The relationship between the Pacific–North American (PNA) teleconnection pattern and Ohio River Valley (ORV) winter precipitation and hydrology is described. The PNA is significantly linked to moisture variability in an area extending from southeastern Missouri, northeastward over states adjacent to the Ohio River through Ohio. Maximum correlation between the PNA index and station precipitation peaks in southern Indiana at r = −0.71, making the circulation/climate teleconnection one of the strongest in the Northern Hemisphere. The North Pacific index (NPI), a Pacific basin sea level pressure index that is highly correlated to the PNA, confirms a strong circulation–ORV precipitation relationship extending back to 1899. In contrast, measures such as the Tahiti–Darwin Southern Oscillation index (SOI) and Niño-3.4 (5°S–5°N, 120°–170°W) sea temperatures are not significantly correlated to ORV winter precipitation. Wettest (driest) winters occur with zonal (meridional) flow with the PNA negative (positive) and North Pacific sea level pressure anomalously high (low). Moisture flux convergence extends much farther north from the Gulf of Mexico in the wet winters, compared to dry, and excess of precipitation over evaporation (moisture budget) is over 100 mm larger throughout much of the ORV. Wet winters, particularly those of 1949 and 1950 changed the ORV hydrology to one of extensive wet conditions, as measured by the Palmer hydrologic drought index (PHDI). Unusually dry winters, however, appear to have less impact on the index; many ORV climate divisions remain moist through the winter despite low precipitation. Winter mean streamflow along the Ohio River and its tributaries varies significantly between PNA extremes, with river discharges up to 100% higher in PNA-negative winters as opposed to PNA-positive winters.
The Ohio River Valley (ORV) exhibits some of the highest regional winter moisture variability in the United States, spatially represented by the leading rotated principal component of winter precipitation (Walsh et al. 1982) and the third component of Palmer drought severity index values (Karl and Koscielny 1982). The ultimate causes of this variability are only partly known although they play a major role in shaping the regional climate and hydrology. High precipitation winters (Table 1) have, for example, led to catastrophic flooding, either along the Ohio River or along the tributaries that feed this river. Flooding has most notably occurred along ORV tributary rivers in 1913 (Marshall 1913) and 1959, while major floods along the Ohio River itself have occurred in 1881, 1937 (Grover 1938), and more recently in 1997. Dry winters associated with low stream flow (Table 1), such as those of the early 1930s and 1980s, are also problematic, potentially causing increases in pollution concentrations, reductions in available biotic nutrients, and shortages in water supplies.
ORV winter precipitation variability has been linked to the large-scale atmospheric circulation. Leathers et al. (1991) show a significant negative correlation in winter months between ORV precipitation and the index of the Pacific–North American (PNA) teleconnection pattern (Wallace and Gutzler 1981). Winter precipitation variability in the ORV is one of several wintertime climatic impacts across North America that has been associated with the PNA (Dickson and Namias 1976; Douglas et al. 1982; Trenberth 1990; Rogers and Rohli 1991; Rogers and Raphael 1992). The ORV tends to experience below-normal winter precipitation when an amplified West Coast ridge and East Coast trough occur (positive-PNA phase) while above normal precipitation occurs during zonal flow in the negative phase. During the positive-PNA phase (illustrated in Fig. 1a), the North Pacific and Gulf Coast areas (northwestern Canada and Hawaii) are situated in regions of negative (positive) 500-hPa height anomalies, establishing a meridional circulation pattern across North America. This meridional flow intensifies the baroclinic zone along the Gulf and East Coasts and below normal ORV winter precipitation occurs (Walsh et al. 1982; Leathers et al. 1991). The zonal phase (Fig. 1b) is characterized by above-normal 500-hPa heights and below-normal vorticity (Walsh et al. 1982) over the eastern United States, and winter precipitation tends to be rain rather than snow (Serreze et al. 1998). The extratropical PNA centers of action (i.e., the North Pacific, northwest Canada, the southeastern United States) are evident in the mean 500-hPa height differences between the negative and positive PNA phases (Fig. 1c).
