A comparative evaluation of principal components-based and information theory methods of precipitation regionalization

  • D. R. Legates
  • C. J. Willmott


Precipitation regionalization techniques alternatively advocated by Johnston [15] and Willmott [30] are reviewed and compared. An implementation by Johnston [15] of the information theory or entropy method was determined to have limited applicability —principally owing to its inability (1) to describe periodicies that are inherent in the seasonal cycle and (2) to account for differences between time-series of precipitation with respect to their means. The entropy method, however, has a practical appeal because the spatial distribution of stations has a relatively minor, deleterious impact on the regionalization [15]. On the other hand, the principal components-based procedure described by Willmott [30] includes more information about precipitation in the evaluation of regional patterns, e.g., station means, and it is well-able to characterize periodicities in the variance. With the appropriate experimental design, the combination of principal components analysis and grouping is regarded as superior to the entropy method in identifying and characterizing precipitation regimes and regions.


Principal Component Analysis Spatial Autocorrelation Entropy Method Hierarchical Cluster Method Monthly Total 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Eine vergleichsweise Auswertung des Regionalniederschlages auf Grund der Hauptkomponenten und der Methoden der Informationstheorie


Die von Johnston [15] und Willmott [30] vorgeschlagenen Methoden zur Berechnung des Gebietsniederschlages werden besprochen und miteinander verglichen. Eine Anwendung der Informationstheorie oder Entropiemethode, von Johnston [15] vorgeschlagen, besitzt nur beschränkte Möglichkeiten, hauptsächlich wegen der Unfähigkeit, 1. die im Jahresgang enthaltenen Periodizitäten zu beschreiben und 2. die Unterschiede zwischen den zeitlichen Reihen des Niederschlages und deren Mittelwerten zu erklären. Die Entropiemethode hat jedoch praktische Vorzüge, da die räumliche Verteilung der Stationen nur einen relativ geringen, ungünstigen Einfluss auf die Gebietswerte hat [15]. Die auf den Hauptkomponenten beruhenden Berechnungen von Willmott [30] beinhalten andererseits mehr Information über den Niederschlag in der Beurteilung der gebietsmässigen Verteilung, z.B. der Stationsmittelwerte, und ist gut in der Lage, Periodizitäten in der Varianz zu charakterisieren. Bei entsprechender Anwendung der Methoden ist eine Kombination von Hauptkomponentenanalyse und Gruppierungen der Entropiemethode bei der Charakterisierung von Niederschlagsregimen und -regionen überlegen.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Anderson, T. W.: Asymptotic Theory for Principal Components Analysis. Ann. Math. Stat.34, 123–148 (1963).Google Scholar
  2. 2.
    Ayoade, J. O.: On the Use of Multivariate Techniques in Climatic Classification and Regionalization. Arch. Met. Geoph. Biokl., Ser. B24, 257–267 (1977).CrossRefGoogle Scholar
  3. 3.
    Cattell, R. B.: The Scree Test for the Number of Factors. Multivariate Behavioral Res.1, 245–276 (1966).CrossRefGoogle Scholar
  4. 4.
    Craddock, J. M.: Problems and Prospects for Eigenvector Analysis in Meteorology. The Statistician22, 133–145 (1973).CrossRefGoogle Scholar
  5. 5.
    Daultrey, S.: Principal Components Analysis. Norwich: Geo Abstracts Ltd. 1976.Google Scholar
  6. 6.
    Davies, W. K. D.: Varimax and the Destruction of Generality. A Methodological Note. Area3, 112–118 (1971).Google Scholar
  7. 7.
    Davies, W. K. D.: Varimax and Generality: A Reply. Area3, 254–259 (1971).Google Scholar
  8. 8.
    Davies, W. K. D.: Varimax and Generality: A Second Reply. Area4, 207–208 (1972).Google Scholar
  9. 9.
    Dyer, T. G. J.: On the Components of Time Series; The Removal of Spatial Dependence. Quart. J. R. Met. Soc.102, 157–165 (1976).CrossRefGoogle Scholar
  10. 10.
