• Alley, W. H., 1984: The Palmer Drought Severity Index: Limitations and assumptions. J. Climate Appl. Meteor, 23 , 11001109.

  • Barnett, T. P., and R. Preisendorfer, 1987: Origins and levels of monthly and seasonal forecast skill for the United States surface air temperature determined by the canonical correlation analysis. Mon. Wea. Rev, 115 , 18251850.

    • Search Google Scholar
    • Export Citation
  • Barnston, A. G., and T. M. Smith, 1996: Specification and prediction of global surface temperature and precipitation from global SST using CCA. J. Climate, 9 , 26602697.

    • Search Google Scholar
    • Export Citation
  • Bonsal, B. R., and R. G. Lawford, 1999: Teleconnections between El Niño and La Niña events and summer extended dry spells on the Canadian prairies. Int. J. Climatol, 19 , 14451458.

    • Search Google Scholar
    • Export Citation
  • Bretherton, C. S., C. Smith, and J. M. Wallace, 1992: An intercomparison of methods for finding coupled patterns in climate data. J. Climate, 5 , 541560.

    • Search Google Scholar
    • Export Citation
  • Cherry, S., 1997: Some comments on singular value decomposition analysis. J. Climate, 10 , 17591761.

  • Cole, J. E., and E. R. Cook, 1998: The changing relationship between ENSO variability and moisture balance in the continental United States. Geophys. Res. Lett, 25 , 45294532.

    • Search Google Scholar
    • Export Citation
  • Cook, E. R., D. M. Meko, D. W. Stahle, and M. K. Cleaveland, 1999: Drought reconstructions for the continental United States. J. Climate, 12 , 11451162.

    • Search Google Scholar
    • Export Citation
  • Czaja, A., and J. Marshall, 2001: Observations of atmosphere ocean coupling in the North Atlantic. Quart. J. Roy. Meteor. Soc, 127 , 18931916.

    • Search Google Scholar
    • Export Citation
  • Deser, C., and M. L. Blackmon, 1995: On the relationship between tropical and North Pacific sea surface temperature variations. J. Climate, 8 , 16771680.

    • Search Google Scholar
    • Export Citation
  • Enfield, D. B., A. M. Mestas-Nuñez, and P. J. Trimble, 2001: The Atlantic multidecadal oscillation and its relation to rainfall and river flows in the continental U.S. Geophys. Res. Lett, 28 , 20772080.

    • Search Google Scholar
    • Export Citation
  • Heim Jr., R. R., 2002: A review of twentieth-century drought indices used in the United States. Bull. Amer. Meteor. Soc, 83 , 11491165.

    • Search Google Scholar
    • Export Citation
  • Hoerling, M., and A. Kumar, 2003: The perfect ocean for drought. Science, 299 , 691694.

  • Horel, J. D., and J. M. Wallace, 1981: Planetary-scale atmospheric phenomena associated with the Southern Oscillation. Mon. Wea. Rev, 109 , 813829.

    • Search Google Scholar
    • Export Citation
  • Hoskins, B. J., and D. Karoly, 1981: The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci, 38 , 11791196.

    • Search Google Scholar
    • Export Citation
  • Huang, J., H. M. Van den Dool, and K. P. Georgakakos, 1996: Analysis of model-calculated soil moisture over the United States (1931–1993) and applications to long-range temperature forecasts. J. Climate, 9 , 13501362.

    • Search Google Scholar
    • Export Citation
  • Hurrell, J. W., 1996: Influences of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature. Geophys. Res. Lett, 23 , 665668.

    • Search Google Scholar
    • Export Citation
  • Isaaks, E. H., and R. M. Srivastava, 1989: An Introduction to Geostatistics. Oxford University Press, 561 pp.

  • Iwasaka, N., and J. M. Wallace, 1995: Large scale air sea interaction in the Northern Hemisphere from a view point of variations of surface heat flux by SVD analysis. J. Meteor. Soc. Japan, 73 , 781794.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 2003: The nature and causes for the delayed atmospheric response to El Niño. J. Climate, 16 , 13911403.

    • Search Google Scholar
    • Export Citation
  • Lau, N-C., and M. J. Nath, 1994: A modeling study of the relative roles of tropical and extratropical anomalies in the variability of the global atmosphere–ocean system. J. Climate, 7 , 11841207.

    • Search Google Scholar
    • Export Citation
  • Lawrimore, J., R. Heim Jr., M. Svoboda, V. Swail, and P. Englehart, 2002: Beginning a new era of drought monitoring across North America. Bull. Amer. Meteor. Soc, 83 , 11911192.

    • Search Google Scholar
    • Export Citation
  • Louie, P. Y. T., 1986: An operational drought severity index program for Canadian synoptic stations. Proc. 16th Symp. on Canadian Hydrology, Regina, SK, Canada, National Research Council, 101–112.

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

    • Search Google Scholar
    • Export Citation
  • Mekis, E., and W. D. Hogg, 1999: Rehabilitation and analysis of Canadian daily precipitation time series. Atmos.–Ocean, 37 , 5385.

  • Mestas-Nuñez, A. M., and D. B. Enfield, 1999: Rotated global modes of non-ENSO sea surface variability. J. Climate, 12 , 27342746.

  • Newman, M., and P. D. Sardeshmukh, 1995: A caveat concerning singular value decomposition. J. Climate, 8 , 352360.

  • Nigam, S., M. Barlow, and E. H. Berbery, 1999: Analysis links Pacific decadal variability to drought and streamflow in the United States. Eos, Trans. Amer. Geophys. Union, 80 , 621625.

    • Search Google Scholar
    • Export Citation
  • Palmer, W. C., 1965: Meteorological drought. U.S. Department of Commerce Weather Bureau Res. Paper 45, 58 pp.

  • Rajagopalan, B., E. Cook, U. Lall, and B. K. Ray, 2000: Spatiotemporal variability of ENSO and SST teleconnections to summer drought over the United States during the twentieth century. J. Climate, 13 , 42444255.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis. J. Climate, 15 , 16091625.

    • Search Google Scholar
    • Export Citation
  • Ropelewski, C. F., and M. S. Halpert, 1986: North American precipitation and temperature patterns associated with El Niño/Southern Oscillation (ENSO). Mon. Wea. Rev, 114 , 23522362.

    • Search Google Scholar
    • Export Citation
  • Sauchyn, D. S., and W. R. Skinner, 2001: A 400 year reconstruction of Palmer Drought Severity Index (PDSI) for the southern Canadian prairies. Can. Water Resour. Assoc. J, 26 , 253272.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., and A. G. Barnston, 1996: Skill of seasonal climate forecasts in Canada using canonical correlation analysis. Mon. Wea. Rev, 124 , 23702385.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., and M. Khandekar, 1996: The impact of El Niño–Southern Oscillation on the temperature field over Canada. Atmos.–Ocean, 34 , 401416.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., B. Bonsal, and M. Khandekar, 1997a: Canadian precipitation patterns associated with the Southern Oscillation. J. Climate, 10 , 30163027.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., K. Higuchi, W. Skinner, and J. L. Knox, 1997b: The association between the BWA index and winter surface temperature variability over eastern Canada and west Greenland. Int. J. Climatol, 17 , 11951210.

    • Search Google Scholar
    • Export Citation
  • Skinner, W. R., B. J. Stocks, D. L. Martell, B. Bonsal, and A. Shabbar, 1999: The association between circulation anomalies in the mid-troposphere and area burned by wildland fire in Canada. Theor. Appl. Climatol, 63 , 89105.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., and R. W. Reynolds, 2003: Extended reconstruction of global sea surface temperatures based on COADS data (1854– 1997). J. Climate, 16 , 14951510.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., R. W. Reynolds, R. E. Livezey, and D. Stokes, 1996: Reconstruction of historical sea surface temperatures using empirical orthogonal functions. J. Climate, 9 , 14031420.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., R. E. Livezey, and S. S. Shen, 1998: An improved method for analyzing sparse and irregularly distributed SST data on a regular grid: The tropical Pacific Ocean. J. Climate, 11 , 17171729.

    • Search Google Scholar
    • Export Citation
  • Thornthwaite, C. W., 1948: An approach toward a rational classification of climate. Geogr. Rev, 38 , 5594.

  • Trenberth, K. E., and D. A. Paolino, 1981: Characteristic patterns of variability of sea level pressure in the Northern Hemisphere. Mon. Wea. Rev, 109 , 11691189.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and G. W. Barnstator, 1992: Issues in establishing causes of the 1988 drought over North America. J. Climate, 5 , 159172.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere–ocean variations in the Pacific. Climate Dyn, 9 , 303309.

  • Vincent, L. A., 1998: A technique for the identification of inhomogeneities in Canadian temperature series. J. Climate, 11 , 10941104.

