Abstract

Most investigations of relationships between tropical Pacific sea surface temperature anomaly (SSTA) events and regional climate patterns have assumed the teleconnections to be linear, whereby the climate patterns associated with cold SSTA events are considered to be similar in structure and morphology but opposite in sign to those linked to warm SSTA events. In contrast, and motivated by early evidence of nonlinearity in the above regard, this study identifies characteristic (i.e., composite) calendar monthly central and eastern North American precipitation patterns separately for warm and cold SSTA events in different regions of the tropical Pacific (central, eastern, west-central“horseshoe,” far western) identified through principal component analysis. The precipitation anomaly patterns are computed from an approximately 1° lat–long set of monthly station data for 1950–92. Their robustness and nonlinearity are established using local, regional, and field statistical significance tests and a variance analysis.

This combination of unique SSTA analyses, resulting composite selection, and characteristic precipitation anomaly determination from a fine-resolution dataset increases our understanding of tropical Pacific–North American precipitation teleconnections in several respects. First, significant linkages to the two SSTA modes related to traditional warm and cold events (central and eastern tropical Pacific) are identified for all months except September and October, with all exhibiting some nonlinear characteristics. The most important of those nonlinearities involve associations with eastern tropical Pacific SSTAs, which affect precipitation near the southern Atlantic and Gulf of Mexico coasts (dry for cold Novembers), around the Great Lakes and in the Ohio River valley (dry, warm, January–February), in the southeastern United States (dry, warm, July–August), and across the northern Great Plains (dry, warm, November–January). Conversely, the regions confirmed to have essentially linear associations with traditional warm and cold events include the Gulf of Mexico coast (positive relation with eastern tropical Pacific, January–March), Ohio River valley (negative, central tropical Pacific, February), and mid-Atlantic coast (negative, eastern tropical Pacific, July–August). However, only nonlinear precipitation teleconnections are associated with SSTAs in tropical Pacific regions largely unrelated to ENSO. These principally involve anomalously dry conditions in much of the eastern half of the United States during January–March and in the central United States in July–October (warm SSTAs in west-central tropical Pacific horseshoe), and in a strip from Texas to New England in January and along the central gulf coast and lower Mississippi valley in April (warm SSTAs in far western tropical Pacific). The results thus demonstrate the sensitivity of central and eastern North American precipitation teleconnections to the location and extent of tropical Pacific SSTAs. In the appendix, the present results are also compared to the observed climate anomalies during the 1997–98 El Niño event.

1. Introduction

During the last 15 years, a number of studies have linked tropical Pacific sea surface temperature anomaly (SSTA) events to global climate patterns on a variety of timescales (reviewed, e.g., in Glantz et al. 1991; Allan et al. 1996; Montroy 1997, henceforth M97). For example, Ropelewski and Halpert (hereafter RH; 1986, 1987, 1989, 1996) used a harmonic dial analysis of the Southern Oscillation index (SOI) to identify regional seasonal precipitation and temperature patterns related to tropical Pacific SSTA events, and documented shifts in the precipitation distributions in each of the areas concerned. Earlier, Kiladis and Diaz (1989, hereafter KD89) delineated regions exhibiting a linear relation with tropical Pacific SSTAs and the SOI by compositing 3-month global precipitation anomalies based on warm El Niño–Southern Oscillation (ENSO) events and subtracting counterpart composite fields based on cold ENSO events. These seasonal studies were recently complemented by the monthly analysis of M97, which correlated three leading modes of tropical Pacific SSTA variability with monthly North American precipitation anomalies. That work identified both the months during which the seasonal relationships of the above studies have the strongest signal, and also additional teleconnection relations that could not be identified using seasonal anomalies.

Whereas many previous studies (including KD89 and M97) have used the assumption that the teleconnection relationships associated with cold tropical Pacific SSTA events are similar in structure and morphology and opposite in sign to those linked to warm SSTA events (i.e., a linear relationship with SSTAs), recent work has demonstrated that removing the constraint of linearity and analyzing warm and cold SSTA events separately can be informative. The RH papers outline different seasonal global precipitation anomaly patterns linked to low-SOI (i.e., warm SSTA) and high-SOI (i.e., cold SSTA) events. Additionally, Livezey et al. (1997, hereafter L97) composited monthly fields of U.S. surface temperature and precipitation and 700-mb height based on warm and cold SSTAs in a small, but highly sensitive, region of the central equatorial Pacific (180°–150°W and 5°N/S), and noted that the fields associated with warm and cold SSTA events there have different structures. Furthermore, Hoerling et al. (1997) have added to the dynamic foundation for nonlinear analyses by showing that observed and modeled 500-mb northern Pacific circulation anomalies linked to warm and cold SSTA events are separated by about 35° longitude.

With the above observational and dynamical motivation for separately examining North American precipitation patterns related to warm and cold tropical Pacific SSTA events, the present study seeks to extend previous research by using a unique precipitation dataset, different analysis techniques, and four contrasting SSTA regions from which to identify associated teleconnections. In particular, the present work complements the contribution of L97, which was largely conducted independently of our effort. While L97 uses the irregularly spaced monthly precipitation averages for U.S. climate divisions, a spatially homogeneous fine-resolution set of monthly station precipitation totals spanning the eastern two-thirds of North America (including southern Canada) is used here. Additionally, the classification methodology for the two investigations are quite complementary. L97 classified months as warm or cold based on one of two possible average SSTA thresholds for a single, small, central equatorial Pacific region to which fluctuations in tropical rainfall (and hence convective forcing) are known to be particularly sensitive. Specifically, they computed composites based on the threshold yielding a sample size of 4–10 yr (specified a priori) or having the highest statistical significance when the sample sizes from both thresholds fell in that range. Here, in contrast, larger-scale modes of SSTA variability that could substantially impact tropical Pacific convection are identified and teleconnections to North America sought using a classification scheme applied to principal component score time series, with no adjustments needed for sample size. Finally, the present study uses more than one SSTA mode (four) to identify associated monthly teleconnections, with each mode emphasizing variability in a different portion of the tropical Pacific. Only M97 has previously performed an analysis of this type, but for linear (not nonlinear) SSTA-precipitation relations. (The appendix compares the results presented in this paper to the observed precipitation anomalies during the 1997–98 El Niño through February 1998.)

2. Data

a. Sea surface temperature

The individual monthly SSTA data used in this study consist of anomalies from calendar monthly means for 1950–92 for 2° × 2° lat–long grid cells between 19°N and 19°S and from 119°E to the west coast of the Americas. The same data were used in M97 (his Fig. 2a), but here the domain stretches 20° longitude farther west to include far western tropical Pacific SSTA variability in the analysis. These data were extracted from the Climate Prediction Center’s global reconstructed sea surface temperature (SST) dataset. Full details on the set’s construction are provided in Smith et al. (1996); the rationale for its present use follows M97. This is the same set L97 used to force a global climate model in an attempt to reproduce monthly teleconnection relationships.

b. North American precipitation

The individual monthly precipitation data used here for 1950–92 were totaled from daily values for 766 sites spanning the eastern two-thirds of North America south of about 56°N at an approximate 1° lat–long resolution (Fig. 1). This dataset, which was also used by M97 and Gong and Richman (1995), constitutes an extension (from the western edge of the Appalachian Mountains to the Atlantic coast, to all months of the year, and through 1992) of a previously developed May–August 1949–80 daily precipitation dataset for central North America. This set’s construction procedures are fully described in Richman and Lamb (1985, 1987). No missing station values are present, due to the substitution of daily values from a nearby secondary station when observations from a primary station were unavailable (less than 15% of the record).

Fig. 1.

Locations of North American precipitation stations used.

Fig. 1.

Locations of North American precipitation stations used.

3. Methodology

a. Identifying SSTA modes

To identify coherent modes of variability in tropical Pacific SSTAs, a principal component (PC) analysis was applied to the data for all calendar months combined. This procedure identified the few robust modes present throughout the year. Following M97, unrotated PCs (UPCs) were first derived from the inter-grid-cell correlation matrix in an S-mode sense, yielding a set of spatial loading patterns and their associated time series of PC scores. Based on truncation tests described and utilized by M97—scree test (Cattell 1966), eigenvalue separation criterion (North et al. 1982), and “point teleconnection patterns” (Richman 1986)—three modes explaining 59% of the SSTA variance were determined to contain nonrandom signal. These modes were retained and rotated to the Varimax criterion (Kaiser 1958), as PC rotation has been demonstrated to yield spatial loading patterns that are typically more representative than their UPC forerunners of the variability in a dataset (Horel 1981; Richman 1986). After examining the three UPCs and the three Varimax-rotated PCs (VPCs), four of those UPC or VPC modes were selected as potential bases for teleconnections with central and eastern North American precipitation. The spatial loading patterns and associated PC scores for these modes are given in Figs. 2 and 3, respectively, and are described below.

