|
|
|
|
Seasonal Prediction of North American Winter Climate Paper about Seasonal Prediction of the North American Winter Climate discussing the approaches, methodology, climate teleconnections used, and suggestions for future research and improvements in seasonal prediction practices. The focus of this paper is on the Pacific-North American sector. |
|
| About WLH | |
| Contact WLH | |
| Site Map | |
| Atmospheric & Oceanic Sites | |
| General Weather Sites | |
| Weather Forecast Tools | |
| Major Forecast Centers | |
| Model Sites | |
| Climate Analysis Tools | |
| Reanalysis Data Sites | |
| Weather Research | |
|
|
|
|
Sections of Paper
I. Introduction to Seasonal
Predictability of North American Winter Climate
III. North Pacific Variability 1. Introduction to Seasonal Predictability of North American Winter Climate Our ability to forecast seasonal climate depends on how much we understand about the evolution of the atmospheric response to surface boundary forcing and the inherent predictability of the atmosphere-ocean system. Examples of surface boundary forcing include sea surface temperatures (SST), snow cover, vegetation type and coverage, and soil moisture. Due to the complexity of this problem, only SSTs will be the focus in this paper. Even with a perfect general circulation model, climate predictability on sub-seasonal to interannual timescales has an internal limit due to the chaotic and non-linear nature of the atmosphere. To further complicate this problem, the global surface boundary forcing is not static. SSTs are always dynamically evolving, changing boundary forcing over time. In order to enhance our ability to predict the climate on seasonal time scales, we must not only understand the atmospheric response to boundary forcing, but also understand how boundary forcing evolves over time. There exists high interest in improving the predictability of the North Hemispheric winter climate, due to the heating, health and transportation needs of the public as well as the requirements of the energy firms, utilities, and agricultural industries worldwide. Operational seasonal climate prediction is an emerging practice with many societal applications, and an ability to anticipate climate fluctuations in advance would benefit decision and policy making in hydrology, agriculture, health, energy, et cetera (Barnston et al, 2005). Being able to predict temperatures and precipitation on scales longer than two weeks requires knowledge of the currently observed atmospheric circulation and their anomalies, the surface boundary forcing, and the evolution of the climate system throughout the season. To help understand the predictability of the climate system on these time scales, both statistical and modeling studies have been conducted. The focus of this paper will be on analyzing the predictability of the North American winter climate with surface boundary forcing of the Pacific Ocean as well as atmospheric circulation anomalies in the vicinity of the Pacific-North American region. a. Characteristics of Seasonal Climate Predictability Predictability of extratropical winter climate depends on the atmospheric response to slowly-varying external surface boundary forcing, and on the internal variability of the atmosphere. Climate predictability is based on the fact that slowly-varying boundary conditions influence the general atmospheric circulation (Barnston et al, 2005). If prediction of the evolution of these boundary conditions can be improved, then seasonal forecasting skill can also be improved. To do this, atmospheric responses to a broad range of anomalous boundary surface forcings must be documented and understood. Although the effects of the El Niño and Southern Oscillation (ENSO) on the general atmospheric circulation is fairly well understood, understanding of global SST forcing is currently incomplete; more needs to be understood about SST forcing by the Indian and Atlantic Oceans as well as the extratropical North Pacific Ocean. This incomplete understanding of global boundary forcing is reflected in the fact that it is not always clear why seasonal predictions succeed in some instances but fail in others (Barnston et al, 2005). Do failures of some of the seasonal climate predictions reflect a natural predictability limit imposed by internal atmospheric variability, aptly known as “climate noise”? Are errors in seasonal climate predictions at least partly contributed to by biases and inaccuracies in the methodologies used? If so, how can these errors be minimized as to achieve an optimal level of skill associated with winter climate forecasting? These questions posed by Barnston (2005) are addressed in his paper discussing the issues related to the practice of making seasonal climate predictions. In addition to these questions, this paper poses a few questions of its own. It has been said in past studies that climate noise and internal variability is not forced and is unpredictable on seasonal timescales (Barnston et al, 2005; Renshaw et al, 1998; Kumar and Hoerling, 2000; Leetmaa, 2003). Therefore, seasonal predictions have been cast as probabilistic due to the ever-present uncertainty from climate noise. Internal variability and climate noise consists of individual weather events and intensities from transient baroclinic eddies, high latitude blocking, and stationary troughs. If possible, how much can the uncertainty be reduced by ascertaining the impact of high latitude blocking, stationary troughs, and transient eddies on the general atmospheric circulation? The answer to this question may not come immediately, but this paper will explore the possible avenues toward arriving at an answer with future research. The most fundamental characteristic of extratropical seasonal predictability is the inherent probabilistic behavior of the atmosphere given a particular state of the ocean (Kumar and Hoerling, 2000), referring to the fact that more than one atmospheric state is possible under the influence of identical surface boundary forcing or SST configuration. These atmospheric states correspond to well-documented patterns of atmospheric low-frequency variability (Barnston and Livezey, 1987). The effect of surface boundary forcing is to alter the probabilities of the occurrence of certain atmospheric states. Kumar and Hoerling (2000) uses the Pacific-North American pattern teleconnection as an example of an atmospheric state sensitive to ENSO. Better quantification of the effects of ENSO on the seasonal winter climate worldwide has helped improve seasonal forecast skill for temperature and precipitation averaged over a period of three months (Barnston et al, 2005). Prediction of seasonal climate anomalies is based on a combination of prediction tools including ensemble general circulation model (GCM) methods. Ensembles of an atmospheric general circulation model (AGCM) are forced with identical boundary conditions, but are initialized with different atmospheric states. A statistical analysis, such as the mean, of the ensemble members forms the basis of the prediction (Kumar and Hoerling, 2000). Theoretically, as the number of cases of the same anomalous boundary forcing increases, the statistical distribution of atmospheric anomalies would converge to an atmospheric state that is the most likely response to that forcing. This involves the usage of a probability distribution function (PDF), whereas the mean anomaly across the ensemble members would be considered the best deterministic forecast and the uncertainty would be reflected in the ensemble spread about the mean anomaly (Barnston et al, 2005). b. Practical Approaches to Seasonal Climate Prediction In the present, there are two approaches for inferring atmospheric responses to the forcings of slowly-varying boundary conditions: an empirical approach and an ensemble modeling approach. The empirical approach consists of statistical use of analogs (near-matches) of historical boundary conditions and the accompanying global circulation and surface climate, as