• Cañon, J., , González J. , , and Valdés J. , 2007: Precipitation in the Colorado River Basin and its low frequency associations with PDO and ENSO signals. J. Hydrol., 333 , 252264.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cayan, D. R., , Redmond K. T. , , and Riddle L. G. , 1999: ENSO and hydrologic extremes in the western United States. J. Climate, 12 , 28812893.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cole, J. E., , Overpeck J. T. , , and Cook E. R. , 2002: Multiyear La Niña events and persistent drought in the contiguous United States. Geophys. Res. Lett., 29 , 1647. doi:10.1029/2001GL013561.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cook, E., , Woodhouse C. , , Eakin C. M. , , Meko D. , , and Stahle D. , 2004: Long-term aridity changes in the western United States. Science, 306 , 10151018. doi:10.1126/science.1102586.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cressie, N. A. C., 1991: Statistics for Spatial Data. Wiley, 900 pp.

  • Gershunov, A., , and Barnett T. P. , 1998: Interdecadal modulation of ENSO teleconnections. Bull. Amer. Meteor. Soc., 79 , 27152725.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gochis, D. J., , Brito-Castillo L. , , and Shuttleworth W. J. , 2007: Correlations between sea-surface temperatures and warm season streamflow in northwest Mexico. Int. J. Climatol., 27 , 883901. doi:10.1002/joc.1436.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hamlet, A. F., , and Lettenmaier D. P. , 2005: Production of temporally consistent gridded precipitation and temperature fields for the continental United States. J. Hydrometeor., 6 , 330336.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hamlet, A. F., , Mote P. W. , , Clark M. P. , , and Lettenmaier D. P. , 2007: Twentieth-century trends in runoff, evapotranspiration, and soil moisture in the western United States. J. Climate, 20 , 14681486.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, S-C., , Shum C. K. , , Jekeli C. , , and Alsdorf D. , 2005: Improved estimation of terrestrial water storage changes from GRACE. Geophys. Res. Lett., 32 , L07302. doi:10.1029/2005GL022382.

    • Search Google Scholar
    • Export Citation
  • Hasan, S., 2009: Terrestrial water storage change from temporal gravity variation. Ph.D. thesis, Wageningen University, 83 pp.

  • Hidalgo, H. G., , and Dracup J. A. , 2003: ENSO and PDO effects on hydroclimatic variations of the upper Colorado River Basin. J. Hydrometeor., 4 , 523.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hirschi, M., , Seneviratne S. I. , , and Schär C. , 2006: Seasonal variations in terrestrial water storage for major midlatitude river basins. J. Hydrometeor., 7 , 3960.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hu, Q., , and Feng S. , 2001: Variations of teleconnection of ENSO and interannual variation in summer rainfall in the central United States. J. Climate, 14 , 24692480.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Johansson, A., 2007: Prediction skill of the NAO and PNA from daily to seasonal time scales. J. Climate, 20 , 19571975.

  • Liang, X., , Lettenmaier D. P. , , Wood E. F. , , and Burges S. J. , 1994: A simple hydrologically based model of land surface water and energy fluxes for general circulation models. J. Geophys. Res., 99 , (D7). 1441514458.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • MacDonald, G. M., , and Case R. A. , 2005: Variations in the Pacific Decadal Oscillation over the past millennium. Geophys. Res. Lett., 32 , L08703. doi:10.1029/2005GL022478.

    • Search Google Scholar
    • Export Citation
  • Maity, R., , and Kumar D. N. , 2008: Basin-scale stream-flow forecasting using the information of large-scale atmospheric circulation phenomena. Hydrol. Processes, 22 , 643650. doi:10.1002/hyp.6630.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mantua, N. J., , and Hare S. R. , 2002: The Pacific Decadal Oscillation. J. Oceanogr., 58 , 3544.

  • Mantua, N. J., , Hare S. R. , , Zhang Y. , , Wallace J. M. , , and Francis R. C. , 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78 , 10691079.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCabe, G. J., , Palecki M. A. , , and Betancourt J. L. , 2004: Pacific and Atlantic Ocean influences on multidecadal drought frequency in the United States. Proc. Natl. Acad. Sci. USA, 101 , 41364141. doi:10.1073/pnas.0306738101.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Newman, M., 2007: Interannual to decadal predictability of tropical and North Pacific sea surface tempperatures. J. Climate, 20 , 23332355.

  • Niu, G-Y., , and Yang Z-L. , 2006: Assessing a land surface model’s improvements with GRACE estimates. Geophys. Res. Lett., 33 , L07401. doi:10.1029/2005GL025555.

    • Search Google Scholar
    • Export Citation
  • Okin, G. S., , and Reheis M. C. , 2002: An ENSO predictor of dust emission in the southwestern United States. Geophys. Res. Lett., 29 , 1332. doi:10.1029/2001GL014494.

    • Search Google Scholar
    • Export Citation
  • Redmond, K. T., , and Koch R. W. , 1991: Surface climate and streamflow variability in the western United States and their relationship to large-scale circulation indices. Water Resour. Res., 27 , 23812399.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Seager, R., and Coauthors, 2007: Model projections of an imminent transition to a more arid climate in southwestern North America. Science, 316 , 11811184. doi:10.1126/science.1139601.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Seneviratne, S. I., , Viterbo P. , , Lüthi D. , , and Schär C. , 2004: Inferring changes in terrestrial water storage using ERA-40 reanalysis data: The Mississippi River basin. J. Climate, 17 , 20392057.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Swenson, S. C., , and Milly P. C. D. , 2006: Climate model biases in seasonality of continental water storage revealed by satellite gravimetry. Water Resour. Res., 42 , W03201. doi:10.1029/2005WR004628.

    • Search Google Scholar
    • Export Citation
  • Tapley, B. D., , Bettadpur S. , , Watkins M. , , and Reigber C. , 2004: The gravity recovery and climate experiment: Mission overview and early results. Geophys. Res. Lett., 31 , L09607. doi:10.1029/2004GL019920.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., 1997: The definition of El Niño. Bull. Amer. Meteor. Soc., 78 , 27712777.

  • Troch, P. A., , Durcik M. , , Seneviratne S. I. , , Hirschi M. , , Teuling A. J. , , Hurkmans R. , , and Hasan S. , 2007: New data sets to estimate terrestrial water storage change. EOS, Trans. Amer. Geophys. Union, 88 .doi:10.1029/2007EO450001.