Fig. 2. Spatial loading patterns for leading modes of tropical Pacific SSTAs: (a) unrotated PC 1, (b) Varimax-rotated PC 1, (c) Varimax-rotated PC 2, and (d) Varimax-rotated PC 3. Contour interval is 0.2, with negative contours dashed. Percent variance explained by mode is given in lower-right corner. Key region used by L97 is shaded in yellow.

Fig. 2. Spatial loading patterns for leading modes of tropical Pacific SSTAs: (a) unrotated PC 1, (b) Varimax-rotated PC 1, (c) Varimax-rotated PC 2, and (d) Varimax-rotated PC 3. Contour interval is 0.2, with negative contours dashed. Percent variance explained by mode is given in lower-right corner. Key region used by L97 is shaded in yellow.

Fig. 3.

Score time series associated with PC loading patterns given in Fig. 2, with year 0 of warm (cold) ENSO events catalogued in Kiladis and van Loon (1988) denoted by “E” (“L”).

Fig. 3.

Score time series associated with PC loading patterns given in Fig. 2, with year 0 of warm (cold) ENSO events catalogued in Kiladis and van Loon (1988) denoted by “E” (“L”).

1) Unrotated PC 1

Of the UPCs, only the first (UPC1) captures a substantial mode of SSTA variation (37.4% of variance). The loading pattern for UPC1 (Fig. 2a) emphasizes variability (loadings ≥ 0.6) spanning 20°S–15°N and stretching from the coast of South America to the international date line. This SSTA pattern reflects traditional ENSO events, as it closely resembles the warm SSTA pattern in Rasmusson and Carpenter’s (1982)“transition” and “mature” phases of El Niño evolution, while the corresponding score time series (Fig. 3a) reaches extrema during warm and cold SSTA events cataloged by Kiladis and van Loon (1988). Additionally, both Hsuing and Newell’s (1983) first unrotated empirical orthogonal function (EOF) of 1949–79 SSTAs and the first UPC of M97’s analysis that did not extend west of 139°E, emphasized central tropical Pacific SSTAs in a region similar to UPC1. The validity of this UPC1 mode is further supported by Richman (1987), who demonstrated that the first UPC may have physical meaning if the scale of the phenomenon captured is comparable to the scale of the domain, as here.

2) Varimax PC 1

In addition to UPC1, the first Varimax PC (VPC1) also represents an SSTA mode associated with traditional ENSO events. Its spatial loading pattern (Fig. 2b), explaining 32.2% of the SSTA variance, is broadly similar to that of UPC1, but with greater emphasis on the eastern Pacific where the loadings are generally larger than those in the UPC1 pattern. This pattern is strongly analogous to Rasmusson and Carpenter’s (1982) “transition” El Niño phase, which depicts a warm pool off the Ecuador–Peru coast and an associated warm tongue stretching to 160°W that weakens by 160°E. This loading pattern also corresponds well to the first nonseasonal, Pacific SSTA EOF identified by Weare et al. (1976) and the first VPC of M97’s analysis, which focused on SSTAs east of 139°E. Moreover, VPC1’s score time series (Fig. 3b) indicates high (low) values occurring during almost all of the documented warm (cold) events, as in the time series associated with SSTA UPC1.

3) Varimax PC 2

Another important SSTA mode that is somewhat related to traditional ENSO events is given by VPC2. Its loading pattern (Fig. 2c), explaining 15.8% of the SSTA variance, emphasizes variability from 15°N, 140°W to 0°, 160°E to 15°S, 160°W (hereafter termed the west-central tropical Pacific “horseshoe”). While VPC2’s spatial structure does not correspond to any of the phases of El Niño evolution identified by Rasmusson and Carpenter (1982), it is similar to that of the second of two rotated EOFs derived from 1982–87 SSTAs by Ropelewski et al. (1992), and the second of six VPCs obtained by M97 for an SSTA domain that extended only to 139°E. Because the associated score time series (Fig. 3c) reaches only extrema during some documented warm (1963, 1969, 1982, 1986) and cold (1964, 1973, 1975) events, it is plausible that the VPC2 mode captures a westward extension of SSTAs occurring in these subsets of events. Moreover, relatively high equatorial loadings (≥+0.6) span 155°–170°E, a region which Fu et al. (1986) showed to have climatological mean SST values around 28.5°C, a value commonly used as an indicator of overlying convection. This suggests the possibility of a close relationship between positive SSTAs and convective forcing in the west-central Pacific horseshoe, which could effect teleconnections with midlatitude weather systems (e.g., Kang and Lau 1987).

4) Varimax PC 3

An SSTA mode related to a trend in far western tropical Pacific SSTA is represented by VPC3. The loading pattern for this mode (Fig. 2d) explains 11.0% of the SSTA variance, primarily in the extreme western tropical Pacific north of Australia and Indonesia (between 15°N/S and 120°–160°E). This region was demonstrated by Gutzler (1995) to have experienced a warming trend in SSTs during 1973–93. Such a warming trend is also implicit in the associated VPC3 score time series between the mid-1970s and late 1980s (Fig. 3d). That time series further demonstrates that this mode is not directly related to ENSO, with no extrema during documented events. However, this mode could have the potential for teleconnections with midlatitude climate, since the long-term mean SSTs in the area of highest PC loading values are around the convection-sensitive 28°–29°C in all calendar months (Fu et al. 1986). No previous study has investigated monthly teleconnection relationships with an SSTA mode in this region.

b. Case selection

Warm and cold event months were identified for different regions of the tropical Pacific using the SSTA modes described in section 3a. Thresholds of +0.97 (−0.97) in the PC score time series were used to select events, thereby designating the warmest (coldest) 1/6 of months in each event category for each SSTA PC. These thresholds reflect the normalized nature of the PC scores. This method of identifying events differs considerably from those used by earlier authors (e.g., KD89;RH papers) who classified entire years as warm or cold, as opposed to the present variable-length sequences of months. Our approach also accommodates the initiation and termination of ENSO events during different calendar months. Furthermore, while L97’s similar identification of warm or cold event months was based on one key SSTA region, this work broadens the focus to several different SSTA regions (of which only UPC1 includes maximum loadings in L97’s key area, see Fig. 2), as M97 showed that the linear teleconnections with North American precipitation vary as a function of the location of the SSTAs.

c. Compositing

Using the sets of warm and cold event months so identified for each SSTA PC, composites of calendar monthly station precipitation anomalies were formed based on 1950–92 calendar monthly means. In addition, composite precipitation patterns were obtained for periods of two and three consecutive months to increase the robustness of the relations when the monthly composite fields varied little from month to month. In the case of the two-month composites, both months had to have been classified as either warm or cold event months according to the above criteria, whereas this condition was only required for two months of the three-month composites. This approach differs from L97’s method, which composited precipitation patterns for two or three consecutive months using their above SSTA thresholds without considering the continuity of the monthly constituents.

d. Significance testing

To assess the significance of the composite fields, a battery of statistical significance tests was employed. The most basic was a field significance test that closely followed those of Livezey and Chen (1983) and L97. This initially involved applying a local t test to the composite values for each precipitation station, to ascertain which values were significantly different from zero at the 90% level. The 43 precipitation fields for each calendar month were then randomly reshuffled 1000 times, the composites recomputed, and local t tests reapplied to the resulting individual station composite means. The field significance was then assessed as the percentage of the reshufflings that had fewer significant stations than the original composite, thus yielding high field significance values when many stations were characterized by locally significant composite values. The composite precipitation results that follow include documentation of both the local (i.e., individual station) and field significance.

To complement these local and field significance assessments, a test was also performed on the nonlinearity of composite precipitation anomalies over regions where those anomalies were especially strong and spatially coherent. For these regions exhibiting substantial signal, spatially averaged composite values based on warm and cold SSTA events were separately computed and then subjected to a t test, giving a “regional significance.” When the spatial composite mean for a region was found to be statistically significant for only one type of SSTA event, or was significant for both events but had the same sign in each, a significant nonlinear relationship was established.

It should be noted that the skewness of the precipitation anomaly distributions can influence the results of these statistical significance tests. The anomalies are of course bounded on the negative side which, in general, reduces the variance of negative precipitation anomaly events and results in higher t values compared to wet events. Consequently, all relations discussed here involving negative precipitation anomalies were required to have either a field significance greater than 90%, a duration of more than one month, or a regional average composite anomaly significant above 95%.

e. Variance explained by composites

In the following presentation of the composite precipitation fields based on each of the SSTA modes described in section 3a, we also document the percent variance of the individual constituent precipitation anomaly patterns that is explained by the composite fields to which they contributed. This quantity is calculated (as in L97) using

 
formula

Here, xi is the observed anomaly during one of the N months used to create the composite, x′ is the composite value for that calendar month, and the angle brackets denote a sum over all stations in the study region. Thus EV complements the statistical significance testing procedures described above.

