Spatiotemporal variability of the precipitation dipole transition zone in the western United States



[1] El Niño-Southern Oscillation-related hydroclimatic variability in the western United States is characterized by a north–south dipole pattern of precipitation anomalies with opposing signs. Here I use a high-resolution dataset to analyze spatiotemporal patterns in the transition zone between the centers of opposite association. Results indicate that the transition zone is spatially limited west of the continental divide, consisting of a narrow zone within the 40–42°N latitude band and portions of eastern Washington, eastern Oregon, and western Colorado. The transition zone was narrower when the Southern Oscillation Index and the Pacific Decadal Oscillation (PDO) were in a constructive phase and during the negative Atlantic Multidecadal Oscillation phase. Although the transition zone has remained remarkably stationary across the Great Basin, shifts (likely PDO-related) have occurred over time in the West Coast states. Implications for water resource planning in river basins underlying the transition zone vary by region.

1. Introduction

[2] Ecologically and economically damaging episodes of drought and severe wet periods in the western United States (West) have been linked to the El Niño-Southern Oscillation (ENSO) cycle, necessitating characterization of the ENSO-precipitation relationship. Studies of ENSO-related precipitation variability across the West have identified a north–south “seesaw” pivoting on a transition zone that separates Pacific Northwest and Desert Southwest centers of opposite association [Redmond and Koch, 1991; Dettinger et al., 1998]. The north and south centers tend to behave in opposition to each other (i.e., when one is anomalously wet, the other is anomalously dry), and the resulting north–south contrast in precipitation patterns (hereafter referred to as a “dipole”) is an important feature of Western streamflow and climate variability [e.g., Meko and Stockton, 1984; Dettinger et al., 1998].

[3] The dipole precipitation pattern was linked to changes in the Southern Oscillation Index (SOI) by Redmond and Koch [1991], who found that the strongest SOI-precipitation connection is a lagged relationship between summer and fall (June–November) SOI conditions and cool-season (October–March) precipitation. Cool-season precipitation is the West's dominant source of water supply [Hamlet et al., 2005], and this lagged relationship has allowed for long lead-time forecasts for areas near the climatic centers of opposite association. In contrast, the transition zone between the northern and southern response regions, thought to be centered at approximately 40°N latitude [Dettinger et al., 1998], has lacked this predictive capacity. Although this zone has been recognized as a pivot point of interannual and decadal precipitation variability [Dettinger et al., 1998], dipole precipitation patterns are thought to be highly variable over time and space [Brown and Comrie, 2004], and the exact location and behavior of the zone have not been well quantified. With the exception of Dettinger et al. [1998], previous studies examined the dipole pattern using climate divisions, which cover large areas in the West, obscuring the boundary between dipole centers [e.g., Redmond and Koch, 1991; Brown and Comrie, 2004].

[4] The objective of this study is to analyze the spatial boundaries, temporal behavior, and controlling factors of the transition zone between the centers of opposite association using a high-resolution dataset. The lagged ENSO-precipitation relationship has been a useful forecasting tool for regions near the dipole centers. Better understanding of this relationship's spatiotemporal variability and sharper delineation of the boundaries of the transition zone will extend this predictive capacity into regions previously considered part of the transition zone and identify regions with an inconsistent ENSO response. The region between the dipole centers has experienced rapid population growth over recent decades, straining the region's water resources, which include the Sacramento and Upper Colorado river basins. Water resource management and climate forecasting in this arid, highly variable region would benefit from enhanced spatial characterization of the ENSO-precipitation response pattern.

2. Data and Methods

[5] Analyses for this study are based on monthly PRISM precipitation data [Daly et al., 2002]. These data were upscaled to an 8 km grid and restricted to the 1926–2007 time period due to data quality concerns in the early part of the record. All precipitation data were converted to monthly standardized values for further analyses.