explained by Barnston (2005). A glaring disadvantage of this method is the short historical record of globally adequate observational analyses. The sampling size is consequently too small to resolve differences in the relationships between SST forcing and climate. Due to the short historical record, a major shortcoming of the empirical method to seasonal forecasting, a dynamical method involving atmospheric general circulation modeling (AGCM) has been used to investigate the broad spectrum of atmospheric responses to boundary forcings observed in the past. Several different models and their ensemble members would be forced with an identical global SST state, but initialized with different atmospheric states. In some cases, the atmospheric response to identical boundary forcings would be different from one ensemble member to another, owing to different initial values across the ensembles. In other cases, the atmospheric response would be biased towards a certain atmospheric state despite differences in initial values, owing to the identical forcing of the boundary in each ensemble run. In nature, the real atmospheric response would be somewhere in between, due to both the physics of the response to the boundary forcing and the chaotic nature of internal atmospheric variability. Ensemble modeling, the second approach, can rectify the disadvantage of the empirical method when defining the climate signals forced by boundary conditions using dynamical approaches. AGCM simulations can produce a large sample of climate states for each and every observed boundary state in the historical record (Barnston et al, 2005). These AGCMs can better reproduce the atmospheric signals forced by SST anomalies than deriving empirically from the short historical records (Kumar and Hoerling, 2000). The ensemble size also plays a role in seasonal prediction skill; the more ensemble members, the smaller the sampling error and the higher the prediction skill up to the natural limit of predictability (Kumar et al, 2000; Kumar and Hoerling, 2000). A high degree of agreement between AGCMs and their ensembles indicates the prominence of the boundary forced signal compared with internal atmospheric variability. But if the model-produced signal turns out to be a poor match to observations, Barnston (2005) asks whether this is due to poor performance by the AGCMs marred by their own biases, or that the observed anomalies have been influenced by atmospheric internal variability while still being consistent with the PDF of seasonal mean atmospheric states associated with SST forcing. It is therefore difficult to draw a definitive conclusion if the observed result is on the tail of a PDF, albeit within the range of possibilities postulated by the PDF. There also exists the question of whether internal atmospheric variability constructively or destructively interferes with a signal. For these reasons, the dynamical modeling method is not without its own set of disadvantages, one particularly being model biases. To minimize this shortcoming, multiple AGCMs are used rather than just one. It remains unknown to what extent dynamical methods may improve upon information gathered using empirical methods alone for climate prediction. For example, if ENSO forcing alone is present, empirical approaches have adequate historical archives to define the mode of variability, but the question is whether inter-ENSO variations of tropical SST anomalies affect the atmosphere, and whether there are atmospheric response patterns to non-ENSO SST variations unexplored in the observed archive (Barnston et al, 2005). It is believed in this paper that inter-ENSO variations of tropical SST anomalies are important and that the response patterns to non-ENSO forcing do exist, because there are no two atmospheric states in response to ENSO that are identical. Therefore, non-ENSO forcing, such as the Pacific Decadal Oscillation (PDO), Indian Ocean Dipole (IOD) and non-SST forcing, must also play a role in seasonal climate variability. This paper examines recent ideas about the relationships between ENSO, PDO, and the Pacific-North American (PNA) pattern, including high-latitude blocking regimes and their effect on North American winter climate. 2. Relationship between ENSO and the Pacific-North American Sector a. Overview of the El Niño and Southern Oscillation The effects of equatorial Pacific SST anomalies (Figure 1) known as the El Niño and Southern Oscillation (ENSO) on extratropical winter climate have been well-studied in the past few decades. Several studies including Renshaw et al (1998), Kumar and Hoerling (1998), and Kumar et al (2000) discuss the effects of ENSO on the atmospheric general circulation in the seasonal winter climate. Mason and Goddard (2001) and Goddard et al (2006) also diagnosed temperature and precipitation anomalies associated with ENSO. Two more studies, Quiroz (1983) and Quiroz (1984) also offer a simple and deterministic look at the effects of the strong El Niño event in 1982-83 and the subsequent La Niña event in 1983-84, respectively, on the winter climate in North America. Kumar et al (2000) conducted a modeling study that compared simulations of ENSO-induced atmospheric variability between four AGCMs, and the regression patterns (not shown) in each AGCM were found to be very similar with minor differences (Kumar et al, 2000).
When an El Niño event occurs, the atmospheric response usually presents itself in the form of a strong westerly jet across the Pacific basin, and a strengthened subtropical jet across the Gulf of Mexico displaced equatorward from climatology. A negative height anomaly in the northeast Pacific and a subtropical pacific ridge southeast of Hawaii strengthens the meridional height contrast (Figure 2a). The justification for using the 200-hPa level for analyzing ENSO influence on the extratropics lies in the fact that the 500-hPa level is near the level of non-divergence in the tropics, hence not the level of choice for analyzing tropical-extratropical interactions. During warm ENSO winters, a ridge situates itself over Canada resulting in anomalously warm temperatures in that region; wetter conditions occur across the southern tier of the United States as a result of an equatorward shift of the subtropical jet. These anomalous atmospheric circulations have been observed by Quiroz (1983 and 1984), who conducted case studies of the strong 1982-83 El Niño event and the subsequent strong 1983-84 La Niña winter. The 1983-84 La Niña winter was a season of strong high-latitude blocking and severe cold in North America (Quiroz, 1984). During that winter, cold air outbreaks originating from Alaska flowed southeastward through western Canada towards central US (Figure 2b) as a result of a strong blocking anticyclone between the Gulf of Alaska and the Bering Sea. It is during neutral and La Niña winters that mid- to high-latitude blocking over the North Pacific is more frequent than during El Niño winters. Frequency of blocking events during La Niña versus El Niño winters will be discussed more in depth in the later sections of this paper.
b. Effects of ENSO on Seasonal Predictability It must be cautioned that the atmospheric response to a La Niña event is not exact inverse of that to a warm ENSO event. Predictability of winter climate associated with cold ENSO is, unfortunately, quite low compared to that of warm ENSO. The asymmetry of the atmospheric response between cold and warm ENSO further complicates things. However, it is found by some studies (Renwick and Wallace, 1996; Chen and Yoon, 2002; Huang et al, 2004; Carrera et al, 2004) and in this paper that both neutral and cold ENSO winters consist of more mid- and high-latitude blocking over the North Pacific than do warm ENSO events (Figures 3 and 4). The strength and frequency of blocking episodes are quite similar between ENSO-neutral and La Niña winters, raising the possibility that an atmospheric response to a cold ENSO event is not as climatologically anomalous as that to a warm ENSO event.
