Long-Term Trends and Variability of Precipitation in Kentucky

Southeastern Geographer Vol. 29, No. 1, May 1989, pp. 55-65 LONG-TERM TRENDS AND VARIABILITY OF PRECIPITATION IN KENTUCKY David A. Howarth INTRODUCTION. Recent dry summers over much of the United States and the well-publicized "greenhouse effect" have generated considerable interest in climatic change and variability in both the scientific and lay presses. (7) Of particular importance are short-term fluctuations in climate at the local and regional scales where, for example, the effects of an abnormally hot and dry summer can have an immediate impact on the population. One can argue that precipitation and its variability are the most critical climatic elements in the Southeastern United States. (2) Nevertheless , the literature on climate change is dominated by studies of temperature variability; studies of precipitation variability certainly exist but are decidedly fewer in number. Given steadily increasing populations and economic development and the concurrent demand for fresh water in the Southeast, any shortage of precipitation over periods of a few weeks or longer can create serious problems for communities, businesses , and agriculture. As such, it may be instructive to investigate the long-term precipitation records for states or regions within the Southeast to provide a framework whereby extremes and fluctuations can be evaluated . The intent of this study is to analyze the precipitation record for the state of Kentucky during this century, and its primary emphasis is to identify trends and variability in the seasonal and annual time series for the state as a whole. DATA AND METHODOLOGY. The precipitation data used in this work are monthly totals by climatic divisions for the period 1895—1983 and are available from the National Climatic Data Center (NCDC) on magnetic tape. Precipitation totals for the four climatic divisions in Kentucky were combined to create monthly time series for the state as a whole. The composite series are not areally weighted because the four divisions are nearly equal in area. (3) Monthly series were then combined Dr. Howarth is Associate Professor ofGeography at the University ofLouisville in Louisville, KY 40292. 56Vol. XXIX, No. 1 into seasonal and annual series, for which long-term means and standard deviations were calculated. Standard deviations are appropriate measures of variability in this particular case since the data constituting the time series are approximately normally distributed. (4) In the five plots of annual and seasonal precipitation totals (Figs. 1—5), the resulting time series are connected by a solid line. The long-term means and plus/minus two standard deviations are represented as thick solid lines and dotted lines, respectively. Two additional statistical procedures have been applied to the original time series in order to extract information. First, a trend line (calculated by least squares methods) is shown in the five plots as a thin solid line. Tests of significance are performed against the null hypothesis that no trends exist in the original data. Second, the original time series have KENTUCKY ANNUAL 65 ? 50 o. 45 30 25 1900 1910 1920 1930 1940 1950 1960 1970 1980 YEAR Fig. 1. Time series ofannual precipitation for Kentucky. The heavy solid line is the long-term mean; the dotted lines, ±2 standard deviations; and the thin solid line, the least squares trend line. Plus signs represent the smoothed values derived from a low-pass filter. Vol. XXIX, No. 1 57 Fig. 2. KENTUCKY WINTER 1920 1930 [ItIIIIiIi] ? t ? ? r ? r ? ? [ ? ? ? 196019701980 YEAR Time series of winter precipitation for Kentucky. The heavy solid line is the long-term mean; the dotted lines, ±2 standard deviations; and the thin solid line, the least squares trend line. Plus signs represent the smoothed values derived from a low-pass filter. been smoothed with a low-pass, nine-point binomial filter. This smoothing function effectively removes fluctuations of approximately eight years or less. (5) The smoothed series are represented by the plus signs in the first five figures. Additional information about the variability within the five time series is gained by calculation of Suckling's absolute-value climate departure index. (6) For the annual time series (Fig. 6), both the individual magnitudes (solid line) of the index as well as the smoothed values (open squares) are shown. Smoothing was accomplished with the same...