Summary of Kentucky Temperature and Precipitation Trends
Climate is an important influence upon our economy and the lifestyles of people in communities throughout Kentucky. However, climate is not constant. Examination of temperature and precipitation data since 1895 shows both periods of warming and cooling, along with periods when precipitation was relatively more and less abundant.
Kentucky's climate has been warming since the most recent cool period of the 1960s and 1970s. The average annual temperature in the Western, Central, and Bluegrass divisions now exceeds that of the prior warm period during the 1930s and 1940s. The average temperature in the Eastern division is also rising but remains below the earlier peak.
Average annual precipitation has been trending upward in the Western, Central, and Bluegrass divisions and is at or near its highest level since 1895. The historical record includes a dry period commencing in the 1930s and persisting in the early 1960s, followed by a wet period that peaked in the 1970s to early 1980s. In contrast, the Eastern division has experienced a decline in average precipitation since the mid 1990s and is near the minimum reached in the late 1930s.
Click on the thumbnails to explore patterns of variability and change that characterize Kentucky's climate since 1895. These graphs, portraying annual and monthly time series, provide historical perspective to current climate trends. While the focus is on climate change, it should be noted that the annual variability of both temperature and precipitation far outweigh the changes in their respective averages.
More information about the data and the method used to track changing averages of temperature and precipitation are provided below.
Graphs and Data
Click the appropriate divisions on Kentucky below for data graphs.(Graphs may take a few seconds to load)
Data are obtained from the National Climatic Data Center's Time Bias Corrected Divisional Temperature-Precipitation-Drought Index dataset (DSI-9640) containing monthly time series dating to 1895. Monthly values are calculated by averaging data from all stations that report both temperature and precipitation. Adjustments are made to correct for temperature biases that occur when observations of maximum and minimum temperatures are made for 24-hour periods ending at times other than midnight.
Note that while these are the best available data, they are not without limitations. Climate division averages give equal weight to each reporting station in a division. No accounting is made for the spatial distribution of stations. Hence, calculated averages may be sensitive not only to climatic influences, but also to changes in the number and location of observing stations over time. Further, many sites are known to have exposures where observed temperature and precipitation values are contaminated by effects of topography, vegetation, and anthropogenic influences. For example, observing stations are often situated near trees, buildings, parking lots, and other features that are known to influence observations of temperature and precipitation. Here, the assumption is made that such biases, when averaged over entire climate divisions, do not qualitatively affect the representation and interpretation of patterns of climate change.
Finally, temperature data used here represent the average of daily maximum and minimum temperatures. From these data alone, it is not possible to determine the extent to which observed trends reflect changes associated with daily maximum temperatures, daily minimum temperatures, or both. Likewise, it is not possible to determine whether trends in precipitation totals represent a change in the frequency of heavy precipitation events, light precipitation events, or both.
Climatic time series of temperature and precipitation are characterized by a high degree of year-to-year variability. As a result, important patterns of climate change that evolve gradually may be difficult to detect. For this reason, the choice of a method for tracking climatic averages over time is crucial.
Climate normals have traditionally been calculated as 30-year moving averages updated each decade. While relatively simple to implement and understand, this approach produces a pattern of change that is characterized by a step change every ten years. A global regression model (i.e., one that is fit to the entire time series) fits a straight line or higher-order function through a time series, thereby eliminating regular step changes. However, a global model assumes an unchanging trend or pattern of climate change. Other approaches for tracking climate change involve fitting curves locally (i.e., fitting a model to subsets of the time series). Here, a locally weighted regression model is fit using a 30-year window. The position of the window is then shifted forward, and the procedure is repeated until the window has progressed from the beginning of the series to the end of the series. Prior to applying the local regression model, the distributions of temperature and precipitation are winsorized to damp the influence of extreme observations. The result is a smooth curve of locally-accurate averages that better tracks climate change. While many alternative smoothing models are available, the model presented is designed to capture patterns of change that are more persistent and not to be excessively influenced by a subset of unusual data values. Smoothing models are not designed to capture sudden shifts in climate.