4. Results

The major North American precipitation relationships associated with the SSTA modes described in section 3a are now presented in sections 4a,b. These are confined to associations that encompass more than one calendar month, which are examined using all of the statistical significance measures and EV described in sections 3d,e. Section 4c deals with single-month relations that are assessed in the same manner.

a. SSTA UPC1 and VPC1

Four regions were found with precipitation exhibiting substantial nonlinear relationships with SSTA UPC1 (central tropical Pacific, Fig. 2a) and VPC1 (eastern tropical Pacific, Fig. 2b): 1) Atlantic–Gulf of Mexico coasts and contiguous inland regions (November–April); 2) northern Great Plains (November–January); 3) southern United States and Great Lakes (May–June);and 4) southeastern United States, Atlantic coast, and Great Plains (July–August). The evidence for these associations is now summarized.

1) Atlantic–Gulf of Mexico coasts and contiguous inland regions (November–April)

Figure 4 provides strong indications that November–March precipitation in the southern, southeastern, and central United States is nonlinearly related to SSTA VPC1 (eastern tropical Pacific, Fig. 2b). For example, in the November composites (Figs. 4a,b; Table 1), while the average precipitation anomalies along the Gulf of Mexico coast are locally/regionally significant and opposite in warm (positive) and cold (negative) events, those near the south Atlantic coast (negative) and in an interior region stretching from Oklahoma to southern Wisconsin (negative) are only strong (and locally/regionally significant) during cold events. Furthermore, for the December composites (Figs. 4c,d), the only locally significant features in the southern part of the domain are clusters of significant stations from Texas to Nebraska for the cold event composite, over which the average composite anomaly, though relatively small, is highly regionally significant (99.3%, Table 1). In January–March, the warm VPC1 composite patterns (Figs. 4e,g,i; Table 1)—which are field significant above 96.7% for January and February, persist into March, and explain substantial fractions (24%–27%) of the domain variance of their constituents—contain well-delineated and locally significant positive coherencies along the Gulf of Mexico and Atlantic coasts (and over Kansas–Nebraska in March) and negative coherencies from north and west of the Appalachian Mountains into the Great Lakes region. While the monthly cold event composites for January–March (Figs. 4f,h,j) are strongly field significant (92.1%–99.8%) and explain a moderate amount (10%–20%) of their constituents’ domain variance, the inland positive coherencies are much smaller, weaker, and less locally significant than the negative anomalies in the corresponding warm event composites, especially for February and March. Figures 4e–j thus suggests a strong nonlinear VPC1 SST–precipitation relationship in this inland region. Further in this regard, the average composite anomalies over northern Wisconsin–Michigan and the Lake Erie region are strongly locally/regionally significant during warm (but not cold) January and February VPC1 events (Figs. 4e,g; Table 1). Although a negative March precipitation anomaly coherency in the central plains for VPC1 cold events has a counterpart positive region in the warm composite, the latter does not extend as far east and is not as locally/regionally significant (Figs. 4i,j; Table 1). This suggests that the central plains March relationship (i.e., for an important agricultural region–month) identified in M97’s linear analysis is most consistent during cold events.

Fig. 4.

Composite monthly precipitation anomalies (mm) for warm and cold SSTA VPC1 events for November–March. Color shading follows legend. An “×” is plotted at stations where composite value is significantly different from zero at 90% level. Field significances (FS, see section 3d) and explained variances (EV, section 3e) appear at bottom of each panel. Constituent years of composites are listed at right. Small maps at sides locate numbered regions in Table 1.

Fig. 4.

Composite monthly precipitation anomalies (mm) for warm and cold SSTA VPC1 events for November–March. Color shading follows legend. An “×” is plotted at stations where composite value is significantly different from zero at 90% level. Field significances (FS, see section 3d) and explained variances (EV, section 3e) appear at bottom of each panel. Constituent years of composites are listed at right. Small maps at sides locate numbered regions in Table 1.

Table 1.

Statistics for average precipitation anomalies over selected regions in composites based on SSTA VPC1, as presented in the indicated figures and sections. Region numbers pertain to shaded areas in small maps along sides of Figs. 4; 6; 7a,b; and 8. For warm and cold events, “CA” denotes the composite regional precipitation anomaly (mm), “FR” the fraction of composite members for which the regional mean anomaly has the same sign as the overall composite anomaly (CA), and “SIG” is the regional significance level (see section 3d). Boldface entries have regional significance above 95%.

Statistics for average precipitation anomalies over selected regions in composites based on SSTA VPC1, as presented in the indicated figures and sections. Region numbers pertain to shaded areas in small maps along sides of Figs. 4; 6; 7a,b; and 8. For warm and cold events, “CA” denotes the composite regional precipitation anomaly (mm), “FR” the fraction of composite members for which the regional mean anomaly has the same sign as the overall composite anomaly (CA), and “SIG” is the regional significance level (see section 3d). Boldface entries have regional significance above 95%.
Statistics for average precipitation anomalies over selected regions in composites based on SSTA VPC1, as presented in the indicated figures and sections. Region numbers pertain to shaded areas in small maps along sides of Figs. 4; 6; 7a,b; and 8. For warm and cold events, “CA” denotes the composite regional precipitation anomaly (mm), “FR” the fraction of composite members for which the regional mean anomaly has the same sign as the overall composite anomaly (CA), and “SIG” is the regional significance level (see section 3d). Boldface entries have regional significance above 95%.

The November–April precipitation relationships based on SSTA UPC1 (central tropical Pacific, Fig. 2a) are generally similar to those for SSTA VPC1 in Fig. 4 for November–March, except during cold February–April events (Figs. 5b,d,f). In the February and March cold UPC1 composite fields (Figs. 5b,d)—which have moderate-to-high field significances (84.8%, 98.9%) and explain 24.8% and 16.3% of the anomaly variance of their constituents—the positive anomalies over Lake Erie and the Ohio River valley (February only) are broadly the inverse of negative anomalies linked to warm UPC1 events (Figs. 5a,c) and are regionally significant in both warm and cold events (Table 2). These cold UPC1 anomalies are larger and stronger than those in their cold VPC1 counterparts (Figs. 4h,j), suggesting much greater UPC1 than VPC1 February–March linearity (cf. Figs. 4g–j). However, Figs. 5c–f and Table 2 also indicate three locally/regionally significant nonlinear relations involving warm UPC1 events and precipitation over northern Wisconsin–Michigan (February, warm = dry/cold = normal), the St. Lawrence valley (March, cold = wet/warm = normal), and the northern and mid-Atlantic coast (April, cold = dry/warm = wet, but not significant).

Fig. 5.

Same as Fig. 4 but for cold SSTA UPC1 events for February, March, and April, and with small maps at sides locating numbered regions in Table 2.

Fig. 5.

Same as Fig. 4 but for cold SSTA UPC1 events for February, March, and April, and with small maps at sides locating numbered regions in Table 2.

Table 2.

As in Table 1 but for SSTA UPC1 events and with region numbers pertaining to shaded areas in Figs. 5 and 7c.

As in Table 1 but for SSTA UPC1 events and with region numbers pertaining to shaded areas in Figs. 5 and 7c.
As in Table 1 but for SSTA UPC1 events and with region numbers pertaining to shaded areas in Figs. 5 and 7c.

2) Northern Great Plains (November–January)

Figures 4 and 6 identify other nonlinear relationships between SSTA VPC1 and November–January precipitation, involving the southern Canadian prairies and Montana–North Dakota. The monthly composite fields for November, December, and January during warm VPC1 events (Figs. 4a,c,e) show locally significant (albeit small) negative anomalies at a number of stations in Montana, North Dakota, and across southern Alberta–Saskatchewan–Manitoba, with December and January having the strongest associations. This linkage is further supported when the precipitation anomalies for these months are composited for warm VPC1 events on a seasonal basis (Fig. 6a). The resulting pattern has high field significance (98.5%), explains a substantial amount (18.8%) of the domain variance of its constituents, and exhibits locally/regionally significant negative coherencies in eastern Montana and the southern Canadian prairies (99.7% and 95.8%, Table 1). Although the composite anomalies over these regions are relatively small (around 11 mm), they represent nearly a 20% decrease from the long-term seasonal mean values here. While the inverse of this relationship is only slightly less characteristic of cold VPC1 events for the southern Canadian prairies (Fig. 6b; regional significance of 92.1% vs 95.8% for warm VPC1 events, Table 1), this is not true for eastern Montana, suggesting that the similar relation in M97 is nonlinear for the latter area. Furthermore, these relations suggest that the January (December–January) coherencies near Montana in L97’s warm (cold) SSTA composites may originate as early as November and may be more related to eastern tropical Pacific SSTAs than those in their key area. Interestingly, the above northern Great Plains associations coincide with a dominant storm track identified by Whitaker and Horn (1984) and Richman et al. (1992), possibly suggesting a link between positive eastern tropical Pacific SSTAs and cyclogenesis in south-central Canada.

