A Comparison of Daily Temperature Averaging Methodologies: Spatial Variability, Seasonal Patterns, and Related Climate Variables in the CONUS

Traditionally, daily average temperature is computed by taking the mean of two values- the maximum temperature over a 24-hour period and the minimum temperature over the same period. These data form the basis for numerous studies of long-term climatologies (e.g. 30-year normals) and recent temperature trends and changes. However, many first-order weather stations (e.g. airports) also record hourly temperature data. Using an average of the 24 hourly temperature readings to compute daily average temperature should provide a more precise and representative estimate of a given day’s temperature. These two methods of daily temperature averaging (max+min/2, average of 24 hourly temperature values) are computed and mapped for first-order weather stations across the United States for the 30-year period 1981-2010. There is a statistically significant difference between the two methods, as well as an overestimation of temperature by the traditional method (Tmax + Tmin /2), particularly in southern and coastal portions of the Continental U.S. The likely explanation for the long-term difference between the two methods is the underlying assumption of the twice-daily method that the diurnal cycle of temperature follows a symmetrical pattern. There is a relationship of these differences to the regional synoptic climatology, along with related factors such as atmospheric moisture, land use-land cover, and snow cover, which help to explain both seasonal patterns and spatial variability. Therefore, these climate variables are used to create multiple linear regression models to explain the observed patterns. Additional analysis of station normals based on hourly data should have important applications, such as improved assessment of recent anthropogenic climate change.