• Ackerley, D., and Coauthors, 2011: Using synoptic type analysis to understand New Zealand climate during the Mid-Holocene. Climate Past, 7, 11891207, https://doi.org/10.5194/cp-7-1189-2011.

    • Search Google Scholar
    • Export Citation
  • Ackerley, D., A. Lorrey, J. A. Renwick, S. J. Phipps, S. Wagner, and A. Fowler, 2013: High-resolution modelling of mid-Holocene New Zealand climate 6000 yr BP. Holocene, 23, 721285, https://doi.org/10.1177/0959683613484612.

    • Search Google Scholar
    • Export Citation
  • Arthur, D., and S. Vassilvitskii, 2006: K-means++: The advantages of careful seeding. SODA ′07: Proc. 18th Annual ACM-SIAM Symp. on Discrete Algorithms, ACM Digital Library, New Orleans, LA, 10271035.

  • Bettolli, M. L., O. C. Penalba, and W. M. Varga, 2010: Synoptic weather types in the south of South America and their relationship to daily rainfall in the core crop-producing region in Argentina. Aust. Meteor. Oceanogr. J., 60, 3748, https://doi.org/10.22499/2.6001.004.

    • Search Google Scholar
    • Export Citation
  • Cassou, C., 2008: Intraseasonal interaction between the Madden–Julian oscillation and the North Atlantic oscillation. Nature, 455, 523527, https://doi.org/10.1038/nature07286.

    • Search Google Scholar
    • Export Citation
  • Charles, S. P., B. C. Bates, I. N. Smith, and J. P. Hughes, 2004: Statistical downscaling of daily precipitation from observed and modelled atmospheric fields. Hydrol. Processes, 18, 13731394, https://doi.org/10.1002/hyp.1418.

    • Search Google Scholar
    • Export Citation
  • Cheng, C. S., G. Li, Q. Li, and H. Auld, 2010: A synoptic weather typing approach to simulate daily rainfall and extremes in Ontario, Canada: Potential for climate change projections. J. Appl. Meteor. Climatol., 49, 845866, https://doi.org/10.1175/2010JAMC2016.1.

    • Search Google Scholar
    • Export Citation
  • Cheng, C. S., G. Li, Q. Li, and H. Auld, 2011: A synoptic weather-typing approach to project future daily rainfall and extremes at local scale in Ontario, Canada. J. Climate, 24, 36673685, https://doi.org/10.1175/2011JCLI3764.1.

    • Search Google Scholar
    • Export Citation
  • Cohen, L., S. Dean, and J. Renwick, 2013: Synoptic weather types for the Ross sea region, Antarctica. J. Climate, 26, 636649, https://doi.org/10.1175/JCLI-D-11-00690.1.

    • Search Google Scholar
    • Export Citation
  • Dean, S. M., and P. A. Stott, 2009: The effect of local circulation variability on the detection and attribution of New Zealand temperature trends. J. Climate, 22, 62176229, https://doi.org/10.1175/2009JCLI2715.1.

    • Search Google Scholar
    • Export Citation
  • DeLaFrance, A., and S. A. McAfee, 2019: Evaluation of synoptic-scale patterns during extreme temperature and precipitation events in Alaska. Int. J. Climatol., 39, 31343146, https://doi.org/10.1002/joc.6006.

    • Search Google Scholar
    • Export Citation
  • Fauchereau, N., B. Pohl, and A. Lorrey, 2016: Extratropical impacts of the Madden–Julian oscillation over New Zealand from a weather regime perspective. J. Climate, 29, 21612175, https://doi.org/10.1175/JCLI-D-15-0152.1.

    • Search Google Scholar
    • Export Citation
  • Frey, B. J., and D. Dueck, 2007: Clustering by passing messages between data points. Science, 315, 972976, https://doi.org/10.1126/science.1136800.

  • Gallant, A. J. E., S. J. Phipps, D. J. Karoly, A. B. Mullan, and A. M. Lorrey, 2013: Nonstationary Australasian teleconnections and implications for paleoclimate reconstructions. J. Climate, 26, 88278849, https://doi.org/10.1175/JCLI-D-12-00338.1.

    • Search Google Scholar
    • Export Citation
  • Gibson, P. B., S. E. Perkins-Kirkpatrick, and J. A. Renwick, 2016: Projected changes in synoptic weather patterns over New Zealand examined through self-organizing maps. Int. J. Climatol., 36, 39343948, https://doi.org/10.1002/joc.4604.

    • Search Google Scholar
    • Export Citation
  • Goodwin, I. D., and Coauthors, 2014: A reconstruction of extratropical Indo-Pacific sea-level pressure patterns during the medieval climate anomaly. Climate Dyn., 43, 11971219, https://doi.org/10.1007/s00382-013-1899-1.

    • Search Google Scholar
    • Export Citation
  • Griffiths, G., 2011: Drivers of extreme daily rainfalls in New Zealand. Wea. Climate, 31, 2449, https://doi.org/10.2307/26169716.

  • Hartigan, J. A., and M. A. Wong, 1979: Algorithm AS 136: A k-means clustering algorithm. J. Roy. Stat. Soc., 28C, 100108, https://doi.org/10.2307/2346830.

    • Search Google Scholar
    • Export Citation
  • Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 19992049, https://doi.org/10.1002/qj.3803.

    • Search Google Scholar
    • Export Citation
  • Hope, P. K., W. Drosdowsky, and N. Nicholls, 2006: Shifts in the synoptic systems influencing southwest Western Australia. Climate Dyn., 26, 751764, https://doi.org/10.1007/s00382-006-0115-y.

    • Search Google Scholar
    • Export Citation
  • Huang, W. T. K., A. Charlton-Perez, R. W. Lee, R. Neal, C. Sarran, and T. Sun, 2020: Weather regimes and patterns associated with temperature-related excess mortality in the UK: A pathway to sub-seasonal risk forecasting. Environ. Res. Lett., 15, 124052, https://doi.org/10.1088/1748-9326/abcbba.

    • Search Google Scholar
    • Export Citation
  • Jenkinson, A. F., and F. P. Collinson, 1977: An initial climatology of gales over the North Sea. Synoptic Climatology Branch Memo. 62, Meteorological Office, Bracknell, Berkshire, United Kingdom,18 pp.

  • Jiang, N., 2011: A new objective procedure for classifying New Zealand synoptic weather types during 1958-2008. Int. J. Climatol., 31, 863879, https://doi.org/10.1002/joc.2126.

    • Search Google Scholar
    • Export Citation
  • Jiang, N., J. E. Hay, and G. W. Fisher, 2004: Classification of New Zealand synoptic weather types and relation to the southern oscillation index. Wea. Climate, 23, 323, https://doi.org/10.2307/26169666.

    • Search Google Scholar
    • Export Citation
  • Jiang, N., G. Griffiths, and A. Lorrey, 2013: Influence of large-scale climate modes on daily synoptic weather types over New Zealand. Int. J. Climatol., 33, 499519, https://doi.org/10.1002/joc.3443.

    • Search Google Scholar
    • Export Citation
  • Jones, P. D., C. Harpham, and K. R. Briffa, 2013: Lamb weather types derived from reanalysis products. Int. J. Climatol., 33, 11291139, https://doi.org/10.1002/joc.3498.

    • Search Google Scholar
    • Export Citation
  • Jones, P. D., T. J. Osborn, C. Harpham, and K. R. Briffa, 2014: The development of Lamb weather types: From subjective analysis of weather charts to objective approaches using reanalyses. Weather, 69, 128132, https://doi.org/10.1002/wea.2255.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437471, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kidson, J. W., 1994: An automated procedure for the identification of synoptic types applied to the New Zealand region. Int. J. Climatol., 14, 711721, https://doi.org/10.1002/joc.3370140702.

    • Search Google Scholar
    • Export Citation
  • Kidson, J. W., 2000: An analysis of New Zealand synoptic types and their use in defining weather regimes. Int. J. Climatol., 20, 299316, https://doi.org/10.1002/(SICI)1097-0088(20000315)20:3<299::AID-JOC474>3.0.CO;2-B.

    • Search Google Scholar
    • Export Citation
  • Kidson, J. W., and J. A. Renwick, 2002: Patterns of convection in the tropical Pacific and their influence on New Zealand weather. Int. J. Climatol., 22, 151174, https://doi.org/10.1002/joc.737.

    • Search Google Scholar
    • Export Citation
  • Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247267, https://doi.org/10.1175/1520-0477(2001)082<0247:TNNYRM>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lachniet, M. S., C. M. Moy, C. Riesselman, H. Stephen, and A. M. Lorrey, 2021: Climatic and topographic control of the stable isotope values of rivers on the South Island of New Zealand. Paleoceanogr. Paleoclimatol., 36, e2021PA004220, https://doi.org/10.1029/2021PA004220.

    • Search Google Scholar
    • Export Citation
  • Lamb, H. H., 1950: Types and spells of weather around the year in the British Isles: Annual trends, seasonal structure of the year, singularities. Quart. J. Roy. Meteor. Soc., 76, 393429, https://doi.org/10.1002/qj.49707633005.

    • Search Google Scholar
    • Export Citation
  • Lamb, H. H., 1972: Meteorological office: Geophysical memoirs. Vol. 116, British Isles Weather Types and a Register of Daily Sequence of Circulation Patterns, 1861–1971. H.M. Stationery Office, 85 pp.

  • Lanfredi, I. S., and R. de Camargo, 2018: Classification of extreme cold incursions over South America. Wea. Forecasting, 33, 11831203, https://doi.org/10.1175/WAF-D-17-0159.1.

    • Search Google Scholar
    • Export Citation
  • Lefèvre, J., P. Marchesiello, N. C. Jourdain, C. Menkes, and A. Leroy, 2010: Weather regimes and orographic circulation around New Caledonia. Mar. Pollut. Bull., 61, 413431, https://doi.org/10.1016/j.marpolbul.2010.06.012.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., and P. R. Chappell, 2016: The “dirty weather” diaries of reverend Richard Davis: Insights about early colonial-era meteorology and climate variability for northern New Zealand, 1839-1851. Climate Past, 12, 553573, https://doi.org/10.5194/cp-12-553-2016.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., and N. C. Fauchereau, 2018: Southwest Pacific atmospheric weather regimes: Linkages to ENSO and extra-tropical teleconnections. Int. J. Climatol., 38, 18931909, https://doi.org/10.1002/joc.5304.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., A. M. Fowler, and J. Salinger, 2007: Regional climate regime classification as a qualitative tool for interpreting multi-proxy palaeoclimate data spatial patterns: A New Zealand case study. Palaeogeogr. Palaeoclimatol. Palaeoecol., 253, 407433, https://doi.org/10.1016/j.palaeo.2007.06.011.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., P. Williams, J. Salinger, T. Martin, J. Palmer, A. Fowler, J. Zhao, and H. Neil, 2008: Speleothem stable isotope records interpreted within a multi-proxy framework and implications for New Zealand palaeoclimate reconstruction. Quat. Int., 187, 5275, https://doi.org/10.1016/j.quaint.2007.09.039.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., and Coauthors, 2012: Palaeocirculation across New Zealand during the Last Glacial Maximum at ∼ 21 ka. Quat. Sci. Rev., 36, 189213, https://doi.org/10.1016/j.quascirev.2011.09.025.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., G. Griffiths, N. Fauchereau, H. J. Diamond, P. R. Chappell, and J. Renwick, 2014a: An ex-tropical cyclone climatology for Auckland, New Zealand. Int. J. Climatol., 34, 11571168, https://doi.org/10.1002/joc.3753.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., and Coauthors, 2014b: The Little Ice Age climate of New Zealand reconstructed from Southern Alps cirque glaciers: A synoptic type approach. Climate Dyn., 42, 30393060, https://doi.org/10.1007/s00382-013-1876-8.

    • Search Google Scholar
    • Export Citation
  • Mackintosh, A. N., B. M. Anderson, A. M. Lorrey, J. A. Renwick, P. Frei, and S. M. Dean, 2017: Regional cooling caused recent New Zealand glacier advances in a period of global warming. Nat. Commun., 8, 14202, https://doi.org/10.1038/ncomms14202.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., H. van Loon, and J. M. Arblaster, 2017: The role of the Southern Hemisphere semiannual oscillation in the development of a precursor to central and eastern Pacific Southern Oscillation warm events. Geophys. Res. Lett., 44, 69596965, https://doi.org/10.1002/2017GL073832.

    • Search Google Scholar
    • Export Citation
  • Michelangeli, P. A., R. Vautard, and B. Legras, 1995: Weather regimes: Recurrence and quasi stationarity. J. Atmos. Sci., 52, 12371256, https://doi.org/10.1175/1520-0469(1995)052<1237:WRRAQS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Moron, V., I. Gouirand, and M. Taylor, 2016: Weather types across the Caribbean basin and their relationship with rainfall and sea surface temperature. Climate Dyn., 47, 601621, https://doi.org/10.1007/s00382-015-2858-9.