[6] The SOI-precipitation connection is the central relationship under study in this paper. SOI, which measures the atmospheric component of the ENSO system, was chosen over other indices of tropical Pacific conditions for its strong lag relationship with winter precipitation in the West [Cayan et al., 1999; Brown and Comrie, 2004]. To establish the strength and stability of the dipole pattern, I also examined modification of the SOI-precipitation relationship due to changing northern Pacific and Atlantic Ocean conditions. The PDO [Mantua et al., 1997], a measure of sea-surface temperature (SST) variability in the northern Pacific Ocean, is an important control on Western climate [Higgins et al., 2007] and is thought to have a modulating effect on the SOI-precipitation relationship at decadal time scales [Gershunov and Barnett, 1998]. There is increasing evidence that the Atlantic Multidecadal Oscillation (AMO) [Enfield et al., 2001], defined by anomalies in northern Atlantic Ocean SSTs, also influences hydroclimatic variability in the United States [Enfield et al., 2001; McCabe et al., 2004], possibly though interactions with Rossby waves and/or circulation patterns [Dong et al., 2006; McCabe et al., 2007]. Although primarily considered an influence on warm-season climate conditions, it has been suggested that the AMO can also modify the SOI-winter precipitation relationship [Enfield et al., 2001].

[7] Monthly values of SOI, PDO, and AMO were obtained from the University of East Anglia's Climatic Research Unit, the University of Washington's Joint Institute for the Study of the Atmosphere and Ocean, and the National Oceanic and Atmospheric Administration's Climate Prediction Center, respectively. Correlations between seasonally-averaged SOI and precipitation data were calculated for 3–6 month season combinations and lags of 0–6 months. The strongest and most significant correlations were found between Jun–Nov SOI and Oct–Mar precipitation, consistent with findings of Redmond and Koch [1991] that were based on a different data set and a 30-yr shorter data record. SOI-precipitation correlations in the remainder of this paper are based on these sets of months. For additional analyses, the SOI was divided into terciles of SOI+, SOI neutral, and SOI− years, with every year assigned to one of these three categories. PDO and AMO analyses are based on sets of years representing phases rather than on individual yearly or seasonal values. The period of analysis includes three PDO phases (1926–1943 [PDO+], 1944–1976 [PDO−], and 1977–1998 [PDO+]) and two phases of the AMO (1926–1963 [AMO+] and 1964–1994 [AMO−]).

[8] I define the transition zone as the area dividing the dipole centers of opposite association and lacking an ENSO-precipitation relationship based on correlation and precipitation anomaly patterns. The correlations between Jun–Nov SOI and Oct–Mar precipitation were calculated for each grid point over 1926–2007, for the PDO+ and PDO− phases, and over AMO+ and AMO− phases. Precipitation anomalies were calculated for each of the SOI terciles and the PDO and AMO phases, as well as for SOI/PDO and SOI/AMO phase combination subsets. Statistical significance of the anomaly patterns was tested using a permutation resampling method. Sets of Oct–Mar precipitation anomaly values were randomly sampled without replacement from the set of all Oct–Mar periods in the 82-year record and averaged together to calculate mean anomaly values. The number of years sampled corresponded to the number of years in the phase combination to which it was compared. This process was repeated 1000 times for each SOI/PDO/AMO combination. Anomalies from the original data set were considered significant at the 0.1 level if their magnitude exceeded the 95th or 5th percentile of the random permutations.

[9] The centerline of the transition zone was delineated by the grid cells with lowest correlation between precipitation and SOI or least anomalous precipitation during SOI+ and SOI− years. In order to examine the north–south movement of the SOI-precipitation relationship and changes in intensity over time, 31-yr moving correlations were calculated over the 1926–2007 time period (center years from 1941 to 1992) and averaged over three 5° longitudinal bands.

3. Results

[10] There is a surprisingly narrow boundary between the northern and southern sections of the dipole (Figure 1). Cool-season precipitation across most of the West, particularly west of the continental divide, is significantly correlated with SOI. The main exception is the line delineating the transition zone that begins around 40°N in California and shifts poleward to approximately 42°N at the continental divide. The Southwest is more strongly and cohesively correlated with SOI than the Northwest. Two noteworthy anomalous areas exist outside of the main transition zone in areas of unusual topography: a low-lying region in eastern Washington where the leeward side of the Cascades meets the Columbia Basin, and a high-elevation area primarily on the windward side of the Colorado Rocky Mountains (Figure 1).

Figure 1.

Correlation coefficients between Jun-Nov SOI and Oct-Mar precipitation, 1926–2007. Correlations significant at the 0.1 level are highlighted and contoured. Continental divide shown by thick, black line.