To further complicate matters, inter-ENSO SST variability (such as positions of the maximum temperature anomalies in ENSO sub-regions 1.2, 3, 3.4, and 4) as well as SST gradients can cause small but important differences in the atmospheric response. These differences must be documented and accounted for to improve seasonal forecasting for ENSO winters. In addition, differences in global SSTs between two otherwise identical ENSO events will also result in important differences in the atmospheric response. Other ocean basins, such as Indian, Atlantic, and North Pacific Oceans have SST anomalies that can adjust or bias the atmospheric responses to certain circulation states that may either constructively or destructively interfere with each other and that of ENSO. The more they constructively interfere with each other, the higher the predictability because the resultant atmospheric anomalies would be more in phase. However, there have been many cases of destructive interference, resulting in much lower seasonal predictability unless it is clear which signal dominates. Kumar and Hoerling (1998) showed that the North American climate is most predictable during late winter and early spring of warm ENSO years, stemming from the fact that the SST-forced signal in El Niño years is much stronger for winter and spring than for summer and autumn (Figure 5). Potential predictability was estimated to be twice as strong for winter-spring as it is for summer-autumn, revealing an unexplained asymmetry between these two equinoxes. Kumar and Hoerling (1998) also pointed out that the potential predictability for moderate to strong El Niños far exceeds that for the moderate to strong La Niñas. For La Niña events, the signal-to-noise ratio rarely exceeds 0.5 for all years in the historical record.
Renshaw et al (1998) and other studies have shown that warm SST anomalies in the equatorial Pacific enhance local convergence and precipitation, giving rise to upper level divergence and advection of vorticity by divergent flow. Theoretically, this generates a Rossby wave train emanating from the western subtropical Pacific, carrying energy into the extratropics along a large circle path arcing across North Pacific and North America, as predicted by barotropic models. This wave train would result in a characteristic pattern in the mid-latitude jet stream (Hoerling and Ting, 1994). The latitudinal shifts in storm tracks in warm and cold ENSO events have a role in influencing the winter climate in North America. However, Renshaw et al (1998) cautioned that the relationship between ENSO events and the PNA pattern is less deterministic than previously suggested. The yearly correlation between PNA and ENSO, as calculated by the Climate Diagnostics Center (CDC) at the National Oceanic and Atmospheric Administration (NOAA), is only around 0.1, albeit a little higher in the winter months. Figure 6 shows the correlation between the wintertime DJF-averaged PNA index and oceanic SSTs, where there is a noticeable local minimum (about 0.2-0.3) in the equatorial Pacific between 5°S and 5°N, and there are maximums in the North Pacific resembling a positive PDO SST configuration.
The suggested weak link between ENSO and PNA is corroborated by a paper by Nigam about teleconnections published in the Encyclopedia of Atmospheric Sciences in 2003, where the relationship between ENSO and PNA was questioned. Contrary to previous research, many positive PNA episodes throughout DJF 1958-1998 have occurred particularly during weak La Niñas (Nigam, 2003). Conversely, more negative PNA episodes have occurred during weak El Niños (Figure 7).
It is difficult to draw a definitive conclusion about the relationship between moderate-strong ENSO events and PNA variability due to sampling issues. Kumar and Hoerling (1998) suggested that this is due to the extratropical climate noise being stronger than the ENSO signal in the overall interannual variability. Only for the strongest El Niños, does the signal exceed the noise, and as a result there is a noticeable increase in occurrences of positive PNA episodes. Nigam (2003) also has pointed out that, in the weekly evolution of the PNA (view PNA Figures A and B), the significant upper-level divergence anomalies in the equatorial Pacific are notably absent. This suggests a more limited role of tropical convection and Rossby wave trains originating from the tropics in exciting PNA variability than suspected in previous studies. PNA analyses in Figure 8 show an absence of divergence over the equatorial Pacific between 5°S and 5°N at the 0.2 sigma level (near 200 hPa), thus corroborating what Nigam had suggested. Rather, significant upper-level convergence is centered just southeast of Hawaii between 10°N and 25°N; and upper-level divergence is centered just south of the Gulf of Alaska between 25°N and 55°N. This divergence-convergence dipole appears to be related to ageostrophic circulation about the subtropical Pacific jet exit at around 140°W. It is for this reason that warm and cold ENSO events do not always give the expected PNA response, observed or modeled (Renshaw et al, 1998).
Unfortunately, the marginal positive correlation between ENSO and the seasonally averaged PNA does not elucidate the relationship between the strength of ENSO and the frequency of positive or negative PNA episodes in each winter season, as implied by Figure 7. Although ENSO appears to bias the atmospheric state towards positive PNA on seasonal time scales, averaging the PNA index over an entire winter season masks strong weekly fluctuations and frequent phase changes of the PNA. Even the monthly averaged PNA changes sign at least once for nearly 73% of the 1958-1993 winters (Figure 9). However, the PNA weekly lead-lag autocorrelation (Figure 10) shows that the PNA fluctuates on weekly, not monthly, time scales with each PNA episode lasting about 2-3 weeks. Therefore, it is deduced that the PNA changes sign at least once, if not twice, nearly every winter on weekly time scales. Hence, weekly reanalysis data would be much more suited to analysis of PNA variability than monthly data, especially when ascertaining its evolution and what contributes to the onset or decay of PNA episodes. ENSO, on the other hand, does not readily change phase throughout the winter season (Nigam, 2003). Hence, it is highly unlikely that the intraseasonal geopotential height variations during the winter months are associated with ENSO; it is more likely that these subseasonal variations are related to the PNA, among other teleconnections.