    • Search Google Scholar
    • Export Citation
  • Pedregosa, F., and Coauthors, 2011: Scikit-learn: Machine learning in Python. J. Mach. Learn. Res., 12, 28252830, https://doi.org/10.48550/arXiv.1201.0490.

    • Search Google Scholar
    • Export Citation
  • Pepler, A. S., J. Fong, and L. V. Alexander, 2017: Australian east coast mid-latitude cyclones in the 20th century reanalysis ensemble. Int. J. Climatol., 37, 21872192, https://doi.org/10.1002/joc.4812.

    • Search Google Scholar
    • Export Citation
  • Pepler, A. S., A. J. Dowdy, and P. Hope, 2021: The differing role of weather systems in southern Australian rainfall between 1979–1996 and 1997–2015. Climate Dyn., 56, 22892302, https://doi.org/10.1007/s00382-020-05588-6.

    • Search Google Scholar
    • Export Citation
  • Pohl, B., A. Lorrey, A. Sturman, H. Quénol, J. Renwick, N. Fauchereau, and J. Pergaud, 2021: “Beyond weather regimes”: Descriptors monitoring atmospheric centers of action—A case study for Aotearoa New Zealand. J. Climate, 34, 8341–8360, https://doi.org/10.1175/JCLI-D-21-0102.1.

    • Search Google Scholar
    • Export Citation
  • Porhemmat, R., H. Purdie, P. Zawar-Reza, C. Zammit, and T. Kerr, 2021: The influence of atmospheric circulation patterns during large snowfall events in New Zealand’s Southern Alps. Int. J. Climatol., 41, 23972417, https://doi.org/10.1002/joc.6966.

    • Search Google Scholar
    • Export Citation
  • Prebble, J. G., and Coauthors, 2017: Evidence for a Holocene Climatic Optimum in the southwest Pacific: A multiproxy study. Paleoceanogr. Paleoclimatol., 32, 763779, https://doi.org/10.1002/2016PA003065.

    • Search Google Scholar
    • Export Citation
  • Ramos, A. M., D. Barriopedro, and E. Dutra, 2015: Circulation weather types as a tool in atmospheric, climate, and environmental research. Front. Environ. Sci., 3, 44, https://doi.org/10.3389/fenvs.2015.00044.

    • Search Google Scholar
    • Export Citation
  • Raphael, M. N., 2004: A zonal wave 3 index for the Southern Hemisphere. Geophys. Res. Lett., 31, L23212, https://doi.org/10.1029/2004GL020365.

    • Search Google Scholar
    • Export Citation
  • Raphael, M. N., 2007: The influence of atmospheric zonal wave three on Antarctic sea ice variability. J. Geophys. Res., 112, D12112, https://doi.org/10.1029/2006JD007852.

    • Search Google Scholar
    • Export Citation
  • Raynaud, D., B. Hingray, G. Evin, A.-C. Favre, and J. Chardon, 2020: Assessment of meteorological extremes using a synoptic weather generator and a downscaling model based on analogues. Hydrol. Earth Syst. Sci., 24, 43394352, https://doi.org/10.5194/hess-24-4339-2020.

    • Search Google Scholar
    • Export Citation
  • Renwick, J. A., 2011: Kidson’s synoptic weather types and surface climate variability over New Zealand. Wea. Climate, 31, 323, https://doi.org/10.2307/26169715.

  • Robertson, A. W., N. Vigaud, J. Yuan, and M. K. Tippett, 2020: Toward identifying subseasonal forecasts of opportunity using North American weather regimes. Mon. Wea. Rev., 148, 18611875, https://doi.org/10.1175/MWR-D-19-0285.1.

    • Search Google Scholar
    • Export Citation
  • Schroeter, S., H. Richter, C. Arthur, D. Wilke, M. Wehner, and E. Ebert, 2021: Forecasting the impacts of severe weather. Aust. J. Emerg. Manage., 36, 7683, https://doi.org/10.47389/36.1.76.

    • Search Google Scholar
    • Export Citation
  • Stashevsky, P. S., I. N. Yakovina, T. M. Alarcon Falconi, and E. N. Naumova, 2019: Agglomerative clustering of enteric infections and weather parameters to identify seasonal outbreaks in cold climates. Int. J. Environ. Res. Public Health, 16, 2083, https://doi.org/10.3390/ijerph16122083.

    • Search Google Scholar
    • Export Citation
  • Suriano, Z. J., 2020: Synoptic and meteorological conditions during extreme snow cover ablation events in the Great Lakes basin. Hydrol. Processes, 34, 19491965, https://doi.org/10.1002/hyp.13705.

    • Search Google Scholar
    • Export Citation
  • Tait, A., R. Henderson, R. Turner, and X. Zheng, 2006: Thin-plate smoothing spline interpolation of daily rainfall for New Zealand using a climatological rainfall surface. Int. J. Climatol., 26, 20972115, https://doi.org/10.1002/joc.1350.

    • Search Google Scholar
    • Export Citation
  • Tait, A., J. Sturman, and M. Clark, 2012: An assessment of the accuracy of interpolated daily rainfall for New Zealand. J. Hydrol., 51, 2544, https://www.jstor.org/stable/43944886.

    • Search Google Scholar
    • Export Citation
  • Theobald, A., H. McGowan, J. Speirs, and N. Callow, 2015: A synoptic classification of inflow-generating precipitation in the snowy mountains, Australia. J. Appl. Meteor. Climatol., 54, 17131732, https://doi.org/10.1175/JAMC-D-14-0278.1.

    • Search Google Scholar
    • Export Citation
  • van der Wiel, K., H. C. Bloomfield, R. W. Lee, L. P. Stoop, R. Blackport, J. A. Screen, and F. M. Selten, 2019: The influence of weather regimes on European renewable energy production and demand. Environ. Res. Lett., 14, 094010, https://doi.org/10.1088/1748-9326/ab38d3.

    • Search Google Scholar
    • Export Citation
  • Vautard, R., 1990: Multiple weather regimes over the North Atlantic: Analysis of precursors and successors. Mon. Wea. Rev., 118, 20562081, https://doi.org/10.1175/1520-0493(1990)118<2056:MWROTN>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Yiou, P., and M. Nogaj, 2004: Extreme climatic events and weather regimes over the North Atlantic: When and where? Geophys. Res. Lett., 31, L07202, https://doi.org/10.1029/2003GL019119.

  • View in gallery
    Fig. 1.

    (left) The spatial domain used for identifying the optimal number of weather regimes for New Zealand and the statistically significant trends (all colored areas) in NCEP––NCAR1 1000-hPa geopotential height (z1000) gridded data (1 Jan 1948–31 Dec 2020). All linear trends were removed prior to K-means clustering. (right) Main regions of New Zealand (standard font) and key towns and cities (bold) within those regions that are referred to in the text. Abbreviations for New Zealand regions, districts, cities, and towns can be found in Table 3. In some cases, a region or district can also have the same name as a city (e.g., Auckland, Gisborne, Wellington).

  • View in gallery
    Fig. 2.

    Affinity propagation results for detrended 1000-hPa geopotential height (z1000) in the New Zealand domain objectively indicating nine primary weather regimes should be ascribed for K-means clustering of once-daily synoptic weather patterns. See prior work (Frey and Dueck 2007) for more details about affinity propagation and its use in synoptic weather regime classification (Lorrey and Fauchereau 2018).

  • View in gallery
    Fig. 3.

    Atmospheric pressure anomalies and associated flow patterns for the nine once-daily New Zealand synoptic weather types that were defined using affinity propagation and K-means clustering on detrended and deseasonalized NCEP–NCAR1 z1000 data for 1979–2020, which overlaps with and that were independently verified using ERA5. LSE—low to the southeast; HS—high to the south; LSW—low to the southwest; HSE—high to the southeast; LNE—low to the northeast; L—low; HW—high to the west; LNW—low to the northwest; and H—high. The total number of days and percentage of frequency are noted next to the synoptic-type label at the top of each plot.

  • View in gallery
    Fig. 4.

    Frequency of occurrence of nine synoptic weather regimes for New Zealand determined using affinity propagation followed by K-means clustering of 1000-hPa geopotential height (z1000) NCEP–NCAR1 data spanning 1948–2020. The WR patterns determined from deseasonalized and detrended data were then re-projected back onto the untreated z1000 data to obtain frequency changes for each month. The red line is the mean annual frequency of occurrence (also indicated in each panel along with the amplitude of change related to the seasonal cycle).

  • View in gallery
    Fig. 5.

    Evolving changes in percent frequency occurrence per decade of nine New Zealand synoptic types as determined by affinity propagation. The year indicates the start of the decadal average window (i.e., 2010 = average percentage of occurrence for a type for 2010–19). Time series are shown for annual, spring (SON), summer (DJF), autumn (MAM), and winter (JJA), with trends shown only for the seasons that had significant long-term changes. Relative stepwise percentage change per decade and absolute percentage change of total synoptic-type occurrence for all days since 1948 are shown in the top-right corner.

  • View in gallery
    Fig. 6.

    Three LSE subsidiary types: low to the southeast with southwesterly flow (LSE-sw); low to the southeast with southwesterly flow intensified (LSE-swi); and low to the southeast (LSE-x), rare occurrence. Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

  • View in gallery
    Fig. 7.

    Three HS subsidiary types: high to the south with southeasterly flow (HS-se); high to the south, with disturbed circulation (HS-d); and high to the south with southeasterly flow intensified (HS-sei). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

  • View in gallery
    Fig. 8.

    Two LSW subsidiary types: low to the southwest with westerly flow (LSW-w) and low to the southwest with northwesterly flow (LSW-nw). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

  • View in gallery
    Fig. 9.

    Three HSE subsidiary synoptic types: high to the southeast with northeasterly flow (HSE-ne), high to the southeast with northerly flow (HSE-n), and high to the southeast with northerly flow intensified (HSE-ni). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

  • View in gallery
    Fig. 10.

    Two LNE subsidiary synoptic types: low to the northeast with easterly flow (LNE-e) and low to the northeast with disturbed circulation (LNE-d). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

  • View in gallery
    Fig. 11.

    Three low (cyclonic) subsidiary types: low with disturbed circulation (L-d), low with northeasterly flow (L-ne), and low with northeasterly flow intensified (L-nei). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

  • View in gallery
    Fig. 12.

    HW type: high to the west with westerly flow (HW-w). No subsidiary pattern was identified relative to the primary WR. Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

  • View in gallery
    Fig. 13.

    Two LNW subsidiary types: low to the northwest with easterly flow (LNW-e) and low to the northwest, rare occurrence (LNW-x). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

  • View in gallery
    Fig. 14.

    Two high (anticyclonic) subsidiary types: high, calm conditions (H-c) and high, rare occurrence (H-x). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 265 265 57
PDF Downloads 211 211 41

An Objective Weather Regime Classification for Aotearoa New Zealand Using a Two-Tiered K-Means Clustering Approach

Neelesh RampalaNational Institute of Water and Atmospheric Research, Auckland, New Zealand

Search for other papers by Neelesh Rampal in
Current site
Google Scholar
PubMed
Close
,
Andrew LorreyaNational Institute of Water and Atmospheric Research, Auckland, New Zealand

Search for other papers by Andrew Lorrey in
Current site
Google Scholar
PubMed
Close
, and
Nicolas FauchereaubNational Institute of Water and Atmospheric Research, Hamilton, New Zealand

Search for other papers by Nicolas Fauchereau in
Current site
Google Scholar
PubMed
Close
Open access

Abstract

Weather regimes (WRs), also known as synoptic types, are defined as recurrent patterns that have been used to categorize variability in atmospheric circulation. However, defining the optimal number of patterns can often be arbitrary, and there are common shortcomings when oversimplifying a wide range of synoptic conditions and weather outcomes. We build on previous work that has defined regional WRs and objectively ascribe an optimal number of once-daily weather patterns for Aotearoa New Zealand (ANZ) using affinity propagation combined with K-means clustering. Nine primary WRs for ANZ were classified based on once-daily geopotential height spatial patterns, but these patterns still retained a wide degree of spatial variability. Subsidiary clusters were subsequently defined within each primary WR by applying affinity propagation and K-means clustering to reveal the largest within-cluster differences based on joint daily temperature and precipitation anomalies. Up to three subsidiary patterns in each of the primary regimes were revealed, with a total of 21 unique daily patterns emerging from the two-tier classification. Subsidiary WRs reveal subtle differences in the location and intensity of regional-scale pressure anomalies, pressure gradients, and wind flow over both main islands that lead to large differences in surface weather anomalies. Impacts of atmospheric variability related to each subsidiary WR are exemplified by different spatial outcomes for rainfall and temperature (including intensity of anomalies) at regional and subregional levels. The approach presented in this study has utility for enhancing prediction of weather outcomes, including extreme weather, and can also be applied more widely over a range of time scales to improve understanding of weather and climate linkages.