[11] ENSO conditions impact precipitation patterns through shifting pressure patterns and storm track position. Negative SOI values are associated with the development of a deep Aleutian low, leading to ridging over the West and a south-shifted storm track position [Cayan and Redmond, 1994]. This typically occurs when the tropical Pacific is in its warm “El Niño” phase (negative SOI) and results, on average, with a relatively dry Northwest and a wet Southwest. In contrast, positive SOI values are associated with a more northerly storm track, an anomalously wet Northwest and dry Southwest, a weak Aleutian low, and lower pressure over the West [Cayan and Redmond, 1994]. Results of this study show that the precipitation anomaly patterns in SOI+ and SOI− years are remarkably close to being mirror images of each other, although important differences exist (see auxiliary material Figures S1–S3). The transition zone between positive and negative anomalies moves southward in SOI+ years in California and parts of Nevada, but there is very little change over Utah. Southern Oregon, southern Idaho, and parts of Wyoming tend to have a stronger ENSO response during SOI+ years (Figures S2S3). The dipole pattern is not characteristic of neutral SOI years (Figure S2).

[12] The PDO+ phase is associated with a deep Aleutian low and ridging over the West (see Figure S4) [Gershunov and Barnett, 1998], similar to the SOI− pattern. Research suggests that the West's dipole is stronger when PDO and SOI are in a constructive phase (SOI+/PDO− or SOI−/PDO+) [Gershunov and Barnett, 1998], and results from the current analysis (Figure 2) are consistent with this hypothesis. As shown in Figures S4S5, the synoptic pressure patterns resembled the overall SOI+ or SOI− patterns during these constructive phases but indicate an eastward shift in the pattern during the destructive phases (SOI+/PDO+ and SOI−/PDO−). This change is reflected in the SOI+/PDO+ anomaly pattern, as the positive precipitation anomalies have also shifted to the east (Figure 2). In the two PDO+ phases, the transition zone generally shifted north relative to the PDO− phase, particularly in the West Coast states (Figure S1).

Figure 2.

Standardized Oct-Mar precipitation anomalies by SOI/PDO and SOI/AMO phase. Anomalies significant at the 0.1 level are highlighted and contoured.

[13] The dipole pattern was weak in the 1926–1963 AMO+ phase for both ENSO phases, while the AMO− phase (1964–1994) was characterized by a strong dipole pattern for both SOI+ and SOI− years (Figure 2). The anomaly pattern indicates a stronger connection between SOI and precipitation during the AMO− phase, along with a larger area of significant anomalies stretching across Idaho and Montana (Figure 2). Geopotential height patterns show that the pressure patterns characteristic of SOI dominated during the SOI+/AMO− and SOI−/AMO− years. However, during SOI+/AMO+ and SOI−/AMO+ conditions, the AMO+ synoptic pressure pattern dominated, with high pressure extending across much of the United States (Figures S4S5).

[14] Examined temporally across longitude bands, the most striking north–south transition zone movement occurred in the western most band (119–124°W) over a period centered in the 1950s, with a southward shift in the transition zone of 4–5° latitude, as well as a strengthening of positive SOI-precipitation correlations at higher latitudes (Figure 3). Although temporally blurred due to the application of a moving correlation filter, the time period of these changes roughly corresponds with a period of extremely negative PDO values in the 1950s. This is consistent with the influence of negative PDO conditions on the SOI-precipitation relationship (Figure S1).

Figure 3.

31-yr moving correlations (assigned to center year of the window) between Jun-Nov SOI and Oct-Mar precipitation as a function of latitude (y-axis) and time (x-axis) and corresponding to the longitudinal bands in map inset. Black line indicates center (zero-line) of transition zone. Correlations significant at the 0.1 level are contoured with gray line.

[15] Although there was a slight transition zone shift south and a strengthening of northern positive correlations near the middle of the record in the 114–119°W and 109–114°W longitude bands, there was little overall change through time. The center location of the transition zone across Nevada and Utah has remained stationary, particularly in the last 30 years of the record, at approximately 42°N latitude. The 109–114°W band does indicate change over time in western Montana, from negative or no correlation in the beginning of the record to positive correlation in later years (Figure 3). Statistical significance of the correlations shown in Figure 3 is generally low due to the short (31 year) moving average window and the fact that each encompasses a 5° longitude band.