The PNA pattern is a feature of atmospheric low-frequency variability partly generated by thermal forcing from the equatorial Pacific (Shukla and Wallace 1983; Kawamura et al. 1995). A sizeable component of its variability can, however, be attributed to atmospheric–oceanic changes in the midlatitudes, possibly linked to the Pacific decadal oscillation (Rodionov and Assel 2001). ORV winter precipitation variability has also been linked to the occurrence of warm and cold events in the equatorial Pacific Ocean. Kiladis and Diaz (1989) identified a tendency for warm (cold) equatorial Pacific events to be significantly dry (wet) at six stations in and around the ORV, and Gerushunov (1998) finds that ORV extreme precipitation events have a lower (higher) frequency in the Ohio–Kentucky–Tennessee region. Montroy (1997) links ORV February and March precipitation and January Kentucky–Tennessee precipitation to the leading rotated and unrotated principal components of tropical Pacific SST variability. Montroy et al. (1998) further suggest that the relationship is stronger in warm events than during cold events. ORV regional precipitation does not, however, appear as part of the significant response to the Southern Oscillation when it is defined by the Tahiti–Darwin pressure difference (Ropelewski and Halpert 1986, 1989); in contrast, responsive areas include the Gulf Coast and Florida.
The purpose of this paper is to expand our understanding of the atmospheric circulation linkages to ORV winter precipitation variability and to show the significant impacts of this variability in elements of the regional hydrology. We will focus on the atmospheric circulation links to precipitation, measures of surface moisture variability, as well as streamflow on the Ohio River and some of its tributaries.
Monthly mean precipitation totals for 5038 stations are obtained from the Global Historical Climate Network (GHCN, version 2) for the years 1899–1998. Stations analyzed from the GHCN inventory range from 0° to 90°N and 50° to 165°W, encompassing North and Central America, Baffin Island, Hawaii, the Caribbean, as well as northernmost South America. Outliers that exceeded three standard deviations were identified and excluded if the anomaly was not in agreement with that occurring at nearby stations. Months with missing data values were also removed from the analysis. State climate division precipitation data (1899–1998) from the National Climatic Data Center (NCDC) are also used. Climate divisions, delineated according to topographical, hydrological, and climatological characteristics of each state, are useful in assessing large-scale climatic anomalies (e.g., seasonal moisture deficiencies) and have spatial and spectral coherence (Guttman and Quayle 1996). Precipitation data consist of unweighted monthly means from all representative stations (excluding outliers) within each division, serving as a filter for erroneous observations that may have been occurring at the station level. As illustrated in Fig. 2, the ORV study area consists of 32 climate divisions covering portions of seven states (Arkansas, Illinois, Indiana, Kentucky, Missouri, Ohio, and Tennessee).
The Palmer drought severity index (PDSI) and the Palmer hydrologic drought index (PHDI) are used here as measures of surface moisture variability and are available for climatic divisions from the NCDC. Developed by Palmer (1965), the PDSI measures the deviation of moisture supply from climatologically appropriate conditions using the supply and demand concept of the water balance equation. The PDSI responds rapidly to changes in weather conditions without considering soil moisture levels, and registers the end of a drought (wet spell) when its moisture index becomes positive (negative) for the first time. The cessation of dry and wet spells, according to the PDSI, is determined by meteorological factors alone (Guttman and Quayle 1996). The PHDI differs by providing a means by which to estimate when runoff, recharge, and demand have returned to climatologically appropriate levels (Karl et al. 1987). The PHDI does not register the termination of a drought until the last month occurs in a series of months having sufficient moisture with which to recharge the soil (Karl 1983) and is not affected by spurious intermittent monthly episodes of adequate precipitation (Karl et al. 1987). Consequently, the PHDI exhibits a lagged response to changes in moisture conditions. The PDSI and PHDI each numerically range generally between −4.0 and +4.0, with negative (positive) values indicative of dry (wet) conditions. The indices are standardized so that comparisons can be made across various North American climate regimes (Wilhite and Glantz 1985).