    • Search Google Scholar
    • Export Citation
  • Vincent, L. A., and D. W. Gullett, 1999: Canadian historical and homogeneous temperature datasets for climate change analyses. Int. J. Climatol, 19 , 13751388.

    • Search Google Scholar
    • Export Citation
  • von Storch, H., and F. W. Zwiers, 1999: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.

  • Wallace, J. M., C. Smith, and C. S. Bretherton, 1992: Singular value decomposition of wintertime sea surface temperature and 500-mb height anomalies. J. Climate, 5 , 561576.

    • Search Google Scholar
    • Export Citation
  • Wheaton, E. E., and L. M. Arthur, 1989: Some environmental and economic impacts of the 1988 drought. SRC Rep. E-2330-4-E-89, 362 pp.

  • Woodruff, S. D., H. F. Diaz, J. D. Elms, and S. J. Worley, 1998: COADS Release 2 data and metadata enhancements for improvements of marine surface flux fields. Phys. Chem. Earth, 23 , 517527.

    • Search Google Scholar
    • Export Citation
  • Zhang, X., L. A. Vincent, W. D. Hogg, and A. Niitsoo, 2000: Temperature and precipitation trends in Canada during the 20th century. Atmos.–Ocean, 38 , 395429.

    • Search Google Scholar
    • Export Citation
  • Zhang, Y., J. M. Wallace, and D. S. Batista, 1997: ENSO-like interdecadal variability: 1900–93. J. Climate, 10 , 10041020.

  • View in gallery

    (a) North America monthly PDSI site locations as calculated from Canadian historical database stations (red) and the United States Climate Division centroids (blue) monthly mean air temperature and monthly total precipitation data. Data are gridded at approximate 250 km × 250 km. (b) Summer (JJA) PDSI for 2001; with the exception of southern Manitoba and parts of northwestern Ontario, most of southern Canada experienced prolonged drought

  • View in gallery

    National summary of Canadian summer (JJA) PDSI, 1940– 2002. The line indicates the Canada-wide mean PDSI and the bars represent the spatial variability (standard deviation)

  • View in gallery

    (a) Composite Canada PDSI map for the five driest summers (JJA) 1940–2002. The five driest summers in descending order are 2002, 1941, 1998, 1961, and 1940. (b) Composite Canada PDSI map for the five wettest summers (JJA) 1940–2002. The five wettest summers in ascending order are 1992, 1996, 1947, 1974, and 1972

  • View in gallery

    (a) Global winter (DJF) SST anomaly composite for the five driest Canadian summers (JJA). (b) Same as (a), but for the five wettest Canadian summers (JJA). Thick-black dashed line delineates regions where the values are significant at the 5% level

  • View in gallery

    (a) The first SVD pattern (S1) of PDSI and standardized amplitude based on data for 63 summer (JJA) seasons 1940–2002. (b) Same as (a) except for the second SVD pattern (S2) of PDSI. (c) Same as (a) except for the third SVD pattern (S3) of PDSI

  • View in gallery

    (a) The first SVD pattern (S1) of SST and standardized amplitude based on data for 63 winter (DJF) seasons 1940–2002. (b) Same as (a) except for the second SVD pattern (S2) of SST. (c) Same as (a) except for the third SVD pattern (S3) of SST

  • View in gallery

    (a) Heterogeneous correlation patterns for the first mode in the direct SVD expansion. The temporal correlation coefficient between the corresponding expansion coefficients r(PDSIk SSTk), where k refers to the mode number, and the SCF (%) are shown. (b) Same as (a), but for the second mode. (c) Same as (a), but for the third mode

  • View in gallery

    Heterogeneous correlation patterns for the first mode in the direct SVD expansion using Atlantic SSTs. The temporal correlation coefficient between the corresponding expansion coefficients r(PDSIk, SSTk), where k refers to the mode number, and the SCF (%) are shown

  • View in gallery

    (a) Winter (DJF) AMO and winter S1 (SST) Atlantic, 1940–2002. (b) Correlation pattern between the winter (DJF) AMO and the following summer (JJA) PDSI. (c) Difference field between the positive and negative PDSI composites taken from the standardized amplitude for S1 (SST). Regions of Canada with significant correlation and differences at the 5% level are shown by dashed lines

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 363 331 23
PDF Downloads 194 179 17

Summer Drought Patterns in Canada and the Relationship toGlobal Sea Surface Temperatures

View More View Less
  • 1 Meteorological Service of Canada, Environment Canada, Toronto, Ontario, Canada
© Get Permissions
Full access

Abstract

Canadian summer (June–August) Palmer Drought Severity Index (PDSI) variations and winter (December– February) global sea surface temperature (SST) variations are examined for the 63-yr period of 1940–2002. Extreme wet and dry Canadian summers are related to anomalies in the global SST pattern in the preceding winter season. Large-scale relationships between summer PDSI patterns in Canada and previous winter global SST patterns are then analyzed using singular value decomposition (SVD) analysis. The matrix for the covariance eigenproblem is solved in the EOF space in order to obtain the maximum covariance between the singular values of the SST and the PDSI. The robustness of the relationship is established by the Monte Carlo technique, in which the time expansion of the primary EOF analysis is shuffled 1000 times.

Results show that the leading three SVD-coupled modes explain greater than 80% of the squared covariance between the two fields. The interannual El Niño–Southern Oscillation (ENSO), the Pacific decadal oscillation (PDO), and the interrelationship between the two play a significant role in the determination of the summer moisture availability in Canada. These Pacific Ocean processes are reflected in the second and third SVD modes, and together explain approximately 48% of the squared covariance. It is found that the warm ENSO (El Niño) events lead to a summer moisture deficit in the western two-thirds of Canada. Conversely, cold ENSO (La Niña) events produce an abundance of summer moisture, mainly in extreme western Canada and in the southeastern portions of the Canadian Prairies.

The first SVD mode strongly relates to the trend in global SSTs and multidecadal variation of the Atlantic SST, explaining approximately one-third of the squared covariance. It is reflective of both the warming trend in the global southern oceans and the influences of the Atlantic multidecadal oscillation (AMO) variability.

The 6-month lag relationship between the PDSI and large-scale SSTs provides a basis for developing long-range forecasting schemes for drought in Canada. A two-tier forecast scheme, in which the SST is predicted by an ocean model or a coupled climate model, can potentially further increase the lead time of drought forecasting.

Corresponding author address: Amir Shabbar, Meteorological Service of Canada, Environment Canada, 4905 Dufferin Street, Toronto, ON M3H 5T4, Canada. Email: amir.shabbar@ec.gc.ca

Abstract

Canadian summer (June–August) Palmer Drought Severity Index (PDSI) variations and winter (December– February) global sea surface temperature (SST) variations are examined for the 63-yr period of 1940–2002. Extreme wet and dry Canadian summers are related to anomalies in the global SST pattern in the preceding winter season. Large-scale relationships between summer PDSI patterns in Canada and previous winter global SST patterns are then analyzed using singular value decomposition (SVD) analysis. The matrix for the covariance eigenproblem is solved in the EOF space in order to obtain the maximum covariance between the singular values of the SST and the PDSI. The robustness of the relationship is established by the Monte Carlo technique, in which the time expansion of the primary EOF analysis is shuffled 1000 times.

Results show that the leading three SVD-coupled modes explain greater than 80% of the squared covariance between the two fields. The interannual El Niño–Southern Oscillation (ENSO), the Pacific decadal oscillation (PDO), and the interrelationship between the two play a significant role in the determination of the summer moisture availability in Canada. These Pacific Ocean processes are reflected in the second and third SVD modes, and together explain approximately 48% of the squared covariance. It is found that the warm ENSO (El Niño) events lead to a summer moisture deficit in the western two-thirds of Canada. Conversely, cold ENSO (La Niña) events produce an abundance of summer moisture, mainly in extreme western Canada and in the southeastern portions of the Canadian Prairies.

The first SVD mode strongly relates to the trend in global SSTs and multidecadal variation of the Atlantic SST, explaining approximately one-third of the squared covariance. It is reflective of both the warming trend in the global southern oceans and the influences of the Atlantic multidecadal oscillation (AMO) variability.

The 6-month lag relationship between the PDSI and large-scale SSTs provides a basis for developing long-range forecasting schemes for drought in Canada. A two-tier forecast scheme, in which the SST is predicted by an ocean model or a coupled climate model, can potentially further increase the lead time of drought forecasting.

Corresponding author address: Amir Shabbar, Meteorological Service of Canada, Environment Canada, 4905 Dufferin Street, Toronto, ON M3H 5T4, Canada. Email: amir.shabbar@ec.gc.ca

1. Introduction

Prolonged drought can have a serious impact on the Canadian economy, natural environment, and society, especially in western Canada where drought-related losses were in excess of 1.8 billion dollars in 1988 (Wheaton and Arthur 1989). Economic losses and other impacts due to the 2001–02 drought are estimated at 6.14 billion dollars (S. Kulshreshtha 2003, personal communication). The uncertainty of drought prediction contributes to substantial crop insurance payouts every year. A better understanding of the spatial distribution of drought, and its frequency, intensity, and duration in Canada is, thus, required. Increased knowledge of the relationship of drought intensity, duration, and spatial extent to large-scale ocean–atmosphere forcing is necessary for predicting seasonal drought severity, as well as for planning for impacts due to future climate change.