Further evidence is presented here that the PNA is not as strongly influenced by ENSO as previously thought by past studies. Figure 11a shows a comparison of 200-hPa height anomalies regressed on Nino 3.4 and PNA indices using DJF-averaged data. Nino 3.4 is used here for consistency because it is widely known and used by other studies and by the NOAA Climate Prediction Center as a measure of ENSO. It is clear from Figure 11a that atmospheric variability associated with the PNA is much larger than that associated with ENSO. A significant subtropical ridge associated with ENSO is centered southeast of Hawaii, while the PNA-associated ridge is centered west of Hawaii. The Aleutian Low influenced by the PNA is deeper and centered over the Aleutian Islands, while a comparatively weaker negative anomaly associated with ENSO is centered further southeast near the Gulf of Alaska towards the Pacific Northwest. A strong ridge associated with the PNA is centered over the Yukon territories and Northwest Canada, while a relatively weaker ENSO-influenced ridge is centered further east over the Hudson Bay. Figure 11b shows the same comparison as in Figure 11a, but for the sea level pressure (SLP) anomalies. The positions of the SLP anomalies show the same relative strength of the PNA influence compared to that of ENSO. Negative SLP anomalies resulting from the PNA-induced Aleutian Low are also deeper and positioned further north and west than the relatively weaker ENSO-induced negative SLP anomaly just off the west coast of Canada.
From the 200-hPa and SLP regressions on both indices, one can infer two things: 1) PNA variability is a much more dominant factor in influencing the North American winter climate than ENSO, and 2) upper tropospheric height anomalies influenced by both indices are not exactly in phase. That the PNA- and ENSO-associated height anomalies are not in phase but not entirely out of phase either may seem like a minor detail, but an important one when the exact positions of storm tracks are considered. The 500-hPa height anomalies regressed onto both ENSO and PNA (Figure 11c) also support the aforementioned hypothesis. In addition, the subtropical ridge at the 500-hPa level is noticeably much weaker especially for the ENSO-related height variability, further supporting the hypothesis that tropical-extratropical interactions are more important at the upper troposphere than at the mid-levels. Finally, Figure 12 shows an overlay of 500-hPa height anomalies induced by ENSO and PNA (Nigam, 2003). The shifts in positions of the anomaly centers are nearly 90°, with ENSO-related height anomalies further east-southeast from the PNA-related height anomalies. These seemingly insignificant details in the position and strength of height anomalies are important when determining the storm track, as well as any temperature and precipitation deviations from climatology in the interest meeting the needs of the public and various communities.
These results suggest that the PNA pattern is a natural mode internal to the atmosphere, and that ENSO only attempts to bias the probability of a given PNA mode occurring in its positive or negative phase. Averaged over seasonal to interannual time scales, ENSO may bias the atmospheric state towards a positive PNA due to its influence on the wintertime Northern Hemisphere Hadley Cell. When warm ENSO SST anomalies occur, the Hadley circulation becomes more concentrated at the Inter-Tropical Convergence Zone (ITCZ). Given the weakened equatorial trades, air parcels moving towards the subtropics possess increased angular momentum. Angular momentum conservation requires that parcels gain westerly speed as they approach the subtropics. Since the trades have been weakened during a warm ENSO event, the Pacific jet is expected to be stronger and the storm track is shifted further southeast from its normal position, as shown in Figure 11 and 12. However, on weekly time scales, both phases of the PNA may be as likely to occur in nearly any given ENSO state, except for a very strong El Niño event. Although the ideas discussed in this subsection may be somewhat new, it is not necessarily proof that the ENSO-PNA relationship is in fact weak. a. Overview of North Pacific variability The leading mode of variability of sea surface temperatures (SST) in the North Pacific is the Pacific Decadal Oscillation. It is a prominent slowly-varying, low-frequency SST pattern in the North Pacific. The PDO is positive (Figure 13) when anomalously warm SSTs hug the coast of North America and stretch poleward into the Gulf of Alaska, while anomalously cold SSTs nose into the central North Pacific from the west, superposed by anomalously strong surface westerlies. Due to the presence of a strong Aleutian Low, strong westerly winds cool the central North Pacific SSTs due to entrainment of cold subsurface water into the mixed layer, turbulent heat fluxes, Ekman currents, and cold advection from the north (Schneider and Cornuelle, 2005). As these westerlies approach the North American continent, the flow steers poleward, and these southwesterly winds assist in the downwelling of coastal waters. The resulting SST configuration, resembling a horseshoe (need figures), is considered the positive phase of the PDO (Schneider and Cornueke, 2005). Perturbations in these components contributing to PDO variability are associated with anomalies of the Pacific storm track. Schneider and Cornuelle (2005) raised the possibility that the PDO is not a dynamical mode, but it arises from a superposition of atmospheric forcing onto SST fluctuations with differing dynamical origins, including ENSO.
Whether the PDO is linked with ENSO is a matter of debate (Newman et al, 2003). The simultaneous correlation between ENSO and PDO for the months of November to March, as an intriguing local minimum in the equatorial Pacific (Figure 13), is somewhat weak with values around 0.3-0.4 for all seasons over decadal time scales, according to the NOAA Climate Diagnostics Center (CDC). This indicates that the frequency of warm ENSO events is slightly elevated when the PDO decadal phase is positive. The decadal variability of the PDO is apparent in its time series (http://jisao.washington.edu/pdo/img/pdo_latest.png); being mostly negative from the 1950s up until 1976-77, and positive throughout the 1980s and 1990s. The spatial and temporal patterns of SST variability in the Pacific differ between ENSO and PDO, therefore low simultaneous correlation between ENSO and PDO can be misleading. Anomalous tropical convection induced by ENSO affects global atmospheric circulation and alters surface fluxes over the North Pacific, forcing North Pacific SST anomalies that peak about 2-3 months after the peak of an ENSO event (Figure 14) (Newman et al, 2003). Hence, the decadal phase of the PDO may be dependent on the frequency of ENSO events. Therefore, climate shifts of the PDO such as that of 1976-77 will not be apparent until years after they have occurred, because ENSO may not be predictable beyond about 1-2 years (Newman et al, 2003).
