Recent rainfall cycle in the Intermountain Region as a quadrature amplitude modulation from the Pacific decadal oscillation



[1] Precipitation in the Intermountain Region has experienced a pronounced 10–20 yr cycle during the last four decades. We present evidence of a quadrature phase coupling of these 10–20 yr cycles between the Pacific Decadal Oscillation (PDO) and the Intermountain rainfall. The PDO has been linked to precipitation anomalies in the Northwest and Southwest U.S., while the Intermountain Region remains marginal. During the transition phase of a 10–20 yr cycle in the PDO, an anomalous circulation dipole over the Gulf of Alaska is formed. This circulation dipole modulates the synoptic activity that produces rainfall in the Intermountain Region. As a result, the Intermountain rainfall cycle has repeatedly lagged the PDO cycle by a quarter phase, amounting to about 3 to 4 years. Future understanding of the 10–20 yr cycle embedded in the PDO and its quadrature phase modulation may help predict the Intermountain climate.

1. Introduction

[2] The Intermountain Region experiences large temporal variability in precipitation and periodic drought is endemic. This region receives its yearly majority of precipitation during the cold season, while most of the precipitation falls over the windward side of mountain ranges (Figure 1a) [e.g., Leung et al., 2003]. In addition to the orographic effects, the rainfall climatology in this region is also governed by synoptic weather systems associated with the large-scale circulation of the atmosphere and oceans. Previous studies [Dettinger et al., 1998; Jain and Lall, 2000; Hidalgo and Dracup, 2003] have pointed out that the Intermountain precipitation is significantly correlated with the El Niño-Southern Oscillation (ENSO) that has a 2–7 year recurrence interval and the PDO that changes sign about every 25 years [Mantua et al., 1997] or longer [Minobe, 1999].

Figure 1.

(a) Cold season precipitation (Oct–Apr; contours) superimposed with terrain (shadings). Precipitation contours begin from 0.9 inch · month-1 (yellow) until 4 (red) in 0.3 inch · month−1 increments and continue with purple contours in 1 inch · month−1 increments. (b) Monthly station rainfall at Utah State University (USU) in Logan, UT (station ID 425186, marked by a star in Figure 1a) filtered by the 18-month lowpass (thin line) and 10–20 yr bandpass (thick line) frequencies. Power spectra of the unfiltered USU rainfall is plotted on the right (thick line) with the 99% confidence level (dashed line). (c) and (d) Same as Figure 1b but for rainfall and the PDSI, respectively, averaged in the Intermountain domain outlined in Figure 1a (thick dashed lines). (e) Same as Figure 1b but for the PDO index. (f) Cross-correlation functions between the Intermountain rainfall and the PDO timeseries of the two frequencies. Orange dashed lines indicate the 95% confidence limits.

[3] During the past four decades, precipitation near the central Intermountain Region has experienced a pronounced increase in temporal variability, as shown in the monthly rainfall time series observed at Utah State University (USU) in Logan, UT (Figure 1b). Visual inspection indicates that the precipitation time series from 1970 onwards undergoes an amplified cycle with a frequency of about 15 years, a time scale that falls between the 2–7 yr ENSO cycle and the ≥25 yr PDO variation. Such a precipitation cycle echoes Hidalgo and Dracup [2003] noting a distinct 10–16 yr fluctuation of precipitation in the Colorado River Basin after 1977, as well as Jain and Lall [2000] uncovering the decadal signal of streamflow near Logan, UT. Various studies [e.g., McCabe and Dettinger, 1999] attributed the periodicity change of rainfall variations during the 1970s to the global-scale “climate regime shift” found in the ENSO activity [Trenberth and Hoar, 1997] and the PDO evolution [Mantua et al., 1997; Minobe, 1999]. However, the mechanism causing the increasingly periodic rainfall variation in the Intermountain Region has not been well documented. This study examines the apparent rainfall periodicity in the Intermountain Region and its association with the circulation system.

2. Data, Domain and the PDO Index

[4] The Intermountain Region is defined as the domain roughly encircling the area with elevation higher than 1500 m (Figure 1a). The gauge-based monthly precipitation data is obtained from the University of Delaware with a 0.5° spatial resolution [Legates and Willmott, 1990]. The grid points in which the cold-season mean precipitation (October–April) is larger than 0.5 inch·month−1 within this domain are averaged to form the “Intermountain rainfall” time series (Figure 1c; thin line). Despite some year-to-year fluctuations, the low frequency rainfall variations in Figures 1b and 1c are consistent, suggesting that such periodic variations are at the regional scale. Drought conditions are represented by the gridded monthly Palmer Drought Severity Index (PDSI) [Dai et al., 2004]. The PDSI time series (Figure 1d) is compiled by averaging all grids within the Intermountain domain. All monthly data, including the following, are applied with an 18-month lowpass filtering to remove the seasonal cycle.