    • Search Google Scholar
    • Export Citation
  • Wolter, K., , and Timlin M. S. , 1998: Measuring the strength of ENSO events: How does 1997/98 rank? Weather, 53 , 315323.

  • View in gallery

    Location of and elevations in the CRB. The basin outline represents the discretization of the basin in VIC at a resolution of 0.25°. The Colorado River is shown as a black line.

  • View in gallery

    Time series of spatially averaged anomalies of (a) storage components and total storage in VIC, (b) outflow (surface runoff + baseflow) and precipitation, and (c) four climate indices (MEI, PDO, Niño-3.4, and PNA). For clarity, 12-month running averages were calculated for each variable.

  • View in gallery

    ACFs for all (a) anomalies and (c) indices displayed in Fig. 2, up to a time lag of 180 months (15 yr). Gray shaded areas indicate autocorrelations not significantly different from zero. Power spectra are also shown for (b) anomalies and (d) climate indices. Numbers in (d) represent the return period in years of the strongest spectral peaks of Niño-3.4 and PDO. Both (a) and (b) show six lines [shallow soil moisture (SM), deep SM, SWE, TWS, discharge, and precipitation) for which the legend is split between (a) and (b).

  • View in gallery

    Correlation between 6-month moving averages of five hydrologic anomalies and monthly values of four climate indices (on the x axes). The y axes therefore represent twelve 6-month moving averages: (bottom)–(top) February–July (F–J), April–September (A–S), June–November (J–N), August–January (A–J), October–March (O–M), December–May (D–M). Correlations that are significant at α = 0.05 are denoted by a black dot.

  • View in gallery

    Time series of (left) monthly Niño-3.4, (right) PDO, and 6-month averaged shallow SM, similar to the data used to create Fig. 4. Four pixels from the matrices plotted in Fig. 4 are used: (a),(b) the M indices vs M–S averaged shallow SM; (c),(d) M indices vs S–F averaged shallow SM; (e),(f) S indices vs M–S averaged shallow SM; and (g),(h) S indices vs S–F averaged shallow SM. Also the cross-correlation coefficients (without time lag) are shown in every panel. Gray shaded areas indicate El Niño (positive) and La Niña (negative) events.

  • View in gallery

    Interannual variability investigated through time series of 24-month running averages of (a) deep SM, (b) shallow SM, (c) SWE, (d) precipitation, and (e) discharge. Niño-3.4 and PDO are plotted in all plots. In addition, the storage anomaly of Lake Mead is plotted in (a). To illustrate the correlation between deep SM and reservoir storage and PDO phases, the 10-yr running means of the deep SM anomaly (black) and that of the storage anomaly of Lake Mead (gray) are plotted as well in (a) as thick dashed lines. Black vertical lines denote the shifts from warm to cool and again to warm PDO phase in all panels, and the dashed line at 1998 denotes a possible third phase change.

  • View in gallery

    Maps of correlation coefficients between anomalies of three variables (shallow SM, discharge, and precipitation) and the climate index Niño-3.4. Only significant correlations (α = 0.05) are shown. Time series were aggregated similar to Fig. 4; calendar months of climate indices are correlated with 6-month averages of anomalies. Here, the climate index values for January (Jan) and July (Jul) are shown, correlated with anomaly values averaged over January–June (winter), or July–December (summer).

  • View in gallery

    Same as Fig. 7 but for the climate index PDO.

  • View in gallery

    Maps of temporal standard deviations of anomalies of three storage components (shallow SM, deep SM, and snowpack), total storage, outflow, and precipitation. Storage components are in mm, fluxes in mm month−1.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 36 36 3
PDF Downloads 7 7 0

Effects of Climate Variability on Water Storage in the Colorado River Basin

View More View Less
  • 1 Hydrology and Quantitative Water Management Group, Wageningen University, Wageningen, Netherlands
  • | 2 Department of Hydrology and Water Resources, University of Arizona, Tucson, Arizona
  • | 3 Hydrology and Quantitative Water Management Group, Wageningen University, Wageningen, Netherlands
  • | 4 Sustainability of Semi-Arid Hydrology and Riparian Areas, Department of Hydrology and Water Resources, University of Arizona, Tucson, Arizona
© Get Permissions
Full access

Abstract

Understanding the long-term (interannual–decadal) variability of water availability in river basins is paramount for water resources management. Here, the authors analyze time series of simulated terrestrial water storage components, observed precipitation, and discharge spanning 74 yr in the Colorado River basin and relate them to climate indices that describe variability of sea surface temperature and sea level pressure in the tropical and extratropical Pacific. El Niño–Southern Oscillation (ENSO) indices in winter [January–March (JFM)] are related to winter precipitation as well as to soil moisture and discharge in the lower Colorado River basin. The low-frequency mode of the Pacific decadal oscillation (PDO) appears to be strongly correlated with deep soil moisture. During the negative PDO phase, saturated storage anomalies tend to be negative and the “amplitudes” (mean absolute anomalies) of shallow soil moisture, snow, and discharge are slightly lower compared to periods of positive PDO phases. Predicting interannual variability, therefore, strongly depends on the capability of predicting PDO regime shifts. If indeed a shift to a cool PDO phase occurred in the mid-1990s, as data suggest, the current dry conditions in the Colorado River basin may persist.

Corresponding author address: Ruud Hurkmans, Hydrology and Quantitative Water Management Group, Wageningen University, P.O. Box 47, 6700 AA, Wageningen, Netherlands. Email: ruud.hurkmans@wur.nl

Abstract

Understanding the long-term (interannual–decadal) variability of water availability in river basins is paramount for water resources management. Here, the authors analyze time series of simulated terrestrial water storage components, observed precipitation, and discharge spanning 74 yr in the Colorado River basin and relate them to climate indices that describe variability of sea surface temperature and sea level pressure in the tropical and extratropical Pacific. El Niño–Southern Oscillation (ENSO) indices in winter [January–March (JFM)] are related to winter precipitation as well as to soil moisture and discharge in the lower Colorado River basin. The low-frequency mode of the Pacific decadal oscillation (PDO) appears to be strongly correlated with deep soil moisture. During the negative PDO phase, saturated storage anomalies tend to be negative and the “amplitudes” (mean absolute anomalies) of shallow soil moisture, snow, and discharge are slightly lower compared to periods of positive PDO phases. Predicting interannual variability, therefore, strongly depends on the capability of predicting PDO regime shifts. If indeed a shift to a cool PDO phase occurred in the mid-1990s, as data suggest, the current dry conditions in the Colorado River basin may persist.