c. Atmospheric forcing of the PDO of North Pacific origin The North Pacific index (NPI), a measure of the strength of the Aleutian Low at sea level, leads the PDO on monthly and annual time scales (Newman et al, 2003; Schneider and Cornuelle, 2005). In this paper, the NPI is linked with the PNA by virtue of the Aleutian Low; they are anti-correlated at a value of about -0.8 averaged over the DJF winter season (Figure 15). In the mid-latitudes, the North Pacific SSTs have a long-term memory during the cold season. The oceanic mixed layer temperature anomalies from one winter decouple from the surface during summer, when westerlies weaken and insolation strengthens, and reemerges the following winter as these anomalies are entrained back into the mixed layer. This results in a yearly cycle such that a growing ENSO forces the PDO towards the winter season. Consequently, the North Pacific SST anomalies persist well into spring and early summer.
Other modes of atmospheric variability, such as the PNA, can also force the PDO. While the simultaneous DJF correlation between PNA and PDO is about 0.8, the highest PDO-PNA correlations are found when the PDO lags the PNA (Figure 16). This lead-lag finding confirms that the PNA and its evolution is one of the drivers of the PDO on subseasonal time scales, as shown by Nigam (2003) that a PNA event pumps a positive PDO with surface winds around the Aleutian Low - see figure of the impact on North Pacific SST by PNA evolution (Nigam, 2003).
d. Prediction of the PDO on various time scales Well-known factors contributing to PDO variability on interannual time scales include atmospheric noise, ENSO, and the NPI by virtue of the PNA. ENSO and NPI both yield about equal weights for forcing of PDO variability, but with opposite signs. Newman et al (2003) also suggests that the atmospheric component of the North Pacific decadal variability (NPI) is weaker than the SST component (PDO). The variability of the PDO is well modeled as the sum of direct forcing by ENSO, the reemergence of North Pacific SSTs in subsequent winters, and atmospheric forcing from climate noise. However, Schneider and Cornuelle (2005) have added the zonal extension of the Kuroshio-Oyashio Current as an equal contributor along with ENSO and NPI/PNA to PDO variability on decadal time scales, and having about one-third of the weight compared to ENSO and NPI on interannual time scales. According to Schneider and Cornuelle (2005), a strong Kuroshio-Oyashio current leads to warm SST advection into the central North Pacific, leading to negative PDO values. To forecast changes in the PDO, predictions of the NPI, ENSO, and the strength of the Kuroshio-Oyashio current are required. The relative importance of each forcing process depends on the frequency and time scale (Schneider and Cornuelle, 2005). For time scales shorter than one year, internal atmospheric variability dominates, so the PNA could be used to predict intraseasonal changes in the PDO by using lead-lag analyses. At interannual time scales on the order of 1-2 years, ENSO variability and changes in the Aleutian Low become most important because when the Aleutian Low deepens (negative NPI), it pumps a PNA ridge that contributes to a positive PDO. Predictive skill on decadal time scales for wintertime North Pacific SSTs has been documented for the Kuroshio-Oyashio current and the propogation of oceanic Rossby waves, but it only accounts for a small part of interannual PDO variability. ENSO forecasts have skill at lead times of 1-2 years (Barnston et al, 1999; Chen et al, 2004). However, the component of NPI variability unaffected by ENSO has very limited predictability due to the chaos of internal atmospheric variability. Remote forcing from the western Pacific or the Indian Ocean has yet to be determined (Deser et al, 2004; Schneider and Cornuelle, 2005). Thus, using ENSO, it is possible to predict PDO variability with a lead time of up to two years, but is much more difficult to create a skillful forecast of the PDO on decadal time scales. 4. High Latitude Blocking in the North Pacific a. Overview of North Pacific Blocking North Pacific blocking is another factor involved in influencing temperature anomalies over the North American continent during winter. Examination of blocking episodes is required to determine what influences their strength and frequency. Preferred regions for mid-latitude blocking include the northeast boundaries of the Pacific and Atlantic Oceans, to the north and east of the storm tracks. However, since the influence of the North Pacific on North American winter climate is the focus of this paper, only North Pacific blocking events will be considered in this section. The frequency and impact of North Pacific mid-latitude blocking had been examined with respect to ENSO and PNA by several studies (Renwick and Wallace, 1996; Chen and Yoon, 2002; Huang et al, 2004; Carrera et al, 2004). North Pacific blocking refers to the interruption of zonal flow by strong and persistent meridional flow both upstream and downstream of the blocking ridge, and an equatorward shift of westerly flow in the subtropical North Pacific (Carrera et al, 2004). A downstream trough usually forms over western North America and a ridge forms off the U.S. southeast coast (Figure 17). When the atmosphere forms and maintains a large ridge or anticyclone, zonal flow and its associated storm tracks upstream is obstructed by the “blocking anticyclone” or “blocking high”. A blocking high obstructs the eastward progression of synoptic disturbances, forcing weather systems to the north or south of the block. This results in an anomalous storm track or jet stream, sometimes resulting in a split-flow configuration. The presence of a blocking anticyclone results in anomalous temperature and precipitation patterns. Warmer and wetter weather would occur on the western flank of the block due to poleward advection of warm air and rising motion. Conversely, colder and drier weather would occur on the eastern flank of the block due to equatorward advection of cold air and sinking motion.