© 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Neelesh Rampal and Andrew Lorrey contributed equally to this work.

This article is licensed under a Creative Commons Attribution 4.0 license (http://creativecommons.org/licenses/by/4.0/).

Corresponding author: Andrew Lorrey, a.lorrey@niwa.co.nz

Abstract

Weather regimes (WRs), also known as synoptic types, are defined as recurrent patterns that have been used to categorize variability in atmospheric circulation. However, defining the optimal number of patterns can often be arbitrary, and there are common shortcomings when oversimplifying a wide range of synoptic conditions and weather outcomes. We build on previous work that has defined regional WRs and objectively ascribe an optimal number of once-daily weather patterns for Aotearoa New Zealand (ANZ) using affinity propagation combined with K-means clustering. Nine primary WRs for ANZ were classified based on once-daily geopotential height spatial patterns, but these patterns still retained a wide degree of spatial variability. Subsidiary clusters were subsequently defined within each primary WR by applying affinity propagation and K-means clustering to reveal the largest within-cluster differences based on joint daily temperature and precipitation anomalies. Up to three subsidiary patterns in each of the primary regimes were revealed, with a total of 21 unique daily patterns emerging from the two-tier classification. Subsidiary WRs reveal subtle differences in the location and intensity of regional-scale pressure anomalies, pressure gradients, and wind flow over both main islands that lead to large differences in surface weather anomalies. Impacts of atmospheric variability related to each subsidiary WR are exemplified by different spatial outcomes for rainfall and temperature (including intensity of anomalies) at regional and subregional levels. The approach presented in this study has utility for enhancing prediction of weather outcomes, including extreme weather, and can also be applied more widely over a range of time scales to improve understanding of weather and climate linkages.

© 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Neelesh Rampal and Andrew Lorrey contributed equally to this work.

This article is licensed under a Creative Commons Attribution 4.0 license (http://creativecommons.org/licenses/by/4.0/).

Corresponding author: Andrew Lorrey, a.lorrey@niwa.co.nz

1. Introduction

Synoptic weather types (or weather regimes; WRs) are an important diagnostic tool that have been harnessed across a wide variety of meteorology and climate research applications. WRs provide a paradigm for categorizing and describing regional-scale atmospheric circulation patterns (Michelangeli et al. 1995; Jones et al. 2013; Pohl et al. 2021), and they mechanistically link atmospheric dynamics with near-surface environmental impacts. WRs can drive both advective and orographic effects, with anomalies of temperature, precipitation, sunshine, etc. arising from transient changes in pressure gradients and the direction of wind flow interacting with topography across a region. Connections of WRs to modes of variability (Cassou 2008; Fauchereau et al. 2016; Lorrey and Fauchereau 2018) and extremes (Cheng et al. 2010, 2011; DeLaFrance and McAfee 2019; Raynaud et al. 2020; Suriano 2020) have also been documented, which highlights the potential to exploit their transition properties and intra/inter-seasonal associations for prediction (Vautard 1990). Improved understanding of WRs can therefore assist the development of weather and climate guidance that can help mitigate impacts on weather- and climate-sensitive socioeconomic sectors (Huang et al. 2020; van der Wiel et al. 2019; Robertson et al. 2020; Schroeter et al. 2021).

Decomposition of regional-scale atmospheric circulation into discrete WRs is well established using a range of techniques and for a wide variety of locations (Ramos et al. 2015). Previous studies have applied clustering algorithms to daily and subdaily regional-scale pressure patterns within continental (Lamb 1950, 1972; Jones et al. 2014) and maritime climate settings (Yiou and Nogaj 2004; Lefèvre et al. 2010; Moron et al. 2016). In the Southern Hemisphere, WRs have been explored across tropical, midlatitude and polar regions (Kidson 2000; Hope et al. 2006; Bettolli et al. 2010; Cohen et al. 2013; Jiang 2011; Renwick 2011; Theobald et al. 2015; Lorrey and Fauchereau 2018; Lanfredi and de Camargo 2018; Pepler et al. 2021) to establish predominant classes of both average and extreme weather. In Aotearoa New Zealand (ANZ), an array of synoptic classification schemes (Kidson 1994, 2000; Jiang et al. 2004, 2013; Gibson et al. 2016) have been harnessed by the meteorology and climate research communities, with the most predominant applied scheme being that of Kidson (2000). Previous ANZ applications of Kidson (2000) include, statistical downscaling (Charles et al. 2004), paleoclimate reconstructions (e.g., Lorrey et al. 2007, 2008, 2012, 2014b; Ackerley et al. 2011, 2013; Goodwin et al. 2014; Prebble et al. 2017), surface weather and climate variability (e.g., Lorrey et al. 2014a; Fauchereau et al. 2016; Mackintosh et al. 2017; Lachniet et al. 2021; Porhemmat et al. 2021), tropical–midlatitude teleconnections (Kidson and Renwick 2002; Gallant et al. 2013) and early instrumental era historic climate (Lorrey and Chappell 2016).

A wide variety of regional synoptic classifications exist for ANZ; however, few studies have explored objective methods to objectively define the optimal number of WR patterns for this region (e.g., Jiang 2011). Application of dimensional reduction using K-means clustering (e.g., Kidson 2000) have well-known issues relating to arbitrary methodological decisions, including a priori assumptions about the expected number of regime clusters and the choice of retained principal components that capture weather variability. Furthermore, previous K-means applications for ANZ WRs have ignored the nonstationary state of modern climate, with synoptic-type classifications undertaken using un-detrended geopotential height data spanning the mid-twentieth to early twenty-first century.

It is also well-known that subsets of days with similar synoptic pressure spatial patterns as the WR archetype can have weather outcomes that appear completely dissimilar to average conditions. Thus, K-means application for ANZ (e.g., Kidson 2000) has known issues related to how daily weather patterns are simply grouped. While the K-means approach has benefits for climatological studies, there are unique subsets of synoptic situations grouped within one main archetypal pattern (and labeled as such) that can be widely divergent from the standard weather outcomes associated with the average pattern. This situation makes application of K-means synoptic-type classification less than ideal for describing extreme events, thereby limiting our interrogation of idiosyncratic conditions and their changes through time. As such, there is a need to revisit how we classify weather patterns objectively for ANZ.

Recent work used affinity propagation (Frey and Dueck 2007) to help ascribe an objective number of synoptic weather regimes for the southwest Pacific (Lorrey and Fauchereau 2018). The affinity propagation algorithm is founded on the concept of “message passing” between data points, and it uses a dissimilarity matrix as an input to find “exemplars” in a multivariate a dataset (i.e., 1000-hPa geopotential height employed to define daily weather regimes). With this approach, there is no a priori understanding of the number of expected clusters in a dataset of interest. While the affinity propagation algorithm is computationally intensive, a Monte Carlo sampling scheme that draws random samples from a larger dataset can be used with it to determine an optimal number of partitions or clusters for reanalysis data (Lorrey and Fauchereau 2018). Once a number of optimal clusters has been established using Affinity Propagation, a computationally efficient K-means clustering algorithm (Kidson 2000) can then be applied to the full dataset to establish archetypal patterns and classify daily weather types (Michelangeli et al. 1995). The K-means clustering algorithm has been previously described. It is capable of dividing a multidimensional dataset [like 4D reanalysis data; (Kistler et al. 2001)] into clusters, where the within-cluster sum of squares is minimized, and solutions are presented that achieve stability for that statistical metric irrespective of any inter-cluster movement of discrete data points (Hartigan and Wong 1979).

Our new two-tier WR scheme demonstrated for ANZ advances the standard application of K-means clustering for synoptic-type classification by illustrating the important nuances for regional pressure gradients (and related direction and intensity of flow) that are inherently hidden by a one-layer K-means clustering application to geopotential height. In doing so, we demonstrate the potential to apply these findings to all the aforementioned efforts that have previously employed Kidson’s (2000) synoptic-type classification as an investigative tool. The added benefit of expanding the ANZ synoptic-type toolbox with this new scheme is highlighted for extreme event attribution and for investigating how WRs cause extreme impacts in distinct regions. From this perspective, we also conclude that the new two-tiered scheme has wide utility and potential for prediction purposes. Links between the new WR types and modes of variability, in addition to the transition matrix for the WRs on daily to multi-week time scales, will be undertaken as follow-on studies.

2. Data and methods

a. Reanalyses

Once-daily WRs were determined using daily 1000-hPa geopotential heights (z1000) from NCEP–NCAR1 (Kalnay et al. 1996; Kistler et al. 2001) over the 1 January 1948–31 December 2020 time span. Daily averages of the zonal (U) and the meridional (V) component of the wind were also obtained for this analysis (http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.html) in order to calculate primary direction of wind flow for the north and south islands related to synoptic weather types. The domain of analysis encompasses 160°E–175°W and extends from 55° to 25°S (see Fig. 1), aligning to the same domain used in previous New Zealand synoptic weather regime classifications (Kidson 2000; Jiang et al. 2013). While this work focuses exclusively on use of the NCEP reanalysis, our work was further validated on ERA5 reanalysis at 2.5° resolution (Hersbach et al. 2020).

Fig. 1.
Fig. 1.

(left) The spatial domain used for identifying the optimal number of weather regimes for New Zealand and the statistically significant trends (all colored areas) in NCEP––NCAR1 1000-hPa geopotential height (z1000) gridded data (1 Jan 1948–31 Dec 2020). All linear trends were removed prior to K-means clustering. (right) Main regions of New Zealand (standard font) and key towns and cities (bold) within those regions that are referred to in the text. Abbreviations for New Zealand regions, districts, cities, and towns can be found in Table 3. In some cases, a region or district can also have the same name as a city (e.g., Auckland, Gisborne, Wellington).

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

b. Surface weather observations

To quantify relationships between atmospheric WR and surface weather outcomes, gridded daily temperature and daily accumulated rainfall were obtained from the NIWA Virtual Climate Station Network (VCSN; Tait et al. 2006). VCSN data cover the New Zealand region on a 5-km2 grid, derived from surface interpolation of station weather data using a thin-plate smoothing spline (Tait et al. 2006, 2012). It is important to note that rainfall and temperature biases are likely to occur in regions where in situ station observation density is particularly low, which exists across complex alpine terrain like the Southern Alps (Tait et al. 2012).

c. Data processing and WR clustering

Daily geopotential anomalies were computed relative to a 5-day rolling climatology for each day relative to the 1981–2010 climatic normal period, and statistically significant trends in geopotential height (Fig. 1) were removed at each grid point. Geopotential height trends at or exceeding the 95% significance are outlined in Fig. 1 (see caption for details). The removal of seasonality and temporal trends is an important data pretreatment ahead of defining atmospheric circulation archetypes. This step allows a more robust clustering baseline to be established that is broadly generalizable to other climate research domains (e.g., climate change and paleoclimate). While the clustering methodology that establishes the archetypal patterns is independent of seasonality and trends, projecting those patterns back on to the unfiltered data can reveal distinct WR frequency changes associated with seasonality and long-term temporal trends.

1) Primary weather regimes

To determine the optimal number of primary WRs, we adopted a Monte Carlo approach in conjunction with affinity propagation (AP) as an a priori application ahead of K-means clustering (Frey and Dueck 2007). We applied AP iteratively across random partitions (P) of sample size N = 1000 (days) on the first five principal components (PCs) of detrended near-surface geopotential height (z1000) anomaly fields. The first five PCs explain up to 90% of the original variance (the retention of five PCs is similar to previous WR classification work; Kidson 2000). Any natural partition should emerge as a peak in the distribution of P as a function of K. Once the optimal number of clusters were derived from AP, a more computationally efficient K-means clustering algorithm was applied to determine the WR patterns (Michelangeli et al. 1995).

Euclidean distance was used to determine the clusters from the K-means method. In addition, the K-means method is known to be highly sensitive to the initial seed selection of cluster centers and also to the initial conditions of the cluster centers. In this study, we used the K-means++ algorithm which overcomes issues related to occasionally poor clustering results from a standard K-means algorithm (Arthur and Vassilvitskii 2006). The K-means++ algorithm assigns the first centroid to the location of a randomly selected data point. Then, the algorithm chooses subsequent centroids from the remaining data points based on a probability that is proportional to the squared distance away from a given data point’s nearest centroid. The effect of this approach pushes centroids as far from one another as possible, covering as much of the occupied data space as possible from the point of initialization. This approach works better than a selection of random data points for assigning initial centroids, which is highly volatile, and also compensates for scenarios where randomly selected centroids are not positioned throughout the entire data space.