Drought is often defined in terms of its primary influences (such as agricultural productivity, soil moisture, reservoir levels, and streamflow) or by its economic impacts. It can be described as a significant deficit in moisture availability over an extended period of several months arising from lower-than-normal rainfall and/or higher-than-normal air temperature. The change in the balance between evaporation and precipitation may result from changes in the atmospheric circulation, which may be teleconnected to long-term ocean variability. This definition is complicated, however, when attempts are made to compare droughts in regions with differing atmospheric moisture regimes. Thus, an index standardized to regional and local climatology is required when approaching studies on a national to continental spatial scale. A number of different indices have been developed to quantify drought, each with its own strengths and weaknesses (Heim 2002).

Two of the most commonly used are the Palmer Drought Severity Index (PDSI) and the Standard Precipitation Index (SPI). The PDSI reflects long-term moisture, runoff, recharge, deep percolation, and evaporation. The index is useful for drought analysis over time spans of months or seasons. Also, it has proven useful for drought reconstruction prior to the instrumental record by using climate proxy data in the conterminous United States (Cook et al. 1999), as well as for summer drought reconstruction from the early eighteenth century from tree rings in the southwestern Canadian Prairies (Sauchyn and Skinner 2001). There have been several drought indices developed (Heim 2002) that are related to different aspects of drought intensity, duration, and persistence. This paper does not discuss the relative merits of each index, but only uses the PDSI because it is the most commonly cited index, and currently constitutes a component of the United States Drought Monitor (Lawrimore et al. 2002).

Severe large spatial scale North American droughts over the past century have had a tendency to be clustered in successive years and are likely related to ocean–atmosphere variability on similar time scales (Hoerling and Kumar 2003). The Dust Bowl drought of the 1930s was a natural disaster that severely affected much of the western United States and western Canada, occurring in three waves: 1934, 1936, and 1939–40. Some regions of the High Plains (from Colorado and Kansas to southern Alberta and Saskatchewan) experienced drought conditions for as many as eight consecutive years. During the 1950s, the southwestern United States and the Great Plains, and the adjacent Canadian Prairies, were subjected to a 5-yr drought. In three of those years, drought conditions stretched from coast to coast. The 1950s drought was characterized by both below-normal rainfall and higher-than-normal air temperatures. The 3-yr drought of the late 1980s (1987–89) resulted in tremendous agricultural losses in western Canada. In addition, approximately 7 million ha of Canadian forest burned during the summer of 1989—the most severe forest fire year since reliable records began in the early 1950s (Skinner et al. 1999). This includes the most recent 2003 fire season, when only approximately 1.6 million ha burned in British Columbia. The drought of 2001–02 was particularly severe in western Canada. As well, it had national extent during the summer of 2001, similar to large-scale drought patterns of the 1980s and 1950s.

A few recent studies have examined the associations among hydrologic drought and climatic conditions leading to drought in the United States and the El Niño– Southern Oscillation (ENSO) phenomenon in the tropical Pacific Ocean (i.e., Cole and Cook 1998; Nigam et al. 1999; Trenberth and Barnstator 1992). The role of ENSO in the initiation of North American drought is further supported in a study by Rajagopalan et al. (2000). Using partial correlation analysis and the singular value decomposition (SVD) technique, they identified the changing relationship between ENSO and the United States drought patterns during the twentieth century. By employing a suite of climate model simulations, Hoerling and Kumar (2003) have recently linked the Northern Hemisphere midlatitude droughts that occurred from 1998 to 2002 in the United States, southern Europe, and southwest Asia to common oceanic influences. Cold sea surface temperature (SST) in the eastern tropical Pacific and warm SSTs in the western tropical Pacific and Indian Oceans, a condition reminiscent of the cold phase of ENSO, persisted throughout the period. Their climate model experiments indicated that the climate anomalies forced by these oceanic regions contributed in unison to large-scale midlatitude drying. Forced by anomalous rainfall in the tropical Pacific, the resulting Rossby waves give rise to one of the principal modes of variability in the Northern Hemisphere circulation—the Pacific–North American (PNA) teleconnection (Horel and Wallace 1981; Hoskins and Karoly 1981). The PNA teleconnection has one of its strongest centers of action over western Canada. It is, therefore, important to investigate and quantify the role of the interannual and interdecadal oceanic anomalies in the initiation and maintenance of extended drought in Canada.

A strong relationship between seasonal Canadian temperature and precipitation, key factors in the formation of drought, and the ENSO cycle has already been established (Shabbar and Khandekar 1996; Shabbar et al. 1997a). As well, Bonsal and Lawford (1999) have related variations in the tropical Pacific SSTs to regional Canadian Prairie dry spells. It is plausible that midlatitude decadal ocean–atmosphere variability could also play an important role in the establishment and persistence of extended dry spells in Canada. Mestas-Nuñez and Enfield (1999) define the Atlantic multidecadal oscillation (AMO) as the first rotated empirical orthogonal function (EOF) of the global SSTs from which intraseasonal ENSO and local trends have been removed. The AMO is a long-time-scale oceanic phenomena with a 65–80 yr time scale and 0.4°C range. Enfield et al. (2001) have found distinct associations between cold and warm phases of the AMO and the United States summer rainfall and river flows. This Atlantic atmosphere–ocean long-term variability may also provide additional information about summer moisture patterns in Canada.

In Canada, the physical and economic impacts of drought are most evident during the warm season. Numerous studies have linked Canadian and United States precipitation and ENSO variability (e.g., Ropelewski and Halpert 1986). Few studies, however, have focussed on summer season moisture conditions and ocean–atmosphere forcing mechanisms during the preceding winter season. Evidence for the lagged relationship between the summer climate variability in Canada and preceding winter tropical SSTs exist in observational data. While examining seasonal temperature and precipitation prediction skill, Shabbar and Barnston (1996) identified the ENSO cycle, which manifests itself most forcefully in winter SSTs, as the main source of variability in producing skillful forecasts for Canada from winter into early summer. As well, Rajagopalan et al. (2000) found a coupled pattern between summer drought over the continental United States and winter SST variability during the twentieth century. For these reasons, summer-season drought was analyzed in order to assess the utility of antecedent winter-season global SST anomalies for drought prediction.

The PDSI is a soil moisture accounting algorithm based on the water balance and is derived from ongoing measurements of precipitation, air temperature, and local soil moisture. The index, in any given season, is a function of climatic and soil moisture conditions occurring in the current and preceding seasons, thus, making it a potentially useful measure to predict. Soil moisture is affected by precipitation, which, in turn, influences current and future surface temperature, primarily through evaporation (Huang et al. 1996).

The purpose of this study is to determine the spatial patterns and temporal variability of summer (June–August) moisture availability in Canada as determined by the PDSI from 1940 to 2002. Composite analysis and SVD of the joint structure between Canadian PDSI and global SSTs are used to quantify the impact of ENSO and other midlatitude ocean–atmosphere variability on extended dry spells. Additionally, structures in PDSI related to global warming, as manifested by long-term changes in global SSTs, are also examined. It is hoped that the coupled patterns obtained by identifying the source of global SST variability in the preceding winter months (December–February) will provide useful insights toward developing a seasonal and possibly a multiyear drought severity prediction scheme in Canada. The rest of the paper is organized as follows: the datasets are described in section 2. The SVD methodology and the teleconnection results are highlighted in section 3. The paper closes with a summary and discussion in section 4.

2. Data

The Palmer Drought Severity Index has been computed for approximately 100 Canadian stations from the homogenized Canadian historical air temperature and precipitation database of climate stations, having collocated monthly mean air temperature and monthly total precipitation records extending from 1940 or before to the end of 2002 (Fig. 1a). Monthly PDSI values for 344 contiguous United States Climate Divisions obtained from the National Climatic Data Center (NCDC) were added to the Canadian data to provide more complete spatial coverage for analysis at the international border.