Although the criteria used to identify blocking events may differ among researchers, each blocking event has been found to last 1-3 weeks (Renwick et al, 1996; Huang et al, 2004; Carrera et al, 2004), but the onset and breakdown of each event may only last 2-3 days. The onset and maintenance of blocking events are thought to involve: 1) large scale low-frequency Rossby waves, 2) high-frequency transient baroclinic waves, and 3) non-linear interactions between the two aforementioned types of waves. Yoon and Chen (2002) concur that two basic factors are required for the formation and maintenance of blocking; large scale environmental flow involving planetary waves, and internal dynamics of the atmosphere including transient eddies. The development of a blocking anticyclone of equivalent barotropic (warm core) structure involves enhanced upstream baroclinic wave activity, leading to an alteration of the background circulation state by an anomalous poleward shift of the Pacific storm track on the upstream side of the developing block (Renwick et al, 1996). The poleward advection of heat and anticyclonic vorticity then serves to build a ridge that may eventually develop into a blocking anticyclone. The maintenance of a blocking high involves poleward advection of anticyclonic vorticity by high-frequency disturbances against dissipitative forces, such as friction, turbulence, and Ekman pumping by the boundary layer. The study done by Huang et al (2003) concurs that blocking can be maintained by internal dynamics of the atmosphere against dissipitative forces due to low sensitivity to the boundary layer. b. Downstream Weather Impacts of North Pacific Blocking As Carrera et al (2004) explores the downstream weather impacts of North Pacific blocking, the evolution of blocking events is considered in detail from onset to decay. His study displays a temporal evolution of 500-hPa geopotential heights and anomalies prior to the onset of a composite North Pacific blocking event (Figure 18). Prior to the onset of the block, an upstream trough situates itself west of Alaska and negative sea level pressure (SLP) anomalies occupy the Gulf of Alaska. This trough deepens as a ridge strengthens and slowly retrogrades westward from western North America into the Gulf of Alaska over the course of five days prior to the onset of the block. Positive SLP anomalies in the northeast Pacific strengthen and drift northward into the Gulf of Alaska (figure not shown), their eventual position being underneath the blocking anticyclone, consistent with the equivalent barotropic warm core structure of blocking events (Carrera et al, 2004). The study done by Carrera et al (2004) has the potential to provide guidance to forecasters at the NOAA Climate Prediction Center (CPC) for producing 6-10- and 8-14-day probabilistic outlooks for large-scale areas given a blocking scenario forecast on numerical models, so that they are able to express possible deviations of temperatures and precipitation from the climatological forecast of 33.3% of each tercile category; below, neutral, and above climatology. As the block strengthens and matures over the Gulf of Alaska, a longwave trough deepens over central North America while a flat ridge develops over the southeast U.S. The development of a trough over the subtropical Pacific is noteworthy due to the equatorward shift of the Pacific storm track and its associated southwesterly moisture transport into the U.S. southwest. This particular anomalous moisture transport in this region is known as the “Pineapple Express” due to the moisture originating near Hawaii (Figure 19). This sudden influx of moisture can lead to heavy precipitation events along the U.S. west coast. Also, due to the ridge off the U.S. southeast coast, anomalous moisture flux originating from the Gulf of Mexico converges into the Ohio Valley and U.S. southeast. Regions such as southern California, U.S. Rockies, Ohio Valley, and the southwest U.S. all acquire at least a 15% chance of seeing precipitation greater than the 90th climatological percentile using a threshold of 1 mm day-1 (Carrera et al, 2004), meaning these regions have an elevated chance of heavy precipitation during blocking events regardless of the precipitation threshold used. Blocking events not only impact precipitation over the United States, but also impact temperatures over most of the North American continent. Colder temperatures become widespread via northwesterly winds over northwestern Canada southeastward across south-central US, while western Alaska and northeast Siberia experience anomalously warm temperatures due to southerly winds around the blocking high (Figure 19). During about 70% of blocking days, anomalously cold temperatures occur from the British Columbia region to the Northern Plains with a raised possibility of extremely cold temperatures, compared with a 33.3% chance given by a null hypothesis. During blocking events, southeast Alaska, Yukon territories, and south-central U.S. also has a 50% chance of seeing temperatures below climatology.
As the block decays, both the blocking ridge and the downstream trough retrograde westward. The retrogressing trough dips back into the Alaskan region, building a renewed downstream ridge over the west coast of North America as well as a trough over the eastern U.S (Figure 20). This decay phase involves a deepening negative SLP anomaly in the Gulf of Alaska involving explosive cyclogenesis, hence shifting the Pacific storm track poleward out of the anomalous equatorward position. Moisture influx returns to its more poleward trajectory from the U.S. southwest to the Pacific Northwest and western Canada (Carrera et al, 2004), around the building anticyclone over the western U.S. The main positive SLP anomaly over Alaska splits in two, one high pressure system retrograding westward towards the northwest Pacific, and the other spreading southeastward over the central United States. The high pressure systems that spread southeastward across the United States when a North Pacific block decays are arctic anticyclones associated with cold air outbreaks in the central United States. The renewed Aleutian Low and western US ridge after the blocking high decays brings the PNA index from negative to positive values. This implies that strong blocking winters may have as many positive PNA episodes as negative ones, whereas during weak blocking winters less negative-to-positive PNA transitions occur due to a smaller number of blocking episodes. This may be the reason why ENSO-neutral and weak La Nina winters see more positive PNA episodes than negative ones (Figure 7), despite the weakly positive PNA-ENSO correlation on seasonal time scales.
It also must be understood that the aforementioned downstream weather impacts of North Pacific blocking depend on the location of the blocking high and downstream trough (Renwick et al, 1996; Carrera et al, 2004). Hence, the exact location of a blocking anticyclone is as important as the event itself, so the factors that influence the frequency of blocking events and the ultimate location of a blocking high will now be considered. c. ENSO influence on North Pacific Blocking Several studies have examined the influence of ENSO on mid- and high-latitude blocking in the North Pacific (Renwick and Wallace, 1996; Chen and Yoon, 2002; Huang et al, 2004; Carrera et al, 2004). Statistical evidence supports the conclusion agreed upon many studies that the frequency of blocking events each winter is negatively correlated with the ENSO phase. During La Nina and ENSO neutral winters, more blocking events occur and each event lasts longer on average as well. The frequency of blocking events and the duration of each decrease drastically during El Nino winters (Figure 21). Carrera et al (2004) looked at 37 blocking events for the winters over the period of 1979-2000, and they lasted an average of 11.3 days each (ranging from 8 to 25 days). El Nino winters of the study period had an average of only twelve days of blocking, while La Nina winters had an average of 27 days of blocking. Interestingly, it was ENSO neutral winters that had the most days of blocking (31 days), indicating asymmetry and perhaps nonlinearity within the relationship between ENSO and blocking (Figure 22). Although the numbers differed slightly due to the criteria for selecting blocking events and the time range of study, the fact that more blocking occurs during La Nina and ENSO neutral winters than El Nino winters has also been corroborated by other studies (Figure 23) (Renwick and Wallace, 1996; Chen and Yoon, 2002; Huang et al, 2004).