Corresponding author address: Ruud Hurkmans, Hydrology and Quantitative Water Management Group, Wageningen University, P.O. Box 47, 6700 AA, Wageningen, Netherlands. Email: ruud.hurkmans@wur.nl

1. Introduction

Recently, the Colorado River basin (CRB) experienced a severe multiyear drought that is unprecedented in the hydroclimatic record (Cook et al. 2004). Because of the temperature rise associated with climate change, similar drought episodes are predicted to occur more often (Seager et al. 2007). For water management operations in the basin, understanding and predictive capacity of terrestrial water storage (TWS) dynamics and its associated hydrologic fluxes is crucial (Troch et al. 2007). Precipitation in the southwestern United States has been linked to oceanic interannual variability (e.g., Redmond and Koch 1991; Okin and Reheis 2002; McCabe et al. 2004; Hidalgo and Dracup 2003), mostly involving El Niño–Southern Oscillation (ENSO; Trenberth (1997)) and Pacific decadal oscillation [PDO; Mantua et al. (1997), a special case of Pacific decadal variability (PDV)]. ENSO is considered the most important process of interannual variability of water availability in the southwestern United States, and the most reliable one in terms of prediction (Cayan et al. 1999). Warm (El Niño) and cold (La Niña) events typically occur every 3–7 yr and last 8–16 months. These events have been linked to precipitation, floods, and droughts across the western United States (Cayan et al. 1999; Cañon et al. 2007; Hamlet et al. 2007). La Niña conditions lead to wet conditions in the Northwest and dry conditions in the Southwest and vice versa for El Niño conditions (Redmond and Koch 1991). PDO shows similar effects in the southwestern United States (MacDonald and Case 2005) and can, when in phase with ENSO, amplify El Niño or La Niña effects (Gershunov and Barnett 1998; Hamlet et al. 2007; Cole et al. 2002). PDO regime shifts, however, are much less predictable than ENSO cycles. Although its power spectrum was shown to be most energetic at two ranges of return periods (one from 15 to 25 yr and the other from 50 to 70 yr; Mantua and Hare 2002), the physical processes driving PDV and PDO remain unclear. The abrupt changes between cool and warm phases (each lasting 20–30 yr), therefore, are not yet predictable with sufficient lead times (Newman 2007).

Much of previous research has focused on precipitation and, to a lesser degree, on streamflow (e.g., Maity and Kumar 2008). Much less attention has been paid to regional water availability or TWS (Troch et al. 2007). Understanding the effects of climate variability on TWS is, however, relevant because it is a hydrologic state variable and therefore integrates hydrologic processes, such as snow accumulation, evapotranspiration, infiltration, and recharge. There are several techniques available to estimate TWS, which are described in detail in Troch et al. (2007). The first method is the basin-scale water balance (BSWB) method, used by Seneviratne et al. (2004) and Hirschi et al. (2006), which combines the atmospheric and terrestrial water balance to estimate changes in TWS from atmospheric moisture convergence, changes in atmospheric water vapor storage, and streamflow data. A second method is the Gravity Recovery and Climate Experiment (GRACE) satellite mission (Tapley et al. 2004), which has been also applied successfully to estimate water storage variability (e.g., Swenson and Milly 2006; Han et al. 2005). A third method is hydrological modeling of TWS components. Among all these methods, estimated storage dynamics in the CRB agree reasonably well with each other and with in situ data (Troch et al. 2007; Hasan 2009).

Here, we focus on TWS dynamics simulated using the Variable Infiltration Capacity (VIC) hydrological model (Liang et al. (1994)) forced by a meteorological dataset of interpolated observations, spanning the period 1915–2003. Hydrological modeling has several advantages compared to the other methods mentioned earlier. First, simulations are based on observed meteorological forcings (precipitation, temperature, air pressure, short- and longwave incoming radiation, vapor pressure, and wind speed), whereas, for example, the BSWB method applies model reanalysis data, which are constrained by radiosonde observations. Especially for hydrologic variables, balance problems can occur in such datasets (Seneviratne et al. 2004). Second, TWS estimates from the VIC model span a longer period (1915–2003) compared to the other methods. Third, TWS can be investigated in a distributed manner. Finally, TWS can be analyzed for each of its components (such as soil moisture, snow, and groundwater) separately. Troch et al. (2007) and Hasan (2009) indicated that the amplitudes of modeled storage anomalies do not always agree well with that of other methods, and Niu and Yang (2006) found that they are generally too low in tropical regions and too high at high latitudes. Trends and variability, however, agree well (see also http://voda.hwr.arizona.edu/twsc/sahra/).

We analyze the interannual variability of individual components of TWS: shallow and deep soil moisture, snow water equivalent (SWE), as well as precipitation and discharge. Using several climate indices describing oceanic and atmospheric variability in the Pacific, we attempt to explain this variability. With enhanced understanding of this variability, the predictability of hydrologic anomalies in the CRB may be improved. As already mentioned, an advantage of obtaining TWS from hydrological modeling is its relatively high spatial resolution. Storage dynamics will, therefore, be analyzed both in a spatially averaged way and a distributed way.

2. Study area and datasets

The CRB is located in the southwestern United States, and it has a drainage area of about 637 000 km2 (Fig. 1). Much of the basin is semiarid; in the southern part, rainfall is generally concentrated in the summer monsoon. The high-elevation snowpack in the Rocky Mountains contributes about 70% of the total annual runoff.