Detrended and deseasonalized daily z1000 anomalies were used for all of the empirical orthogonal analysis (EOF) decomposition (following prior work; see Lorrey and Fauchereau 2018) and WR analyses. The average geopotential height anomalies for all days grouped by cluster were used to define archetypal WR patterns. Seasonality and trends were then calculated based on projecting the primary WR patterns onto the original (“raw”) data by ascribing each day to the closest archetype derived using the aforementioned AP and K-means clustering approaches. Codes for AP and K-means clustering were adopted from previous work (Lorrey and Fauchereau 2018), built from Scikit learn machine learning tools in Python (Pedregosa et al. 2011; see https://scikit-learn.org).

2) Subsidiary weather regimes

Subsidiary clustering, or subclustering, is a process of determining key variants within a dominant WR cluster, which has not been thoroughly explored for ANZ. Hierarchical clustering algorithms, such as agglomerative clustering (Stashevsky et al. 2019), can be used to determine additional subsidiary branches and associations within a set of pre-established primary clusters. While those types of algorithms used for subsidiary clustering can provide significant insight about the hierarchy of atmospheric flows related to variable synoptic weather patterns, our approach has targeted definition of subsidiary WR structure through variable outcomes of precipitation and temperature that relate to a simplified WR classification.

Establishing subsidiary, within-regime hierarchies for WRs supports a further evaluation of the representation of how each main WR archetypal pattern depicts day-to-day atmospheric flows and impacts. It is also a useful approach for quantification of extreme daily weather occurrences associated within and between the main WR clusters. While consistent, regional-scale relationships between temperature and precipitation anomalies for New Zealand are often poorly defined, there are some recurrent spatially unique outcomes that are often arise directly from the direction/intensity of wind flow and orographic effects (Griffiths 2011). As such, the subsidiary clustering we have undertaken quantifies joint relationships between temperature and precipitation anomalies related to discrete subsidiary regime types and their frequency of occurrence.

The approach for hierarchal clustering proceeded as follows:

  • For each individual WR the detrended anomalies in temperature and precipitation relative to the 1981–2010 climatic normal period for each VCSN grid point were grouped.

  • Dimensionality reduction was performed on the joint temperature and precipitation anomalies for each grid point (to reduce the dimensionality associated with 11 491 national VCSN grid points) using principal component analysis (PCA) to explain 90% of the variance in the temperature and precipitation anomalies.

  • For each WR, affinity propagation was run to determine the number of subsidiary clusters.

While the number of grid points in the VCSN is greater than 11 000, approximately 6–10 components are typically required to explain 90% of the variance in the outcomes. We describe the outcomes of the subsidiary clustering using temperature anomalies (°C) relative to the 1981–2010 climatic normal period and percentage of normal precipitation (a ratio of average daily precipitation rate relative to the precipitation rate during the occurrence of the synoptic weather regime) for the same period.

d. Weather regime names and label conventions

Geopotential height anomalies in the domain can be used to describe each regime and its flow across New Zealand, as implemented in previous WR studies (Kidson 2000). Because WRs are relevant to operational forecasting and other climate science applications, we have constructed a consistent naming scheme and labeling convention for the primary WRs and subsidiary regimes (next section). These conventions draw from objective typologies consistent with previous work (Jenkinson and Collinson 1977; Kidson 2000; Jiang et al. 2013).

The WR naming convention and associated labels were constructed as follows:

  1. Location of the prominent geopotential height anomaly (e.g., low or high) within the NZ domain. A WR label prefix begins with either “H” or “L” to describe the occurrence of a “high” or “low” within the domain.

  2. Location of the high or low pressure anomaly feature with respect to the center of the analysis domain. An additional prefix was added directly after either the H or L label indicating the relative position of the anomaly with respect to New Zealand (e.g., N, NE, E, SE, S, SW, W, NW). If the z1000 anomaly is broad and covers the entire country, then noting the relative location of it within the NZ WR domain was dropped from the label and descriptor.

Additional qualitative descriptors were used to differentiate subsidiary types within each primary WR (see results for more detail). These additional labels were added onto the primary WR label and were used to note the following for each synoptic type:

  1. 3) Primary direction of wind flow. Mean wind flow grouped by cardinal direction (e.g., “n,” “ne,” “e,” “se,” “s,” “sw,” “w,” “nw”) were calculated using NCEP–NCAR1 u-wind and υ-wind data for the North Island and South Island for each day, and then used to indicate the average direction of wind flow across the North Island and South Island for each subsidiary type. A principal wind direction and spread for the main direction of flow across each main island was calculated for each day labeled as a subsidiary type as to describe conditions associated with each WR. The direction ascribed for the primary wind direction was ascribed based on closeness of the mean value to each cardinal point (using a 360° compass segmented into 45° portions). The standard deviation of mean flow was used to further ascribe winds in each regime as tight (within ±30° of the mean direction), constrained (within ±60° of the mean direction) or unconstrained (greater than ±60° from the mean direction). In the description of wind patterns over each island for the subsidiary types (see section 3e), a cardinal direction is used to describe a “tight” spread of flow (e.g., “S”), a cardinal direction along with the term “quarter” was used to describe relatively constrained direction of flow (e.g., S-qtr), and a multiquadrant descriptor (e.g., E–S) was used to describe unconstrained winds where flow directions exceeded a 120° range. The range of flows were qualitatively assessed for commonalities between both islands for each WR, and a direction suffix was added to the primary synoptic-type label. A calm (“c”) or disturbed (“d”) label was used where predominant wind directionality across both islands (see vector winds in each archetype in section 3e) included nominal or chaotic flow in specific situations.
  2. 4) Relative intensification of flow within a regime. An additional suffix was added to the primary direction of wind flow (see point C) that indicates whether wind flow was intensified (“i”) relative to another subsidiary pattern within the same WR that looked similar; this extra notation was applied if the location for highs and lows appeared akin to another subsidiary WR within the same WR type, but only when an obvious increased pressure gradient existed.
  3. 5) Rarity. An extreme (“x”) label was used for all subsidiary types with less than 100 days occurrence (∼1% or less within total daily WR population).

Absolute percentage changes for synoptic-type occurrence were calculated with respect to high monthly rainfall (categorized as 120% or more of normal) and low monthly rainfall (categorized as 80% or less of normal) experienced for the main cities and towns of New Zealand (Fig. 1). Rainfall climatological values were calculated using the 1972–2020 period using NIWA’s VCSN data that corresponded to each location. The percent difference from normal for synoptic-type occurrence (recalculated for the same interval) was obtained by determining the difference between the average monthly synoptic-type value corresponding to high and low rainfall months for 1972–2020 and the climatological value, then dividing by the climatological value and multiplying the result by 100. For purposes of simplicity, this result was then rounded to the nearest whole number.

3. Results and discussion

a. Primary New Zealand WRs determined from affinity propagation and K-means

A Monte Carlo–type approach provides a distributional measure of synoptic organization. When applying AP to a random selection of NCEP–NCAR1 z1000 daily data (N = 1000), a distribution peak emerges at nine clusters for ANZ (Fig. 2), and that result is also consistent when using the ERA5 reanalysis which overlapped for 1979–2020 (the time period of focus for establishing the main archetypal patterns). The K-means clustering that prescribed nine archetypes was applied on the first five EOFs derived from detrended z1000 anomalies over the ANZ domain to produce WR labels for each daily geopotential height pattern. Composite daily geopotential height anomalies and their associated surface wind anomalies from all days labeled as one of the nine WRs are shown in Fig. 3, along with the number of days and percentage of occurrence over the period 1979–2020 (the common overlap period between NCEP1 and ERA5). While our analysis suggests that there are fewer than 12 synoptic weather regimes than previously outlined (Kidson 2000) there are some strong similarities with antecedent work, especially for WR spatial patterns, intensity and orientation of regional pressure gradients, and wind flow directions.

Fig. 2.
Fig. 2.

Affinity propagation results for detrended 1000-hPa geopotential height (z1000) in the New Zealand domain objectively indicating nine primary weather regimes should be ascribed for K-means clustering of once-daily synoptic weather patterns. See prior work (Frey and Dueck 2007) for more details about affinity propagation and its use in synoptic weather regime classification (Lorrey and Fauchereau 2018).

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

Fig. 3.
Fig. 3.

Atmospheric pressure anomalies and associated flow patterns for the nine once-daily New Zealand synoptic weather types that were defined using affinity propagation and K-means clustering on detrended and deseasonalized NCEP–NCAR1 z1000 data for 1979–2020, which overlaps with and that were independently verified using ERA5. LSE—low to the southeast; HS—high to the south; LSW—low to the southwest; HSE—high to the southeast; LNE—low to the northeast; L—low; HW—high to the west; LNW—low to the northwest; and H—high. The total number of days and percentage of frequency are noted next to the synoptic-type label at the top of each plot.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

The nine WRs produced using AP and the K-means method for New Zealand were labeled according to the conventions (see section 2d). Each WR descriptor (full name, label, frequency of occurrence in the detrended and deseasonalized geopotential anomalies) is summarized in Fig. 3. The “low to the southeast” (LSE) type (Fig. 3, top-left panel) is used as an example of how we applied WR labeling conventions. The LSE primary label is based on the occurrence of a strong low pressure anomaly positioned to the southeast of the country. Most of the New Zealand domain is also occupied by grid cells with anomalously low geopotential height anomalies that occur during this regime. Some of the nine WRs have two key atmospheric pressure anomalies of opposite sign that were prominent (e.g., both a high and a low pressure anomaly occurred within the New Zealand domain box) or partly overlapped with the domain external border [such as “high to the south” (HS), “low to the northwest” (LNW) WRs; Fig. 3]. In those situations, the WR naming convention relied on the most well-defined/consistent pressure anomaly that was located within the domain (as with LNW) for the majority of days, or relied on the pressure anomaly feature that was consistently positioned in the same location in a majority of instances within the domain (as with HS). This tactic also lent to consistent naming for the subsidiary WR types described later (see section 3e).

b. Seasonality of primary weather regimes

To quantify WR seasonality and trends, we applied our classification scheme [as determined using detrended, de-seasonalized geopotential height data; see section 2c(1)] to the original z1000 data that retained long-term trends and the seasonal cycle. Seasonality among all WRs is significant. However, there are some clear differences in both the amplitude of each WR frequency of occurrence and their timing of predominance during the year (Fig. 4) that can be attributed to the underlying geopotential height seasonal cycle in the southern midlatitudes. LSE, LSW, “high to the southeast” (HSE), and “high to the west” (HW) types all have relatively small seasonal amplitudes (<5.5%), with a very weak seasonal cycle. On the other hand, HS, L, and H tend to have more distinct seasonal cycles, and interannual amplitudes between 6.1% and 8.3% of occurrence. The seasonality of HS and H is distinct and maximizes during the winter. In comparison, the L type tends to maximize during September. The HW type has a semiannual peak in austral autumn and spring. The seasonality of the LNE and LNW types is very strong, both with amplitudes > 12.5% and peaking during the summer. Collectively the “low to the northeast” (LNE) and the LNW types account for >50% of synoptic occurrences during the summer months.

Fig. 4.
Fig. 4.

Frequency of occurrence of nine synoptic weather regimes for New Zealand determined using affinity propagation followed by K-means clustering of 1000-hPa geopotential height (z1000) NCEP–NCAR1 data spanning 1948–2020. The WR patterns determined from deseasonalized and detrended data were then re-projected back onto the untreated z1000 data to obtain frequency changes for each month. The red line is the mean annual frequency of occurrence (also indicated in each panel along with the amplitude of change related to the seasonal cycle).

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

Seasonal patterns are evident for some of the WRs (Fig. 4). During the summer months there is an increase in frequency of the LNE and LNW types. During the winter months, pressure tends to be lower directly over New Zealand, but higher than normal both north and south of the country, leading to increased high pressure synoptic-type frequency during that time of year.

c. Long-term trends for primary weather regimes

Archetypal patterns described above were applied to undetrended z1000 data to examine the long-term changes in WR frequency of occurrence through time [see section 2c(1)]. A calculation of the rolling average for synoptic-type frequency of occurrence (Fig. 5; the year indicates the start of the decade when the average was calculated) shows significant changes for all nine WRs. However, the changes are not consistent across seasons (Fig. 5), and they are more notable for some types than others. The most consistent trends are observed for austral spring and autumn WRs (all WR types show significant changes), followed by summer (6 of 9 types) and winter (5 of 9 types). The largest frequency change is observed for the LNW (about −9%) for austral summer, with other types registering significant changes on the order of ±2%–8% changes since 1948. The HW type was the only WR that exhibited positive trends in all of the main seasons.