Monthly mean streamflow values are obtained for six ORV rivers (available online at http://water.usgs.gov/pubs/wri/wri934076). These include the Monongahela River at Braddock, Pennsylvania; Scioto River at Chillicothe, Ohio; Great Miami River at Dayton, Ohio; Kentucky River at Frankfort, Kentucky; Wabash River at Vincennes, Illinois; and the Ohio River at Louisville, Kentucky. Of these rivers the Wabash and Ohio are free flowing and the other rivers, with the exception of Kentucky, have locks or dams more than 10 miles upstream of the gauging station. The Kentucky has a lock and a dam on both sides of the gauging station.
Atmospheric circulation indices used include the PNA index, calculated for the period 1947–98 using the Wallace and Gutzler (1981) method:
where z* represents the 500-hPa winter geopotential height anomaly for the four pattern centers. The North Pacific index (NPI), as devised by Trenberth and Hurrell (1995), uses a weighted mean sea level pressure for an area extending from 30° to 65°N and 160°E to 140°W, typically characterized by the pressure anomaly of the mean winter Aleutian low. The correlation between the NPI and PNA index for 1947–98 is −0.93, indicating a strengthened (weakened) Aleutian low for positive (negative) PNA index values. This relationship exhibits considerable stationarity throughout time and is never lower than r = −0.91 in any 20-yr period since 1947. Sea level pressure data are readily accessible since 1899 for computation of the NPI, thus making the NPI a useful alternative to the PNA index that has no data prior to 1946. NPI data are from the National Center for Atmospheric Research (NCAR) gridded sea level pressures based on daily historical weather maps while PNA data are calculated from the NCAR gridded 500-hPa geopotential heights.
Niño-3.4 sea surface temperature, obtained from the National Centers for Environmental Prediction (NCEP), extend over the central equatorial Pacific from 5°N to 5°S and 170°W to 120°W, overlapping the eastern and western halves of the Niño-4 (5°N–5°S, 150°W–160°E) and Niño-3 (5°N–5°S, 90°–150°W) regions, respectively. Measures of the El Niño–Southern Oscillation (ENSO) used here also include the Tahiti–Darwin pressure difference [Southern Oscillation index (SOI)].
Elements of the atmospheric moisture budget are compared during wet and dry ORV winters using NCEP–NCAR reanalysis data. The atmospheric moisture budget is represented as follows:
where E is the surface evaporation rate, P is the precipitation rate, g is the earth's gravitational constant, V is the horizontal wind, q is the specific humidity, and p is the pressure integrated from the surface (s) to the top of the atmosphere (300 hPa), the upper boundary of available humidity data. The precipitable water w is given as
Moisture fluxes were calculated using the monthly mean and the eddy fluxes integrated from the surface to 300 hPa. When P − E is positive (negative), precipitation (evaporation) exceeds evaporation (precipitation) and surface conditions will be wet (dry). The data are available on a 2.5° × 2.5° latitude–longitude grid and we use a T21 truncation in smoothing the fields. Results are presented below for both the moisture budget (P − E) and the 850-hPa moisture flux components and .
a. Winter precipitation and the Pacific teleconnection indices
The sign of the statistically significant correlation coefficients between mean winter (DJF) PNA index values and the GHCN station winter precipitation totals (1947–98) are shown in Fig. 3. The signs (+, −) of the coefficients are shown for stations with correlations r ≥ |0.4|. These coefficients are significant at the 95% confidence interval providing the station had a minimum 26 years of data (a criterion for the analysis). Most stations have over 40 degrees of freedom and correlations would be significant at values below r = 0.4. The highest PNA correlation of any individual GHCN station is r = −0.71 at Bloomington, Indiana. Figure 3 clearly identifies distinct regions on the North American continent in which winter precipitation responds to phase shifts (positive or negative) in the PNA pattern, most notably the ORV. The inverse PNA precipitation relationship also exists for a linear alignment of stations windward of the Rocky Mountains, consistent with Leathers et al. (1991).