The meteorological data used in this study were compiled by Hamlet and Lettenmaier (2005) and span the period from 1915 to 2005. On the basis of comparisons with observed streamflow and snow cover, Hamlet and Lettenmaier (2005) concluded that the dataset is temporally consistent and suitable for trend analysis of hydrologic variables at the macroscale. The Hamlet and Lettenmaier (2005) dataset is thus used to force the VIC (Liang et al. 1994) model (version 4.0.5) for the CRB at a spatial resolution of 0.25° and a temporal resolution of three hours. The VIC model is designed as a land surface model to provide the land surface boundary conditions for climate models. Evaporation is calculated by solving the coupled water and energy balance, and the soil column is represented by three layers. The soil moisture contents of the top two layers [the highest layer is generally very thin (0.1 m)], determine the amounts of evaporation and infiltration. The soil moisture content of the lowest layer determines the amount of baseflow, or groundwater discharge. Vertical transport of moisture through the soil column is assumed to be driven by gravity only, and no lateral moisture transport is allowed. Calibration parameter values for the VIC model were obtained from A. Wood (2006, University of Washington, personal communication). The model was calibrated to match naturalized streamflow at the basin outlet. Resulting modeling efficiencies at the outlet of the upper Colorado basin and the entire CRB were 0.88 and 0.85, respectively, where a value of 1 indicates a perfect match. From the modeling results, the following variables were extracted: outflow (surface runoff plus baseflow), SWE, soil moisture in the lowest layer (hereafter deep soil moisture), and soil moisture in the top two layers (hereafter shallow soil moisture). The latter three, as well as interception storage of the canopy (both snow and water), are summed to calculate total TWS. In addition, precipitation from the forcing dataset is included in the analyses. Because of the dominance of the summer monsoon and spring snowmelt, all variables show a very strong annual cycle compared to the interannual variability. To remove seasonal influences, and because all climate indices are also based on a monthly time step, all time series are aggregated to monthly averages (sums for precipitation and outflow) and converted to anomalies by subtracting the mean annual cycle. Smoothed time series (12-month running means) of all anomalies are shown in Figs. 2a,b, their autocorrelation functions are shown in Fig. 3a and the corresponding power spectrum in Fig. 3b. Because storage anomalies in the first 15 yr were very high compared to the rest of the period, whereas precipitation anomalies were not, we discarded the first 15 yr. The period of analysis is, therefore, 1930–2003 hereafter.

Precipitation over the CRB has been linked to both tropical (ENSO) and extratropical (PDO) Pacific variability. Therefore, four indices that together represent both oceanic and atmospheric dynamics in the tropical and extratropical Pacific were selected. The multivariate ENSO index (MEI; Wolter and Timlin 1998) was selected because it is a composite of six variables over the tropical Pacific and includes atmospheric anomalies in addition to sea surface temperatures (SST). Because MEI is only available starting from 1950, we use the Niño-3.4 index (which is available for the entire period) as a proxy because it shows the highest correlation with MEI (0.90) compared with other ENSO indices (Niño-1.2, Niño-3, and Niño-4). The difference between these ENSO indices is the area in the Pacific that is used to calculate them. Niño-3.4 is calculated over the area bounded by 5°N–5°S and 170°–120°W. For Niño-3 and Niño-4, these bounding boxes have the same latitudes; however, in longitude the ranges are 150°–90°W for Niño-3 and 160°E–150°W for Niño-4. Niño-1.2 is bounded by 0°–10°S latitude and 90°–80°W longitude. To represent PDO, the PDO index proposed by Mantua et al. (1997) is used, defined as the anomaly of sea surface temperature in the area of the Pacific north of 20°N. In addition, the Pacific–North America (PNA) pattern is included because it is one of the most important modes of variability in the Northern Hemisphere extratropical atmosphere and is largely independent of ocean temperatures (Johansson 2007). In addition, the PNA pattern was shown to have an influence on precipitation and discharge in the western United States by Redmond and Koch (1991). In this way we cover both ocean and atmospheric dynamics in the tropics (MEI/Niño-3.4) and the extratropics (PDO/PNA). Figure 2c shows 12-month running mean time series of all four indices, and Fig. 3b shows both their autocorrelation functions (ACFs) and power spectra. MEI and Niño-3.4 have their most important spectral peaks at return periods of about 4–6 yr, whereas PDO shares the peak at 6 yr with ENSO but has its most important peaks at 25 yr and longer. However, the time series that is employed is too short (74 yr) to capture this low-frequency behavior of PDO. Cross correlations and cross power spectra between anomalies and climate indices were analyzed as well. Because in many signals there is significant autocorrelation (Figs. 3a,c), they were found to provide little information and are therefore not shown in this article.

3. Results

a. Analysis of spatial averages

As can be seen in Fig. 3, the autocorrelation function for deep soil moisture remains significant for very long time lags, up to about 11 yr. The large autocorrelation propagates in the total TWS, causing the autocorrelation function for total TWS to be significant up to about six years, although small after one year. Therefore, in the analyses described hereafter, dynamics of total TWS are not examined as such, but storage dynamics for deep soil moisture, shallow soil moisture, and SWE are investigated separately. For these latter variables, the autocorrelation function diminishes after a few months; therefore, by calculating averages of six months, autocorrelation is almost filtered out and correlation patterns between climate indices and anomalies can be investigated, including time lags. Figure 4 shows a matrix of correlation coefficients between monthly values of the climate indices and 6-month moving averages of anomalies of shallow soil moisture, SWE, discharge, and precipitation.

In Fig. 4, correlation patterns are similar for precipitation, shallow soil moisture, and discharge, although correlation coefficients are generally slightly lower for discharge. MEI and Niño-3.4 show a nearly identical pattern, which is not surprising given their high mutual correlation (0.90). ENSO indices early in the year (January–May) give the highest correlations with hydrologic anomalies throughout the year, except for the period that is dominated by the summer monsoon. This is consistent with the findings of, for example, Okin and Reheis (2002) and Gochis et al. (2007), who found low correlations between winter ENSO and summer precipitation as well. ENSO indices in the second half of the year, on the other hand, seem to be correlated with the second half of the year (the monsoon and the period just after that). These findings are consistent with previously reported correlations. Hidalgo and Dracup (2003) report high correlations of warm-season ENSO with warm-season precipitation in the upper Colorado basin, whereas in the lower Colorado basin correlations are generally found to be higher between winter ENSO and winter precipitation (e.g., Okin and Reheis 2002). PDO shows a slightly different pattern, with significant correlations for all but the winter months and generally slightly lower correlations. The amount of snow in the winter season, however, seems to be uncorrelated to any climate index. There is some correlation between SWE and MEI in summer but that is only after the main snowmelt season, when there is hardly any snow left. The lack of correlation between SWE and ENSO is consistent with the result that correlations between ENSO and winter precipitation are limited to the lower Colorado basin because SWE is concentrated in the upper Colorado basin. The PNA index in some specific months (January and April) is also correlated with SWE, especially in the cold season. It should be noted, however, that because of limited availability of the climate indices, MEI and PNA only cover the period 1950–2003, whereas Niño-3.4 and PDO cover the entire period from 1930 on.