The physical and dynamic explanation for less blocking during El Nino winters is straightforward. The Aleutian Low and the Pacific Jet both tend to be stronger during El Nino winters, reducing the probability of a blocking high entrenching itself over the Gulf of Alaska. At this point, it should be noted from Figure 22 that strong El Nino winters of 1982-83, 1986-87, and 1991-92 saw no blocking events according to the criteria used by Carrera et al (2004). Otherwise, any blocking during El Nino winters would be shifted eastward so that blocking highs would be over the British Columbia coast rather than over the Aleutian Islands, where geopotential heights tend to be suppressed as shown in Figure 21. This suggests that ENSO and PNA are positively correlated, albeit marginally, only by their respective links with wintertime blocking frequency, as one can infer from the time series in Figure 24. During La Nina and ENSO neutral winters, the Pacific Jet becomes weaker and the atmospheric circulation more meridional while the Pacific storm track splits in two; one track shifting equatorward from the jet’s climatological position owing to blocking highs over the North Pacific between Hawaii and California. Using a time range of 1950-1999, Huang et al (2004) created a list of 10 strong blocking winters and compared it against a list of 11 weak blocking winters, and five of the 10 strong blocking winters were La Nina years, and the rest were ENSO-neutral. Using these lists created by Huang et al (2004), Figure 25 shows the composite 500-hPa geopotential height anomalies, as well as temperature and precipitation anomalies, for strong blocking winters and weak blocking winters, confirming that more blocking occurs during La Nina and ENSO-neutral winters and less so for El Nino winters. Only one El Nino winter (1951-52) was a strong blocking one, but the PDO had been negative that winter. Conversely, nine of the 11 weak blocking winters were El Nino winters. Hence, the extratropical atmospheric responses to warm and cold ENSO events are different and somewhat asymmetrical.
d. Decadal Trend of North Pacific Blocking North Pacific SSTs also have an influence in the frequency and duration of blocking events over the North Pacific. Carrera et al (2004) found that the effect of North Pacific SST is to shift the preferred location of blocking without any influence on the dynamics involved and the persistence of each blocking event. Huang et al (2003) has corroborated this conclusion by Carrera by showing the results of a general circulation modeling study ascertaining whether SSTs influence blocking events. North Pacific SSTs do not affect the frequency and strength of blocking events; however, they do influence the preferred location of the blocking anticyclone. When the equatorial Pacific SSTs are warm (as during an El Nino) along with a positive PDO, the blocking is shifted eastward from the Aleutians to the British Columbian coast, and the downstream trough is also shifted eastward resulting in cold weather over the eastern US. During ENSO-neutral and La Nina winters, the blocking anticyclone’s preferred position remains in the vicinity of the Aleutian Islands. There has been a noticeable decadal trend of an increasing number of blocking days each winter. Chen and Yoon (2002) used a 42 year dataset of all winters within the 1954-1997 time range to find the increasing trend of blocking as well as a spatial shift in the position of blocking anticyclones. From 1954 to 1997, the number of blocking days has increased by 6.5 days, while the preferred position of North Pacific blocking shifted eastward by 8.7 degrees longitude, from the Aleutian Islands to near the British Columbia coast (Figures 26 & 27). This eastward shift is consistent with the higher frequency of warm ENSO events and positive PDO winters, especially since the 1976-77 climate regime shift. It was thought by Chen and Yoon (2002) that the PDO modulates North Pacific blocking by assisting dynamic processes that maintain the wave train arcing through the North Pacific and North America. However, section 3 of this paper shows that this is not the case, because the PDO is not its own dynamical mode but is forced by the PNA as well as the surface reflection of North Pacific blocking itself. One caveat must be made here; when the ENSO and PDO are out of phase, they tend to conflict in terms of the preferred position of blocking. Consequently, in these cases it is difficult to predict the resultant temperature and precipitation anomalies in North America as the exact location of a blocking high heavily influences sensible weather downstream of the block.