Fig. 5.
Fig. 5.

Evolving changes in percent frequency occurrence per decade of nine New Zealand synoptic types as determined by affinity propagation. The year indicates the start of the decadal average window (i.e., 2010 = average percentage of occurrence for a type for 2010–19). Time series are shown for annual, spring (SON), summer (DJF), autumn (MAM), and winter (JJA), with trends shown only for the seasons that had significant long-term changes. Relative stepwise percentage change per decade and absolute percentage change of total synoptic-type occurrence for all days since 1948 are shown in the top-right corner.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

d. Average weather and climate impacts of primary weather regimes

Small frequency changes for some of the primary synoptic types are associated with strong positive and negative monthly rainfall and temperature anomalies for different towns and cities (Tables 1 and 2). The sign of the synoptic-type frequency change associated with either above or below-normal rainfall commonly indicates a response to a change in the direction of wind flow. The interaction of the primary wind direction that arises for each type with mountainous terrain, in addition to the location of high or low pressure centers of action, results either in relatively homogenous conditions arising across the country (as with anomalies associated with the L type and H type) or spatially heterogeneous conditions occurring (as with the LNW type that is characterized by strong North Island–South Island contrasts). The fact such strong anomalies arise from only very small changes in frequency reinforces previous work that New Zealand’s weather and climate as a whole are hypersensitive to changes in synoptic-type occurrence.

Table 1

Absolute percentage change anomalies (rounded to nearest percent) for synoptic-type occurrence (detrended and deseasonalized) associated with monthly above-normal rainfall (at least 120% of normal or higher) for prominent NZ cities and towns for 1972–2020. See Fig. 1 for locations, and descriptions of rainfall outcomes in Figs. 614. Climatological percentages for monthly synoptic-type occurrence differs slightly from Fig. 4, with calculations undertaken for the 1972–2020 interval (dictated by the NIWA VCSN temporal coverage). As an example, a positive change of 6% frequency of occurrence for the LSW type would be more than 1.5 times its typical frequency (>150% of normal), and that situation is associated with above-normal monthly rainfall (>120% or higher) for Westport and Queenstown. Climatological occurrences are noted below each type label and also apply to Table 2.

Table 1
Table 2

Absolute percentage change anomalies (rounded to nearest percent) for synoptic-type occurrence (detrended and deseasonalized) associated with monthly below-normal rainfall (at least 80% of normal or lower) for prominent NZ cities and towns for 1972–2020. See Fig. 1 for locations, and descriptions of rainfall outcomes in Figs. 614. As an example, a negative change of 5% frequency occurrence for the L type would be nearly half of its typical frequency (about 45% of normal), and that situation is associated with below-normal monthly rainfall (<80% or lower) for Hamilton, New Plymouth, and Nelson.

Table 2

During months of above-normal rainfall for most towns and cities, there is a categorical increase of the L type and a concurrent decrease in frequency of the H and HW types (and vice versa for months with low rainfall; see Tables 1 and 2). In addition, below-normal rainfall is associated with an increase of the HW type for most locations on both main islands (except Queenstown and Invercargill, for which changes in frequency of the L type has a small effect on below and above-normal rainfall). Aside from a few notable exceptions, the frequency anomalies for the L type tend to be largest in magnitude for the major towns and cities of New Zealand in comparison to other WRs. Furthermore, the magnitude of the frequency anomalies (the decrease) associated with above-normal rainfall for the HW type appears to be the most negative among all the WR.

Strong WR impacts are noted for cities and towns that are exposed to northerly and easterly quarter flow associated with the LNW type (Tables 1 and 2), with a positive association between rainfall and frequency of occurrence for Kaitaia, Whangarei, Auckland, Hamilton, Tauranga, Rotorua, and Gisborne (see Fig. 1 for locations) and a negative association for towns and cities located in central, western, and southern localities of both main islands. The effects from small frequency changes linked to the LSE type, in terms of the spatial pattern, are almost opposite of that from LNW type during months of above- and below-normal rainfall for most towns and centers across New Zealand (see Tables 1 and 2). For Invercargill, months consisting of above-normal rainfall had a 3% absolute percentage occurrence increase (or a relative increase of 27% from the annual average frequency) for the LSE type. Similarly, during months of below normal there was a 2% frequency decrease in LSE type.

Overall, there is an increased frequency of the HS-type associated with above-normal rainfall for cities and towns exposed to easterly flows, with particularly large increases (>0.04% or >35% from the annual average) for Christchurch, Dunedin, and Gisborne. In comparison, western and interior towns and cities on the South Island, such as Queenstown, Westport, and Invercargill experience months with above (below) normal rainfall associated with a reduction (increase) in frequency of the HS type.

Towns and cities in the western half of ANZ (particularly Westport, Queenstown, and Invercargill) have months with above-normal rainfall associated with 5% increase in absolute frequency of the LSW type (nearly 50% relative decrease from climatology). The same synoptic pattern has an opposite impact for eastern coastal sites on the North Island (Gisborne, Tauranga, Whangarei), which have associations with below-normal monthly rainfall when that pattern increases. During months of above-normal rainfall, there is generally a corresponding reduction in frequency of the LNE type, which seems to have the largest influence on cities and towns in the western half of the North Island. A similar but opposite effect is associated with months of above-normal rainfall.

Fluctuations in the frequency anomalies of the HSE type during months of above- and below-normal rainfall appear to be very small in magnitude for all cities and towns in ANZ. However, this type has a much more consistent spatial association related to small frequency changes that impacts on below-normal monthly rainfall rather than the occurrence of above-normal monthly rainfall. The fact that only a very small percentage change for this type is seen to affect most towns and cities across ANZ suggest a high sensitivity to the frequency of occurrence of the HSE type.

e. Subsidiary WR clustering

Based on the nine primary clusters that were established using affinity propagation and K-means [see section 2c(2)] clustering algorithms, we established subsidiary clusters based on dual outcomes for temperature and precipitation solely within each regime. The number of subsidiary clusters for each WR varied between one and three, which appears to strongly be linked to the variability in the temperature and precipitation outcomes related to each primary WR. For example, if the precipitation and temperature outcomes for New Zealand are generally distinct there will be a dominant type or fewer subsidiary types. In contrast, if the outcomes are more variable, the domain WR is likely to have a small frequency of occurrence. Each of the subsidiary clusters for the main WRs (see Figs. 614) have an associated set of average geopotential height, vector wind, temperature and precipitation outcomes based on all days that fell into that regime. We describe the spatial patterns related to each regime below and expand on the main differences between the weather variables for each subsidiary cluster. Each figure that summarizes the main impacts of the regimes on daily weather mentions key cities or towns in New Zealand (see Table 3 for abbreviations). See section 2d for details about the naming and labeling convention for the subsidiary WR types that are defined below.

Fig. 6.
Fig. 6.

Three LSE subsidiary types: low to the southeast with southwesterly flow (LSE-sw); low to the southeast with southwesterly flow intensified (LSE-swi); and low to the southeast (LSE-x), rare occurrence. Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

Table 3

Abbreviations for regions, cities, and town used in descriptions contained in Figs. 614.

Table 3

1) Low to the southeast (LSE) subsidiary types

Subsidiary clustering on the LSE regime (Fig. 3) produced a total of three WR subtypes (Fig. 6), and thus an indication of high variability in the dual outcomes of precipitation and temperature. The dominant subsidiary regime strongly resembles the original synoptic weather pattern with a deep low pressure anomaly to the southeast of the country and occurs 72.7% (7.9%) of the time within the LSE regime (all time). The “sw” suffix attached to the main WR label for LSE-sw indicates the predominate southerly to westerly quarter flow directions that occur across the country, with temperature anomalies ranging mostly between −1° and −2.5°C (Fig. 6). The rainfall anomaly is drier than normal for most of the country, with isolated regions being wetter than normal, more complex descriptions are provided in Fig. 6. The two other subsidiary LSE types have subtly different flow patterns, with one representing an intensification of the LSE-sw type (LSE-swi) and the other representing an extreme type (LSE-x) with a deepened low pressure anomaly. The LSE-swi type has the center of low pressure positioned similarly to the LSE-sw type (to the east of the South Island), but it occurs less frequently 18.2% (2.0%) of the time (all time). There are associated changes in the wind flow direction and across both main islands for LSE-swi, with a more intense westerly and southerly flow across the North Island and South Island, respectively. The subtle alteration in the geopotential height pattern makes temperature anomalies cooler relative to LSE-sw over the South Island and near normal over the North Island. This outcomes for precipitation for LSE-swi are also more pronounced relative to LSE-sw, with much of the county receiving well above-normal rainfall (>300%), including all of the North Island, the Tasman District, and the eastern half of the South Island (see Fig. 6 for details specific to each subsidiary type). The LSE-x type has temperature and precipitation outcomes that are more similar to the LSE-swi type than the LSE-sw type that are related to stronger zonal flow across the North Island creating slightly warmer temperatures in northern and eastern regions. The precipitation outcomes are much the same as the LSE-swi type, but with some key differences for sites along the southeast coast of the North Island and western South Island. While the LSE-sw is clearly the most predominant subsidiary type in terms of frequency of occurrence, the outcomes of temperature and particularly rainfall appear far more significant for the LSE-swi and LSE-x types.

2) High to the south (HS) subsidiary types

Subsidiary clustering on the HS regime (Fig. 3) produced a total of three WR subtypes (Fig. 7). The three subsidiary regimes strongly resemble the undifferentiated HS synoptic weather pattern with a high pressure anomaly (ridge) to the south and southeast of the country and a low pressure anomaly in the northeast quadrant, which generates more frequent easterly and southerly flows across the country. The main HS subsidiary cluster (HS-se) occurs 74.5% of the time within the HS regime (8.2% of all days). HS-se has a predominately southeasterly flow over the majority of New Zealand that produces temperature anomalies between −1° and −2.5°C for the country. Rainfall outcomes for this regime are much drier than normal conditions for majority of the country, aside from isolated regions on the east-coast (including Gisborne, Hawkes Bay, and Wairarapa), which experienced wetter-than-normal rainfall (Fig. 7). The two other subsidiary WRs (HS-d and HS-sei) are less frequent, with occurrences of 14.6% (1.6%) and 10.9% (1.2%) within the HS regime (all days). The ridge of high pressure to the south of the country is the most spatially consistent part of all three subsidiary patterns, lending support for the naming and labeling convention that has been applied. While the most predominant subsidiary pattern for HS has a low pressure center to the east of the North Island of New Zealand, HS-d and HS-sei also have low pressure systems within close proximity to the east coast of New Zealand. As a result, there is a tendency for all three WRs to produce elevated precipitation anomalies in the central latitudes of New Zealand on the eastern side of the country (see Fig. 7 for more details).

Fig. 7.
Fig. 7.

Three HS subsidiary types: high to the south with southeasterly flow (HS-se); high to the south, with disturbed circulation (HS-d); and high to the south with southeasterly flow intensified (HS-sei). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

3) Low to the southwest (LSW) subsidiary types

Subsidiary clustering on the LSW regime (Fig. 3) produced a total of two WR subtypes (Fig. 8). The dominant subsidiary regime strongly resembles the original synoptic weather pattern with a deep low pressure anomaly located to the south and southwest of the country generating a more frequent occurrence of northwesterly and westerly flows across the North Island and South Island, respectively. The main LSW subsidiary cluster (LSW-w) occurs 74.3% of the time within the LSW regime (and 7.1% of all days). The other subsidiary regime, LSW-nw, occurs 25.7% of the time within the LSW regime and 2.4% of all days. The temperature anomalies associated with both of these subsidiary types are mostly positive (see Fig. 8), but they are stronger for the LSW-nw type (except for exposed areas of the western South Island, which have normal to cool temperatures). The degree of warm anomalies ranging from +1° to +3°C is also on the highest end of that range in eastern regions (particularly for Marlborough, Canterbury, and Dunedin), likely related to the Foehn effect that arises due to interaction of northwesterly wind flow with southwest–northeast-orientated axial ranges. For the LSW-nw subsidiary type, the intensified pressure gradient relative to the LSW-w pattern means wetter precipitation anomalies above 200% of normal occur for western and southern areas. More “spillover” of rainfall east of the main divide also occurs in the northern South Island and more intense dry conditions arise on the southeastern side of the North Island (i.e., in Gisborne and Hawkes Bay) for the LSW-nw type (see Fig. 8 for more details).

Fig. 8.
Fig. 8.