By calculating the 6-month averages, the effect of autocorrelation should be filtered out. To further investigate the correlations that appear in Fig. 4, the corresponding (scaled) time series to some of the pixels in Fig. 4 are shown in Fig. 5. Here, only PDO and Niño-3.4 are shown because they cover the entire period, and only shallow soil moisture is shown because precipitation and discharge show a similar correlation pattern in Fig. 4. Figure 5 shows that both the Niño-3.4 and PDO indices are indeed related to anomalies of shallow soil moisture in winter and spring (Fig. 5): nearly all the El Niño (La Niña) events (shaded in gray) correspond to a positive (negative) shallow soil moisture anomaly. Here, the definition of such events as given by Trenberth (1997) is followed: an event is a consecutive period of six months in which the Niño-3.4 index is higher than 0.4°C (for El Niño) or lower than −0.4°C (for La Niña). The same events are also visible for the PDO index, which explains the nearly identical correlation coefficients for both indices. Climate indices in March are correlated to shallow soil moisture throughout the year (Figs. 5a–d), whereas the climate indices in September are only correlated with cold-season shallow soil moisture (Figs. 5g,h) and much less (not at all for Niño-3.4) with warm-season shallow soil moisture (Figs. 5e,f). In all plots, it appears that especially the extreme El Niño and La Niña events are generally well correlated. For example, the El Niño event of 1940, which also shows a high PDO peak, is related to high shallow soil moisture anomalies throughout the year. Also, the El Niño events of around 1972 and 1982 and the La Niña event of 1989 led to extremely wet (dry for La Niña) conditions. However, they were not so distinct throughout the year; that is, they do not show up in all panels of Fig. 5. It should be noted, however, that although Fig. 5 shows the dynamics of climate indices and shallow soil moisture anomalies, the amplitudes of all signals (also with respect to each other) have no meaning because all time series were scaled by dividing each time series by the difference between its maximum and minimum value.

For some of the variables and indices in Fig. 2, the 24-month running average is plotted in Fig. 6 to show interannual–decadal variability more clearly. The PDO phases, each about 30-yr long, clearly stand out. PDO phases were identified by several independent studies [see Mantua and Hare (2002) for a review on PDO] as follows: warm (positive) PDO regimes dominated from 1925 to 1946 and from 1977 to about the mid-1990s, whereas cool (negative) regimes prevailed between 1947 and 1976. The 74 yr used in this study, therefore, only cover one complete PDO cycle. In Fig. 6, regime shifts as they were indicated by Mantua et al. (1997) are indicated by the vertical black lines.

In Fig. 6a the sign of the anomaly of deep soil moisture is distinctly different between PDO phases. This is illustrated by the 10-yr running mean (the dashed–dotted line in Fig. 6a), of which the positive and negative episodes exactly coincide with the PDO phases. Apparently, during negative PDO episodes conditions are generally dryer, although this does not show that clearly in the other records. In addition, the storage anomaly of Lake Mead is plotted in Fig. 6a. Lake Mead is a large surface water storage reservoir in the lower Colorado basin, accounting for 45% of the total surface storage capacity in the CRB (Troch et al. 2007), and data are available from 1941 onward. Another large reservoir, Lake Powell (accounting for 42% of the total storage capacity), is only operational from 1963 and is therefore not included in the present analysis. The 24-month running average of the storage anomaly of Lake Mead follows the deep soil moisture anomaly with a time lag, although the signal is somewhat flattened. In addition, the 10-yr running average of the storage anomaly of Lake Mead is shown. Again, the positive and negative episodes exactly coincide with the PDO regime phases as indicated in Fig. 6. These findings are consistent with earlier research by McCabe et al. (2004), who find that a large part of the variability in drought frequency is explained by PDO. Some sources indicate that a shift of PDO to a new negative phase has occurred in the 1990s (e.g., Mantua and Hare 2002). If that is really the case and the length of the PDO phases would stay in the same order of magnitude as it has been throughout the twentieth century, this implies that for the coming decades, dry conditions might prevail in the CRB.

Another striking feature in Fig. 6 is the fact that the amplitude of the signal, or the size of the anomaly, seems to be larger in positive PDO phases for shallow soil moisture and precipitation. Table 1 shows the mean absolute anomaly over the entire period (A), and positive (P) and negative (N) PDO phases. A t test was performed to test whether the averaged absolute anomalies between the PDO phases and the entire period are different. The results of two types of t tests are displayed. First, a difference of means test for independent samples, because this test has the advantage that the time series do not need to be of equal length. This test was carried out to test the difference between A and P, A and N, and P and N. The test does assume independence but not equal variance between the samples. For completeness, a paired samples (PS) test was also carried out, which is more suitable for dependent samples. This test, however, needs samples of equal length; therefore, a part of P was selected of similar length to N. In all the tests, the length of the time series was corrected for autocorrelation by calculating an “effective sample size,” which depends on the first-order autocorrelation (Cressie 1991):
i1525-7541-10-5-1257-e1
where n is the length of the time series, ρ is the first-order autocorrelation, and neff is the effective (corrected) length of the time series. It should be noted, however, that this method only corrects for linear correlation and does not take into account nonlinear and higher-order correlation effects and is therefore relatively crude. The results of the tests are displayed in Table 1 in terms of their p values. We find differences between the PDO phases for shallow soil moisture, SWE, and discharge, although it depends on the method of testing and correcting for autocorrelation whether they can be considered significant or insignificant, which is the reason for presenting the results in terms of p values. The differences in amplitude (although small) seem consistent with previous research. Gershunov and Barnett (1998) found that ENSO is enhanced by PDO [or North Pacific Oscillation (NPO) as Gershunov and Barnett (1998) call it]; that is, during positive PDO phases, El Niño patterns are stronger and more stable, whereas during negative PDO phases this is typically the case for La Niña patterns. More extreme and stable ENSO patterns could lead to more extreme hydrological conditions through the correlations displayed in Fig. 4. This also explains the generally dryer conditions during the negative PDO phase. The higher amplitude for shallow soil moisture and discharge in positive PDO phases can also be explained by the higher saturated storage: more saturated storage means less buffering capacity. Runoff and shallow soil moisture content of the upper soil will, therefore, react faster and stronger to precipitation events, whereas during negative PDO phases more water can percolate to the saturated storage reservoir and thus surface runoff and shallow soil moisture are buffered to some extent. In addition, Cayan et al. (1999) also found that the effect of climate variability is amplified in streamflow with respect to precipitation due to the nonlinear response of surface runoff to precipitation. This is consistent with our finding that differences in the amplitudes of discharge and shallow soil moisture anomalies are significant where those of precipitation anomalies are not.