The winter precipitation–PNA index relationship among climate divisions is strongest along a southwest–northeast axis centered just north of the lower (westernmost) Ohio River but then branching across the state of Ohio northeastward from Cincinnati (Fig. 4). The spatial distribution of correlation coefficients between the PNA and climate division precipitation changes to some extent when the analysis is separated by decades. Coefficients are as low as −0.4 < r < 0.0 in up to one-half of the ORV climate divisions during the 1970s and 1990s, especially in Ohio and portions of Kentucky. Based on the pattern of significant correlation coefficients in Figs. 3 and 4, the ORV region (Fig. 2) was spatially defined as consisting of stations (and climate divisions) exhibiting r ≥ |0.5| between winter precipitation and the PNA index. A sample of stations, show as dots in Fig. 2, was selected in order to create a representative ORV winter precipitation index. The station sample was chosen according to 1) the strength of the precipitation–PNA index correlation, 2) the station period of record length (n ≥ 50), and 3) the position of the station within a central ORV “core.” In creating the ORV index the monthly mean precipitation is standardized by subtracting the long-term 1899–1998 monthly mean from each observation and dividing by the station standard deviation. The final ORV index value is the mean of the 12 standardized station values.
Table 2 displays the Pearson correlation coefficients between the PNA, NPI, and ENSO teleconnection indices and the 12 ORV index stations. NPI coefficients are lower than those of the PNA index but consist of more cases (up to n = 100). Although Leathers et al. (1991) suggested some early spring (i.e., Mar) influence by the PNA on ORV precipitation, the DJFM correlations are significant but somewhat lower than those for DJF winters, which are used in all subsequent analyses. While the sign of the SOI and Niño-3.4 coefficients indicate that El Niño (La Niña) accompanies dry (wet) winters, the SOI and Niño-3.4 have less covariability with ORV winter precipitation fluctuations compared to the PNA index and NPI. An analysis conducted using the nonparametric Spearman rank correlation coefficients, based on the ranks as opposed to actual values, confirmed the results presented in Table 2.
To illustrate the degree of stationarity of the reported relationships (Table 2), the study period (1899–1998) was divided into 20-yr segments that had a 15-yr overlap (e.g., 1899–1918, 1904–23, etc.). A Pearson correlation analysis was performed among the teleconnection indices and the ORV index. This revealed that the PNA–ORV index correlation is never below r = −0.59 in any 20 yr period while the Niño-3.4–ORV index correlation never exceeded r = −0.43. The NPI–ORV index correlations exhibit some of their lowest values prior to 1928, as do the Niño-3.4–ORV index correlations. Concerns have been raised about the reliability of North Pacific sea level pressure data in certain years of the pre-1925 era (Trenberth and Paolino 1980) and this may partly account for the lower correlations in this period, although the possibility of an early century change in the circulation/precipitation relationship cannot be ruled out.
Time series of the ORV index (Fig. 5) reveal it is highest in the 1948/49 and 1949/50 winters and lowest in 1962/63 and 1976/77. Both the ORV index and NPI (Fig. 5) tend to be low from 1925 to 1945 and from 1977 to 1988 while the 1950s are distinctly higher in mean index values. The NPI–ORV index correlation (1899–1998) is r = +0.54 while the PNA–ORV index correlation (1947–98) is r = −0.67, with both coefficients significant at the 99.9% confidence level.
The winter ORV index values were ranked, sorted into quintiles, and compared to the teleconnection indices. Table 3 shows the mean and standard deviation for each precipitation-based quintile. The first two quintiles consist of below-normal winter precipitation years (i.e., negative ORV index), while the last two quintiles consist of above-normal years (i.e., positive ORV index); the third quintile embodies years of near-normal winter precipitation. The PNA index generally tends to be positive in Table 3, reflecting the typical standing wave orientation over North America, and basinwide averaged NPI quintile means range from 1006.5 to 1010.9 hPa. The quintile-mean Niño-3.4 values are inconsistent with a strong statistical signal, the highest (lowest) ocean temperatures occurring in quintile 2 (4) instead of 1 (5). A two-tailed t test is performed on comparisons of the first and fifth quintiles and the second and fourth quintiles. Significant differences exist for the PNA index and NPI values of quintiles 1 and 5 at the 99.9% confidence level (t values exceeding |4|). Between quintiles 2 and 4, the confidence level decreases to 95% for the PNA index and NPI values are no longer significantly different.
The mean winter 500-hPa height differences, when ORV quintile 1 (dry; Table 1) is subtracted from ORV quintile 5 (wet) in post-1946 years (see Fig. 1d), further illustrate the close link between the ORV winter precipitation differences and those of the extremes in the PNA index (Fig. 1c).
