85th AMS Annual Meeting

Thursday, 13 January 2005: 3:45 PM
Improved "Climate Divisions" for monitoring, assessing, and predicting climate in the U.S.
Klaus Wolter, NOAA/ERL/CDC, Boulder, CO; and R. Bigley, J. K. Eischeid, and D. Allured
So-called "Climate Divisions" have been used for the purposes of near-realtime climate monitoring and assessment of recent climate anomalies (most prominently in the U.S. drought monitor), as well as climate predictions in their broadest sense (CPC seasonal forecasts, ENSO composites). However, the original climate divisions were not statistically optimized to serve these goals.

Using the full U.S. COOP station network, as well as automated high-elevation SNOTEL data for the last quarter century, we attempt to create new climate divisions that better serve the purposes listed above. Here we present an interim version of these new climate divisions, based on three different multivariate statistical techniques of analyzing seasonal precipitation data. These three techniques are: (1) Average Linkage -, (2) Ward Linkage Cluster Analyses, as well as (3) Rotated Principal Component Analyses. New climate divisions are created via optimization of the regionalizations derived by these three techniques.

The new divisions are assessed for robustness by repeating the analysis for earlier periods, in particular 1948 through 1978, and comparing the solutions against each other. Improvement over traditional climate divisons is quantified by comparing the explained variances at the station level as derived from the traditional and new divisional averages. Since old climate divisional time series did NOT incorporate SNOTEL data, new divisional averages improve their representativeness the most over the higher terrain of the Western U.S. Anticipated work includes analogous analyses of seasonal TEMPERATURE data in order to create new climate divisions that are based on both precipitation and temperature, as well as higher temporal resolution (pentads) analyses in order to assess their dependence on time filtering.

An immediate application of climate divisions is the assessment of recent drought episodes over the Western U.S. Prior to 1930, existing divisional averages are not based on station data within each division, but on state-wide averages that were regressed back to the divisional level based on modern data. Using representative and long-lived stations within each new climate division, it is possible to compute recent precipitation (and temperature) anomalies on the divisional level and compare them against historic drought episodes during at least the last century. This approach will be illustrated for the 1999-2004 drought in different parts of Colorado.

Supplementary URL: