Electronic Supplementary Material for Historical Perspective on the Dust Bowl Drought in the Central United States Climatic Change Dorian J. Burnette1 and David W. Stahle2 1 Department of Earth Sciences, 109 Johnson Hall, University of Memphis, Memphis, TN, 38152 2 Department of Geosciences, 113 Ozark Hall, University of Arkansas, Fayetteville, AR, 72701 Corresponding Author: Dorian J. Burnette Department of Earth Sciences 109 Johnson Hall University of Memphis Memphis, TN 38152 Phone: 901-678-4452 Fax: 901-678-2178 E-Mail: djbrntte@memphis.edu 1 1. Data collection The most continuous weather records in Kansas prior to 1870 are located in the northeastern part of the state, and three clusters of stations that were the most continuous from the 19th century to present were selected for the reconstruction (Darter 1942). The primary stations within each cluster were used as a baseline and then supplemented with surrounding station data to fill gaps and extend the history of the station cluster. The Lawrence cluster was supplemented with additional data from Burlingame, Globe, Morse, Olathe, Ottawa, and Topeka, Kansas. Fort Leavenworth and Kansas City were used to augment the record from the city of Leavenworth, and Manhattan was supplemented with nearby station data available from Fort Riley and Wamego, Kansas. The daily precipitation records from Oregon and Miami, Missouri, were also included to provide additional overlap with the longest stations from northeastern Kansas, and were supplemented with nearby station data for continuity. The Oregon, Missouri, record was kept by William Kaucher, but only portions of the daily record were available on microfilm from the National Archives beginning in 1867 (Darter 1942). Kaucher began systematic daily temperature and precipitation records upon his arrival in Oregon in 1855 (Hackett 1903), but most of these data have not been found. In this paper, we use Kaucher’s daily observations available in the microfilms and an additional one and one-half years of data that were published in the Missouri Valley Times from 1874-76 (total daily record for Oregon = 1 January 1867 to 31 December 1892, with missing values). These daily data were then supplemented with daily precipitation observations from Atchison, Kansas, and St. Joseph, Missouri. Daily precipitation data for Miami, Missouri, running from 1 January 1850 to 31 December 1892 were obtained from the Amos H. Sullivan papers (Sullivan 1934). Sullivan was 2 a surgeon who had an interest in the history of the Miami area and kept a diary of precipitation data beginning in 1847. Daily data were only available back to January 1850, so this study uses his daily observations beginning on 1 January 1850. Miami does not have a long modern record, so nearby Brunswick, Missouri, was substituted for the modern Miami record and supplemented with Carrollton and Marshall, Missouri. All of these daily precipitation records were screened for legibility and metadata as each station was digitized. The Web Search Store Retrieve Display system of NOAA’s Climate Database Modernization Program was consulted for some of these data (Dupigny-Giroux et al. 2007), but the quality of the scanned microfilms can be poor. Therefore, the historical data were primarily digitized directly from the source (i.e., microfilms and hand-written documents). Any observations that were too illegible were recorded as missing. All modern records for each station were obtained from the Global Historical Climatology Network (GHCN, Peterson and Vose 1997). The augmented station data files available online at www.djburnette.com/research/kansas/precip/ list the stations that were used for each day. The degree to which an average of these five-station clusters represents the spatial variability of warm season rainfall across the Kansas-Missouri study area was examined by correlating the 1950-2006 seasonal precipitation totals with the 0.5° gridded precipitation amounts from the Climatic Research Unit (TS3.0; Mitchell and Jones 2005). The growing season is illustrated in Figure 1, and similar results were found in the spring and summer (not shown). The primary data used for precipitation in the TS3.0 gridded dataset were those compiled by Mike Hulme, and these data have stations in common with the GHCN (Mitchell and Jones 2005). Thus, some correlation is expected, but the strong correlation extends beyond the stations used in this study (northeastern Kansas and northwestern Missouri area vs. Fig. 1). 