Figures 6a,b show the surface to 300 hPa total column moisture budget (P − E) for the ORV wet and dry quintiles cases occurring since 1949. The wet cases (Fig. 6a) are characterized by an area of positive P − E extending into the ORV region compared to the dry winters (Fig. 6b). The mean differences, wet minus dry (Fig. 6c), are centered on the Ohio River over southern Indiana and northwestern Kentucky, the ORV core area, and have the same southwest–northeast orientation shown in the precipitation–PNA correlations of Fig. 4. The P − E differences achieving statistical significance at the 95% level have values higher than 40 mm.
The zonal and meridional components of the moisture flux are shown in Fig. 7 for ORV quintiles 1 and 5. In cold dry winters (Fig. 7c) qu has an axis along the Gulf Coast extending eastward to a maximum over the Atlantic, whereas in wet ORV winters (Fig. 7a) it is shifted 5° of latitude northward with a stronger zonal flux of moisture over the eastern United States. The meridional moisture flux (qυ) is much stronger to the north and east in ORV wet winters (Fig. 7b), with a southwest–northeast orientation, whereas there is little meridional moisture transport in ORV dry winters (Fig. 7d). The mean moisture flux differences (Figs. 7e,f) show a southwest–northeast-oriented maxima toward the ORV. These differences and those of Fig. 6c are spatially collocated with regions where the anomalous geostrophic wind component would have the greatest impact in association with the southeastern United States PNA center in Fig. 1c.
b. Hydrological variability in the ORV
The mean ORV hydrological response in the PNA, NPI, and ENSO extremes is evaluated using the PDSI and PHDI. Changes in mean moisture indices over the winter were obtained by subtracting an ending month (Mar, Apr, or May) from a starting month (Nov, Dec) over the period 1899–1998. The mean PHDI and PDSI value for each precipitation-based quintile was obtained for the six combinations of starting and ending months (Table 4). Despite methodological differences in their calculation, relating to the lags between start and end of dry and wet periods, the March–November time period for both the PDSI and PHDI yields the most significant differences in magnitude of ORV moisture conditions taking place over the December–January–February (DJF) winter (see Table 4). The strength of the moisture indices in indicating dry or wet spells deteriorates when later starting (i.e., Dec) and ending months (i.e., Apr, May) are used. The differences in mean PHDI values between quintiles 1 and 5 are especially large for November and March, ranging from −0.85 (incipient drought) to +1.36 (moist spell), and the standard deviations about the mean are comparatively the lowest.
Two examples, one from the wettest (1949/50) and driest (1976/77) ORV index winters are provided to show typical characteristics of PHDI changes between November and March. The change in the PHDI values between November and March are compared to the winter percentage of normal precipitation for the 32 ORV climate divisions (as well as surrounding climate divisions of each state for continuity). During the 1949/50 winter, zonal flow brought Gulf moisture into the ORV. Winter precipitation was 150%–250% of normal for all ORV climate divisions (Fig. 8a). The November 1949 PHDI ORV soil moisture is a mixture of moist and dry conditions, with drought occurring in much of Ohio and eastern Kentucky (Fig. 8b). PHDI values by March 1950 are uniformly moist throughout the ORV, with the greatest positive PHDI values (≥4.0) occurring along the western Ohio River corridor (Fig. 8c). The largest net ORV hydrological moisture changes (Fig. 8d) occur throughout the drier areas (Fig. 8b) in Indiana, Ohio, and Kentucky. The second wettest winter, 1948/49 (not shown; see Coleman 2000), also produced widespread moist conditions by March 1949, erasing neutral to dry conditions from the preceding November in several ORV climate divisions. All ORV climatic divisions became moist by the following March in both winters.