As Fig. 5a, and to some degree also Fig. 6a suggest, correlations between climate indices on the one hand and hydrological anomalies on the other hand are stronger in some periods compared to others. Examples are the periods 1930–50 and 1988–99. Such periods of stronger and weaker correlations were also identified for summer precipitation by Hu and Feng (2001) and are related to PDO-like variability. To investigate whether this is also the case in this study, Table 2 lists correlation coefficients as calculated over each PDO phase separately, based on 24-month running averages of both data and indices. In Table 2, we see that although correlation coefficients for Niño-3.4 are nearly equal in both PDO phases (except perhaps for SWE, which does not show very strong correlations altogether), correlations with PDO during the negative PDO phase are significantly stronger than during the positive PDO phase. This is also visible to some extent in Figs. 5b, 6. Correlation coefficients are relatively high because, as in Fig. 6, correlation coefficients are based on 24-month running averages. Correlations based on monthly values (not shown) show the same pattern but are less clear because overall correlation coefficients are lower.

b. Analysis of distributed data

A similar analysis as is shown in Fig. 4, which represents spatial averages over the entire CRB, can be carried out in a distributed manner. For some of the data points in Fig. 4, the corresponding map is shown in Figs. 7, 8. In these figures, only PDO and Niño-3.4 are shown because these are available for the entire period. In addition, MEI and Niño-3.4 are nearly identical. Because not much structural correlation was visible for SWE in Fig. 4, SWE is not displayed in Figs. 7, 8. Moreover, if any significant correlations are present for SWE, the spatial pattern is rather predictable and limited to the mountainous areas (see Fig. 1). For comparison purposes, standard deviations of the various anomalies are shown in Fig. 9 to indicate the areas that show the highest variability. Most of this variability seems to be related to orographic precipitation: the areas of highest variability in precipitation coincide with the areas with the most pronounced topography (Fig. 1).

Correlation coefficients between climate indices for January and July and 6-month averages of shallow soil moisture, precipitation, and discharge anomalies are plotted in Figs. 7, 8, each for a time lag of 0 and 6 months. As appeared from Fig. 4 as well, Niño-3.4 is mainly correlated to the hydrologic anomalies for time lags of 0 months, especially in the southern part of the basin. At time lags of six months, however, almost all correlation disappears except for a low mountain range along the northern rim of the Little Colorado basin (northeast Arizona; Fig. 1). Niño-3.4 values for July are correlated with summer anomalies over nearly the entire basin, except for two regions (far north and far south) in which very little variability is present according to Fig. 9. This is consistent with Hidalgo and Dracup (2003), who also find high correlations in summer throughout the basin. Especially for precipitation, correlation coefficients are highest in the western part of the basin. It is not clear yet why this is the case. The PDO index for January is not correlated to any anomaly, which is consistent with Fig. 4. The July PDO index, on the other hand, is highly correlated with winter/spring precipitation, shallow soil moisture, and discharge anomalies—especially in the southern part of the basin. This suggests that, for this area, PDO could be used in the prediction of precipitation anomalies in the next cool season, with about a six-month lead time, as opposed to Niño-3.4, which is more synchronously correlated in the southern part of the basin—that is, without any lead time.

The spatial distribution of runoff and shallow soil moisture are to a large extent governed by soil properties, which in VIC depend strongly on model calibration. To correctly represent the spatial variability of hydrological states, separate parameter values should be assigned to every pixel in the VIC model. In the calibration that is employed in this study, however, the spatial variability in calibration parameters is relatively small, as is also pointed out by Hasan (2009). Because discharge and shallow soil moisture show a similar pattern as precipitation, the spatial correlation pattern is most likely dominated by precipitation and not so much by calibration parameters. Moreover, the spatial distribution of the calibration parameter values is entirely different from these patterns, although not shown here.

4. Summary and conclusions

In this study, we investigated the interannual–decadal variability of terrestrial water storage in the Colorado River basin. TWS data were obtained from the modeling results of the Variable Infiltration Capacity (VIC) model, forced by a meteorological dataset of interpolated observations, spanning the period 1915–2003. The resulting TWS estimates have advantages compared to those obtained by other methods in that they 1) span a relatively long period (we analyzed the period 1930–2003), 2) are spatially distributed, and 3) offer the possibility to analyze the various storage components separately. As already mentioned, the model was forced by observations and calibrated to match naturalized streamflow at the basin outlet. The gridded observations that were used to force the model were, however, downscaled and interpolated (Hamlet and Lettenmaier 2005). The hydrological variables that were analyzed should therefore be considered as “model re-analysis” data. We compare the time series of monthly anomalies (i.e., the mean climatologies are removed from the signals) of deep soil moisture, shallow soil moisture, precipitation, “discharge” (the sum of surface runoff and baseflow), and snow water equivalent (SWE) with four climate indices, describing the variability of ocean temperature and atmospheric pressure levels in the tropical and extratropical Pacific. Where previous studies in the CRB mainly investigated precipitation and discharge, we thus also take into account dynamics of deep and shallow soil moisture and snow.

Because autocorrelations extend over very large time lags for deep soil moisture and because this autocorrelation propagates in the TWS signal, cross correlations with climate indices cannot be calculated reliably. Therefore, we focus on separate storage components, mainly shallow soil moisture, and snow water equivalent. Precipitation, shallow soil moisture, and discharge are all related to ENSO, mainly in winter and spring. For these seasons, correlations occur exclusively in the southern part of the basin. ENSO in summer is also correlated to summer precipitation, moisture, and discharge, as was also found by Hidalgo and Dracup (2003), throughout the CRB. The relation between these variables and Niño-3.4 and PDO are strongest during extreme El Niño and La Niña events. PDO is correlated quite strongly with ENSO (0.58), showing the same extreme events. Apart from the PDO mode that is correlated with ENSO, PDO exhibits a periodicity at much lower frequencies (multiple decades), generally known as PDO phases, of which two (maybe three) transitions have taken place during the period of study. Deep soil moisture appeared to be closely related to these PDO phases. Because the time scales at which changes in deep soil moisture occur are much larger than those at which changes in, for example, shallow soil moisture or precipitation occur, the correlation between the low-frequency periodicity of PDO and deep soil moisture is more pronounced than the correlation with other variables. In addition, the high-frequency mode (which is related to ENSO) of the PDO index appeared to be more closely related to these anomalies during the negative episode of PDO.

This study aimed at improving the understanding of the interannual–decadal variability of the hydrologic state of the Colorado River basin. Although there is certainly some correlation with ENSO, as was found by many studies before, it is not always consistent and therefore difficult to employ for prediction purposes. The PDO index in summer, however, appears to be correlated with winter precipitation and shallow soil moisture for the lower Colorado basin and therefore offers the potential for prediction with lead times of about six months. The low-frequency mode of PDO seems to have an important effect on the hydrological conditions: during the negative PDO phase (in the record of study this corresponds to 1946–76), generally dryer conditions occurred. The physics behind the PDO and especially the irregularity of this periodicity is not well understood to date. Prediction of PDO is, therefore, extremely difficult (Newman 2007; Mantua and Hare 2002). If the periodicity as it was seen throughout the twentieth century continues and if indeed a regime shift to a negative PDO phase occurred during the mid-1990s, which is still uncertain (Mantua and Hare 2002), then the current dry conditions in the CRB may persist for several more years.

Acknowledgments

This research was financially supported by the Water Sustainability Program of the University of Arizona, the European Commission through the FP6 Integrated Project NeWater, and the BSIK ACER project of the Dutch Climate Changes Spatial Planning Programme. Andy Wood from the University of Washington is kindly acknowledged for sharing the calibration parameters for the hydrological model and for providing the observed dataset. We thank the three reviewers for their constructive comments.

REFERENCES

  • Cañon, J., , González J. , , and Valdés J. , 2007: Precipitation in the Colorado River Basin and its low frequency associations with PDO and ENSO signals. J. Hydrol., 333 , 252264.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cayan, D. R., , Redmond K. T. , , and Riddle L. G. , 1999: ENSO and hydrologic extremes in the western United States. J. Climate, 12 , 28812893.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cole, J. E., , Overpeck J. T. , , and Cook E. R. , 2002: Multiyear La Niña events and persistent drought in the contiguous United States. Geophys. Res. Lett., 29 , 1647. doi:10.1029/2001GL013561.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cook, E., , Woodhouse C. , , Eakin C. M. , , Meko D. , , and Stahle D. , 2004: Long-term aridity changes in the western United States. Science, 306 , 10151018. doi:10.1126/science.1102586.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cressie, N. A. C., 1991: Statistics for Spatial Data. Wiley, 900 pp.

  • Gershunov, A., , and Barnett T. P. , 1998: Interdecadal modulation of ENSO teleconnections. Bull. Amer. Meteor. Soc., 79 , 27152725.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gochis, D. J., , Brito-Castillo L. , , and Shuttleworth W. J. , 2007: Correlations between sea-surface temperatures and warm season streamflow in northwest Mexico. Int. J. Climatol., 27 , 883901. doi:10.1002/joc.1436.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hamlet, A. F., , and Lettenmaier D. P. , 2005: Production of temporally consistent gridded precipitation and temperature fields for the continental United States. J. Hydrometeor., 6 , 330336.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hamlet, A. F., , Mote P. W. , , Clark M. P. , , and Lettenmaier D. P. , 2007: Twentieth-century trends in runoff, evapotranspiration, and soil moisture in the western United States. J. Climate, 20 , 14681486.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, S-C., , Shum C. K. , , Jekeli C. , , and Alsdorf D. , 2005: Improved estimation of terrestrial water storage changes from GRACE. Geophys. Res. Lett., 32 , L07302. doi:10.1029/2005GL022382.

    • Search Google Scholar
    • Export Citation
  • Hasan, S., 2009: Terrestrial water storage change from temporal gravity variation. Ph.D. thesis, Wageningen University, 83 pp.

  • Hidalgo, H. G., , and Dracup J. A. , 2003: ENSO and PDO effects on hydroclimatic variations of the upper Colorado River Basin. J. Hydrometeor., 4 , 523.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hirschi, M., , Seneviratne S. I. , , and Schär C. , 2006: Seasonal variations in terrestrial water storage for major midlatitude river basins. J. Hydrometeor., 7 , 3960.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hu, Q., , and Feng S. , 2001: Variations of teleconnection of ENSO and interannual variation in summer rainfall in the central United States. J. Climate, 14 , 24692480.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Johansson, A., 2007: Prediction skill of the NAO and PNA from daily to seasonal time scales. J. Climate, 20 , 19571975.

  • Liang, X., , Lettenmaier D. P. , , Wood E. F. , , and Burges S. J. , 1994: A simple hydrologically based model of land surface water and energy fluxes for general circulation models. J. Geophys. Res., 99 , (D7). 1441514458.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • MacDonald, G. M., , and Case R. A. , 2005: Variations in the Pacific Decadal Oscillation over the past millennium. Geophys. Res. Lett., 32 , L08703. doi:10.1029/2005GL022478.

    • Search Google Scholar
    • Export Citation
  • Maity, R., , and Kumar D. N. , 2008: Basin-scale stream-flow forecasting using the information of large-scale atmospheric circulation phenomena. Hydrol. Processes, 22 , 643650. doi:10.1002/hyp.6630.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mantua, N. J., , and Hare S. R. , 2002: The Pacific Decadal Oscillation. J. Oceanogr., 58 , 3544.

  • Mantua, N. J., , Hare S. R. , , Zhang Y. , , Wallace J. M. , , and Francis R. C. , 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78 , 10691079.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCabe, G. J., , Palecki M. A. , , and Betancourt J. L. , 2004: Pacific and Atlantic Ocean influences on multidecadal drought frequency in the United States. Proc. Natl. Acad. Sci. USA, 101 , 41364141. doi:10.1073/pnas.0306738101.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Newman, M., 2007: Interannual to decadal predictability of tropical and North Pacific sea surface tempperatures. J. Climate, 20 , 23332355.

  • Niu, G-Y., , and Yang Z-L. , 2006: Assessing a land surface model’s improvements with GRACE estimates. Geophys. Res. Lett., 33 , L07401. doi:10.1029/2005GL025555.

    • Search Google Scholar
    • Export Citation
  • Okin, G. S., , and Reheis M. C. , 2002: An ENSO predictor of dust emission in the southwestern United States. Geophys. Res. Lett., 29 , 1332. doi:10.1029/2001GL014494.

    • Search Google Scholar
    • Export Citation
  • Redmond, K. T., , and Koch R. W. , 1991: Surface climate and streamflow variability in the western United States and their relationship to large-scale circulation indices. Water Resour. Res., 27 , 23812399.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Seager, R., and Coauthors, 2007: Model projections of an imminent transition to a more arid climate in southwestern North America. Science, 316 , 11811184. doi:10.1126/science.1139601.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Seneviratne, S. I., , Viterbo P. , , Lüthi D. , , and Schär C. , 2004: Inferring changes in terrestrial water storage using ERA-40 reanalysis data: The Mississippi River basin. J. Climate, 17 , 20392057.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Swenson, S. C., , and Milly P. C. D. , 2006: Climate model biases in seasonality of continental water storage revealed by satellite gravimetry. Water Resour. Res., 42 , W03201. doi:10.1029/2005WR004628.

