Climate Zonation in Puerto Rico Based on Principal Components Analysis and an Artificial Neural Network

Björn A. Malmgren Department of Earth Sciences—Marine Geology, Earth Sciences Center, University of Göteborg, Goteborg, Sweden

Search for other papers by Björn A. Malmgren in
Current site
Google Scholar
PubMed
Close
and
Amos Winter Department of Marine Sciences, University of Puerto Rico, Mayaguez, Puerto Rico

Search for other papers by Amos Winter in
Current site
Google Scholar
PubMed
Close
Restricted access

Abstract

The authors analyzed climate data, seasonal averages of precipitation, and maximum, mean, and minimum temperatures over the years 1960–90, from 18 stations spread around the island of Puerto Rico in the Caribbean, to determine whether these distinguish the existence of climate zones in Puerto Rico. An R-mode principal components analysis (PCA), with varimax rotation to the seasonal data in order to reduce their dimensionality, was applied. The first five principal components, found by cross validation to be statistically significant, account for 99% of the variability in the 16 variables included in the analysis. These five components are related to annual variation in mean and minimum temperature (first PC), annual maximum temperature (second PC), and spring, summer, and fall precipitation (third through fifth PCs). A self-organizing map, an artificial neural network algorithm, was then employed to classify the first five PC scores in an optimal fashion. The scores were classified by the neural network into four climatic zones, each with a distinct geographic coverage in Puerto Rico. One zone, marked by the highest mean and minimum annual temperatures, is located along the northern, eastern, and southern coasts of Puerto Rico. The stations referred to the second zone are also from relatively low altitudes in the northern and eastern parts of the island, but they are not located along the immediate coastline. Intermediately high mean and minimum temperatures mark this zone. The third zone consists of stations from high altitudes in the central mountain range and is characterized by the lowest annual mean and minimum temperatures. To the south of the third zone, a fourth zone is identified, which is marked by the highest annual maximum temperatures.

Corresponding author address: Professor Björn Malmgren, Dept. of Earth Sciences—Marine Geology, Earth Sciences Center, University of Göteborg, Box 460, SE405 30 Göteborg, Sweden.

Email: bjorn.malmgren@marine-geology.gu.se

Abstract

The authors analyzed climate data, seasonal averages of precipitation, and maximum, mean, and minimum temperatures over the years 1960–90, from 18 stations spread around the island of Puerto Rico in the Caribbean, to determine whether these distinguish the existence of climate zones in Puerto Rico. An R-mode principal components analysis (PCA), with varimax rotation to the seasonal data in order to reduce their dimensionality, was applied. The first five principal components, found by cross validation to be statistically significant, account for 99% of the variability in the 16 variables included in the analysis. These five components are related to annual variation in mean and minimum temperature (first PC), annual maximum temperature (second PC), and spring, summer, and fall precipitation (third through fifth PCs). A self-organizing map, an artificial neural network algorithm, was then employed to classify the first five PC scores in an optimal fashion. The scores were classified by the neural network into four climatic zones, each with a distinct geographic coverage in Puerto Rico. One zone, marked by the highest mean and minimum annual temperatures, is located along the northern, eastern, and southern coasts of Puerto Rico. The stations referred to the second zone are also from relatively low altitudes in the northern and eastern parts of the island, but they are not located along the immediate coastline. Intermediately high mean and minimum temperatures mark this zone. The third zone consists of stations from high altitudes in the central mountain range and is characterized by the lowest annual mean and minimum temperatures. To the south of the third zone, a fourth zone is identified, which is marked by the highest annual maximum temperatures.

Corresponding author address: Professor Björn Malmgren, Dept. of Earth Sciences—Marine Geology, Earth Sciences Center, University of Göteborg, Box 460, SE405 30 Göteborg, Sweden.

Email: bjorn.malmgren@marine-geology.gu.se

Save
  • Baldwin, J. L., D. N. Otte, and C. L. Wheatley, 1989: Computer emulation of human mental process: Application of neural network simulations to problems in well log interpretation. SPEJ, Soc. Pet. Eng. J.,19619, 481–493.

  • Calvesbert, R. J., 1970: The Climate of Puerto Rico and the U.S. Virgin Islands. Vol. 1, Climates of the States. Water Information Center, 451–478.

  • DeGaetano, A. T., 1996: Delineation of mesoscale climatic zones in the northeastern United States using a novel approach to cluster analysis. J. Climate,9, 1765–1782.

  • Demuth, H., and M. Beale, 1994: Neural Network Toolbox for Use with MATLAB. MathWorks, 158 pp.

  • Diaconis, P., and B. Efron, 1983: Computer-intensive methods in statistics. Sci. Amer.,248, 96–108.

  • Etter, P. C., P. J. Lamb, and D. H. Potis, 1987: Heat and freshwater budgets of the Caribbean with revised estimates of the Central American seas. J. Phys. Oceanogr.,17, 1232–1248.

  • Fovell, R. G., and M.-Y. C. Fovell, 1993: Climate zones of the conterminous United States defined using cluster analysis. J. Climate,6, 2103–2134.

  • Gadgil, S., and N. V. Joshi, 1983: Climatic clusters of the Indian region. J. Climatol.,3, 47–53.

  • Gower, J. C., 1966: Some distance properties of latent root and vector methods used in multivariate analysis. Biometrika,53, 325–338.

  • Grossberg, S., 1974: Classical and instrumental learning by neural networks. Prog. Theor. Biol.,3, 51–141.

  • Jackson, J. E., 1991: A User’s Guide to Principal Components. Wiley-Interscience, 569 pp.

  • Joliffe, I. T., 1986: Principal Component Analysis. Springer-Verlag, 271 pp.

  • Kohonen, T., 1988: Self-Organization and Associative Memory. Springer-Verlag, 312 pp.

  • Krzanowski, W. J., 1987: Cross-validation in principal component analysis. Biometrics,43, 575–584.

  • Malmgren, B. A., and U. Nordlund, 1996: Application of artificial neural networks to chemostratigraphy. Paleoceanography,11, 505–512.

  • Penn, B. S., A. J. Gordon, and R. F. Wendlandt, 1993: Using neural networks to locate edges and linear features in satellite images. Comput. Geosci.,19, 1545–1565.

  • Raiche, A., 1991: A pattern recognition approach to geophysical inversion using neural nets. Geophys. J. Int.,105, 629–648.

  • Rogers, S. J., J. H. Fang, C. L. Karr, and D. A. Stanley, 1992: Determination of lithology from well logs using a neural network. Amer. Assoc. Pet. Geol. Bull.,76, 731–739.

  • Wasserman, P. D., 1989: Neural Computing—Theory and Practice. Van Nostrand Reynold, 230 pp.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 1347 442 44
PDF Downloads 804 250 33