The PDSI has been a commonly used drought index in North America and was developed to measure the intensity, duration, and spatial extent of drought. It is a soil moisture accounting algorithm based on the water balance. The computational procedure is described by Palmer (1965) and Alley (1984). PDSI computation begins with a climatic water balance, which is the same as that developed by Thornthwaite (1948). It requires temperature, precipitation, and soil water-holding capacity as input parameters. The procedure for estimating potential evapotranspiration (PE) is based on temperature and the length of day, and a hierarchy for satisfying water demands (e.g., first satisfy PE, then recharge the soil, and then allow for surface runoff). The Palmer model divides the soil into two layers, assuming that 25 mm can be stored in the surface layer and that moisture cannot be removed or recharged to the underlying layer (75 mm) until the surface layer has been depleted or saturated. There are three additional potential terms in the model of importance to soil moisture: potential recharge (PR; the amount of moisture necessary to bring the soil to its water holding capacity), potential loss [PL; the amount of moisture that can be withdrawn from the soil for evapotranspiration (ET) during a zero-precipitation period], and potential runoff (PRO; the soil water-holding capacity minus potential recharge). The model is initially calibrated to normal levels prior to calculating the monthly index. This involves simulating the water balance over a historical time series of climate data in order to derive several coefficients and parameters that are dependent on the area being analyzed. As a result, PDSI values are normalized with respect to time and location, and, thus, allow for comparison of the index in time and space.

The PDSI has become the primary tool for describing and monitoring prevailing drought (Louie 1986). The PDSI model allows measurement of prolonged abnormal dryness, or wetness, across a region and can be related directly to past weather conditions. Similar to the streamflow data, the PDSI reflects a relatively long-term memory that is modulated by seasonal influences. Summer PDSI values reflect moisture inputs and balances, not only during the current season but also over the previous several months, including winter snowfall and storage. It is, thus, absolutely necessary to have accurate winter precipitation (mainly snowfall) measurements for the soil moisture account.

a. Canadian historical climate database and PDSI

The homogenized Canadian historical air temperature and precipitation database is the result of several years of research at the Climate Research Branch of the Meteorological Service of Canada. This database of homogenized and long-term temperature time series of monthly maximum, minimum, and mean temperatures (Vincent 1998; Vincent and Gullett 1999) has been specifically designed for climate change research. Missing values were estimated using highly correlated neighboring stations. It was necessary in some cases to join short-term-station segments to create long-term series and to test for “relative homogeneity” with respect to surrounding stations. The methodology involves the identification of inhomogeneities in the temperature series, which are often nonclimatic steps due to station alterations, including changes in site exposure, location, instrumentation, observer, observing program, or a combination of the above. Monthly adjustment factors were derived from regression models, and adjustments were applied to bring each homogeneous segment into agreement with the most recent homogenous part of the series. Whenever possible, the main causes of the identified inhomogeneities were retrieved through historical evidence, such as station maintenance reports.

Similar to air temperature, long-term precipitation datasets have also been prepared for climate change analyses in Canada (Mekis and Hogg 1999). However, the precipitation network density is insufficient to allow for widespread use of surrounding station data to identify and adjust inhomogeneities. Instead, the primary goal has been to correct daily rain and snow measurements for all known inhomogeneities. Adjustments were applied on the daily level for rain and snow separately. For each rain gauge type, corrections to account for wind undercatch and evaporation were implemented. Gauge-specific wetting loss corrections were also applied for each rainfall event. For snowfall, ruler measurements were used throughout the time series to minimize potential discontinuities introduced by the adoption of Nipher-shielded snow gauge measurements in the mid-1960s. Density corrections based upon coincident ruler and Nipher measurements were mapped for Canada and applied to all ruler measurements. Where necessary, and where possible, records from neighboring stations were joined by employing a technique based on a simple ratio of observations. Overlapping periods were used to minimize possible inhomogeneities.

In the absence of detailed, site-specific soil type and soil moisture characteristics, all stations in the Canadian analysis were given the same 100-mm field capacity at the drought model initiation. Negative PDSI values indicate dry conditions and positive values indicate wet conditions. PDSI values generally fall between −6 and +6 (Table 1), although extreme values of −10 and +10 can occur. Values near zero represent near-normal conditions.

All North American data were gridded using the Kriging method (Isaaks and Srivastava 1989) at approximately 250 km2, and data representing grid squares north of 40°N were further analyzed (Fig. 1a). The kriging method was used because it is flexible in the degree of smoothing, which allows for larger datasets over large and spatially diverse areas. Averages were calculated for the summer season. As an example, the PDSI field for summer 2001 is shown in Fig. 1b.

b. Global SST data

Extended reconstructed global sea surface temperature (Smith and Reynolds 2003) anomalies are analyzed for the 1940–2002 period, where the anomalies are calculated relative to the 1960–90 base period. For the 1940–97 period, in situ observations from a new version of the Comprehensive Ocean–Atmosphere Data Set (COADS) release 2 (Woodruff et al. 1998) are used together with an eigenvector reconstruction (Smith et al. 1998). The reconstruction of the data involves a rigorous quality control procedure and a statistical analysis methodology, which is an improvement over their previous version (Smith et al. 1996). Data from 1998 to 2002 are based on in situ observations and satellite estimates, which are combined using the optimum interpolation method as described by Reynolds et al. (2002). At high latitudes, an improved sea ice–to-SST conversion algorithm is used. The joined SST data are then analyzed at 2° resolution. The anomaly reconstruction is performed separately for the low- and high-frequency components, and the sum of these components constitute the total SST anomaly. Smith and Reynolds (2003) show that the resulting SSTs are capable of resolving dominant modes of climate variability. In order to extract dominant modes of SST variability, EOFs of the cross-covariance matrix are carried out on the winter averages.

3. Methods and results

Summary statistics of the Canadian summer PDSI data were extracted to enable ranking of national moisture conditions, and also to determine the years of extreme moisture conditions and relate them to current global SSTs anomaly patterns. Large-scale relationships between winter patterns of global SSTs and the following summer drought patterns in Canada are analyzed for the 63-yr period of 1940–2002 by SVD analysis (also known as maximum covariance analysis in the climate literature; von Storch and Zwiers 1999).

The SVD method aims to relate the two fields by decomposing their covariance matrix into singular values and two sets of paired-orthogonal vectors—one for each field. The covariance between the expansion coefficients of the leading pattern in each field is maximized. The singular values give the magnitude of the squared covariance fraction (SCF) as accounted for by the various singular values (Bretherton et al. 1992; Wallace et al. 1992). The square of any singular value between two fields for a given mode is indicative of the fraction of the total squared covariance accounted for by that mode. The teleconnections between the two fields is discerned by the spatial patterns of the heterogeneous correlation, which is defined as the serial correlation between the expansion coefficients of one field and the grid point anomaly values of the other field r[PDSIk(t), SSTi(t)] and r[SSTk(t), PDSIi(t)], where k and i refer to the mode and grid number, respectively. The heterogeneous correlation patterns for the nth mode in the SVD expansion indicate how well the pattern of anomalies in the PDSI (SST) field can be specified on the basis of the nth expansion coefficient of the SST (PDSI) field. In the spatial domain, the loading maps for a given field are mutually orthogonal. Heterogeneous correlation patterns for the first three modes in the direct SVD expansion are analyzed for statistical evidence of teleconnections between the two fields. The homogeneous correlation pattern indicates that the spatial pattern of the variations are associated with the time series of the mode in the same field. The structure of both heterogeneous and homogeneous correlation patterns tends to be the same if the correlation between the amplitude time series of the given mode is high and significant. In this study, we will focus on the heterogeneous correlation patterns between the PDSI and global SSTs. The SVD method is similar to canonical correlation analysis (CCA) where two sets of orthogonal time series are produced, along with their corresponding spatial patterns. CCA aims to maximize correlation between variables, while SVD aims to maximize covariance between variables. Barnston and Smith (1996) provide an in-depth analysis of CCA and its relation with SVD.

a. Extreme moisture conditions

The time series of average summer PDSI for Canada is shown in Fig. 2. Also shown is the standard deviation of summer PDSI. Conditions were generally dry during the 1940s and 1950s, wet during the 1960s to the mid-1990s, and dry again since that time. The driest years occurred in the 1940s and since 1995. Although there is a tendency toward drier conditions since the 1970s, two of the wettest years on record are 1992 and 1996. This highlights the existence of strong interannual variability in the last 13 yr compared to the previous years. There is no evidence of changes in the spatial variability (standard deviation) of PDSI over the period of record.

Figures 3a and 3b show the composite PDSI patterns for the five driest and the five wettest Canadian summers since 1950, as derived from Fig. 2. The dry composite identifies generally dry conditions across Canada with extremely dry conditions in the central areas of the three Prairie provinces. The wet composite identifies extremely wet conditions in most regions of Canada, although there are areas in southern British Columbia and on the Pacific coast that show somewhat drier conditions. The extremes of moisture have a strong tendency to be clustered over a span of successive years, for example, 1972–75 (wet), and 1987–89 and 1998–2002 (both dry), which is suggestive of larger-scale forcings on similar temporal scales (Zhang et al. 1997; Nigam et al. 1999; Rajagopalan et al. 2000).