4. Discussion

[16] This study's high-resolution analysis of the Western precipitation dipole's spatial boundaries and movement over time demonstrates the importance of teleconnections in the West outside the Pacific Northwest and Desert Southwest centers of opposite association. Winter precipitation over most of the West is significantly correlated with SOI (Figure 1). The transition zone between the centers of opposite association is spatially limited, particularly west of the continental divide, consisting of a narrow zone within the 40–42°N latitude band and portions of eastern Washington, eastern Oregon, and western Colorado.

[17] The strength and location of the dipole transition zone are modulated by ocean-atmosphere conditions and have varied through time (Figure 2). The core Southwest region generally maintained a coherent, negative correlation between SOI and winter precipitation. Most changes occurred along the transition zone and in the Northwest, which responded more consistently to SOI when SOI and PDO were in a constructive phase and during the negative AMO phase. The dipole pattern was stronger when AMO was negative (Figure 2). This is based on only two AMO phases, however, and it is possible that the strengthened pattern in the later (AMO−) phase represents a temporal trend rather than a phase association.

[18] Results of this study indicate relatively little change in the location of the transition zone between SOI− and SOI+ conditions, but there are some regions that respond consistently to only one of the phases. The 42–45°N belt showed a stronger response with SOI+, while central California, the Four Corners region, northern New Mexico, and areas north of 45°N responded more consistently to SOI− (Figure S3). There were large differences in the location of the transition zone boundary between PDO phases, with a northward movement along the West Coast under PDO+ conditions (Figure S1). An eastward shift in synoptic pressure patterns and precipitation anomalies was observed when SOI and PDO were in destructive phases (Figure S5).

[19] The center of the transition zone between the northern and southern portions of the dipole has remained remarkably stationary through time across the Great Basin. The transition zone boundary across Nevada and Utah has ranged from approximately 40–42°N over the study period (Figure 3). Larger shifts occurred in other areas, particularly the West Coast states. In California, the center of the transition zone has ranged from 37–42°N. The Sacramento River Basin, which underlies this transition area, has shifted between the northern and southern sides of the dipole over time and is one of the few regions with no significant precipitation anomalies in any of the eight phase combinations examined in this study (Figure 2 and Figure S6).

[20] The large Upper Colorado River Basin (UCRB) has a complex ENSO response. Hidalgo and Dracup [2003] and others have noted a lack of cold-season response to ENSO signals in the UCRB, with the exception of a few high-elevation precipitation stations. This study has shown that the transition zone boundaries have shifted in the Colorado and Wyoming portions of the UCRB, and the overall proportion of the basin exhibiting a Northwest versus Southwest response has changed over time. Although most of the basin had significant precipitation anomalies in one or more phase combinations (Figure 2 and Figure S6), there has been a great deal of variability over time and space and the UCRB has never belonged entirely to one side of the dipole or the other. The spatial delineation of this pattern could aid in studying the basin's SOI response by allowing the separation of similarly-responding watersheds.

[21] There are limitations hindering the application of this information to climate forecasting. ENSO forecasts are of little use in regions that lack a consistent precipitation response in any of the phase combinations. This includes the northern coastal and Sierra Nevada regions of California, northern Nevada, southeast Oregon, and central Washington (Figure S6). Even in regions with a consistent response, SOI only explains a portion of precipitation variability. Year-to-year fluctuations in the transition zone's boundary and lack of predictability in the phase shifts of important modulating conditions like the PDO make forecasting difficult. Projected climate changes add an additional level of uncertainty; for instance, it is uncertain whether warming temperatures will push tropical Pacific conditions towards a more El Niño- or La Niña-like state [Vecchi et al., 2008]. Despite these limitations, sharper delineation of transition zone boundaries and identification of regions with an inconsistent or phase-dependent ENSO response are useful steps towards formulating high resolution, long-lead time climate forecasts.


[22] This research was supported by the United States Environmental Protection Agency (EPA) under the Science to Achieve Results (STAR) Graduate Fellowship Program. EPA has not officially endorsed this publication and the views expressed herein may not reflect the views of the EPA.