North Pacific blocking exhibits some predictability on seasonal-interannual time scales in a statistical sense, though individual blocking events are difficult to predict (Renwick et al, 1996). However, numerical weather prediction (NWP) modeling is poorly skilled at simulating the onset and evolution of blocking events. NWP models frequently underestimate the strength of the blocking anticyclone as well as the frequency and duration due to difficulties transitioning from zonal flow to a blocked state with relatively stronger meridional flow (Carrera et al, 2004). However, NWP models seem to perform better in maintaining blocked flow when the blocking high is present in the initial conditions used for the model runs, pointing to problems forecasting atmospheric pattern changes. Ensemble forecasting has been able to somewhat improve forecasting of blocking events. Interdecadal changes in North Pacific blocking have implications for assessing the predictability of wintertime blocking events. Most precipitation in the Pacific Northwest occurs in winter, and a decreasing trend precipitation in precipitation has emerged for this region in the last four decades. This alarming trend is consistent with the eastward shift of the preferred position of North Pacific blocking towards the Pacific Northwest and the Western Canadian coast, which has become warmer and drier in recent decades. Proper model simulation of planetary waves is also important for successful forecasts of blocking events. Since the processes that modulate the PDO may also modulate North Pacific blocking over decadal time scales, forecasts of blocking episodes may therefore be affected by model biases of the PDO and its related atmospheric variability. The effects of the deepening of the Aleutian Low on the PDO may enhance cyclogenesis over the North Pacific-Alaskan region, thus blocking events may channel storms further north or south, away from the climatological west-east storm track. The presence of blocking highs implies stormier conditions for regions southward of the Pacific Northwest. The effects of PNA variability and ENSO variability on the Pacific jet regulate the Pacific storm track. Internal atmospheric variability and transient baroclinic eddies also contribute to maintaining the PNA structure as well as to the frequency and duration of blocking events. Being able to quantify contributions by both low-frequency and high-frequency forcing of blocking events hopefully will advance understanding and modeling of atmospheric variability on weekly time scales. Coming soon... Anderson J, van den Dool H, Barnston A, Chen W, Stern W, et al. (1999) Present-Day Capabilities of Numerical and Statistical Models for Atmospheric Extratropical Seasonal Simulation and Prediction. Bulletin of the American Meteorological Society: Vol. 80, No. 7 pp. 1349–1361 Barnston AG, He Y, Glantz MH (1999) Predictive Skill of Statistical and Dynamical Climate Models in SST Forecasts during the 1997—98 El Niño Episode and the 1998 La Niña Onset. Bulletin of the American Meteorological Society: Vol. 80, No. 2 pp. 217–243 Barnston AG, Kumar A, Goddard L, Hoerling MP (2005) IMPROVING SEASONAL PREDICTION PRACTICES THROUGH ATTRIBUTION OF CLIMATE VARIABILITY. Bulletin of the American Meteorological Society: Vol. 86, No. 1 pp. 59–72 Barnston AG, Livezey RE (1987) Classification, Seasonality and Persistence of Low-Frequency Atmospheric Circulation Patterns. Monthly Weather Review: Vol. 115, No. 6 pp. 1083–1126 Carrera ML, Higgins RW, Kousky VE (2004) Downstream Weather Impacts Associated with Atmospheric Blocking over the Northeast Pacific. Journal of Climate: Vol. 17, No. 24 pp. 4823–4839 Chen TC, Yoon Jh (2002) Interdecadal Variation of the North Pacific Wintertime Blocking. Monthly Weather Review: Vol. 130, No. 12 pp. 3136–3143 Chen WY (2004) Significant Change of Extratropical Natural Variability Associated with Tropical ENSO Anomaly. Journal of Climate: Vol. 17, No. 10 pp. 2019–2030 Deser C, Phillips AS, Hurrell JW (2004) Pacific Interdecadal Climate Variability: Linkages between the Tropics and the North Pacific during Boreal Winter since 1900. Journal of Climate: Vol. 17, No. 16 pp. 3109–3124 Goddard L, Kumar A, Hoerling MP, Barnston AG (2006) Diagnosis of Anomalous Winter Temperatures over the Eastern United States during the 2002/03 El Niño. Journal of Climate: Vol. 19, No. 21 pp. 5624–5636 Gutzler DS, Kann DM, Thornbrugh C (2002) Modulation of ENSO-Based Long-Lead Outlooks of Southwestern U.S. Winter Precipitation by the Pacific Decadal Oscillation. Weather and Forecasting: Vol. 17, No. 6 pp. 1163–1172 Hoerling MP, Ting M (1994) Organization of Extratropical Transients during El Niño. Journal of Climate: Vol. 7, No. 5 pp. 745–766 Huang F, Zhou F, England MH (2004) Atmospheric Circulation Associated with Anomalous Variations in North Pacific Wintertime Blocking. Monthly Weather Review: Vol. 132, No. 5 pp. 1049–1064 Kumar A, Barnston AG, Peng P, Hoerling MP, Goddard L (2000) Changes in the Spread of the Variability of the Seasonal Mean Atmospheric States Associated with ENSO. Journal of Climate: Vol. 13, No. 17 pp. 3139–3151 Kumar A, Hoerling MP (2000) Analysis of a Conceptual Model of Seasonal Climate Variability and Implications for Seasonal Prediction. Bulletin of the American Meteorological Society: Vol. 81, No. 2 pp. 255–264 Leetmaa A (2003) Seasonal Forecasting: Innovations in Practice and Institutions. Bulletin of the American Meteorological Society: Vol. 84, No. 12 pp. 1686–1691 Lejenäs H, Madden RA (1992) Traveling Planetary-Scale Waves and Blocking. Monthly Weather Review: Vol. 120, No. 12 pp. 2821–2830 Mason SJ, Goddard L (2001) Probabilistic Precipitation Anomalies Associated with ENSO. Bulletin of the American Meteorological Society: Vol. 82, No. 4 pp. 619–638 Newman M, Compo GP, Alexander MA (2003) ENSO-Forced Variability of the Pacific Decadal Oscillation. Journal of Climate: Vol. 16, No. 23 pp. 3853–3857 Nigam S (2003) Teleconnections. Encyclopedia of Atmospheric Sciences. pp. 2243-2269 Quiroz RS (1983) The Climate of the “El Niño” Winter of 1982–83—A Season of Extraordinary Climatic Anomalies. Monthly Weather Review: Vol. 111, No. 8 pp. 1685–1706 Quiroz RS (1984) The Climate of the 1983–84 Winter—A Season of Strong Blocking and Severe Cold in North America. Monthly Weather Review: Vol. 112, No. 9 pp. 1894–1912 Renshaw AC, Rowell DP, Folland CK (1998) Wintertime Low-Frequency Weather Variability in the North Pacific–American Sector 1949–93. Journal of Climate: Vol. 11, No. 5 pp. 1073–1093 Renwick JA, Wallace JM (1996) Relationships between North Pacific Wintertime Blocking, El Niño, and the PNA Pattern. Monthly Weather Review: Vol. 124, No. 9 pp. 2071–2076 Schneider N, Cornuelle BD (2005) The Forcing of the Pacific Decadal Oscillation*. Journal of Climate: Vol. 18, No. 21 pp. 4355–4373 |
|
|
Home |
About
WLH |
Contact WLH
|
Site
Map |
|
Figure 8
– Upper level divergence associated with DJF-averaged PNA (1959-1996); the
0.2 sigma level corresponds to about 200 hPa. CI = 0.25. Left: Correlation between PNA
and 0.2 sigma divergence. Right: Regression of 0.2 sigma divergence on the
DJF-averaged PNA index. 