In contrast, meridional flow prevailed in the 1977 winter with a high amplitude ridge situated in western North America and a deep trough located in the southeastern United States. The diversion of winter storm tracks to the Gulf and East Coasts is evident in the ORV winter precipitation records, where precipitation is only between 25% and 50% of normal along the core climate divisions along the Ohio River and into Kentucky and Ohio (Fig. 9a). Prior to the onset of winter, the 1976 November PHDI values exhibit a distinct soil moisture gradient (Fig. 9b) with dry (negative PHDI) climate divisions west and north of the ORV core area and moist divisions to the south and east. By March 1977 the soil moisture gradient disappears but four ORV climate divisions still have moist conditions (Fig. 9c). There is little net PHDI change over the winter (Fig. 9d) over northern Indiana, while Missouri and southern Illinois ORV climate divisions (see Fig. 2) even become slightly wetter. The largest changes (Fig. 9d) toward dry PHDI conditions occur in the southern and eastern climate divisions, all of which were moist in November. Although the percentage of normal precipitation (Fig. 9a) for the 1977 winter is uniformly low for the entire ORV, the PHDI response is inconsistent. Analysis of the second driest winter, 1962/63 (not shown; see Coleman 2000), exhibited similar tendencies. Winter ORV precipitation was 25%–75% of normal and 13 climate divisions were moist in November 1962. By March 1963, 11 divisions were still moist while fully 17 (of 32) ORV climate divisions exhibit near 0 PHDI change, or a net PHDI increase, despite the lack of winter precipitation.
Based on the PNA and ORV time series (Fig. 5), the winters of 1949–52 were unusually wet while those of 1977–88 were unusually dry. Differences in the ORV index between these periods coincide well with NPI variations (Fig. 5), with above (below) precipitation years generally corresponding to a weakened (intensified) Aleutian low. The post-1976 period is one in which the PNA index became persistently positive (e.g., Trenberth 1990). Using seasonal percentages of normal precipitation, averaged over the 32 ORV climate divisions (Fig. 2), winter precipitation in 1949 and 1950 was 160% and 220% of normal (Fig. 10). The seasons following the wet winters generally exhibit near-normal precipitation except in 1952 when a 5-yr drought begins in the Midwest. The PHDI, however, is continually high through the 1949–52 period (Fig. 11). It is possible that the abundant winter precipitation of 1949 and 1950 may have sustained soil and ground moisture levels into succeeding seasons of 1950 and 1951 as long as nonwinter precipitation remained near normal.
Winter ORV precipitation was well below normal in 1977 and subsequent winters through the late 1980s, with the exceptions of 1979, 1982, 1985, and 1988. Once again the winter precipitation departures do not necessarily persist into subsequent seasons (Fig. 12). PHDI-based soil moisture conditions do not reflect the predominantly dry conditions of the winters (Fig. 13), generally signaling instead the individual nonwinter season's precipitation anomaly. The nonwinter seasonal PHDI values seem to suggest slight to moderate wetness, especially between 1978 and 1986. When the 1986 and 1987 winters are followed by dry springs (Fig. 13), the PHDI moves toward prevailing dry conditions, culminating in the 1988 drought. The pattern for 1977–88 suggests that dry winters are not necessarily followed by dry nonwinter seasons either in precipitation or the PHDI. This is potentially in keeping with the lower spatial regional coherence of the effects of dry winters on the early spring PHDI (Fig. 9).
Mean winter streamflow discharge for selected ORV rivers is given for the highest and lowest PNA and NPI quintile winters (Table 5). Low river discharge occurs during high PNA–low NPI winters while high discharge occurs for low PNA–high NPI winters. Discharge in PNA-negative–NPI-positive winters exceed that of PNA-positive–NPI-negative winters by about 50% beginning in western Pennsylvania on the Monongahela River near the origin of the Ohio River. The southward-flowing Scioto, Great Miami, and Wabash Rivers generally have about twice the mean discharge in PNA-negative–NPI-positive winters although the standard deviations are above 50% of the mean values, reducing the t values in the significance tests. Large highly significant differences occur for the Kentucky River and the Ohio River at Louisville, which have discharges fully 70% and 100% higher in PNA-negative–NPI-positive winters.
