,
15
,
39
68
, doi:.
Ting
,
M.
, and
L.
Yu
,
1998
:
Steady response to tropical heating in wavy linear and nonlinear baroclinic models
.
J. Atmos. Sci.
,
55
,
3565
3582
, doi:.
U.S. Drought Monitor
,
2011
: Drought Monitor Archive for April 26, 2011. Accessed 31 August 2011. [Available online at http://droughtmonitor.unl.edu/MapsAndData/MapArchive.aspx (use the date 26 April 2011).]
Wang
,
H.
,
S.
Schubert
,
R.
Koster
,
Y.-G.
Ham
, and
M.
Suarez
,
2014
:
On the role of SST forcing in the 2011 and 2012 extreme U.S. heat and drought: A study in contrasts
.
J. Hydrometeor.
,
15
,
1255
1273
, doi:.
Wu
,
W.-S.
,
R. J.
Purser
, and
D. F.
Parrish
,
2002
:
Three-dimensional variational analysis with spatially inhomogeneous covariances
.
Mon. Wea. Rev.
,
130
,
2905
2916
, doi:.

Footnotes

1

Further details of the perturbations are as follows. Only the atmospheric states were perturbed. For experiment A, 44 unique perturbations were produced, as described above, based on 22 pairs of atmospheric states, and both adding and subtracting the scaled difference fields. For Ac, a subset of six of the experiment A perturbations was used: the same six perturbations were used for each year (2000–10). Similarly, for experiment Bc, a subset of two of the experiment A perturbations was used, and these were repeated for each year (2000–10). For experiments B and D, the 22 perturbations are a subset of the 44 perturbations used in experiment A.

2

The amplitude of the NAO was, in fact, the largest on record for April, based on the values of the NAO index tabulated by NOAA’s Climate Prediction Center for the period 1950–present. Their NAO index is based on a rotated principal component analysis of the 700-mb height field (Barnston and Livezey 1987) (CPC 2016; http://www.cpc.ncep.noaa.gov/products/precip/CWlink/pna/norm.nao.monthly.b5001.current.ascii.table).

3

We note that the SST anomalies with respect to the 2000–10 climatology (shown in Fig. 3) are overall quite similar to those based on a longer climatology (e.g., 1981–2010).

4

It turns out that the eighth mode is required to reproduce the shortwave pattern across the North Pacific. This unforced pattern is very likely unpredictable and the fact that it happens to project on the eighth REOF almost certainly occurs purely by chance. We do not claim that this REOF is in any way statistically distinct.

5

The inverse Laplacian provides a filtered view of what is otherwise a rather noisy field.

6

Based on an F test for the ratio of the variances at the 90% significance level.

7

While this is based on runs that also included realistic global land initial conditions, our runs that isolated the role of North American land conditions indicate that the land cannot account for the relatively large April ensemble mean precipitation response.