[5] The atmospheric and sea surface temperature (SST) datasets are the 2.5° resolution NCEP/NCAR Global Reanalysis [Kalnay et al., 1996] and the 5° resolution Kaplan Extended SST (version 2) [Kaplan et al., 1998], respectively. Using the Kaplan SST dataset, this study follows Mantua et al. [1997] to define the PDO index as the first principal component of monthly SST variation in the North Pacific. However, to better compare with the large-scale circulation associated with the SST forcing, we adopt Zhang et al.'s [1997] global domain (25°S–60°N) to construct the PDO index. This domain includes the tropical SST variations that are absent in Mantua et al.'s [1997] analysis domain of 20°N poleward. Except for the USU station obtained from, all data are provided by the NOAA/OAR/ESRL PSD (

3. Results

[6] Power spectral analysis on the USU rainfall (unfiltered; Figure 1b right) shows an elevated frequency band at 10–20 yrs with a peak at 12 yrs. Equivalent analysis on the unfiltered Intermountain rainfall and PDSI also reveals distinct 10–20 yr signals (Figures 1c and 1d, right), though the PDSI has a stronger signal in this frequency zone. The large fluctuation of the PDSI cycles highlights the magnified effect of the rainfall cycle on drought conditions. The PDO index (Figure 1e) also reveals a clear 10–20 yr variation becoming more pronounced after the 1970s. This observation agrees with Mantua et al. [1997] noting increasing fluctuations with “a few years in length” in the post-1977 PDO, as well as Minobe [1999] showing a particularly energetic “bidecadal” PDO periodicity in the late 20th century. The 10–20 yr signal is statistically significant in the spectrum among all time series. As a result, the 10–20 yr bandpass filtered time series (Figures 1b1e) appear to fit the lowpass time series quite well. There is a consistent time lag between the phases of the PDO and the Intermountain rainfall, with the PDO cycle leading rainfall by about 3 to 4 years. Cross-correlation function (CCF; Figure 1f) of the bandpass filtered PDO and Intermountain rainfall depict a clear oscillation pattern repeating about every 12 yrs. The ∼3 yr lead time of the PDO appears to be a quarter phase of this ∼12 yr cycle. CCF of the 18-month lowpass filtered PDO and rainfall (thin line) shows significant oscillations with the same periodicity and phase lag, supporting the ∼12 yr cycle.

[7] When the Kaplan SST is regressed upon the Intermountain rainfall, with a three year lead in rainfall, the spatial SST pattern (Figure 2a) depicts a distinct positive PDO phase with a broad, positive area in the tropical central Pacific and strong, negative area in the central North Pacific [Mantua et al., 1997; Zhang et al., 1997]. Note that the rainfall and SST time series, as well as all variables in Figures 2 and 3, are filtered with the 10–20 yr frequency band. The instantaneous regression map (Figure 2b) delivers a very different SST pattern from the PDO with the major positive areas shifting to the eastern tropical Pacific and the southern Atlantic, accompanied by zonally elongated and meridionally stratified SST patterns in the central North Pacific. The tropical SST pattern coincides with that formed by the quasidecadal oscillation of ENSO [White and Liu, 2008].

Figure 2.

Kaplan SST regressed upon the standardized Intermountain rainfall (a) leading by 3 yrs and (b) without lead time, (c) EOF1 and (d) EOF2 as in Figure 3 for the period of record 1950–2007. Contour interval is 0.1°C. Dark red (blue) contours depict positive (negative) SST, while yellow shadings outline the 99% confidence level. All variables are filtered with the 10–20 yr frequency band.

Figure 3.

Rotated EOF analysis of 200mb eddy streamfunction (i.e., removing the zonal mean) with (a) the first leading mode (EOF1) and (b) the second leading mode (EOF1) with time series on the right (thick lines), superimposed with the Intermountain rainfall (shadings). (c) The EOF dial of EOF1 (y-axis) vs. EOF2 and (d) the PDO index (x-axis) are shown, superimposed with the Intermountain rainfall anomalies (triangles) every other month. The legend of rainfall anomalies is given. All variables are filtered with the 10–20 yr frequency band.