Composites of previous winter SST anomaly patterns are shown in Figs. 4a and 4b, corresponding to Canadian summers as identified in Figs. 3a and 3b, respectively. Dry conditions in Canada coincide with warmer-than-normal SSTs in the equatorial eastern Pacific, along the North American west coast, and in the North Atlantic. Previous winter SST anomalies along the equatorial Pacific are warmer than normal, indicative of the warm phase of ENSO. The warm SST anomalies in the North Atlantic resemble the AMO pattern. Dashed lines delineate regions where the composite values are statistically significant at the 5% level. Wet summer conditions in Canada coincide with colder-than-normal SSTs in the previous winter, in the central equatorial Pacific, and along the North American west coast, indicative of the cold phase of ENSO. Present also in the North Pacific is the SST pattern resembling the North Pacific mode as identified by Deser and Blackmon (1995), or the ENSO-like interdecadal variability mode (Zhang et al. 1997). Previous winter global SST anomaly patterns are distinctly different prior to Canadian summers with extreme wet or dry conditions. Regions with values statistically different from zero at the 5% level are shown by dashed lines. These SST composites suggest the influences of large-scale global ocean teleconnection patterns in determining seasonal extreme moisture conditions over Canada. In this study, these teleconnections are explored using the SVD analysis.

b. SVD analysis of Canadian drought data and global SST data

Prior to the SVD analysis, the dimensionality of drought and SST datasets was reduced by the EOF analysis. The first 10 EOF modes of the summer drought and winter SST were chosen as input into the SVD procedure. These 10 EOFs capture just over 70% of the total variance in their respective datasets. Most of the variance, however, is concentrated in the first three EOFs, with 41.2% of the total variability for summer drought and 43.7% of the total variability for winter SST.

The SVD analysis is used to identify and compare the coupled modes of variability in the PDSI and SST anomaly data. A similar technique has been used by Wallace et al. (1992) and Iwasaka and Wallace (1995) to identify large-scale relationships between SST and 500-hPa height anomalies and heat flux and atmospheric circulation, respectively. A mathematical description of SVD analysis can be found in Iwasaka and Wallace (1995). SVD methodology always gives a mathematical solution, and it is recognized that there is a chance that the resulting pair of coupled patterns may be nothing more than a mathematical artifact (Cherry 1997; Newman and Sardeshmukh 1995). The geophysical interpretation of the coupled patterns, however, can be aided by comparison of the results from the principal components analysis (PCA), where the time expansion of summer PDSI EOFs are correlated with the winter anomaly field of the SSTs and vice versa.

c. SVD pattern of Canadian summer drought

Figure 5a shows that the spatial pattern associated with S1(PDSI), explaining 12.7% of the total variance of Canadian PDSI, has a northwest–southeast dipole in western and central Canada, with opposing extreme conditions over the southern Prairie provinces and northwestern Canada. There is a third, albeit weaker, center located over the lower Great Lakes. The time series for S1(PDSI) shows both interannual and decadal variability with a positive trend.

Figure 5b shows the spatial pattern associated with S2(PDSI) (12.0% of the total variance), identifying a gradient to more severe conditions in western and central Canada, centered on the southern Prairie provinces. The time series for S2(PDSI) shows a decreasing trend until the mid-1970s, followed by an abrupt increase toward higher values starting in 1976–77. It will be shown that this mode of the PDSI has its origin in the ENSO phenomenon.

Figure 5c shows that the spatial pattern associated with S3(PDSI) (8.8% of the total variance) has strong positive loadings in western and northwestern Canada and negative loadings in northeastern Canada. Additionally, there is a weak positive center over the southeastern portions of the Prairie provinces. The time series for S3(PDSI) shows little evidence of trend but considerable interannual variability prior to 1960 and after 1990 with extreme conditions in the early 1990s.

d. SVD pattern of global SST

The spatial pattern associated with S1(SST), explaining 10.6% of the total variance of global winter SST variability, is shown in Fig. 6a. It identifies the warming trend signal in the SSTs with strong positive loadings throughout the southern oceans. A similar SST loading pattern has been reported by Smith and Reynolds (2003). In their second rotated EOFs of the global SSTs, they obtained a similar SST pattern, and identified it as the trend mode. The AMO pattern with negative loadings in the North Atlantic is also clearly evident. The accompanying time series shows considerable interannual variability with a strong positive trend, as well as an apparent shift toward higher values circa the early 1970s. This sharp change in the time coefficient is likely associated with the variability of the AMO. The role of the atmosphere in the North Atlantic sector in determining the underlying oceanic variability on an interannual time scale has been shown in a number of studies (e.g., Czaja and Marshall 2001; Hurrell 1996). While examining variability of the 500-hPa geopotential heights in the western Atlantic, Shabbar et al. (1997b) found a similar change in their Baffin–west Atlantic circulation index in the early 1970s.

Figure 6b exhibits the spatial pattern associated with S2(SST) (16.3% of the total variance). It identifies strong loadings in the tropical Pacific Ocean, with a weaker center in the North Pacific. Zhang et al. (1997) found a similar mode while examining wintertime variability in the high-pass (interannual) filtered SSTs. Deser and Blackmon (1995) also report an interannual ENSO pattern as their leading mode of Pacific basin SSTs, with a similar structure as the one shown in Fig. 6b. The accompanying time series shows extremes in ENSO years. Warm El Niño (positive) and cold La Niña (negative) ENSO events are also identified. The time series exhibits interannual variability with a long-term change toward higher values in 1976–77. The relative warmth of the tropical Pacific since 1976, as noted by Trenberth and Hurrell (1994) and Lau and Nath (1994), is reflected in the S2(SST) time series.

The spatial pattern associated with S3(SST) (12.5% of the total variance) is shown in Fig. 6c. The extratropical SST fluctuations in the central North Pacific, resembling the coupled ocean–atmosphere mode known as the Pacific decadal oscillation (PDO; Mantua et al. 1997), are prominent in this mode. This mode also includes a component of the interannual variability as reflected by a broader and weaker center in the eastern tropical Pacific. Compared to S2 (Fig. 6b), the tropical loadings in S3 (Fig. 6c) are located considerably further west, just east of the international date line. Deser and Blackmon (1995) identify this pattern in the Pacific basin SSTs as their second EOF. This pattern also bears a striking resemblance to the low-pass (interdecadal) filtered SSTs in the Pacific basin as identified by Zhang et al. (1997). The time series associated with this mode also identifies some warm (negative) and cold (positive) years of the ENSO phenomenon. Additionally, a change point in the time series is also evident corresponding to the 1976–77 regime change. This shift is reflective of the climate shift in the mean sea level pressure in the North Pacific as identified by Trenberth and Paolino (1981).

The two series S2(SST) and S3(SST) are correlated with one another at a modest level (r = −0.56, where r denotes the correlation coefficient). The differences in the North Pacific features between these two patterns have profound implications on the PDSI pattern in western Canada. Whereas the strong positive loadings in the North Pacific have a northeast–southwest orientation at 30°N in S3, the concomitant negative and weaker loading center is located north of 30°N in S2. These changes in the SSTs are associated with rather distinct midtropospheric circulation patterns. The difference in the composites of 500-hPa geopotential heights between the positive and negative phases of S3 has a very strong anomaly center in the central North Pacific (not shown). The same composite difference for S2 shows a weaker North Pacific center northward over the Aleutian Islands. These circulation changes have strong impacts on the temperature and moisture regimes over the west coastal areas of Canada. The stronger S3(SST) in the central North Pacific leads to higher PDSI over British Columbia and the southern Yukon (cf. Figs. 5b and 5c).

The mathematics of SVD separates the Pacific Ocean processes of ENSO and the interdecadal PDO as two separate modes. Together, these processes account for nearly 50% of the squared covariance fraction, thus, making ENSO and the ENSO-like interdecadal variability in the Pacific basin the most significant factors in explaining the summer moisture variability over Canada.

In order to discern geophysical relevancy of the SVD-coupled patterns, Cherry (1997) suggests a comparison of the SVD results with those from the PCA analysis. When the leading three principal components (PCs) of summer PDSI are correlated with the previous winter global SST anomalies, the salient features of Fig. 6, namely the ENSO-related SST pattern in the Pacific basin, the warming of the southern oceans and the multidecadal AMO dipole in the Atlantic basin, are all recovered (not shown). To highlight the separability of S2 and S3 modes, the correlation between the Pacific basin SSTs and the individual PCs of the PDSI was calculated. The results show the interannual ENSO and the interdecadal PDO variability as two separate entities (not shown). This comparison lends credence to the dynamical significance of the SVD-coupled patterns found in this study. The statistical significance of the coupled patterns will be shown below.

e. Teleconnection between Canadian drought and global SST patterns

Table 2 provides the three main summary statistics of the SVD analysis. These statistics provide a measure of the strength of the relationship between the two fields. The first statistic, the squared covariance fraction (SCFk), where k is the mode number, provides the percentage of the total squared covariance between the two fields explained by the SVD mode, and is proportional to the square of its singular value. This is a measure of the relative importance of the SVD mode in the relationship between the two fields. It is clear from Table 2 that the squared covariance is concentrated in the first three modes (approximately 80% of the squared covariance). Thereafter, the squared covariance statistic drops off sharply, thus, signifying the importance of the first three modes in determining PDSI variability. The second statistic is the correlation coefficient (rk) between the two time series that represent the temporal variations of the mode in the two fields. It is a measure of the similarity between the time variations of the patterns of the two fields, or how strongly the two fields are related to each other. The values remain near 0.5 (significant at the 5% level) for the first five modes. The third statistic, the normalized root-mean-squared covariance (NCFk), is the ratio of the squared singular value of the mode to the greatest possible total squared covariance of the matrix. It is a measure of the absolute importance of the SVD mode in the relationship between the two fields. Nearly 19% of NCF is concentrated in the first three modes of the SVD expansion. In the first three modes, the values are about equally distributed and drop off in higher modes, again emphasizing the importance of these modes in relation to higher modes. While examining relationships between surface heat flux over the North Pacific and 500-hPa geopotential heights and SSTs, Iwasaka and Wallace (1995) found that the significant NCFs were in the range of 8%–14%.