    Dyer, T. G. J.: The Assignment of Rainfall Stations Into Homogeneous Groups: An Application of Principal Components Analysis. Quart. J. R. Met. Soc.101, 1005–1013 (1975).CrossRefGoogle Scholar
  11. 11.
    Gregory, S.: On the Delimitation of Regional Patterns of Recent Climatic Fluctuations. Weather30, 276–287 (1975).Google Scholar
  12. 12.
    IMSL (International Mathematical and Statistical Libraries, Inc.): IMSL Library 1, Edition 6, 1977 (FORTRAN IV) IBM S/270-360, Xerox Sigma S/6-7-9-11-560, Houston: IMSL, Inc.Google Scholar
  13. 13.
    Johnston, R. J.: Multivariate Statistical Analysis in Geography. New York: Longman Group Ltd. 1978.Google Scholar
  14. 14.
    Johnston, R. J.: Principal Components Analysis and Factor Analysis in Geographical Research: Some Problems and Issues. S. Afric. Geogr. J.59, 30–44 (1977).Google Scholar
  15. 15.
    Johnston, R. J.: Regarding the Delimitation of Regions According to Climatic Fluctuations. Arch. Met. Geoph. Biocl., Ser. B29, 215–228 (1981).CrossRefGoogle Scholar
  16. 16.
    Kaiser, H. F.: The Varimax Criterion for Analytic Rotation in Factor Analysis. Psychometrika23, 187–200 (1958).CrossRefGoogle Scholar
  17. 17.
    Karl, T. R., Koscielny, A. J., Diaz, H. F.: Potential Errors in the Application of Principal Component (Eigenvector) Analysis to Geophysical Data. J. Appl. Met.21, 1183–1186 (1982).CrossRefGoogle Scholar
  18. 18.
    Koschmieder, H.: Methods and Results of Definite Rain Measurements. Mon. Weath. Rev.62, 5–7 (1934).CrossRefGoogle Scholar
  19. 19.
    MacDonald, D. K. C.: Information Theory and its Application to Taxonomy. J. Appl. Phys.23, 529–531 (1952).CrossRefGoogle Scholar
  20. 20.
    Mather, P. M.: Varimax and Generality: A Comment. Area3, 252–254 (1971).Google Scholar
  21. 21.
    Mather, P. M.: Varimax and Generality: A Second Comment. Area4, 27–30 (1972).Google Scholar
  22. 22.
    Morrison, D. F.: Multivariate Statistical Methods, 2nd ed. New York: McGraw-Hill 1976.Google Scholar
  23. 23.
    Richman, M. B.: Obliquely Rotated Principal Components: An Improved Meteorological Map Typing Technique? J. Appl. Met.20, 1145–1159 (1981).CrossRefGoogle Scholar
  24. 24.
    Semple, R. K., Youngmann, C. E., Zeller, R. E.: Economic Regionalization and Information Theory: An Ohio Example. Discussion Paper 28, Department of Geography, Ohio State University, Columbus, Ohio, 1972.Google Scholar
  25. 25.
    Shih, S. F.: Rainfall Variation Analysis and Optimization of Gaging Systems. Water Res. Res.18, 1269–1277 (1982).CrossRefGoogle Scholar
  26. 26.
    Walsh, J. A.: Information Theory and Classification in Geography. Discussion Paper 4, Department of Geography, McMaster University, Hamilton, Ontario, Canada, 1976.Google Scholar
  27. 27.
    Walsh, J. E., Richman, M. B., Allen, D. A.: Spatial Coherence of Monthly Precipitation in the United States. Mon. Weath. Rev.110, 272–286 (1982).CrossRefGoogle Scholar
  28. 28.
    Ward, J. H., Jr.: Hierarchical Grouping to Optimize an Objective Function. J. Amer. Star. Ass.58, 236–244 (1963).CrossRefGoogle Scholar
  29. 29.
    Willmott, C. J.: A Component Analytic Approach to Precipitation Regionalization in California. Arch. Met. Geoph. Biokl., Ser. B24, 269–281 (1977).CrossRefGoogle Scholar
  30. 30.
    Willmott, C. J.: P-mode Principal Components Analysis, Grouping and Precipitation Regions in California. Arch. Met. Geoph. Biokl., Ser. B26, 277–295 (1978).CrossRefGoogle Scholar
  31. 31.
    Willmott, C. J., Vernon, M. T.: Solar Climates of the Conterminous United States: A Preliminary Investigation. Solar Energy24, 295–302 (1980).CrossRefGoogle Scholar
  32. 32.
    Wishart, D.: CLUSTAN 1C: User Manual. London: Univ. College Computer Center 1975.Google Scholar

Copyright information

© Springer-Verlag 1983

Authors and Affiliations

  • D. R. Legates
    • 1
  • C. J. Willmott
    • 1
  1. 1.Center for Climatic Research, Department of GeographyUniversity of DelawareNewarkUSA

Personalised recommendations