    • Search Google Scholar
    • Export Citation
  • Tapley, B. D., , Bettadpur S. , , Watkins M. , , and Reigber C. , 2004: The gravity recovery and climate experiment: Mission overview and early results. Geophys. Res. Lett., 31 , L09607. doi:10.1029/2004GL019920.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., 1997: The definition of El Niño. Bull. Amer. Meteor. Soc., 78 , 27712777.

  • Troch, P. A., , Durcik M. , , Seneviratne S. I. , , Hirschi M. , , Teuling A. J. , , Hurkmans R. , , and Hasan S. , 2007: New data sets to estimate terrestrial water storage change. EOS, Trans. Amer. Geophys. Union, 88 .doi:10.1029/2007EO450001.

    • Search Google Scholar
    • Export Citation
  • Wolter, K., , and Timlin M. S. , 1998: Measuring the strength of ENSO events: How does 1997/98 rank? Weather, 53 , 315323.

Fig. 1.
Fig. 1.

Location of and elevations in the CRB. The basin outline represents the discretization of the basin in VIC at a resolution of 0.25°. The Colorado River is shown as a black line.

Citation: Journal of Hydrometeorology 10, 5; 10.1175/2009JHM1133.1

Fig. 2.
Fig. 2.

Time series of spatially averaged anomalies of (a) storage components and total storage in VIC, (b) outflow (surface runoff + baseflow) and precipitation, and (c) four climate indices (MEI, PDO, Niño-3.4, and PNA). For clarity, 12-month running averages were calculated for each variable.

Citation: Journal of Hydrometeorology 10, 5; 10.1175/2009JHM1133.1

Fig. 3.
Fig. 3.

ACFs for all (a) anomalies and (c) indices displayed in Fig. 2, up to a time lag of 180 months (15 yr). Gray shaded areas indicate autocorrelations not significantly different from zero. Power spectra are also shown for (b) anomalies and (d) climate indices. Numbers in (d) represent the return period in years of the strongest spectral peaks of Niño-3.4 and PDO. Both (a) and (b) show six lines [shallow soil moisture (SM), deep SM, SWE, TWS, discharge, and precipitation) for which the legend is split between (a) and (b).

Citation: Journal of Hydrometeorology 10, 5; 10.1175/2009JHM1133.1

Fig. 4.
Fig. 4.

Correlation between 6-month moving averages of five hydrologic anomalies and monthly values of four climate indices (on the x axes). The y axes therefore represent twelve 6-month moving averages: (bottom)–(top) February–July (F–J), April–September (A–S), June–November (J–N), August–January (A–J), October–March (O–M), December–May (D–M). Correlations that are significant at α = 0.05 are denoted by a black dot.

Citation: Journal of Hydrometeorology 10, 5; 10.1175/2009JHM1133.1

Fig. 5.
Fig. 5.

Time series of (left) monthly Niño-3.4, (right) PDO, and 6-month averaged shallow SM, similar to the data used to create Fig. 4. Four pixels from the matrices plotted in Fig. 4 are used: (a),(b) the M indices vs M–S averaged shallow SM; (c),(d) M indices vs S–F averaged shallow SM; (e),(f) S indices vs M–S averaged shallow SM; and (g),(h) S indices vs S–F averaged shallow SM. Also the cross-correlation coefficients (without time lag) are shown in every panel. Gray shaded areas indicate El Niño (positive) and La Niña (negative) events.

Citation: Journal of Hydrometeorology 10, 5; 10.1175/2009JHM1133.1

Fig. 6.
Fig. 6.

Interannual variability investigated through time series of 24-month running averages of (a) deep SM, (b) shallow SM, (c) SWE, (d) precipitation, and (e) discharge. Niño-3.4 and PDO are plotted in all plots. In addition, the storage anomaly of Lake Mead is plotted in (a). To illustrate the correlation between deep SM and reservoir storage and PDO phases, the 10-yr running means of the deep SM anomaly (black) and that of the storage anomaly of Lake Mead (gray) are plotted as well in (a) as thick dashed lines. Black vertical lines denote the shifts from warm to cool and again to warm PDO phase in all panels, and the dashed line at 1998 denotes a possible third phase change.

Citation: Journal of Hydrometeorology 10, 5; 10.1175/2009JHM1133.1

Fig. 7.
Fig. 7.

Maps of correlation coefficients between anomalies of three variables (shallow SM, discharge, and precipitation) and the climate index Niño-3.4. Only significant correlations (α = 0.05) are shown. Time series were aggregated similar to Fig. 4; calendar months of climate indices are correlated with 6-month averages of anomalies. Here, the climate index values for January (Jan) and July (Jul) are shown, correlated with anomaly values averaged over January–June (winter), or July–December (summer).

Citation: Journal of Hydrometeorology 10, 5; 10.1175/2009JHM1133.1

Fig. 8.
Fig. 8.

Same as Fig. 7 but for the climate index PDO.

Citation: Journal of Hydrometeorology 10, 5; 10.1175/2009JHM1133.1

Fig. 9.
Fig. 9.

Maps of temporal standard deviations of anomalies of three storage components (shallow SM, deep SM, and snowpack), total storage, outflow, and precipitation. Storage components are in mm, fluxes in mm month−1.

Citation: Journal of Hydrometeorology 10, 5; 10.1175/2009JHM1133.1

Table 1.

Average of the absolute values of hydrologic anomalies over A (A) as well P and N episodes of PDO. In the right four columns, p values are given for a t test that was performed to test whether averaged absolute anomalies are different between 1) A vs P, (2) A vs N, and the P vs N. A t test for independent samples was used for differences between A vs P, A vs N, and P vs N. Here, P vs N was also tested using a PS test for dependent samples; results are displayed in the rightmost column. All values were calculated using an effective sample size—that is, corrected for autocorrelation.

Table 1.
Table 2.

Correlation coefficients between anomalies (shallow SM, deep SM, SWE, precipitation, and discharge) and climate indices (PDO and Niño-3.4), calculated for the entire period (All) and for positive (PDO+) and negative (PDO−) PDO phases separately, based on 24-month running averages of both data and indices. Significant values (α = 0.05) are indicated in bold.

Table 2.
Save