The Switch to MMTS Has Changed Extreme Temperature Trends

In the 1980s, the National Weather Service’s Cooperative Observer Network (COOP) began to change the daily maximum (T_max) and minimum temperature (T_min) instrumentation from a liquid-in-glass thermometer in a Cotton Region Shelter (CRS/LIG) to an electronic Maximum-Minimum Temperature System (MMTS). Since the changeover, simultaneous side-by-side observations form the CRS/LIG and MMTS systems have been made at the Colorado State University and Rutgers University COOP stations. These simultaneous observations, now available through 2015 and exceeding 20 years in length, were analyzed to estimate the impact of this changeover on trends in extreme temperatures.

The CRS/LIG T_max observations are about 0.4K warmer than MMTS in the central part of the distribution. However, there are differences on the extreme tails. At Fort Collins, for extreme low values of T_max, specifically below the 5^th percentile threshold, the CRS/LIG observations average nearly 1.0K warmer than the MMTS values. Observations of snow depth indicate that snow cover is on average highest on days with the lowest T_max values, indicating that snow cover is likely one of the causes for the larger differences. At the Rutgers station, for extreme high values of T_max, above the 95^th percentile threshold, the CRS/LIG observations average about 0.6K warmer than the MMTS values. For extreme values of T_min, the differences between CRS/LIG and MMTS are small (around 0.1K or less) for both extreme high Tmin and extreme low Tmin at Fort Collins and slightly larger (0.1-0.2K) at Rutgers. A comparison of time series of extreme metrics between the official record (with the change to MMTS) and a homogeneous record consisting solely of LIG/CRS observations shows substantial differences. Specifically, the changeover to MMTS artificially diminishes both recent upward trends in heat waves and downward trends in cold waves.

A sensitivity study for the entire U.S., using various daily temperature adjustment values applied to daily observations for a set of long-term COOP stations, indicates similar noticeable quantitative effects on extreme temperature metrics for the nation as a whole. Most important, the differences between variations in metrics separately calculated with T_max and T_min are substantially diminished and conclusions about different behavior of extreme nighttime vs daytime extremes may need to be re-examined.