Fig. 6.

Same as Fig. 4 but for November–January (NDJ) seasonal precipitation anomalies during warm and cold SSTA VPC1 events. Years listed refer to December of NDJ. Small maps at side locate numbered regions in Table 1.

Fig. 6.

Same as Fig. 4 but for November–January (NDJ) seasonal precipitation anomalies during warm and cold SSTA VPC1 events. Years listed refer to December of NDJ. Small maps at side locate numbered regions in Table 1.

3) Southern United States and Great Lakes (May–June)

Figure 7 suggests May–June precipitation over portions of the southern and central United States and the Great Lakes is nonlinearly related to eastern tropical Pacific SSTAs. In the May and June composites based on warm SSTA VPC1 events (Figs. 7a,b; Table 1), locally/regionally significant positive (negative) May coherencies in southern Oklahoma–Texas (Illinois–Missouri) precede more extensive June anomalies extending into Nebraska and the southeastern United States (Great Lakes). No significant relations are present in the counterpart VPC1 cold event precipitation composites (Table 1; composite maps not shown). Although the May VPC1 warm composite has low field significance and explains only 13.2% of the domain variance of its constituents, additional support is given by the continuity of its above anomalies into June, for which the field significance (91.3%) and explained variance (29.8%) are much higher. In addition, the sizable negative June anomalies over Wisconsin and northern Illinois are also strongly locally/regionally significant (99.4%, Table 1) for this important month for crop growth. No inverse anomaly feature was characteristic of VPC1 cold events (e.g., Table 1).

Fig. 7.

Same as Fig. 4 but for warm May and June SSTA VPC1 events and cold June UPC1 events. Small maps at left (right) side locate numbered regions in Table 1 (2).

Fig. 7.

Same as Fig. 4 but for warm May and June SSTA VPC1 events and cold June UPC1 events. Small maps at left (right) side locate numbered regions in Table 1 (2).

While the May–June precipitation anomalies associated with warm UPC1 events (not shown) are generally similar to those presented in Figs. 7a,b for VPC1, those based on cold UPC1 events indicate a dry association in the southern Great Plains that is not strongly linked to VPC1. For example, the cold UPC1 composite of June precipitation anomalies (Fig. 7c) exhibits locally significant deficits in central and western Texas that also have high regional significance (Table 2), including below-average precipitation in all seven constituent months. Negative precipitation anomalies in the same region are not as strong in the cold June VPC1 composite (not shown) and have a lower regional significance (Table 1). This relation was not identified in L97’s analysis, suggesting that the cold events having teleconnections with June precipitation in the southern United States are characterized primarily by SSTAs related to the broad central Pacific UPC1 mode rather than only L97’s small target region, which occupies the western extremity of UPC1 (Fig. 2a).

4) Southeastern United States, Atlantic coast, and Great Plains (July–August)

Additional nonlinear relationships exist between SSTA VPC1 and July–August precipitation over much of the United States east of the Rocky Mountains. While the monthly composites of both July and August precipitation anomalies for warm and cold VPC1 events (not shown) suggested the latter are broadly associated with dry and wet conditions, respectively, in the southeastern United States and along the Atlantic coast, some nonlinear characteristics are also present. Those nonlinear features are illustrated well in the composites of July–August total precipitation anomalies for the two SSTA extremes (Fig. 8), which have substantial field significance (89.0%, 95.9%) and explain a high or moderate amount of variance in their constituents (34.7%, 17.2%). For example, regionally significant negative anomalies over the southeastern United States and mid-Atlantic coast are linked to warm VPC1 events (Fig. 8a, Table 1), but the cold VPC1 composite (Fig. 8b) is characterized by an absence of significant values in the southeastern United States and only a smaller region of mid-Atlantic positive anomalies with weaker local significance. Also, regionally significant negative anomalies covering Texas–Nebraska associated with cold July–August VPC1 events do not have corresponding positive anomalies for warm VPC1 events. Although the two above relations linked to warm VPC1 (eastern tropical Pacific) events are evident in L97’s warm SSTA composites for July and August, the negative coherency from Texas to Nebraska in the cold VPC1 composite is not present in L97’s corresponding cold SSTA composites.

Fig. 8.

Same as Fig. 4 but for July–August (JA) seasonal precipitation anomalies during warm and cold VPC1 events. Small maps at side locate numbered regions in Table 1.

Fig. 8.

Same as Fig. 4 but for July–August (JA) seasonal precipitation anomalies during warm and cold VPC1 events. Small maps at side locate numbered regions in Table 1.

b. SSTA VPC2

Four North American regions with significant nonlinear associations between precipitation and SSTA VPC2 (west-central tropical Pacific horseshoe, Fig. 2c) were identified, including 1) Atlantic–Gulf of Mexico coasts and north-central United States (November, January–March), 2) middle and northern Atlantic coast (April–May), 3) southern United States (May–June), and 4) central United States (July–October). The first two of these linkages have not been previously identified, while the second two closely parallel L97’s findings for their cold SSTA composites.

1) Atlantic–Gulf of Mexico coasts and north-central United States (November, January–March)

During warm November and January–March VPC2 events, dry conditions occurred over large portions of the Atlantic–Gulf of Mexico coasts and north-central United States (Fig. 9; Table 3). The absence of counterpart wet conditions in the VPC2 cold composite precipitation fields (Table 3; composite maps not shown) provides further evidence of the nonlinearity of the tropical Pacific SSTA–North American precipitation relationship. In the November warm VPC2 composite (Fig. 9a, 96.5% field significance), locally/regionally significant negative anomalies are present in the north-central United States and from Mississippi to the mid-Atlantic coast (Table 3). No significant precipitation relation is linked to warm December VPC2 events (not shown). Figures 9b–d shows that the January and March warm VPC2 composites contain elements of the more pronounced February anomaly pattern, which is highly field significant (99.8%) and explains 18.8% of the domain variance of its constituents. In all three months, locally/regionally significant negative regions span Louisiana to New England (January), the midwestern and southern United States (February), and the northern Great Plains (March). All of these monthly features are also reflected in the January–March composite precipitation anomaly (Fig. 9e), which is highly field significant (96.0%) and explains 15.7% of the domain variance of its constituents.

Fig. 9.

Same as Fig. 4 but for November and January–March monthly precipitation anomalies and January–March (JFM) seasonal anomalies during warm VPC2 events. Small maps at side locate numbered regions in Table 3.

Fig. 9.

Same as Fig. 4 but for November and January–March monthly precipitation anomalies and January–March (JFM) seasonal anomalies during warm VPC2 events. Small maps at side locate numbered regions in Table 3.

Table 3.

As in Table 2 but for SSTA VPC2 events and with region numbers pertaining to shaded areas in Figs. 9, 10, 12, and 13a.

As in Table 2 but for SSTA VPC2 events and with region numbers pertaining to shaded areas in Figs. 9, 10, 12, and 13a.
As in Table 2 but for SSTA VPC2 events and with region numbers pertaining to shaded areas in Figs. 9, 10, 12, and 13a.

These warm VPC2 November–March associations (Fig. 9) differ substantially from those linked to warm UPC1 (Fig. 5) and VPC1 (Fig. 4) events. While warm SSTAs in the west-central tropical Pacific horseshoe (VPC2) were associated with extensive dry November and January–March anomalies spanning Texas–Georgia–Maine and extending farther inland (Fig. 9), warm SSTAs in the central (UPC1) and eastern (VPC1) tropical Pacific coincided with wet anomalies in parts of a similar near-coastal region, although pronounced dry anomalies were present farther inland during January–March (e.g., Figs. 5a,c; Figs. 4e,g,i). These results support the finding in M97 that the location of tropical Pacific SSTAs influences the distribution and sign of North American precipitation anomalies, and also document additional nonlinear characteristics of these relations.