Two LSW subsidiary types: low to the southwest with westerly flow (LSW-w) and low to the southwest with northwesterly flow (LSW-nw). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

4) High to the southeast (HSE) subsidiary types

Subsidiary clustering on the HSE regime (Fig. 3) produced a total of three WR subtypes (Fig. 9). All the subsidiary HES regimes strongly resemble the original HSE synoptic weather pattern (Fig. 3), with a high pressure anomaly located over and to the east of the country and a low pressure anomaly in the southwest quadrant of the domain. These synoptic pressure situations generate more frequent northerly flows across the country. The main HSE subsidiary cluster (HSE-ne) occurs 59% of the time within the HSE regime (and 6.9% of all days). The HSE-n type occurs 30.1% of the time, while a stronger variation of it, HSE-ni, occurs 10.9% of the time with reference to all days in the HSE regime (and 3.5% and 1.3% of all days, respectively). There are almost universally warmer temperatures that occur across the country during all of these subsidiary regimes (except to the east of the main divide near East Cape, Gisborne and Hawke Bay which have near normal temperatures). The main variation for positive temperature anomalies between the HSE-ne and both the HSE-n and HSE-ni subsidiary types arises due to subtle changes in wind flow, with more direct northerlies occurring in the two minor subsidiary type contributions to the HSE regime. Temperature anomalies from +1.5° to +3°C above normal are most pronounced for southern North Island and all of the South Island, and they range from less to more intense from north to south and from west to east for all three subsidiary types (Fig. 9). The primary rainfall pattern associated with the HSE-ne type is drier-than-normal conditions across all of New Zealand, related to the presence of higher-than-normal pressure on average across all of the country. For the HSE-n and HSE-ni types, exposed western regions of the North Island (including Taranaki and the Kapiti Coast) are wetter than normal, along with much wetter than normal conditions for the western half of the South Island. Intensification of the flow in the HSE-ni regime appears to increase the spatial extent of positive rainfall anomalies much farther east of the main divide (see Fig. 9; Temp anomaly column) and to the south relative to what occurs for the HSE-n type.

Fig. 9.
Fig. 9.

Three HSE subsidiary synoptic types: high to the southeast with northeasterly flow (HSE-ne), high to the southeast with northerly flow (HSE-n), and high to the southeast with northerly flow intensified (HSE-ni). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

5) Low to the northeast (LNE) subsidiary types

Subsidiary clustering on the LNE regime (Fig. 3) produced a total of two WR subtypes (Fig. 10). The dominant subsidiary regime (LNE-e) mimics the original WR pattern, with a low pressure anomaly to the northeast of the North Island and a “ridge” over the southern half of the domain that has a high pressure core to the southeast of the country. This pressure pattern occurs 89.1% of the time within the LNE regime (9.3% of all days), and it generates more frequent southeasterly quarter flows across the North Island and light northerly to easterly flows across the South Island. However, the other subsidiary LNE regime diverges from the primary LNE archetypal pattern (Fig. 3) as well as the LNE-e subsidiary regime (Fig. 10) significantly, only sharing similar pressure anomaly traits for the northern third and eastern quarter of the analysis domain. The main difference with the second subsidiary cluster (LNE-d) is the presence of a low to the southwest of the South Island, which means the direction of flow over the North Island and South Island are essentially opposite for this subsidiary type. The LNE-d pattern occurs 10.9% of the time within the LNE regime (and 1.1% of all days). Temperatures associated within the LNE-e regime are near normal or slightly below normal for northern and eastern regions and near normal or slightly above normal for western and southern South Island, but categorically above normal (ranging from +1.5° to +3°C) over the eastern half of the South Island for the LNE-d subsidiary type. Despite being classed under the same main WR archetype, subtle z1000 anomaly differences between LNE-e and LNE-d means opposite rainfall anomaly outcomes occur for the southern and western South Island and Bay of Plenty (much lower than normal versus much higher than normal) and eastern North Island (higher versus lower than normal, respectively), with notable contrasts especially noted for Gisborne, Hawkes Bay, Westland, Fiordland, Southland, and inland Otago (see Fig. 10 for more details).

Fig. 10.
Fig. 10.

Two LNE subsidiary synoptic types: low to the northeast with easterly flow (LNE-e) and low to the northeast with disturbed circulation (LNE-d). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

6) Low (l; cyclonic) subsidiary types

Subsidiary clustering for the L regime (Fig. 3) produced a total of three WR subtypes (Fig. 11). All of the L subsidiary types closely resemble the original WR pattern that includes a strong low pressure anomaly positioned to the west of the country and a ridge of high pressure off to the south of the South Island. Overall, the regional pressure anomalies related to the L types generate more frequent northerly quarter flows across the North Island and easterly quarter flows across the South Island. The main L subsidiary cluster (L-d) occurs 59.9% of the time (6.7% of all days), with the L-ne (30.2% within type; 3.4% of all days) and L-nei patterns (9.9% within type; 1.1% of all days) appearing more similar in terms of their z1000 pattern (as well as their daily temperature and rainfall outcomes). Moderately warmer-than-normal temperatures for most of the North Island and northwest South Island occur with the L-d type. Accompanying rainfall anomalies for the L-d type are typically 120%–150% of normal across the northern North Island and coastal fringes of northern and eastern South Island, and up to 200% of normal in coastal Otago (Fig. 12). Spatial anomalies for temperature that are linked to the L-ne and L-nei subsidiary types appear similar in that they span all of the North Island and range from +1.5° to +3°C, and strong positive temperature anomalies also observed in the northern third of the South Island for the L-nei type. A subtle change in both the position and strength of the negative z1000 anomaly west of the South Island and high pressure to the east of the country for the L-ne and L-nei types produces a slightly different range of wind flow directions and very different outcomes for eastern North Island rainfall (wet versus dry, respectively; see Fig. 11 for specific locations with strong anomalies).

Fig. 11.
Fig. 11.

Three low (cyclonic) subsidiary types: low with disturbed circulation (L-d), low with northeasterly flow (L-ne), and low with northeasterly flow intensified (L-nei). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

Fig. 12.
Fig. 12.

HW type: high to the west with westerly flow (HW-w). No subsidiary pattern was identified relative to the primary WR. Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

7) High to the west (HW) subsidiary types

Subsidiary clustering on the HW regime (Fig. 3) did not produce any unique WR pattern subtypes, indicating a very close match for the dual temperature and precipitation outcomes for all days within this regime. The dominant flows across the region are zonal (westerly and southwesterly) and this regime accounts for 12.5% of all days, making it the most predominant daily synoptic type in the once-daily classification employed in this study. To remain consistent with the labeling convention for the previous regimes, we term this type HW-w (high to the west, with westerly flows; Fig. 12). Temperatures are near or slightly below normal for the North Island and western half of the South Island, and above normal for the eastern half of the South Island. For the HW-w type, the North Island and northern and eastern South Island are drier than normal (with daily anomalies ranging from 10% to 80% of normal rainfall that have the lowest values with increasing distance east of the main axial ranges).

8) Low to the northwest (LNW) subsidiary types

Subsidiary clustering on the LNW regime (Fig. 3) produced two WR subtypes (Fig. 13). Akin to the situation for the LNE regime, the dominant subsidiary pattern for LNW (LNW-e) strongly resembles the original archetype, with a low pressure anomaly positioned to the northwest of New Zealand and an oblong ridge of anomalously high pressure extending across most of the country. This synoptic pressure pattern creates primarily settled conditions and light easterlies across both islands. The main LNW-e subsidiary cluster occurs 90.8% of the time within the LNW regime (8.8% of all days) and has only moderate temperature anomalies (leaning cool for regions east of the main divide and warmer for western areas; see Fig. 13) and drier-than-normal conditions prevail across the South Island and southern North Island. There also a signature of increased rainfall in the Far North for LNW-e. A much smaller number of days are grouped into the LNW-x type, which had pressure anomalies over and to the west of the South Island that are opposite of LNW-e (Fig. 13). The LNW-x pattern occurs 9.2% of the time within the LNW regime (and 0.9% of all days). Similar rainfall for the Far North, but much drier central and eastern North Island and extremely wet South Island, occurs with this type relative to the main subsidiary pattern, which likely arises from divergent airflow across central and southern New Zealand.

Fig. 13.
Fig. 13.

Two LNW subsidiary types: low to the northwest with easterly flow (LNW-e) and low to the northwest, rare occurrence (LNW-x). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

9) High (H) subsidiary types

Subsidiary clustering on the H regime (Fig. 3) produced a total of two WR subtypes (Fig. 14). The dominant subsidiary regime closely resembles the main WR pattern with a core high pressure anomaly located over and just to the south of the South Island and higher-than-normal geopotential heights across most of the domain. This synoptic pattern generates more frequent light easterlies and southerlies over the North Island and relatively light northerly quadrant flows and/or calm conditions across the South Island (Fig. 14). The main H-c subsidiary cluster occurs 92.2% of the time within the H regime and 12.2% of all days overall, making it the second-most predominant daily synoptic type in this study. Temperature anomalies associated with the H-c synoptic type are mostly near normal, but slightly cooler than normal in eastern regions of both islands, and daily rainfall is usually 10%–50% of normal (most pronounced over the Southern Alps). A more complex synoptic pattern exists for the other subsidiary type in the H regime. The H-x pattern (7.8% of the regime, 1.0% of all days) has a strong high pressure cell located east of New Zealand and another extending onto the country from the southwest. The H-x regime is associated with temperature anomalies ranging from +1° to +3°C (especially prominent in the central portion of the country). The confluence of two high pressure anomalies in this pattern (see Fig. 14, H-x geopotential pattern) creates a zone of low-level convergence that is likely responsible for significantly high daily rainfall anomalies (200%–300% of normal) over the majority of the South Island (see Fig. 14 for more details).

Fig. 14.
Fig. 14.

Two high (anticyclonic) subsidiary types: high, calm conditions (H-c) and high, rare occurrence (H-x). Spatial anomalies for 1000-hPa geopotential height (z1000; blue shades are negative anomalies related to “lows” and red shades are positive anomalies related to “highs”), temperature anomalies (shown in STD deviation units and in 0.5°C increments), and rainfall (as percentage of normal) are a composite of all days that fell into each subcluster (geopotential height data from NCEP–NCAR1 and gridded surface climate data for New Zealand come from the NIWA VCSN). See Fig. 1 for the locations of regions and Table 3 for the abbreviations of towns and cities that are associated with weather regime anomalies.

Citation: Monthly Weather Review 150, 11; 10.1175/MWR-D-22-0059.1

4. Discussion

This paper has advanced weather regime classification for Aotearoa New Zealand relative to previous work (e.g., Kidson 2000) by applying objective and hierarchal methods to cluster once-daily synoptic pressure anomaly patterns. We summarize several main advancements below.

First, the method we used to identify an optimal number of WRs for ANZ (Fig. 2) resulted in fewer primary WRs than what has previously been established (Kidson 2000), which discretized nine primary clusters (Fig. 3). This finding may indicate that previous classifications have over prescribed synoptic variability, and that fewer regimes can be used represent ANZ recurrent circulation regimes. While fewer WRs could be due to the use of a once-daily rather than twice-daily analysis, our reduction to a daily scale (rather than twice daily) diminishes the impact that persistent weather systems have on overall frequency of occurrence. In this light, the once-daily patterns may be more well-suited for climatological analyses because they address a relative imbalance of how persistent weather systems associated with fine conditions are perceived (greater frequency) with respect to more rapidly moving transients (storms) that deliver rainfall on subdaily scales. While this issue has yet be fully explored, better relationships may yet emerge between the once-daily classification system and monthly/seasonal climate statistics.

Second, we used detrended and deseasonalized geopotential height anomalies to establish primary weather regime archetypes, with a secondary step of clustering circulation patterns within each main regime based on surface weather outcomes. This approach lends itself to characterization of weather extremes at different locations arising from seasonal synoptic patterns, and supports an evaluation of how weather may be changing to due long-term atmospheric circulation shifts. Seasonal cycles among the nine synoptic types appear pronounced (Fig. 4), with trade-offs for some regimes being more predominant at certain times of the year and less so during other times. These changes are to be expected as a function of how the general atmospheric circulation (including the subtropical ridge and southern westerlies) fluctuates throughout the year over the southern midlatitudes. The largest seasonal WR amplitudes exist for the LNE and the LNW types (Fig. 4). Both of these types are more frequent during austral summer and early autumn, and likely reflects stronger links with tropical low pressure systems during that time of year (including ex tropical cyclones; Lorrey et al. 2014a). Further work is needed to determine the origin of those synoptic systems, which may be connected to Australian East Coast Lows (Pepler et al. 2017). Concurrent reductions in the L-type frequency during summer, which is a WR linked with elevated rainfall for most cities and towns, reflects greater frequency of interactions with mid- and high-latitude cyclones and the New Zealand domain during winter and spring. Overall, we observe WR anomalies are associated with months of above- and below-normal rainfall (Tables 1 and 2), and their seasonal changes highlight links between average climatic conditions and atmospheric flow (see Figs. 614).