The robustness of these results is established by the Monte Carlo technique, as suggested by Barnett and Preisendorfer (1987). Test results on 1000 Monte Carlo SVD expansions, in which the time coefficient of the PDSI series is randomly shuffled, show that the rk and NCFk statistics in all three modes of the PDSI and global SST anomaly data are statistically significant at the 5% level. The (rk) 5% significance level for r1 = 0.45, r2 = 0.38, and r3 = 0.33. The NCFk 5% significance level for NCF1 = 4.66, NCF2 = 3.40, and NCF3 = 2.66 (see Table 2 for comparison with the actual values).

The heterogeneous correlation patterns show how the two fields are related to one another and how much of the amplitude of the variations is explained by the SVD mode. Figure 7 shows the heterogeneous correlation patterns for each of the first three modes in the SVD expansion for winter global SST anomaly data and for the following Canadian summer drought (PDSI) data from 1940 to 2002. Each map represents the correlation between the expansion coefficients of one field and the grid point anomaly values of the other field. The heterogeneous pattern for the first SVD mode (Fig. 7a) has two prominent features. The primary characteristic is the warming of the southern oceans as indicated by the positive loadings, mainly south of the equator. Secondly, the North Atlantic center of action is indicative of cooling. In their rotated EOF analysis, Smith and Reynolds (2003) have also identified this mode as the trend mode. In the next section, the North Atlantic feature will be related to the influences of the AMO. The accompanying PDSI pattern shows a deficit in the PDSI in an area stretching from the southern Canadian prairies to central Quebec. While examining the summer precipitation trend during the second half of the twentieth century, Zhang et al. (2000) also found a tendency toward drying throughout the same region. In the following section, it will be shown that the centers over northwestern Canada and over the lower Great Lakes and the St. Lawrence valley are more closely related to the variability in the AMO. Statistics calculated in Table 2 mark this mode as the most important singular mode in the relationship between the two fields.

The heterogeneous pattern for the second SVD mode (Fig. 7b) has its origin mainly in the interannual ENSO mode as identified by Zhang et al. (1997) and Deser and Blackmon (1995). This mode shows that the warm phase of ENSO leads to a deficit in summer PDSI over western and central Canada. The center is particularly strong over southwestern British Columbia and over the central Prairie provinces.

The heterogeneous pattern for the third SVD mode (Fig. 7c) mainly reveals the ENSO-like interdecadal variability in the North Pacific, with an indication of a weaker and more diffuse ENSO variability in the equatorial Pacific. As noted earlier, the structure of this mode relates more to the interdecadal variability in the central North Pacific (see Fig. 6c), where the center is stronger and shifted southwestward, compared to the center shown in Fig. 7b. The resulting atmospheric circulation from this pattern is conducive to producing a stronger center of circulation over the North American west coast, which leads to an excessive moisture regime over western Canada. In this configuration, the subtropical jet stream brings moisture-bearing storms to the west coast of Canada. This mode shows that the negative phase of the PDO, along with the cold phase of the interannual ENSO, leads to higher PDSI values in summer over western Canada and over the southeastern areas of the Prairie provinces.

4. Summary and discussion

Recently, Hoerling and Kumar (2003) have linked the droughts that occurred from 1998 to 2002 in certain areas of the midlatitudes to common global oceanic influences. In particular, observation and modeling results relate the cold ENSO-like conditions to drought in the southwestern United States, Europe, and southwest Asia. A long-time-scale oceanic phenomenon referred to as the AMO, the first rotated EOF of Atlantic SSTs, has been found with a 65–80-yr time scale (Enfield et al. 2001). The signal is most intense in the North Atlantic, but is global in scope, with a positively correlated co-oscillation in the North Pacific. An AMO warm phase occurred from 1940 to 1960 with less-than-normal rainfall in the continental United States, and a cool phase occurred from 1970 to 1990 with greater-than-normal rainfall in the continental United States. In this study, coupled modes between the winter global SSTs and the following summer PDSI are documented for Canada. It is found that the interannual and interdecadal variability in the Pacific basin, as well as the multidecadal variability (AMO), play a prominent role in affecting summer drought conditions.

Figure 8 shows the heterogeneous correlation patterns for the first mode in the SVD expansion for winter Atlantic SST anomaly data and following Canadian summer drought PDSI data from 1940 to 2002. The heterogeneous patterns for this SVD mode retain the features identified in the global SST analysis in Fig. 7a. The primary feature of the SST pattern is the strong center of negative loadings in the North Atlantic. This pattern is similar to the AMO pattern of variability obtained by Mestas-Nuñez and Enfield (1999). The complementary drought map exhibits strong positive centers over northwestern Canada and over the lower Great Lakes and the St. Lawrence valley, and a weak negative center over the extreme southeastern Prairies. This pattern is very similar to the drought pattern shown in Fig. 7a. Thus, the first SVD mode (Fig. 5a) and the heterogeneous correlation (Fig. 7a) of the Canadian PDSI are reflective of both the warming trend in the global southern oceans and the influences of the AMO variability. The percentage of SCF, as explained by the first mode between the winter Atlantic SSTs and the following summer PDSI, is over 56%. The correlation coefficient between their expansion coefficients is 0.63, and over 10% of the NCF, is concentrated in the first mode of the SVD expansion. All three statistics are significant at the 5% level, indicating the presence of a strong coupled mode.

Figure 9a provides further corroborating evidence concerning the AMO structure inherent in the Atlantic sector S1 (Fig. 8). Plotted on this chart are the AMO index and the time coefficient of the Atlantic S1 pattern. The two time series track each other fairly well with a correlation of 0.57, which is significant at the 5% level. The AMO warm phase from the 1950s to early 1960s is associated with dry conditions (see Fig. 2), while the cool phase from the 1960s to the 1990s is related to wet conditions (see Fig. 2). Finally, the return to a warm AMO phase in the late 1990s is linked with drier conditions. Enfield et al. (2001) also found a negative correlation between the AMO time series and the Climate Division rainfall over the northeastern United States. The negative correlations over the lower Great Lakes and the St. Lawrence valley with the AMO index (Fig. 8) could be regarded as the northward extension of their results.

Figure 9b shows the correlation pattern between the winter AMO index and following summer PDSI at each Canadian grid location. The effect of AMO is seen most clearly in the three Prairie provinces, along the west coast, and the lower Great Lakes region of southern Ontario and central Quebec, with the most robust impact evident over the northern Prairie provinces. Significant negative correlation seen along the west coast of Canada likely reflects the positively correlated co-oscillation in North Pacific with the AMO, as suggested by Enfield et al. (2001). The correlation pattern in eastern Canada is quite similar to that found in the heterogeneous correlation drought pattern for the first mode in the direct SVD expansion SSTs (Fig. 7a). Further evidence of this relationship is provided in Fig. 9c, which shows the difference field between the positive and negative composites taken from the standardized amplitude for S1 (SST) Atlantic (Fig. 8). The emerging pattern strongly resembles that shown in Figs. 7a and 9b. Regions of Canada possessing statistically significant values at the 5% level are outlined by dashed lines.

In summary, this study has determined the spatial patterns of summer moisture availability in Canada as represented by the PDSI, and identified the source of variability in global SSTs in the preceding winter months. Large-scale relationships between summer PDSI patterns in Canada and previous winter global SST patterns are then analyzed using SVD analysis. The resulting coupled patterns provide insight into the teleconnection patterns between Canadian drought and global SSTs.

The leading three SVD modes of Canadian summer drought and global SST explain over 80% of the squared covariance fraction between the two fields. The first field is a mixture of global SST trend and variability in the Atlantic Ocean. This mode accounts for approximately one-third of the squared covariance. It identifies a multidecadal scale variation strongly related to summer drought conditions in southeastern and northwestern Canada, as well as a trend component related to southern global ocean variability throughout central Canada. The time series of the winter AMO is correlated with the following summer PDSI grid, and the resulting pattern reproduces the heterogeneous correlation pattern for the first mode in the direct SVD expansion.

