The U.S. Climate Reference Network (CRN) is a National Oceanic and Atmospheric Administration (NOAA)-sponsored weather and climate observing network and research initiative. When fully deployed, the USCRN will consist of approximately one hundred stations nationwide at locations selected to capture both the national and regional climate trends and variations for temperature and precipitation. In this study, we estimate Normals of near-surface air temperature (T_min, T_max, and T_mean) for CRN sites by using CRN measurements combined with station data from the National Weather Service Cooperative Observer (COOP) Network. Normal is defined as the mean of three consecutive decades (1971-2000).  When CRN stations have Normals, either calculated or accurately estimated, current CRN observations will be able to be put into an historical perspective for operational climate monitoring activities which increases CRN usefulness. With CRN estimated Normals, CRN station data can be integrated into other surface networks for a variety of applications.  

The mathematical formula for an estimated CRN Normal is N_crn= T_crn - DT_coop, where T_crn is the current CRN temperature and DT_coop is the temperature anomaly interpolated from surrounding COOP stations.  Naturally, the goal is to estimate Normals with as small an error as possible.  Towards that end, a variety of spatial interpolation schemes are being investigated.  Evaluation of the different procedures is done by estimating Normals for the COOP stations which already have 1971-2000 Normals calculated.  The approach with the smallest error will then applied to estimate N_crn.  Preliminary work indicates that the error in estimating Normals varies by season, region, and length of the data available to be used. Figures 1 and 2 show an example of the error for the estimated Normal for January minimum temperature (T_min).  In this example, Normals were estimated using an approach in which temperature anomalies were spatially interpolated using a weighting scheme involving the inverse of the mean square difference in actual (not anomaly) temperature, which produced the smallest error of eight approaches evaluated so far.

 

 

Figure 1.  Error of estimated Normal of January minimum temperature (T_min). The magnitude of error varies with the number of years of data to used and the distance from from which stations are selected for use in the interpolation.  The mean number of stations used in each estimation increases with increasing distance.  Data of 2001-2003 are used.  The temperature anomaly was spatially interpolated using a weighting scheme involving the inverse of the mean square difference in actual (not anomaly) temperature. 

 

 

Figure 2.  Spatial distribution of the error of estimated Normal calculated using stations within 140 km based on 3-year’s data (the CYAN line of Fig.1)