[8] The atmospheric circulations responding to the SST anomalies are delineated by an empirical orthogonal function (EOF) analysis on the eddy streamfunction. The leading EOF modes with the sum variances exceeding 90% are recombined using a varimax rotation. The rotated EOF is performed using only the 1970–2007 period. We show the 200 mb streamfunction as it has the strongest response to the global divergent circulation that is sensitive to SST anomalies. The first mode (EOF1; Figure 3a) is dominated by a zonal wavenumber-1 stationary eddies in the tropics, mixed with shorter waves in the middle and high latitudes. The Pacific-North America teleconnection pattern [Wallace and Gutzler, 1981] is highly visible, and the higher-latitude short waves resemble those excited during extreme ENSO phases [Chen, 2002] when extensive regions of Pacific SST anomalies are often present [Zhang et al., 1997]. Time series of EOF1 (Figure 3a) appears to lead the Intermountain rainfall by ∼3 yrs, and the SST regressed upon EOF1 (Figure 2c) reveals a typical positive PDO pattern resembling Figure 2a, indicating that the circulation pattern in Figure 3a is closely associated with the PDO. The second mode (EOF2; Figure 3b), exhibiting a stronger wave-form circulation pattern, depicts a regional circulation dipole with a cyclonic cell over the Gulf of Alaska and an anticyclonic cell off the West Coast. Time series of EOF2 (Figure 3b) expresses a very coherent rhythm with the Intermountain rainfall, and the SST pattern regressed on EOF2 (Figure 2d) is nearly identical to Figure 2b. The results suggest that the Intermountain rainfall variation is closely linked to the SST and circulation anomalies associated with the 10–20 yr PDO cycle.

[9] To support the idea of “cycle”, an EOF dial is generated by connecting the time series of EOF1 and EOF2 (Figure 3c). The track of EOF1 and EOF2 completes three revolutions that can only form when EOF1 is temporally in quadrature with EOF2 with the same variation frequency. The dial of EOF2 with the PDO (Figure 3d) again shows three well organized revolutions. Superimposing the Intermountain rainfall with the EOF dial, positive and negative rainfall anomalies are clearly separated by EOF2, suggesting that the Intermountain rainfall is profoundly modulated by circulation patterns associated with the rising and falling transitions of the 10–20 yr PDO cycle.

[10] The anomalous circulation dipole off the Pacific Northwest coast (Figure 3b) is known to lead to anomalies in U.S. climate [e.g., Namias, 1978]: synoptic short waves approaching the West Coast often deepen and produce cold frontogenesis in the Intermountain Region [Shafer and Steenburgh, 2008]. We generated the transient activity at 500 mb from October to April, using the root-mean-square of meridional winds filtered by 2–7 days, to examine the synoptic response to the 10–20 yr PDO cycle. Figure 4a displays the departures of composite 500mb wind vectors and transient activity from the climatology during high-index years of EOF2 (corresponding to the wet period), including the years of 1982–84, 1996–99, and 2006–07. The cyclonic circulation over the Gulf of Alaska, which reflects the 200mb cyclonic cell in Figure 3b, intensifies the synoptic transient activity upstream of the Intermountain Region. During low-index years of EOF2 (corresponding to the dry years of 1976–79, 1988–92, and 2001–04), the anomalous anticyclonic cell (Figure 4b) produces an opposite effect that decreases the transient activity and arguably reduces rainfall resulting from synoptic waves. These circulation patterns echo the circulation anomalies linked to floods near Logan, UT, shown by Jain and Lall [2000].

Figure 4.

Departures of the composite 500 mb wind vectors and transient activity, ΔRMS(v′), from the climatology during (a) high-index years and (b) low-index years of EOF2 (see text). Values of ΔRMS(v′) above the 95% confidence level are shaded, while wind vectors above the 95% confidence level are in black and otherwise in gray.

4. Discussion

[11] Previous studies emphasizing the teleconnection impacts of ENSO and the PDO on precipitation have shown significant correlations over the northwest and southwest U.S. The central Intermountain Region falls into the marginal zone of these direct influences [e.g., Dettinger et al., 1998; Rajagopalan and Lall, 1998]. Our results show that the Intermountain rainfall is linked to the 10–20 yr cycle embedded in the PDO, but with a quadrature phase-shift in which the rainfall cycle lags the PDO cycle by ∼3 yrs. Meteorological analysis suggests that recurring circulation anomalies are formed over the Gulf of Alaska during the rising and falling transitions of this PDO cycle, which modulate the synoptic activity and affect rainfall in the Intermountain Region.

[12] Though it is reasonable to attribute the increased 10–20 yr PDO variability to the widely reported climate regime shift, physical explanations related to the apparent change in climate regime needs further investigation. One suggestion is that the enhanced El Niño pattern after 1975 amplifies the circulation's responses to the PDO [Zhang et al., 1997; Trenberth and Hoar, 1997], because a majority of the PDO variability is linked to ENSO variations [Newman et al., 2003]. Such amplification may be sensitive to the SST and circulation patterns during the transition phases of this 10–20 yr PDO cycle. How the transition-phase SST pattern is formed and how it induces the circulation dipole over the Gulf of Alaska remain open questions. Clarke [2008] provided a comprehensive review on the physics of the upper air response to ENSO heating and teleconnections. Nonetheless, under the assumption that the current climate regime maintains its post-1975 status, it is optimistic that the understanding of this 10–20 yr PDO cycle can help predict the Intermountain climate beyond the seasonal time scale.


[13] This study is supported by the USDA-CSREES funded Drought Management, Utah project, and the Utah Agricultural Experiment Station, Utah State University, and approved as journal paper 8016.