The second and third fields are related to Pacific Ocean processes and the interrelationship between ENSO and the PDO. Together these two modes explain over 48% of the squared covariance, thus, marking interannual ENSO phenomenon and ENSO-related interdecadal (PDO) variability as the most significant processes in drought variability. Whereas Hoerling and Kumar (2003) link the cold ENSO-like conditions to recent droughts in the midlatitude regions of the Northern Hemisphere, the results of this study show that the warm phase of ENSO leads to droughts in the grain-growing areas of Canada. Specifically, the warm phase of ENSO leads to a deficit in summer PDSI over the western two-thirds of Canada. Conversely, the cold phase produces an excess in summer PDSI over western Canada and the southeastern areas of the Prairie provinces.

In order to gain confidence in the mathematical solution given by the SVD methodology, the geophysical relevancy of the coupled patterns should also be investigated (Cherry 1997). It is recognized that the coupled modes found in this study are subject to the constraint of the number of EOFs chosen for the SSTs and PDSI fields as input into the SVD technique, as well as the mathematical constraint of the SVD itself. The PCA analysis performed by correlating the PCs of SSTs with the PDSI and vice versa, and the recovery of the dominant features shown in Figs. 5 and 6 lends credibility to the dynamical significance of the coupled modes. As noted in the introduction, one of the aims of this study is to lay the groundwork for the development of long-range drought prediction in Canada. The SVD technique for finding the coupled pattern is aptly suited for this purpose because it is easy to apply, and by projecting observed SSTs onto the SVD modes, can be easily incorporated into an operational prediction scheme.

It is the atmospheric response to SST anomalies in the equatorial Pacific during ENSO events that determines ocean conditions over the remainder of the global oceans (Kumar and Hoerling 2003). The lagged relationship between winter ENSO–PDO and summer Canadian drought is at least partially supported by appealing to the connection between the Tropics and the North Pacific through the “atmospheric” bridge concept of Lau and Nath (1994). The impact of the ENSO cycle on the Canadian winter precipitation regime has been previously documented by Shabbar et al. (1997a). The role of the resulting snow cover over Canada can be regarded as an integrating factor that determines the variability of the summertime PDSI over Canada. This study further demonstrates the impact of ENSO on the climate variability by showing significant relationship with drought variability in Canada. The 6-month lag relationship between the PDSI and large-scale SSTs provides a basis for developing long-range forecasting schemes for the occurrence of drought in Canada. This predictability may be further enhanced by the inclusion of direct measurements of snow cover as well as regional information, such as soil type and characteristics and vegetation, that provide additional information independent of large-scale SSTs.

Acknowledgments

The reconstructed global SSTs were kindly provided by Tom Smith of the National Climatic Data Center of NOAA. Constructive comments from two anonymous reviewers and the editor are greatly appreciated.

REFERENCES

  • Alley, W. H., 1984: The Palmer Drought Severity Index: Limitations and assumptions. J. Climate Appl. Meteor, 23 , 11001109.

  • Barnett, T. P., and R. Preisendorfer, 1987: Origins and levels of monthly and seasonal forecast skill for the United States surface air temperature determined by the canonical correlation analysis. Mon. Wea. Rev, 115 , 18251850.

    • Search Google Scholar
    • Export Citation
  • Barnston, A. G., and T. M. Smith, 1996: Specification and prediction of global surface temperature and precipitation from global SST using CCA. J. Climate, 9 , 26602697.

    • Search Google Scholar
    • Export Citation
  • Bonsal, B. R., and R. G. Lawford, 1999: Teleconnections between El Niño and La Niña events and summer extended dry spells on the Canadian prairies. Int. J. Climatol, 19 , 14451458.

    • Search Google Scholar
    • Export Citation
  • Bretherton, C. S., C. Smith, and J. M. Wallace, 1992: An intercomparison of methods for finding coupled patterns in climate data. J. Climate, 5 , 541560.

    • Search Google Scholar
    • Export Citation
  • Cherry, S., 1997: Some comments on singular value decomposition analysis. J. Climate, 10 , 17591761.

  • Cole, J. E., and E. R. Cook, 1998: The changing relationship between ENSO variability and moisture balance in the continental United States. Geophys. Res. Lett, 25 , 45294532.

    • Search Google Scholar
    • Export Citation
  • Cook, E. R., D. M. Meko, D. W. Stahle, and M. K. Cleaveland, 1999: Drought reconstructions for the continental United States. J. Climate, 12 , 11451162.

    • Search Google Scholar
    • Export Citation
  • Czaja, A., and J. Marshall, 2001: Observations of atmosphere ocean coupling in the North Atlantic. Quart. J. Roy. Meteor. Soc, 127 , 18931916.

    • Search Google Scholar
    • Export Citation
  • Deser, C., and M. L. Blackmon, 1995: On the relationship between tropical and North Pacific sea surface temperature variations. J. Climate, 8 , 16771680.

    • Search Google Scholar
    • Export Citation
  • Enfield, D. B., A. M. Mestas-Nuñez, and P. J. Trimble, 2001: The Atlantic multidecadal oscillation and its relation to rainfall and river flows in the continental U.S. Geophys. Res. Lett, 28 , 20772080.

    • Search Google Scholar
    • Export Citation
  • Heim Jr., R. R., 2002: A review of twentieth-century drought indices used in the United States. Bull. Amer. Meteor. Soc, 83 , 11491165.

    • Search Google Scholar
    • Export Citation
  • Hoerling, M., and A. Kumar, 2003: The perfect ocean for drought. Science, 299 , 691694.

  • Horel, J. D., and J. M. Wallace, 1981: Planetary-scale atmospheric phenomena associated with the Southern Oscillation. Mon. Wea. Rev, 109 , 813829.

    • Search Google Scholar
    • Export Citation
  • Hoskins, B. J., and D. Karoly, 1981: The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci, 38 , 11791196.

    • Search Google Scholar
    • Export Citation
  • Huang, J., H. M. Van den Dool, and K. P. Georgakakos, 1996: Analysis of model-calculated soil moisture over the United States (1931–1993) and applications to long-range temperature forecasts. J. Climate, 9 , 13501362.

    • Search Google Scholar
    • Export Citation
  • Hurrell, J. W., 1996: Influences of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature. Geophys. Res. Lett, 23 , 665668.

    • Search Google Scholar
    • Export Citation
  • Isaaks, E. H., and R. M. Srivastava, 1989: An Introduction to Geostatistics. Oxford University Press, 561 pp.

  • Iwasaka, N., and J. M. Wallace, 1995: Large scale air sea interaction in the Northern Hemisphere from a view point of variations of surface heat flux by SVD analysis. J. Meteor. Soc. Japan, 73 , 781794.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., and M. P. Hoerling, 2003: The nature and causes for the delayed atmospheric response to El Niño. J. Climate, 16 , 13911403.

    • Search Google Scholar
    • Export Citation
  • Lau, N-C., and M. J. Nath, 1994: A modeling study of the relative roles of tropical and extratropical anomalies in the variability of the global atmosphere–ocean system. J. Climate, 7 , 11841207.

    • Search Google Scholar
    • Export Citation
  • Lawrimore, J., R. Heim Jr., M. Svoboda, V. Swail, and P. Englehart, 2002: Beginning a new era of drought monitoring across North America. Bull. Amer. Meteor. Soc, 83 , 11911192.

    • Search Google Scholar
    • Export Citation
  • Louie, P. Y. T., 1986: An operational drought severity index program for Canadian synoptic stations. Proc. 16th Symp. on Canadian Hydrology, Regina, SK, Canada, National Research Council, 101–112.

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

    • Search Google Scholar
    • Export Citation
  • Mekis, E., and W. D. Hogg, 1999: Rehabilitation and analysis of Canadian daily precipitation time series. Atmos.–Ocean, 37 , 5385.

  • Mestas-Nuñez, A. M., and D. B. Enfield, 1999: Rotated global modes of non-ENSO sea surface variability. J. Climate, 12 , 27342746.

  • Newman, M., and P. D. Sardeshmukh, 1995: A caveat concerning singular value decomposition. J. Climate, 8 , 352360.

  • Nigam, S., M. Barlow, and E. H. Berbery, 1999: Analysis links Pacific decadal variability to drought and streamflow in the United States. Eos, Trans. Amer. Geophys. Union, 80 , 621625.

    • Search Google Scholar
    • Export Citation
  • Palmer, W. C., 1965: Meteorological drought. U.S. Department of Commerce Weather Bureau Res. Paper 45, 58 pp.

