RECONSTRUCTION OF THE MINNEAPOLIS-ST. PAUL AREA PIONEER-ERA FIXED-TIME-SCHEME TEMPERATURE RECORD
Conversions of the 1820-1872 fixed-time temperature readings to daily midnight-to-midnight maxima and minima estimates were derived from analysis of the multivariate statistical relationships between actual 1961-1980 Minneapolis-St. Paul Airport daily maxima, minima, and midnight temperatures vs. temperatures, sky cover, north/south, and east/west wind component information corresponding to the old fixed-time-scheme hours [Fisk, M. S. Thesis in Meteorology at University of Wisconsin-Madison].
Testing of the statistical conversion models on modern data determined that average daily estimating errors were, depending on the fixed-time-format and calendar month involved, between 0.9 F and 1.3 F for the maxima, 0.9 F and 1.9 F for the minima, and 0.9 F to 1.5 F for the daily means. This is comparable or better to observational discrepancies on a given day that would be encountered between different stations in the same city. Moreover, since there was a day-to-day cancelling out effect of the errors (some positive, some negative), average monthly mean estimation error was only 0.2 F. In addition, the fortunate availability of another St. Paul temperature record from the 1860’s with self-registering daily minimum recordings permitted a comparison with the estimated same-day minima derived from the fixed-time-scheme St. Paul Smithsonian observations. Results indicated that the conversion models could produce reasonably accurate results on nineteenth-century fixed-time scheme data.
This reconstruction is believed to be the only one of its kind for a pre-weather bureau-era, fixed-time scheme temperature record in the United States, and it enables the pioneer-era data to be presented in the same graphical format as the Weather Bureau’s. A few data gaps are present in the pre-weather bureau era series, comprising, however, less than 0.5% of the entire history. The most significant of these are October 1820 (most of the month), October 1827 (the entire month), January 1843 (the first ten days) and May to December 1858 (eight months – reflecting the hiatus between the close of the Fort Snelling observational period and the start of the Smithsonian). Using a combination of objective and subjective means, daily temperature estimates for the first three gaps were made using comparative analyses with either Fort Crawford (present-day Prairie Du Chien, WI) or Fort Howard (present-day Green Bay, WI), data of which was available for the periods in question. The 1858 eight-month gap was filled, primarily subjectively, from three-times daily temperature observations that appeared intermittently in a local St. Paul newspaper, together with Smithsonian volunteer observations from Princeton, MN, and military observations at Fort Ripley (near present-day Little Falls, MN).