The HS, H, and L types are generally more frequent during the winter months. The fact both highs and lows are elevated during winter (and shoulder seasons bounding it) suggests more pronounced weather variability is linked to southern westerly wind (SWW) changes. SWW seasonal changes include more northward displacement of flow in addition to more prominent zonal wave 3 activity during austral winter (Raphael 2004, 2007). We note a bimodal distribution also occurs for the HW type, and that its frequency decreases at time of year when the semiannual oscillation (Meehl et al. 2017) strength increases. A similar but temporally offset and more muted pattern occurs for the HSE type (Fig. 4). A hypothetical link, based on the climatology of elevated and reduced occurrence for the HW and HSE types, is that both of these types may be strongly influenced by the position of the circumpolar trough and the corresponding strength of the latitudinal temperature gradient. Further investigations could focus on the set of high types (HS, H, HW, HSE) to determine how seasonal high pressure anomalies and their long term trends relate to larger scale circulation. This issue is of particular importance because of how these types can promote or limit dry conditions for many cities and towns (see Tables 1 and 2).

Previous work focused on decadal WR frequency changes (Renwick 2011) and indicated very small/insignificant changes across multiple time periods. Our study has probed this issue further by assessing decadal trends over 1948–2020 as a function of season, which provides insight as to how weather patterns are changing alongside a changing climate. Contrasting with previous findings, we have shown significant decadal trends exist for many of the nine daily synoptic types across different seasons (Fig. 5). Most notably, there are large reductions in the frequency of LNW type and large increases in the frequency of the LNE type, with more subtle frequency changes observed for other synoptic types. In addition, multiseason reduction of the frequency of the L type, responsible for delivering rainfall to most locations, is apparent in three out of four seasons (Fig. 5). We have not fully explored the implications of the long-term WR changes in this study. However, seasonal-level WR changes occurring over multiple decades suggest the aggregated impacts on below and above-normal rainfall months for the main cities and towns of ANZ (Tables 1 and 2) may be reflected in cumulative rainfall totals. For example, reduced L-type frequency in autumn, winter, and spring (associated with high rainfall for most cities and towns), may have played a significant role in long-term rainfall decreases and protracted multiseason droughts, and this connection should be examined in future work.

Long-term intensification of the subtropical high and southern shift of the SWW over the New Zealand/Tasman Sea region (Dean and Stott 2009) may be reflected in the rising frequency of occurrence for the HW type across all seasons. This change may have also played a role for increasing the frequency of months with below-normal rainfall over southwestern New Zealand locations. Taken together, these findings suggest the synoptic classification we created could be used as a framework to better understand changes in droughts and pluvials. In addition, decadal changes in the frequency of WRs may have contributed to changing spatial distributions of subseasonal rainfall and temperature anomalies, and this elevates them as important diagnostic tools for contextualizing long-term climate change for ANZ cities and towns. Additional exploration of these long-term trends with respect to modes of variability and palaeoclimatology will also be explored and presented in future work.

Third, the two-tiered classification scheme we used employed more closely examined daily atmospheric variability and weather outcomes for ANZ. By way of additional clustering of geopotential height anomalies based on independent conjointly occurring weather variables, we have better outlined subtle differences in the direction and intensity of atmospheric flow within each WR associated with spatially heterogeneous rainfall and temperature outcomes (see Figs. 614). This has helped to better characterize day-to-day WR impacts over a region with highly mountainous terrain. More extreme and spatially heterogeneous outcomes occur for synoptic patterns when they are intensified (e.g., stronger lows or highs) or when there are changes to wind direction and intensity (Figs. 614). As such, the subsidiary synoptic types in this study may have application to weather forecasting, and could help sharpen expected outcomes by directly relating atmospheric pressure patterns in numerical weather prediction models to analog-based weather impacts rather than solely relying on dynamically derived outputs.

5. Conclusions

This study has produced a new objective synoptic weather regime classification for New Zealand using a two-tiered application of K-means clustering assisted by affinity propagation. We produced nine primary synoptic types that were split further into 21 different daily weather regimes. While this analysis has fewer primary weather regime archetypes than previous work (e.g., Kidson 2000; Jiang et al. 2004), increased complexity is shown with subsidiary WRs that differentiate dual temperature and precipitation outcomes relative to using geopotential height alone for clustering. Our findings illustrate subtle variations in both the intensity and directionality of wind flow, in addition to how large scale circulation interacts with mountainous terrain, greatly affects daily weather (and climate) outcomes for ANZ.

Links between subsidiary weather regimes and infrequently occurring (but strong) precipitation and temperature anomalies can improve insights about the spatial variability of weather in different regions. Of significance, the distinct temporal patterns observed for seasonality and the long-term trends for synoptic types (Figs. 4 and 5), in addition to resulting high or low rainfall for main cities and towns (Tables 1 and 2), lends to future work that can establish connections with modes of variability and climate change projections. We expect application of this approach could significantly improve short-term predictability, and enhance systematic evaluation of climate variability and change within our region.

Acknowledgments.

This work and support for NR, AML, and NCF came from the National Institute of Water and Atmospheric Research Strategic Science Investment Fund project “Climate Present and Past” Contract CAOA2201.

Data availability statement.

All of the reanalysis data analyzed in this study are freely available as ongoing updates from The NCEP–NCAR 40-Year Reanalysis Project. They can be accessed from the National Centers for Environmental Prediction/National Weather Service/NOAA/U.S. Department of Commerce Research Data Archive at NOAA/PSL:/data/gridded/data.ncep.reanalysis.html. ERA5 reanalysis can be obtained from the Copernicus Climate Change Service data store at the following address: https://cds.climate.copernicus.eu/#!/search?text=ERA5&type=dataset. Virtual Climate Station Network data from NIWA may be obtained for research purposes on reasonable request.

REFERENCES

  • Ackerley, D., and Coauthors, 2011: Using synoptic type analysis to understand New Zealand climate during the Mid-Holocene. Climate Past, 7, 11891207, https://doi.org/10.5194/cp-7-1189-2011.

    • Search Google Scholar
    • Export Citation
  • Ackerley, D., A. Lorrey, J. A. Renwick, S. J. Phipps, S. Wagner, and A. Fowler, 2013: High-resolution modelling of mid-Holocene New Zealand climate 6000 yr BP. Holocene, 23, 721285, https://doi.org/10.1177/0959683613484612.

    • Search Google Scholar
    • Export Citation
  • Arthur, D., and S. Vassilvitskii, 2006: K-means++: The advantages of careful seeding. SODA ′07: Proc. 18th Annual ACM-SIAM Symp. on Discrete Algorithms, ACM Digital Library, New Orleans, LA, 10271035.

  • Bettolli, M. L., O. C. Penalba, and W. M. Varga, 2010: Synoptic weather types in the south of South America and their relationship to daily rainfall in the core crop-producing region in Argentina. Aust. Meteor. Oceanogr. J., 60, 3748, https://doi.org/10.22499/2.6001.004.

    • Search Google Scholar
    • Export Citation
  • Cassou, C., 2008: Intraseasonal interaction between the Madden–Julian oscillation and the North Atlantic oscillation. Nature, 455, 523527, https://doi.org/10.1038/nature07286.

    • Search Google Scholar
    • Export Citation
  • Charles, S. P., B. C. Bates, I. N. Smith, and J. P. Hughes, 2004: Statistical downscaling of daily precipitation from observed and modelled atmospheric fields. Hydrol. Processes, 18, 13731394, https://doi.org/10.1002/hyp.1418.

    • Search Google Scholar
    • Export Citation
  • Cheng, C. S., G. Li, Q. Li, and H. Auld, 2010: A synoptic weather typing approach to simulate daily rainfall and extremes in Ontario, Canada: Potential for climate change projections. J. Appl. Meteor. Climatol., 49, 845866, https://doi.org/10.1175/2010JAMC2016.1.

    • Search Google Scholar
    • Export Citation
  • Cheng, C. S., G. Li, Q. Li, and H. Auld, 2011: A synoptic weather-typing approach to project future daily rainfall and extremes at local scale in Ontario, Canada. J. Climate, 24, 36673685, https://doi.org/10.1175/2011JCLI3764.1.

    • Search Google Scholar
    • Export Citation
  • Cohen, L., S. Dean, and J. Renwick, 2013: Synoptic weather types for the Ross sea region, Antarctica. J. Climate, 26, 636649, https://doi.org/10.1175/JCLI-D-11-00690.1.

    • Search Google Scholar
    • Export Citation
  • Dean, S. M., and P. A. Stott, 2009: The effect of local circulation variability on the detection and attribution of New Zealand temperature trends. J. Climate, 22, 62176229, https://doi.org/10.1175/2009JCLI2715.1.

    • Search Google Scholar
    • Export Citation
  • DeLaFrance, A., and S. A. McAfee, 2019: Evaluation of synoptic-scale patterns during extreme temperature and precipitation events in Alaska. Int. J. Climatol., 39, 31343146, https://doi.org/10.1002/joc.6006.

    • Search Google Scholar
    • Export Citation
  • Fauchereau, N., B. Pohl, and A. Lorrey, 2016: Extratropical impacts of the Madden–Julian oscillation over New Zealand from a weather regime perspective. J. Climate, 29, 21612175, https://doi.org/10.1175/JCLI-D-15-0152.1.

    • Search Google Scholar
    • Export Citation
  • Frey, B. J., and D. Dueck, 2007: Clustering by passing messages between data points. Science, 315, 972976, https://doi.org/10.1126/science.1136800.

  • Gallant, A. J. E., S. J. Phipps, D. J. Karoly, A. B. Mullan, and A. M. Lorrey, 2013: Nonstationary Australasian teleconnections and implications for paleoclimate reconstructions. J. Climate, 26, 88278849, https://doi.org/10.1175/JCLI-D-12-00338.1.

    • Search Google Scholar
    • Export Citation
  • Gibson, P. B., S. E. Perkins-Kirkpatrick, and J. A. Renwick, 2016: Projected changes in synoptic weather patterns over New Zealand examined through self-organizing maps. Int. J. Climatol., 36, 39343948, https://doi.org/10.1002/joc.4604.

    • Search Google Scholar
    • Export Citation
  • Goodwin, I. D., and Coauthors, 2014: A reconstruction of extratropical Indo-Pacific sea-level pressure patterns during the medieval climate anomaly. Climate Dyn., 43, 11971219, https://doi.org/10.1007/s00382-013-1899-1.

    • Search Google Scholar
    • Export Citation
  • Griffiths, G., 2011: Drivers of extreme daily rainfalls in New Zealand. Wea. Climate, 31, 2449, https://doi.org/10.2307/26169716.

  • Hartigan, J. A., and M. A. Wong, 1979: Algorithm AS 136: A k-means clustering algorithm. J. Roy. Stat. Soc., 28C, 100108, https://doi.org/10.2307/2346830.

    • Search Google Scholar
    • Export Citation
  • Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 19992049, https://doi.org/10.1002/qj.3803.

    • Search Google Scholar
    • Export Citation
  • Hope, P. K., W. Drosdowsky, and N. Nicholls, 2006: Shifts in the synoptic systems influencing southwest Western Australia. Climate Dyn., 26, 751764, https://doi.org/10.1007/s00382-006-0115-y.

    • Search Google Scholar
    • Export Citation
  • Huang, W. T. K., A. Charlton-Perez, R. W. Lee, R. Neal, C. Sarran, and T. Sun, 2020: Weather regimes and patterns associated with temperature-related excess mortality in the UK: A pathway to sub-seasonal risk forecasting. Environ. Res. Lett., 15, 124052, https://doi.org/10.1088/1748-9326/abcbba.

    • Search Google Scholar
    • Export Citation
  • Jenkinson, A. F., and F. P. Collinson, 1977: An initial climatology of gales over the North Sea. Synoptic Climatology Branch Memo. 62, Meteorological Office, Bracknell, Berkshire, United Kingdom,18 pp.

  • Jiang, N., 2011: A new objective procedure for classifying New Zealand synoptic weather types during 1958-2008. Int. J. Climatol., 31, 863879, https://doi.org/10.1002/joc.2126.

    • Search Google Scholar
    • Export Citation
  • Jiang, N., J. E. Hay, and G. W. Fisher, 2004: Classification of New Zealand synoptic weather types and relation to the southern oscillation index. Wea. Climate, 23, 323, https://doi.org/10.2307/26169666.