A highly significant inverse relationship exists between the phase of the PNA teleconnection pattern and ORV winter precipitation, such that PNA-negative (-positive) mode winters, characterized primarily by zonal (meridional) flow, have above- (below-) normal winter precipitation. Some station correlations between precipitation and the PNA index exceed r = −0.7 since 1947 and reach r = +0.6 to North Pacific sea level pressure (NPI) since 1899. The correlations rival those of other teleconnections, for example, between Azores and Iceland pressure (the North Atlantic Oscillation) and the winter temperature seesaw between Greenland and northern Europe (van Loon and Rogers 1978) caused by the pressure oscillation, both ranging between r = −0.6 and r = −0.7. The statistical significance of the PNA–NPI relationship to ORV winter precipitation is in marked contrast to that of the SOI and Niño-3.4 time series. Relatively warm Niño-3.4 events (SOI negative) typically occur in conjunction with PNA positive mode winters and would be relatively dry. However, the Tahiti–Darwin SOI and Niño-3.4 SSTs have station precipitation correlations that generally do not exceed r = +0.3 and r = −0.3, respectively. Overall, these results support those of Leathers et al. (1991) and verify that the longer-term circulation–climate relationship is fairly stable, given the relatively high NPI correlations extending to 1899.
PNA negative winters have a decisive impact on the hydrology of the ORV. The wet winters of 1948/49 and 1949/50 changed the hydrologic drought index in a manner such that marginally moist conditions in late autumn (Nov) became uniformly very moist by the end of winter (Mar). These same wet winters were also associated with persistently moist soil moisture conditions in the subsequent nonwinter seasons through 1951, aided by near normal nonwinter precipitation. The dry PNA-positive winters such as 1962/63 and 1976/77 produce some tendency for drying in the ORV, as measured by the PHDI, but the pattern is less consistent throughout the region. Instead of a net drying in the index, some climate divisions remained moist (positive index) between November and March or undergo an index increase, indicating a net moistening over the dry winter. From 1977 to 1988, when several PNA-positive winters occurred, nonwinter season soil moisture conditions measured by the PHDI were not necessarily dry. The implications from these cases is that the impact of wet PNA-negative winters is much more substantial on the ORV regional hydrology in subsequent nonwinter seasons than are the dry winter conditions of PNA-positive winters. These case studies are a small sample but represent potential avenues for future research, either in the ORV or in other areas of the continent where the PNA (Fig. 3) or other teleconnections have an impact. Despite the wet winters of 1949–52, the 1952–56 United States drought started with a growing season moisture deficit in 1952. Similarly, the dry winters of 1977–88 had little proclivity to produce drought in that period and the 1987–88 drought began during dry growing seasons (indeed the 1987/88 winter had above-normal ORV precipitation). Streamflow throughout the ORV also responds strongly to the winter atmospheric circulation variations. The variability in PNA negative is particularly large, a product generally of unusually high streamflows in some years of the PNA-negative group of winters.
The strong influence of the PNA in winter ORV precipitation joins that of other climate–PNA relationships occurring over North America. Overall the results suggest that our ability for long-range prediction of the circulation, especially the mode and strength of the PNA pattern, will help provide substantial atmospheric and hydrologic predictability throughout the ORV. At present, however, methods for predicting ORV regional winter precipitation based on El Niño and La Niña extremes appear to offer the best hope in long-range forecasting (Gershunov 1998; Montroy et al. 1998). Efforts to distinguish the PNA signal from that of ENSO in North America climate relationships should be expanded. Anomalous tropical SST forcing may initiate an abrupt change in extratropical circulation (e.g., 1976/77 winter), but may only contribute substantially to the geopotential height variability at the 200 hPa and not at midtropospheric levels where the PNA pattern is most recognizable (Kawamura et al. 1995). In light of Rodionov and Assel's (2001) observation that the PNA centers of action, including the trough and ridge positions and intensity, experience some interannual variability, the PNA–ORV precipitation relationship seems nonetheless strong and stable. The notion prevalent in the popular regional media that El Niño (La Niña) is responsible for dry (wet) winters is also called into question here, the PNA's strong impact implies that other factors are influential and contribute to the strength of this circulation–climate teleconnection.
We thank Sheng-Hung Wang for assistance with the moisture budget analyses and with the figures. Comments from the two anonymous reviewers greatly helped in improving this paper.
Corresponding author address: Jill S. M. Coleman, Dept. of Geography, The Ohio State University, 1035 Derby Hall, 154 North Oval Mall, Columbus, OH 43210-1361. Email: firstname.lastname@example.org