2) Middle and northern Atlantic coast (April–May)

Figure 10 suggests that dry April and May conditions occur over the northern Atlantic coast during warm VPC2 events, whereas no significant precipitation relations were associated with cold VPC2 conditions in those months (not shown). Specifically, negative precipitation anomalies span Virginia–Maryland in the April warm VPC2 composite field (Fig. 10a) and extend from Virginia to Maine in the May warm VPC2 composite (Fig. 10b). Although these composite fields possess low field significance and explain ⩽12.5% of the domain variance of their constituents, the average April (May) precipitation anomalies over Virginia (mid-Atlantic–New England) of −25.9 mm (−30.5 mm) are regionally significant at the 99.8% (92.4%) level (Table 3), with below-average precipitation having occurred during eight of nine (five of six) warm events. It is interesting to note that the plot of April precipitation in Virginia versus SSTA VPC2 scores (Fig. 11) reveals a remarkable linear relationship for positive scores only, with but a single exception (1962) out of 22 years. The influence of SSTA location on the distribution and sign of precipitation anomalies is evident here, as these VPC2 (west-central tropical Pacific horseshoe) relations are not evident in the warm April–May VPC1 (eastern tropical Pacific) or UPC1 (central tropical Pacific) composites (not shown).

Fig. 10.

Same as Fig. 9 but for monthly April and May precipitation anomalies during warm VPC2 events. Small maps at side locate numbered regions in Table 3.

Fig. 10.

Same as Fig. 9 but for monthly April and May precipitation anomalies during warm VPC2 events. Small maps at side locate numbered regions in Table 3.

Fig. 11.

Scatterplot of average April precipitation anomaly (mm) over Virginia vs SSTA VPC2 score. Last two digits for each year are plotted, vertical broken lines are at ±0.97, and horizontal broken lines are at ±1 σ.

Fig. 11.

Scatterplot of average April precipitation anomaly (mm) over Virginia vs SSTA VPC2 score. Last two digits for each year are plotted, vertical broken lines are at ±0.97, and horizontal broken lines are at ±1 σ.

3) Southern United States (May–June)

During cold VPC2 events, wet May–June conditions occurred across the extreme southern United States, whereas precipitation was near normal there during warm VPC2 events. In the cold VPC2 May and June precipitation composites (Figs. 12a,b), although locally significant positive anomalies are scattered across Louisiana–Mississippi (May) and the southeastern United States (June) and the field significances are low (May) or primarily due to locally significant anomalies elsewhere (June, Great Lakes to southeastern Canada), Table 3 shows that the average May (June) precipitation anomalies in Louisiana–Mississippi (southeastern United States) of +36.1 mm (+54.4 mm) have moderate regional significances of 85.1% (90.4%). Furthermore, precipitation was above average in those regions in six of seven and six of six cold VPC2 events, respectively (Table 3). No counterpart dry conditions characterized the May–June warm VPC2 composites (Table 3; composite maps not shown). In addition, the dependence of teleconnection relations on SSTA location is again illustrated by the above positive precipitation coherencies linked to cold May–June VPC2 events being broadly similar to those based on warm May–June VPC1 events (Figs. 7a,b).

Fig. 12.

As in Fig. 10 but for monthly May–October precipitation anomalies and seasonal June–August (JJA) and August–October (ASO) precipitation anomalies during cold VPC2 events. Small maps at side locate numbered regions in Table 3.

Fig. 12.

As in Fig. 10 but for monthly May–October precipitation anomalies and seasonal June–August (JJA) and August–October (ASO) precipitation anomalies during cold VPC2 events. Small maps at side locate numbered regions in Table 3.

4) Central United States (July–October)

Figure 12 contains further evidence of the nonlinearity of tropical Pacific SSTA–North American precipitation relations, in this case for important agricultural regions in the central United States that experience anomalously dry (but not wet) conditions during cold (warm) VPC2 events. For July and August, vital reproductive and grainfill/podfill months for corn and soybeans, the cold VPC2 precipitation composites include locally/regionally significant (Figs. 12c,d; Table 3) negative coherencies in important production areas that span northern Illinois–Wisconsin (July) and Kansas–Iowa (August). The July map is qualitatively similar to that for August, which has higher field significance (95.8%), and each explain 14%–19% of the variance of their constituents. In the September and October cold VPC2 precipitation composites (Figs. 12e,f), this July–August relation extends northeastward, with locally/regionally significant negative anomalies over Wisconsin and New Brunswick (September) and Arkansas and New Brunswick (October) (Table 3). With the exception of a somewhat linear October relation for New Brunswick (warm VPC2 composite map not shown), all of the above constitute nonlinear SSTA VPC2-precipitation relations. Note that the strong month-to-month continuity of these anomalies is reflected in the field significant composites of total June–August and August–October (Figs. 12g–h) precipitation anomalies, which explain large and moderate fractions (27.1%, 11.7%) of the variance of their constituents and contain locally/regionally significant anomalies over Kansas–Missouri and Illinois–Iowa–Wisconsin, respectively (Table 3).

The differences between the above cold VPC2 July–October precipitation associations and the corresponding patterns for VPC1 further demonstrate that the location and extent of tropical Pacific SSTAs affect their precipitation teleconnections. While some negative precipitation coherencies in the central United States are associated with cold VPC1 events in July–August (Fig. 8b), no significant counterpart coherencies were identified for September or October (not shown). Moreover, the negative anomaly areas in Figs. 12c–h are similar to those in L97, even though SSTAs in L97’s small key area are not emphasized by VPC2 (Fig. 2c). Thus, our work suggests dry September–October conditions prevail in the central United States when cold SSTAs are limited to the west-central tropical Pacific horseshoe (VPC2) or L97’s key region (around which the horseshoe wraps), but not the eastern (VPC1) tropical Pacific.

c. Single-month associations

In addition to the associations with durations of at least two calendar months presented in the previous sections, five linkages involving individual calendar months were also identified. All of these relations were present for only one type of SSTA event (Tables 3, 4) and are thus nonlinear. Although the short duration of these relations increases the possibility that they may arise by random chance, following section 3d all have either field significances greater than 90% or regional significances above 95%. Accordingly, these relations are presented here briefly and await future confirmation.

Table 4. 

As in Table 2 but for SSTA VPC3 events and with region numbers pertaining to shaded areas in Fig. 14.

As in Table 2 but for SSTA VPC3 events and with region numbers pertaining to shaded areas in Fig. 14.
As in Table 2 but for SSTA VPC3 events and with region numbers pertaining to shaded areas in Fig. 14.

1) Gulf of Mexico and south Atlantic coasts (December, VPC2)

Figure 13 suggests that above-normal precipitation occurs along the Gulf of Mexico coast during cold December VPC2 (west-central tropical Pacific horseshoe) events. In the cold VPC2 December precipitation composite (Fig. 13), locally/regionally significant positive anomalies span Mississippi–North Carolina, with above average precipitation observed during eight of nine cold VPC2 Decembers (Table 3). Interestingly, M97, L97, and the present UPC1 and VPC1 analyses did not identify clear relationships between central and eastern tropical Pacific SSTAs and December precipitation along the south Atlantic and Gulf of Mexico coasts.

Fig. 13.

As in Fig. 12 but for monthly December precipitation anomalies. Small map at side locates numbered region in Table 3.

Fig. 13.

As in Fig. 12 but for monthly December precipitation anomalies. Small map at side locates numbered region in Table 3.

2) Texas to New England (January, VPC3)

Figure 14a indicates that a large area from Texas to New England experiences below normal precipitation during warm January VPC3 (far western tropical Pacific) events. The January precipitation composite for warm VPC3 events (Fig. 14a) has high field significance (96.8%), explains 27.4% of the domain variance of its constituents, and includes extensive areas of locally significant negative anomalies. This coherency is regionally significant only in warm events (98.1%, Table 4), with below normal precipitation having been observed during five out of six warm VPC3 Januarys. Note that five out of the six warm events occur after 1970, suggesting that the associated trend in tropical Pacific SSTAs (section 3a.4) might provide important information on January precipitation trends over the south-central and northeastern United States.

Fig. 14.

Same as Fig. 4 but for monthly January and April (March and August) precipitation anomalies during warm (cold) VPC3 events. Small maps at side locate numbered regions in Table 4.

Fig. 14.

Same as Fig. 4 but for monthly January and April (March and August) precipitation anomalies during warm (cold) VPC3 events. Small maps at side locate numbered regions in Table 4.

3) Central Gulf coast and lower Mississippi River valley (April, VPC3)

Figure 14b suggests that the central Gulf coast and lower Mississippi River valley experience dry conditions during warm April VPC3 (far western tropical Pacific) events. In the warm VPC3 April precipitation composite, locally significant negative anomalies extend from eastern Texas to Alabama, with the average precipitation anomaly of −44.6 mm being highly regionally significant (99.7%) and all five warm VPC3 Aprils being characterized by below average precipitation (Table 4). It is striking that all five of these years are after 1980, further suggesting an association between regional United States dryness and the warming trend in far western tropical Pacific SSTs.