3 2. Screening for data quality Histograms of the frequency of daily precipitation totals can be used to assess the quality of daily precipitation records (e.g., Daly et al. 2007; Burnette 2009). If the quality of the daily precipitation data is high, then the histogram should resemble a smooth, negative exponential curve (i.e., a high frequency of daily 0.01 inch amounts and fewer daily 1.00 inch totals). Failure to record all of the small precipitation events (e.g., 0.01 to 0.05 inches) is revealed by the truncation of the frequency histogram for the smallest daily totals. In addition to undercount, Daly et al. (2007) found that “spikes” can occur at amounts ending with a zero or five, which could bias the precipitation totals. This “5/10 bias” could be due to the coarse scale on the precipitation gauge, but could also arise from rounding the precipitation totals to “even” values and other poor observation practices (e.g., estimating the amount of precipitation that was in the gauge at observation time; Daly et al. 2007). These frequency histograms were constructed for each of the five augmented stations on a seasonal basis (see Burnette 2009). They were also constructed for specific time segments based on changes apparent in the metadata (e.g., for single observers and station locations, when possible). Segments of the historical and modern station data with possible undercount and 5/10 bias were flagged for further analysis. 3. Calculation of the effective moisture index The self-calibrating PDSI program of Wells et al. (2004) was used to calculate the effective moisture values on a monthly basis, which computes the difference between the precipitation and potential evapotranspiration, where potential evapotranspiration is derived 4 using the three possible formulations below from the Thornthwaite method (Thornthwaite 1948; Willmott et al. 1985): PE = 0, when T < 0C (1) PE = 16[(10T)(I)-1 ]a , when 0C T 26.5C (2) PE = -415.85 + 32.24(T) - 0.43(T)2 , when T 26.5C (3) where PE is the monthly potential evapotranspiration in mm and T is the monthly mean temperature in °C. The variable, I, is the heat index, which is a single value computed for a station based on modern normal monthly mean temperatures (T) during the number of months (N) when T ≥ 0°C: N I= [(T)(5) -1 1.514 ] (4) i=1 The variable, a, in Equation 2 is a function of the heat index: a = (0.00000067)(I)3 - (0.0000771)(I)2 + (0.0179)(I) + 0.49 (5) Monthly estimates of PE are then adjusted for the length of the month and duration of daylight hours by: APE = PE[(d)(30)-1 (h)(12)-1 ] (6) where APE is the adjusted monthly potential evapotranspiration in mm, d is the length of the month in days, and h is the duration of daylight hours on the fifteenth day of the month. References Burnette DJ (2009) Reconstruction of the eastern Kansas temperature and precipitation records into the mid-19th century using historical sources. Dissertation, University of Arkansas. Darter LJ (1942) List of climatological records in the National Archives. National Archives, Washington, D.C. 5 Daly C, Gibson WP, Taylor GH, Doggett MK, Smith JI (2007) Observer bias in daily precipitation measurements at United States cooperative network stations. Bull Am Meteorol Soc 88:899-912. Dupigny-Giroux L-A, Ross TF, Elms JD, Truesdell R, Doty SR (2007) NOAA's Climate Database Modernization Program: rescuing, archiving, and digitizing history. Bull Am Meteorol Soc 88:1015-1017. Hackett AE (1903) William Kaucher. Mon Weather Rev 31:142. Mitchell TD, Jones PD (2005) An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int J Clim 25:693-712. Peterson TC, Vose RS (1997) An overview of the Global Historical Climatology Network temperature database. Bull Am Meteorol Soc 78:2837-2849. Sullivan AH (1934) Amos H. Sullivan Papers. Manuscript available at the Western Historical Manuscript Collection, Columbia. Thornthwaite CW (1948) An approach toward a rational classification of climate. Geogr Rev 38:55-94. Wells N, Goddard S, Hayes MJ (2004) A self-calibrating Palmer Drought Severity Index. J Clim 17:2335-2351. Willmott CJ, Rowe CM, Mintz Y (1985) Climatology of the terrestrial seasonal water cycle. Int J Clim 5:589-606. 6 Fig. 1 The five station regional average growing season precipitation totals for 1950-2006 are positively correlated with the 0.5° gridded precipitation data during the growing season (both AMJJA) across the central U.S. (gridded precipitation data from Climatic Research Unit TS3.0; Mitchell and Jones 2005, all shaded areas = P<0.05) 7