The maximum and minimum temperatures are two of the most common climate variables in the historical climate observations. As new surface climate networks are established or upgrades to existing networks are undertaken, the sampling rates used to establish the maximum and minimum temperature must be carefully considered. This paper will present the effect of sampling rate on the observations of maximum and minimum air temperatures in several surface networks including the U. S. Climate Reference Networks (CRN), the Cooperative Observer Network (COOP), and the automated weather station networks (AWS).

A field comparison was conducted by simultaneously using six different sampling rates that ranged from two seconds per sample to five-minute average per sample to generate different realizations of the maximum and minimum temperature. Both manual and automated observations were conducted according to the operational network requirements. The temperature bias introduced by different sampling rates was analyzed for the CRN, COOP, and AWS, respectively. In addition, the current sampling rate used in the latest CRN network is discussed in detail. Our results provide information on the magnitude and sign of the bias due to different sampling rates used in current networks.