4) Southern Great Plains (March, VPC3)

The March precipitation composite for cold VPC3 (far western tropical Pacific) events (Fig. 14c) suggests they coincide with below normal precipitation in the southern Great Plains. Locally significant negative precipitation anomalies spanning Texas–western Kansas are also highly regionally significant (99.4%), with dry conditions having occurred during all five cold VPC3 Marchs (Table 4). Of these five months, two (three) are also part of cold VPC1 (UPC1) events, perhaps suggesting that cold SSTAs in all three of these tropical Pacific regions are associated with dry March conditions in the southern Great Plains.

5) Mid-to-south Atlantic coast (August, VPC3)

Figure 14d suggests that the middle and southern Atlantic coast experiences below normal August precipitation during cold VPC3 (far western tropical Pacific) events. The cold August VPC3 precipitation composite (91.0% field significance, 13.8% explained variance) features a negative Georgia–Pennsylvania coherency that is highly regionally significant (99.7%) and the result of below average precipitation in six out of seven cold August events (Table 4). Except for 1982, all of these dry years preceded 1977, possibly suggesting an association between the aforementioned far western tropical Pacific SST trend and decreasing mid-to-south Atlantic coastal dryness. Note that dry mid-Atlantic conditions are also associated with warm eastern tropical Pacific (VPC1) events (Fig. 8), which do not overlap with the cold August VPC3 events.

5. Summary and discussion

Motivated by recent studies showing that many seasonal precipitation, surface temperature, and geopotential height patterns linked to warm tropical Pacific SSTA events do not simply have similar structures but opposite signs to their cold SSTA counterparts—that is, they are not necessarily linearly related—we identified characteristic monthly central and eastern North American precipitation anomaly patterns separately for warm and cold events. Specifically, through compositing techniques, we delineated regions with monthly precipitation responses that were potentially nonlinearly related to four tropical Pacific SSTA modes—two of which reflected traditional ENSO SSTA patterns in the eastern (VPC1) and central (UPC1) tropical Pacific; one of which involved west-central tropical Pacific SSTA variability in a horseshoe-shaped region (VPC2) that, in approximately half of the extremes (only), constituted westward extensions of traditional ENSO events; and a fourth that emphasized far western tropical Pacific SSTAs (VPC3) that are largely unrelated to ENSO. Significant precipitation teleconnections based on the two SSTA modes related to traditional warm and cold events were identified for all months except September and October, with all exhibiting some nonlinear characteristics. Furthermore, all relations based on west-central tropical Pacific horseshoe (VPC2) and far western tropical Pacific (VPC3) SSTAs were completely nonlinear in nature.

The present study also demonstrated that differences in central and eastern North American precipitation teleconnections with tropical Pacific SSTAs are associated with the location and extent of the SSTAs. This emerging general issue was previously investigated for North American precipitation only in M97’s linear analysis. In this regard, the precipitation anomaly composites vary considerably for the same calendar months as a function of SSTA location and sign, most notably between the traditional ENSO modes (UPC1 and VPC1) for the Great Lakes to the Ohio River valley region for cold February–March events and along the northern Atlantic coast during cold April events. In addition, for all calendar months with coherent precipitation anomaly regions linked to warm or cold SSTA VPC2 and VPC3 events, the resulting composite fields exhibited no similarity with each other or with the patterns linked to UPC1 and VPC1, except during warm January events when dry conditions west of and over the Appalachian Mountains are associated with warm VPC1, VPC2, and VPC3 events.

By using this nonlinear approach and monthly resolution to examine central and eastern North American precipitation teleconnections with four tropical Pacific SSTA regions, the present results substantially extended previous studies that have been limited to linear and/or seasonal relationships. A summary of the new information presented here, vis-à-vis earlier results, follows.

a. Ropelewski and Halpert (1986, 1989, 1996)

The present results extended RH’s work by documenting significant monthly relationships embedded within their seasonal results and identifying additional relations that could not be detected using semestral/seasonal resolutions and only the SOI, including the degree of nonlinearity involved. In particular, the SSTA UPC1 and VPC1 precipitation anomaly composites (Figs. 4, 5) demonstrated that RH86/89’s wet (dry) conditions along the Gulf of Mexico coast for October–March (October–April) during low-SOI (high-SOI) events are present and statistically significant in the individual monthly anomalies for November and January–March (but not December) for both types of events. Furthermore, the positive January–March precipitation anomalies along the southern Atlantic coast, shown only by RH96’s seasonal low-SOI analyses, were found to be statistically significant in the corresponding calendar monthly composites for both warm and cold eastern tropical Pacific SSTA events (VPC1). Additionally, the results in sections 4a(2)–4a(4) linking precipitation in the northern Great Plains (November–January), southern United States and Great Lakes (May–June), and southeastern United States, Atlantic coast, and Great Plains (July–August) to the traditional ENSO modes (UPC1, VPC1) were not present in the RH studies. Also, by focusing on the SOI, RH did not utilize information contained in the west-central tropical Pacific horseshoe (VPC2) and far western tropical Pacific (VPC3) SSTA modes, and thus did not identify the relations in sections 4b,c.

b. Kiladis and Diaz (1989)

The results based on UPC1 and VPC1 also provided important detail concerning the North American precipitation teleconnection results of KD89, by identifying the months for which their seasonal precipitation relations are strongest and the regions that exhibit a nonlinear connection to tropical Pacific SSTAs. While Fig. 4 established the linearity of the KD89 southern Texas (positive, September–November) and Gulf of Mexico coast (positive, December–February) ENSO-related precipitation patterns, as already noted these associations were most evident in the monthly anomalies for November and January–March, respectively. Furthermore, Figs. 4 and 8 revealed important nonlinearities in KD89’s Great Lakes (negative, December–February) relation, which was present for warm January–February eastern tropical Pacific events only, and Atlantic coast (negative, June–August) coherency, which was most significant during July and August during warm eastern tropical Pacific SSTA events. Also, KD89’s southern/central Great Plains (positive, March–May) association was found to be most significant during cold central/eastern Pacific SSTA March events (Figs. 4i,j). The present nonlinear northern Great Plains and southern United States and Great Lakes relationships (sections 4a.2–3) did not appear in the KD89 linear seasonal analyses.

c. Montroy (1997)

The composite anomaly (mm) results in sections 4a,b extended M97’s linear (correlation-based) analysis by documenting the nonlinear characteristics of all six multiple-month precipitation relations obtained in that study for very similar VPC1, UPC1, and VPC2 SSTA modes. For example, M97’s positive relations for the southeastern United States (November and January–March) and Texas (November–March), while broadly linear, were shown to be more significant and extensive in November and February (Texas only) for cold central and eastern tropical Pacific SSTA events (Figs. 4, 5a–d). Also, the M97 Great Lakes–Ohio River Valley signal was here revealed to be highly nonlinear, with the negative relations for the Great Lakes (Ohio River valley) being significant only during warm January–February (January–March) central and eastern tropical Pacific SSTA events (Figs. 4e–j, 5a–d). Furthermore, Figs. 6 and 8 revealed that M97’s negative linkages between eastern tropical Pacific SSTAs and precipitation in the southern Canadian prairies (November–January) and the southeastern United States (July–August) are most significant during warm November–January and July–August events, respectively. The positive northern storm track relation for September–October in that study was here shown to be part of a longer July–October relationship between central United States precipitation and cold (only) west-central tropical Pacific horseshoe SSTA (VPC2) events (Figs. 12c–h). Additionally, Figs. 4i–j and 5e–f suggested that M97’s one-month SSTA-precipitation relationships over the central plains (positive, March) and mid-Atlantic coast (positive, April) were most significant during cold eastern and central tropical Pacific SSTA events, respectively, while those for the southeastern United States (positive, June) and Mississippi River basin (positive, December) were most evident during warm eastern and central tropical Pacific SSTA events, respectively (Figs. 7b, 4c,d). Finally, the present VPC1-based relations for parts of the southern United States [May–June, section 4a(3)] and Great Plains [July–August, section 4a(4)], and all those linked to the west-central tropical Pacific horseshoe (section 4b), were not detected by M97’s linear analysis.

d. Livezey et al. (1997)

Since the present results were obtained for calendar months, with warm and cold tropical Pacific SSTA events being treated separately, they can be directly compared with L97’s United States precipitation anomaly results based on SSTAs in a small but highly sensitive central equatorial Pacific region (180°–150°W, 5°N/S; Fig. 2). While some of the present nonlinear relationships based on warm and cold central (UPC1) and eastern (VPC1) tropical Pacific SSTA events are reflected in L97, other important linkages not identified in that study include

  • January–February precipitation near the Great Lakes (warm = dry/cold = normal; Figs. 4e–h, Tables 1–2);

  • January–March precipitation in the Ohio River valley (warm = dry/cold = wet, but not significant; Figs. 4e–j, Table 2);

  • July–August precipitation from Texas to Nebraska (cold = dry/warm = normal) and over the mid-Atlantic coast (cold = dry/warm = wet, but not as significant; Fig. 8, Table 1);

  • November precipitation along the Atlantic coast (cold = dry/warm = wet, but not significant; Figs. 4a,b, Table 1); and

  • December precipitation in the southern plains (cold = dry/warm = normal; Figs. 4c,d, Table 1).