    • Search Google Scholar
    • Export Citation
  • Jiang, N., G. Griffiths, and A. Lorrey, 2013: Influence of large-scale climate modes on daily synoptic weather types over New Zealand. Int. J. Climatol., 33, 499519, https://doi.org/10.1002/joc.3443.

    • Search Google Scholar
    • Export Citation
  • Jones, P. D., C. Harpham, and K. R. Briffa, 2013: Lamb weather types derived from reanalysis products. Int. J. Climatol., 33, 11291139, https://doi.org/10.1002/joc.3498.

    • Search Google Scholar
    • Export Citation
  • Jones, P. D., T. J. Osborn, C. Harpham, and K. R. Briffa, 2014: The development of Lamb weather types: From subjective analysis of weather charts to objective approaches using reanalyses. Weather, 69, 128132, https://doi.org/10.1002/wea.2255.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437471, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kidson, J. W., 1994: An automated procedure for the identification of synoptic types applied to the New Zealand region. Int. J. Climatol., 14, 711721, https://doi.org/10.1002/joc.3370140702.

    • Search Google Scholar
    • Export Citation
  • Kidson, J. W., 2000: An analysis of New Zealand synoptic types and their use in defining weather regimes. Int. J. Climatol., 20, 299316, https://doi.org/10.1002/(SICI)1097-0088(20000315)20:3<299::AID-JOC474>3.0.CO;2-B.

    • Search Google Scholar
    • Export Citation
  • Kidson, J. W., and J. A. Renwick, 2002: Patterns of convection in the tropical Pacific and their influence on New Zealand weather. Int. J. Climatol., 22, 151174, https://doi.org/10.1002/joc.737.

    • Search Google Scholar
    • Export Citation
  • Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247267, https://doi.org/10.1175/1520-0477(2001)082<0247:TNNYRM>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lachniet, M. S., C. M. Moy, C. Riesselman, H. Stephen, and A. M. Lorrey, 2021: Climatic and topographic control of the stable isotope values of rivers on the South Island of New Zealand. Paleoceanogr. Paleoclimatol., 36, e2021PA004220, https://doi.org/10.1029/2021PA004220.

    • Search Google Scholar
    • Export Citation
  • Lamb, H. H., 1950: Types and spells of weather around the year in the British Isles: Annual trends, seasonal structure of the year, singularities. Quart. J. Roy. Meteor. Soc., 76, 393429, https://doi.org/10.1002/qj.49707633005.

    • Search Google Scholar
    • Export Citation
  • Lamb, H. H., 1972: Meteorological office: Geophysical memoirs. Vol. 116, British Isles Weather Types and a Register of Daily Sequence of Circulation Patterns, 1861–1971. H.M. Stationery Office, 85 pp.

  • Lanfredi, I. S., and R. de Camargo, 2018: Classification of extreme cold incursions over South America. Wea. Forecasting, 33, 11831203, https://doi.org/10.1175/WAF-D-17-0159.1.

    • Search Google Scholar
    • Export Citation
  • Lefèvre, J., P. Marchesiello, N. C. Jourdain, C. Menkes, and A. Leroy, 2010: Weather regimes and orographic circulation around New Caledonia. Mar. Pollut. Bull., 61, 413431, https://doi.org/10.1016/j.marpolbul.2010.06.012.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., and P. R. Chappell, 2016: The “dirty weather” diaries of reverend Richard Davis: Insights about early colonial-era meteorology and climate variability for northern New Zealand, 1839-1851. Climate Past, 12, 553573, https://doi.org/10.5194/cp-12-553-2016.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., and N. C. Fauchereau, 2018: Southwest Pacific atmospheric weather regimes: Linkages to ENSO and extra-tropical teleconnections. Int. J. Climatol., 38, 18931909, https://doi.org/10.1002/joc.5304.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., A. M. Fowler, and J. Salinger, 2007: Regional climate regime classification as a qualitative tool for interpreting multi-proxy palaeoclimate data spatial patterns: A New Zealand case study. Palaeogeogr. Palaeoclimatol. Palaeoecol., 253, 407433, https://doi.org/10.1016/j.palaeo.2007.06.011.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., P. Williams, J. Salinger, T. Martin, J. Palmer, A. Fowler, J. Zhao, and H. Neil, 2008: Speleothem stable isotope records interpreted within a multi-proxy framework and implications for New Zealand palaeoclimate reconstruction. Quat. Int., 187, 5275, https://doi.org/10.1016/j.quaint.2007.09.039.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., and Coauthors, 2012: Palaeocirculation across New Zealand during the Last Glacial Maximum at ∼ 21 ka. Quat. Sci. Rev., 36, 189213, https://doi.org/10.1016/j.quascirev.2011.09.025.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., G. Griffiths, N. Fauchereau, H. J. Diamond, P. R. Chappell, and J. Renwick, 2014a: An ex-tropical cyclone climatology for Auckland, New Zealand. Int. J. Climatol., 34, 11571168, https://doi.org/10.1002/joc.3753.

    • Search Google Scholar
    • Export Citation
  • Lorrey, A. M., and Coauthors, 2014b: The Little Ice Age climate of New Zealand reconstructed from Southern Alps cirque glaciers: A synoptic type approach. Climate Dyn., 42, 30393060, https://doi.org/10.1007/s00382-013-1876-8.

    • Search Google Scholar
    • Export Citation
  • Mackintosh, A. N., B. M. Anderson, A. M. Lorrey, J. A. Renwick, P. Frei, and S. M. Dean, 2017: Regional cooling caused recent New Zealand glacier advances in a period of global warming. Nat. Commun., 8, 14202, https://doi.org/10.1038/ncomms14202.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., H. van Loon, and J. M. Arblaster, 2017: The role of the Southern Hemisphere semiannual oscillation in the development of a precursor to central and eastern Pacific Southern Oscillation warm events. Geophys. Res. Lett., 44, 69596965, https://doi.org/10.1002/2017GL073832.

    • Search Google Scholar
    • Export Citation
  • Michelangeli, P. A., R. Vautard, and B. Legras, 1995: Weather regimes: Recurrence and quasi stationarity. J. Atmos. Sci., 52, 12371256, https://doi.org/10.1175/1520-0469(1995)052<1237:WRRAQS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Moron, V., I. Gouirand, and M. Taylor, 2016: Weather types across the Caribbean basin and their relationship with rainfall and sea surface temperature. Climate Dyn., 47, 601621, https://doi.org/10.1007/s00382-015-2858-9.

    • Search Google Scholar
    • Export Citation
  • Pedregosa, F., and Coauthors, 2011: Scikit-learn: Machine learning in Python. J. Mach. Learn. Res., 12, 28252830, https://doi.org/10.48550/arXiv.1201.0490.

    • Search Google Scholar
    • Export Citation
  • Pepler, A. S., J. Fong, and L. V. Alexander, 2017: Australian east coast mid-latitude cyclones in the 20th century reanalysis ensemble. Int. J. Climatol., 37, 21872192, https://doi.org/10.1002/joc.4812.

    • Search Google Scholar
    • Export Citation
  • Pepler, A. S., A. J. Dowdy, and P. Hope, 2021: The differing role of weather systems in southern Australian rainfall between 1979–1996 and 1997–2015. Climate Dyn., 56, 22892302, https://doi.org/10.1007/s00382-020-05588-6.

    • Search Google Scholar
    • Export Citation
  • Pohl, B., A. Lorrey, A. Sturman, H. Quénol, J. Renwick, N. Fauchereau, and J. Pergaud, 2021: “Beyond weather regimes”: Descriptors monitoring atmospheric centers of action—A case study for Aotearoa New Zealand. J. Climate, 34, 8341–8360, https://doi.org/10.1175/JCLI-D-21-0102.1.

    • Search Google Scholar
    • Export Citation
  • Porhemmat, R., H. Purdie, P. Zawar-Reza, C. Zammit, and T. Kerr, 2021: The influence of atmospheric circulation patterns during large snowfall events in New Zealand’s Southern Alps. Int. J. Climatol., 41, 23972417, https://doi.org/10.1002/joc.6966.

    • Search Google Scholar
    • Export Citation
  • Prebble, J. G., and Coauthors, 2017: Evidence for a Holocene Climatic Optimum in the southwest Pacific: A multiproxy study. Paleoceanogr. Paleoclimatol., 32, 763779, https://doi.org/10.1002/2016PA003065.

    • Search Google Scholar
    • Export Citation
  • Ramos, A. M., D. Barriopedro, and E. Dutra, 2015: Circulation weather types as a tool in atmospheric, climate, and environmental research. Front. Environ. Sci., 3, 44, https://doi.org/10.3389/fenvs.2015.00044.

    • Search Google Scholar
    • Export Citation
  • Raphael, M. N., 2004: A zonal wave 3 index for the Southern Hemisphere. Geophys. Res. Lett., 31, L23212, https://doi.org/10.1029/2004GL020365.

    • Search Google Scholar
    • Export Citation
  • Raphael, M. N., 2007: The influence of atmospheric zonal wave three on Antarctic sea ice variability. J. Geophys. Res., 112, D12112, https://doi.org/10.1029/2006JD007852.

    • Search Google Scholar
    • Export Citation
  • Raynaud, D., B. Hingray, G. Evin, A.-C. Favre, and J. Chardon, 2020: Assessment of meteorological extremes using a synoptic weather generator and a downscaling model based on analogues. Hydrol. Earth Syst. Sci., 24, 43394352, https://doi.org/10.5194/hess-24-4339-2020.

    • Search Google Scholar
    • Export Citation
  • Renwick, J. A., 2011: Kidson’s synoptic weather types and surface climate variability over New Zealand. Wea. Climate, 31, 323, https://doi.org/10.2307/26169715.

  • Robertson, A. W., N. Vigaud, J. Yuan, and M. K. Tippett, 2020: Toward identifying subseasonal forecasts of opportunity using North American weather regimes. Mon. Wea. Rev., 148, 18611875, https://doi.org/10.1175/MWR-D-19-0285.1.

    • Search Google Scholar
    • Export Citation
  • Schroeter, S., H. Richter, C. Arthur, D. Wilke, M. Wehner, and E. Ebert, 2021: Forecasting the impacts of severe weather. Aust. J. Emerg. Manage., 36, 7683, https://doi.org/10.47389/36.1.76.

    • Search Google Scholar
    • Export Citation
  • Stashevsky, P. S., I. N. Yakovina, T. M. Alarcon Falconi, and E. N. Naumova, 2019: Agglomerative clustering of enteric infections and weather parameters to identify seasonal outbreaks in cold climates. Int. J. Environ. Res. Public Health, 16, 2083, https://doi.org/10.3390/ijerph16122083.

    • Search Google Scholar
    • Export Citation
  • Suriano, Z. J., 2020: Synoptic and meteorological conditions during extreme snow cover ablation events in the Great Lakes basin. Hydrol. Processes, 34, 19491965, https://doi.org/10.1002/hyp.13705.

    • Search Google Scholar
    • Export Citation
  • Tait, A., R. Henderson, R. Turner, and X. Zheng, 2006: Thin-plate smoothing spline interpolation of daily rainfall for New Zealand using a climatological rainfall surface. Int. J. Climatol., 26, 20972115, https://doi.org/10.1002/joc.1350.

    • Search Google Scholar
    • Export Citation
  • Tait, A., J. Sturman, and M. Clark, 2012: An assessment of the accuracy of interpolated daily rainfall for New Zealand. J. Hydrol., 51, 2544, https://www.jstor.org/stable/43944886.

    • Search Google Scholar
    • Export Citation
  • Theobald, A., H. McGowan, J. Speirs, and N. Callow, 2015: A synoptic classification of inflow-generating precipitation in the snowy mountains, Australia. J. Appl. Meteor. Climatol., 54, 17131732, https://doi.org/10.1175/JAMC-D-14-0278.1.

    • Search Google Scholar
    • Export Citation
  • van der Wiel, K., H. C. Bloomfield, R. W. Lee, L. P. Stoop, R. Blackport, J. A. Screen, and F. M. Selten, 2019: The influence of weather regimes on European renewable energy production and demand. Environ. Res. Lett., 14, 094010, https://doi.org/10.1088/1748-9326/ab38d3.

    • Search Google Scholar
    • Export Citation
  • Vautard, R., 1990: Multiple weather regimes over the North Atlantic: Analysis of precursors and successors. Mon. Wea. Rev., 118, 20562081, https://doi.org/10.1175/1520-0493(1990)118<2056:MWROTN>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Yiou, P., and M. Nogaj, 2004: Extreme climatic events and weather regimes over the North Atlantic: When and where? Geophys. Res. Lett., 31, L07202, https://doi.org/10.1029/2003GL019119.

Save