  • Rajagopalan, B., E. Cook, U. Lall, and B. K. Ray, 2000: Spatiotemporal variability of ENSO and SST teleconnections to summer drought over the United States during the twentieth century. J. Climate, 13 , 42444255.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis. J. Climate, 15 , 16091625.

    • Search Google Scholar
    • Export Citation
  • Ropelewski, C. F., and M. S. Halpert, 1986: North American precipitation and temperature patterns associated with El Niño/Southern Oscillation (ENSO). Mon. Wea. Rev, 114 , 23522362.

    • Search Google Scholar
    • Export Citation
  • Sauchyn, D. S., and W. R. Skinner, 2001: A 400 year reconstruction of Palmer Drought Severity Index (PDSI) for the southern Canadian prairies. Can. Water Resour. Assoc. J, 26 , 253272.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., and A. G. Barnston, 1996: Skill of seasonal climate forecasts in Canada using canonical correlation analysis. Mon. Wea. Rev, 124 , 23702385.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., and M. Khandekar, 1996: The impact of El Niño–Southern Oscillation on the temperature field over Canada. Atmos.–Ocean, 34 , 401416.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., B. Bonsal, and M. Khandekar, 1997a: Canadian precipitation patterns associated with the Southern Oscillation. J. Climate, 10 , 30163027.

    • Search Google Scholar
    • Export Citation
  • Shabbar, A., K. Higuchi, W. Skinner, and J. L. Knox, 1997b: The association between the BWA index and winter surface temperature variability over eastern Canada and west Greenland. Int. J. Climatol, 17 , 11951210.

    • Search Google Scholar
    • Export Citation
  • Skinner, W. R., B. J. Stocks, D. L. Martell, B. Bonsal, and A. Shabbar, 1999: The association between circulation anomalies in the mid-troposphere and area burned by wildland fire in Canada. Theor. Appl. Climatol, 63 , 89105.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., and R. W. Reynolds, 2003: Extended reconstruction of global sea surface temperatures based on COADS data (1854– 1997). J. Climate, 16 , 14951510.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., R. W. Reynolds, R. E. Livezey, and D. Stokes, 1996: Reconstruction of historical sea surface temperatures using empirical orthogonal functions. J. Climate, 9 , 14031420.

    • Search Google Scholar
    • Export Citation
  • Smith, T. M., R. E. Livezey, and S. S. Shen, 1998: An improved method for analyzing sparse and irregularly distributed SST data on a regular grid: The tropical Pacific Ocean. J. Climate, 11 , 17171729.

    • Search Google Scholar
    • Export Citation
  • Thornthwaite, C. W., 1948: An approach toward a rational classification of climate. Geogr. Rev, 38 , 5594.

  • Trenberth, K. E., and D. A. Paolino, 1981: Characteristic patterns of variability of sea level pressure in the Northern Hemisphere. Mon. Wea. Rev, 109 , 11691189.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and G. W. Barnstator, 1992: Issues in establishing causes of the 1988 drought over North America. J. Climate, 5 , 159172.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere–ocean variations in the Pacific. Climate Dyn, 9 , 303309.

  • Vincent, L. A., 1998: A technique for the identification of inhomogeneities in Canadian temperature series. J. Climate, 11 , 10941104.

    • Search Google Scholar
    • Export Citation
  • Vincent, L. A., and D. W. Gullett, 1999: Canadian historical and homogeneous temperature datasets for climate change analyses. Int. J. Climatol, 19 , 13751388.

    • Search Google Scholar
    • Export Citation
  • von Storch, H., and F. W. Zwiers, 1999: Statistical Analysis in Climate Research. Cambridge University Press, 484 pp.

  • Wallace, J. M., C. Smith, and C. S. Bretherton, 1992: Singular value decomposition of wintertime sea surface temperature and 500-mb height anomalies. J. Climate, 5 , 561576.

    • Search Google Scholar
    • Export Citation
  • Wheaton, E. E., and L. M. Arthur, 1989: Some environmental and economic impacts of the 1988 drought. SRC Rep. E-2330-4-E-89, 362 pp.

  • Woodruff, S. D., H. F. Diaz, J. D. Elms, and S. J. Worley, 1998: COADS Release 2 data and metadata enhancements for improvements of marine surface flux fields. Phys. Chem. Earth, 23 , 517527.

    • Search Google Scholar
    • Export Citation
  • Zhang, X., L. A. Vincent, W. D. Hogg, and A. Niitsoo, 2000: Temperature and precipitation trends in Canada during the 20th century. Atmos.–Ocean, 38 , 395429.

    • Search Google Scholar
    • Export Citation
  • Zhang, Y., J. M. Wallace, and D. S. Batista, 1997: ENSO-like interdecadal variability: 1900–93. J. Climate, 10 , 10041020.

Fig. 1.
Fig. 1.

(a) North America monthly PDSI site locations as calculated from Canadian historical database stations (red) and the United States Climate Division centroids (blue) monthly mean air temperature and monthly total precipitation data. Data are gridded at approximate 250 km × 250 km. (b) Summer (JJA) PDSI for 2001; with the exception of southern Manitoba and parts of northwestern Ontario, most of southern Canada experienced prolonged drought

Citation: Journal of Climate 17, 14; 10.1175/1520-0442(2004)017<2866:SDPICA>2.0.CO;2

Fig. 2.
Fig. 2.

National summary of Canadian summer (JJA) PDSI, 1940– 2002. The line indicates the Canada-wide mean PDSI and the bars represent the spatial variability (standard deviation)

Citation: Journal of Climate 17, 14; 10.1175/1520-0442(2004)017<2866:SDPICA>2.0.CO;2

Fig. 3.
Fig. 3.

(a) Composite Canada PDSI map for the five driest summers (JJA) 1940–2002. The five driest summers in descending order are 2002, 1941, 1998, 1961, and 1940. (b) Composite Canada PDSI map for the five wettest summers (JJA) 1940–2002. The five wettest summers in ascending order are 1992, 1996, 1947, 1974, and 1972

Citation: Journal of Climate 17, 14; 10.1175/1520-0442(2004)017<2866:SDPICA>2.0.CO;2

Fig. 4.
Fig. 4.

(a) Global winter (DJF) SST anomaly composite for the five driest Canadian summers (JJA). (b) Same as (a), but for the five wettest Canadian summers (JJA). Thick-black dashed line delineates regions where the values are significant at the 5% level

Citation: Journal of Climate 17, 14; 10.1175/1520-0442(2004)017<2866:SDPICA>2.0.CO;2

Fig. 5.
Fig. 5.

(a) The first SVD pattern (S1) of PDSI and standardized amplitude based on data for 63 summer (JJA) seasons 1940–2002. (b) Same as (a) except for the second SVD pattern (S2) of PDSI. (c) Same as (a) except for the third SVD pattern (S3) of PDSI

Citation: Journal of Climate 17, 14; 10.1175/1520-0442(2004)017<2866:SDPICA>2.0.CO;2

Fig. 6.
Fig. 6.

(a) The first SVD pattern (S1) of SST and standardized amplitude based on data for 63 winter (DJF) seasons 1940–2002. (b) Same as (a) except for the second SVD pattern (S2) of SST. (c) Same as (a) except for the third SVD pattern (S3) of SST

Citation: Journal of Climate 17, 14; 10.1175/1520-0442(2004)017<2866:SDPICA>2.0.CO;2

Fig. 7.
Fig. 7.

(a) Heterogeneous correlation patterns for the first mode in the direct SVD expansion. The temporal correlation coefficient between the corresponding expansion coefficients r(PDSIk SSTk), where k refers to the mode number, and the SCF (%) are shown. (b) Same as (a), but for the second mode. (c) Same as (a), but for the third mode

Citation: Journal of Climate 17, 14; 10.1175/1520-0442(2004)017<2866:SDPICA>2.0.CO;2

Fig. 8.
Fig. 8.

Heterogeneous correlation patterns for the first mode in the direct SVD expansion using Atlantic SSTs. The temporal correlation coefficient between the corresponding expansion coefficients r(PDSIk, SSTk), where k refers to the mode number, and the SCF (%) are shown

Citation: Journal of Climate 17, 14; 10.1175/1520-0442(2004)017<2866:SDPICA>2.0.CO;2

Fig. 9.
Fig. 9.

(a) Winter (DJF) AMO and winter S1 (SST) Atlantic, 1940–2002. (b) Correlation pattern between the winter (DJF) AMO and the following summer (JJA) PDSI. (c) Difference field between the positive and negative PDSI composites taken from the standardized amplitude for S1 (SST). Regions of Canada with significant correlation and differences at the 5% level are shown by dashed lines

Citation: Journal of Climate 17, 14; 10.1175/1520-0442(2004)017<2866:SDPICA>2.0.CO;2

Table 1.

Severity classes of PDSI

Table 1.
Table 2.

Summary statistics of SVD analysis for Canadian drought (PDSI) data and global SST anomaly data, 1940–2002

Table 2.
Save