Although the present northern Great Plains relation for November–January (section 4a.2) is evident in the L97 January (December–January) warm (cold) SSTA results as a negative (positive) coherency in the north-central United States, its extension into Canada was demonstrated here. Also, our strongly significant VPC1-based nonlinear relations for the southeastern United States and southern Atlantic coast in July–August [section 4a(4)] only partially support L97’s counterpart results—namely, the weakly significant negative precipitation anomalies in both areas for their July–August warm SSTA composites, but not the deficient August precipitation in the southeastern United States for their cold composite. We obtained no significant signal in the latter regard.

The precipitation teleconnections in sections 4b,c, all of which were based on SSTAs in the west-central tropical Pacific horseshoe (VPC2) and far western tropical Pacific (VPC3), were also not evident in L97 because of its narrower central equatorial Pacific SSTA focus. In addition, some of those results [central United States, July–October, section 4b(4)] place L97’s findings in a broader context. Specifically, by identifying the large-scale (VPC2) structure of the cold SSTAs associated with below normal July–October precipitation in the central United States, we established that L97’s similar but less temporally continuous results stemmed from their central equatorial Pacific region reflecting SSTAs in the west-central Pacific horseshoe region rather than the larger central (UPC1) or eastern (VPC1) tropical Pacific domains.

By using the present statistical approach to identify the monthly central and eastern North American precipitation anomaly patterns that are simultaneously associated with SSTAs in different regions of the tropical Pacific, we have established a framework that is now permitting an efficient, focused, physical–dynamical investigation of the teleconnection mechanisms involved. In building on this foundation, we are both utilizing synoptic climatological analyses that exploit the basic daily resolution of the present precipitation dataset (section 2b), and also nesting a mesoscale weather prediction model within a global climate model that is forced by global SSTA patterns. This combined empirical-modeling approach is designed to identify the response mechanisms (including time lags) of central and eastern North American weather systems to tropical Pacific forcing on intraseasonal, interannual, and subdecadal timescales and, accordingly, thus enhance our ability to predict those societally important phenomena in an “ensemble of states” [as opposed to an individual instantaneous state; Peixoto and Oort (1992, 462)] sense on these timescales.

Fig. A1. Observed individual monthly (a–j) and multiple-month (k, l) precipitation anomalies (based on 1950–92 means) for U.S. climate divisions within domain outlined in Fig. 1 for May 1997–February 1998. Color coding is identical to that in Fig. 4.

Fig. A1. Observed individual monthly (a–j) and multiple-month (k, l) precipitation anomalies (based on 1950–92 means) for U.S. climate divisions within domain outlined in Fig. 1 for May 1997–February 1998. Color coding is identical to that in Fig. 4.

Acknowledgments

This work was funded by NSF Grant ATM 92-96119, USEPA Cooperative Agreement CR-819646010, National Oceanic and Atmospheric Administration (NOAA) Cooperative Agreement NA67RJ0150, and an American Meteorological Society graduate fellowship received by the first author and cosponsored by the NOAA Office of Global Programs. The computing was performed at the Pittsburgh Supercomputing Center and the NOAA Climate Diagnostics Center. The authors would like to thank Robert E. Livezey of the NOAA Climate Prediction Center (CPC) for his helpful suggestions on statistical significance testing and insightful comments linking the present study to the work of his group. The contributions of the following individuals are also gratefully acknowledged: Chester F. Ropelewski (CPC) for initial guiding comments, Thomas M. Smith and Richard W. Reynolds (CPC) for information on the SST data used, Claude E. Duchon (University of Oklahoma, OU) for comments on an earlier version of the manuscript, and M. Neil Ward (OU) for statistical advice. Helpful comments from two anonymous reviewers also substantially improved the paper.

REFERENCES

REFERENCES
Allan, R., J. Lindesay, and D. Parker, 1996: El Niño Southern Oscillation and Climatic Variability. CSIRO Publishing, 405 pp.
Cattell, R. B., 1966: The scree test for the number of factors. Multivar. Behav. Res., 1, 245–276.
Fu, C., H. F. Diaz, and J. O. Fletcher, 1986: Characteristics of the response of sea surface temperature in the central Pacific associated with warm episodes of the Southern Oscillation. Mon. Wea. Rev., 114, 1716–1738.
Glantz, M. H., R. W. Katz, and N. Nicholls, Eds., 1991: Teleconnections Linking Worldwide Climate Anomalies. Cambridge University Press, 535 pp.
Gong, X., and M. B. Richman, 1995: On the application of cluster analysis to growing season precipitation data in North America east of the Rockies. J. Climate, 8, 897–931.
Gutzler, D. S., 1995: Decadal and interannual variability in the tropical western Pacific. Proc. 20th Annual Climate Diagnostics Workshop, Seattle, WA, Climate Prediction Center/NCEP, 199–202.
Hoerling, M. P., A. Kumar, and M. Zhong, 1997: El Niño, La Niña, and the nonlinearity of their teleconnections. J. Climate, 10, 1769–1786.
Horel, J. D., 1981: A rotated principal component analysis of the interannual variability of the Northern Hemisphere 500 mb height field. Mon. Wea. Rev., 109, 2080–2092.
Hsuing, J., and R. E. Newell, 1983: The principal nonseasonal modes of variation of global sea surface temperature. J. Phys. Oceanogr., 13, 1957–1967.
Kaiser, H. F., 1958: The Varimax criterion for analytic rotation in factor analysis. Psychometrika, 23, 187–200.
Kang, I. N., and N. C. Lau, 1987: Principal modes of atmospheric variability in model atmospheres with and without sea surface temperature forcing in the tropical Pacific. J. Atmos. Sci., 43, 2719–2735.
Kiladis, G. N., and H. van Loon, 1988: The Southern Oscillation. Part VII: Meteorological anomalies over the Indian and Pacific sectors associated with the extremes of the oscillation. Mon. Wea. Rev., 116, 120–136.
——, and H. F. Diaz, 1989: Global climatic anomalies associated with the Southern Oscillation. J. Climate, 2, 1069–1090.
Livezey, R. E., and W. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev., 111, 46–59.
——, A. Leetmaa, M. Masutani, H. Rui, M. Ji, and A. Kumar, 1997:Teleconnective response of the Pacific–North American region atmosphere to large central equatorial Pacific SST anomalies. J. Climate, 10, 1787–1820.
Montroy, D. L., 1997: Linear relation of central and eastern North American precipitation to tropical Pacific sea surface temperature anomalies. J. Climate, 10, 531–548.
North, G. R., T. L. Bell, R. F. Calahan, and F. J. Moeng, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110, 699–706.
Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp.
Rasmusson, E. M., and T. H. Carpenter, 1982: Variations in tropical sea surface temperature and surface wind fields associated with the Southern Oscillation/El Niño. Mon. Wea. Rev., 110, 354–384.
Richman, M. B., 1986: Rotation of principal components. J. Climatol., 6, 293–333.
——, 1987: Rotation of principal components: A reply. J. Climatol., 7, 511–520.
——, and P. J. Lamb, 1985: Climatic pattern analysis of three- and seven-day summer rainfall in the central United States: Some methodological considerations and a regionalization. J. Climate Appl. Meteor., 24, 1325–1343.
——, and ——, 1987: Pattern analysis of growing season precipitation in southern Canada. Atmos.–Ocean, 25, 137–158.
——, ——, A. Zangvil, and J. R. Angel, 1992: Multivariate statistical analysis of precipitation associated with North American storm tracks. Preprints, 12th Conf. on Probability and Statistics in the Atmos. Sciences, Toronto, ON, Canada, Amer. Meteor. Soc., 277–283.
Ropelewski, C. F., and M. S. Halpert, 1986: North American precipitation and temperature patterns associated with the El Niño–Southern Oscillation. Mon. Wea. Rev., 114, 2352–2362.
——, and ——, 1987: Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev., 115, 1606–1626.
——, and ——, 1989: Precipitation patterns associated with the high index phase of the Southern Oscillation. J. Climate, 2, 268–284.
——, and ——, 1996: Quantifying Southern Oscillation–precipitation relationships. J. Climate, 9, 1043–1050.
——, M. Chelliah, T. Smith, and D. Marsico, 1992: What can COADS tell us? Potential and limitations. Proc. Int. COADS Workshop, Boulder, CO, U.S. Department of Commerce, 295–302.
Smith, T. M., R. W. Reynolds, R. E. Livezey, and D. C. Stokes, 1996:Reconstruction of historical sea surface temperatures using empirical orthogonal functions. J. Climate, 9, 1403–1420.
Weare, B. C., A. R. Navato, and R. E. Newell, 1976: Empirical orthogonal analysis of Pacific sea surface temperatures. J. Phys. Oceanogr., 6, 671–678.
Whitaker, L. M., and L. H. Horn, 1984: Northern Hemisphere extratropical cyclone activity for four mid-season months. J. Climatol., 4, 297–310.

APPENDIX

The 1997–98 El Niño Event (added in proof)

By the time this paper was accepted for publication, an exceptionally strong El Niño event had developed in the eastern and central tropical Pacific. When projected onto the above SSTA modes (Fig. 2), the monthly optimally interpolated SSTAs (obtained from the NOAA Climate Prediction Center; ftp://nic.fb4.noaa.gov/pub/ocean/clim1/oimonth/) from May 1997 to the latest available month (February 1998) all yielded VPC1 and UPC1 scores exceeding the warm event threshold used in section 3b (+0.97). The VPC2 scores obtained in the same way were more negative than the counterpart cold event threshold (−0.97) for December 1997–February 1998. Thus, to this point, the 1997–98 El Niño has closely resembled a blending of the warm VPC1/UPC1 (all months) and cold VPC2 (December–February) types of events. This suggests the 1997–98 observed precipitation patterns should reflect the anomaly structures evident in Figs. 4–8 (warm event panels) and 13. In fact, monthly precipitation anomalies based on preliminary U.S. climate division (CD) data (from the NOAA National Climatic Data Center; http://www.nedc.noaa.gov/onlineprod/drought/ftppage.html; shown in Fig. A1 for May 1997 through February 1998) and Canadian climate maps (from Environment Canada;http://www.mb.cc.gc.ca/Nino/Prairies.html; not shown) do compare very favorably with the results in section 4. Due to the similarity in the precipitation anomaly patterns linked to warm VPC1 and UPC1 events noted in section 4a, and the prevalence of warm VPC1 event relationships presented there, only the VPC1 (and not UPC1) relations are here compared to the precipitation anomalies for May 1997–February 1998. Comparison is also made with VPC2 relations for December 1997. These comparisons are summarized below, with emphasis on the months for which the above results identified significant associations (May–August and November–February).

May–June 1997

Consistent with Fig. 7a, May 1997 precipitation (Fig. A1a) was below normal in all South Carolina CDs, six of nine Georgia CDs, and most CDs within the Illinois–Missouri anomaly region (no. 13 in Table 1), with average anomalies of −33.9 mm, −14.3 mm, and −8.7 mm, respectively. (However, the May precipitation anomalies in northern Texas were not positive, as suggested in Fig. 7a, but slightly negative, with a regional average of −6.4 mm.) Furthermore, the three principal areas of above normal June 1997 precipitation (Fig. A1b) compare favorably with the average anomalies over the following regions identified for June in Fig. 7b and Table 1: western Oklahoma to Nebraska (no. 14 in Table 1; June 1997 regional average anomaly of +26.4 mm), central and western Texas (no. 16; +57.2 mm), and the southeastern United States (no. 17; +60.8 mm). Also, while the June 1997 regional average anomalies for the entire region 15 (Wisconsin–northern Illinois) were slightly positive (+8.1 mm), rather than negative as suggested by Fig. 7b, below normal precipitation was observed in the southern and northern parts of this region, namely in northwestern Illinois (CDs 1 and 3; −27.4 mm) and northwestern Wisconsin (CDs 1–4; −12.0 mm).

July–August 1997

Negative July–August 1997 precipitation anomalies (Fig. A1k) in the eastern United States correspond well with the large negative region delineated in Fig. 8a, with 1997 average anomalies over the entire southeastern United States (no. 18 in Table 1) and mid-Atlantic coast anomaly regions (no. 19) being −18.1 mm and −29.8 mm, respectively. While these regional averages are smaller than the corresponding values in Table 1 (cf. −72.2 and −65.0 mm), some CD clusters did record larger 1997 negative anomalies, including North Carolina CDs 1–4 and 6–8 (−70.2 mm average), Georgia CDs 1–5 and 7–9 (−61.2 mm), and South Carolina CDs 1–2 and 4–7 (−56.2 mm). Additionally, although not discussed in section 4a(4), the positive coherency near the border of Minnesota and Wisconsin in Fig. 8a was reflected in July–August 1997 as an average anomaly for southeastern Minnesota (CDs 5, 6, 8, and 9) and northwestern Wisconsin (CDs 1 and 4) of +81.0 mm, with the two-month anomalies being strongly determined by the July anomalies (Figs. A1c,d).

September–October 1997

Since the above results did not include any locally or regionally significant associations for September and October, additional work is necessary to determine whether the most pronounced 1997 precipitation anomalies during these months can be associated with the warm tropical Pacific SSTAs. The major precipitation anomalies occurred along the Gulf of Mexico coast in September (negative), and in the southeastern (strong positive) and northeastern (strong negative) United States in October (Figs. A1e,f).

November 1997–February 1998

In the southeastern United States, positive November 1997 precipitation anomalies (Fig. A1g) in southern Alabama (CDs 7–8; +76.5 mm), southern Georgia (CDs 7–9; +64.3 mm), and South Carolina (CDs 3–7; +24.1 mm) parallel the locally significant values in Fig. 4a. While the positive December coherency in the south-central United States in Fig. 4c was not found to be locally significant there, it was broadly reflected in positive 1997 anomalies (Fig. A1h) over northeastern Texas (CDs 3–4; average of +60.3 mm) and Oklahoma (+47.6 mm). Also, the positive January–February precipitation anomalies suggested for the Texas through North Carolina region by Figs. 4e,g were evident in 1998 (Figs. A1i,j). Specifically, the 1998 anomalies averaged +84.9 mm (January) and +54.4 mm (February) for all CDs extending from southern Louisiana to North Carolina, and +20.3 mm for southeastern Texas (CDs 4 and 7–10) in February. While only weakly evident in the January 1998 anomalies (Fig. A1i) as a small area of below normal precipitation in western Kentucky (CDs 1–2; −19.9 mm) and southern Indiana (CDs 7–9; −18.7 mm), the pronounced Ohio River valley negative coherency in Figs. 4e,g was more strongly reflected in the February 1998 precipitation anomalies (Fig. A1j) over Kentucky (−43.1 mm) and Indiana (−21.2 mm). This Ohio valley anomaly was actually better developed during November–December 1997 (Figs. A1g,h,l) and, thus, commenced earlier in the cool season than was characteristic of previous VPC1/UPC1 type events (cf. Figs. 4a,c).

In the northern Great Plains, the indication of below normal November–January precipitation by Fig. 6a and Table 1 was mirrored in the 1997–98 anomalies (Fig. A1l), as the regional average 1997–98 anomalies were negative over eastern Montana (no. 10, −21.5 mm), North Dakota (−17.3 mm), and South Dakota (−21.8 mm). Also, according to Environment Canada climate maps (not shown), the entire southern Canadian prairies received below normal (over 50% less in many areas) November–December precipitation, while much of Alberta and the southernmost portion of Saskatchewan experienced below normal January 1998 precipitation.

In addition to the foregoing VPC1/UPC1-based comparisons, another important parallel between the present results and 1997–98 observations involves precipitation patterns linked to SSTA VPC2 (west-central tropical Pacific horseshoe). Recall that positive December precipitation anomalies in the southeastern United States were more strongly linked to cold VPC2 events (Fig. 13) than warm VPC1 (Fig. 4c) or warm UPC1 [not shown; see section 4a(1)] events. As already noted, the December 1997 SSTA VPC2 score classified that month as a cold VPC2 event, suggesting wet conditions should have occurred in much of the southeastern United States (Fig. 13). This association was indeed characteristic of December 1997 (Fig. A1h), when precipitation exceeded normal (by an average of +22.0 mm) in 17 of 22 CDs within the locally significant region in Fig. 13.

This appendix has described how the observed precipitation anomaly patterns for May 1997–February 1998 have corresponded to those historically associated (section 4) with warm eastern and central tropical Pacific SSTA events and (for December 1997 only) cold west-central tropical Pacific horseshoe events. However, other notable 1997–98 precipitation anomalies occurred that were not evident in the present work, including relatively wet conditions in the central Great Plains (August), southeastern United States (October), and the northeastern United States (January), and dry conditions in the northeastern United States (October) and the midwestern–south-central United States (November–December). Further investigation will therefore be needed to determine whether these anomalies were also somehow associated with tropical Pacific SSTAs.

Footnotes

Corresponding author address: David L. Montroy, CIMMS, School of Meteorology, University of Oklahoma, 100 E. Boyd, Room 1110, Norman, OK 73019-0628.