Multiscale evaluation of NCEP and CRUNCEP data sets at 90 large U.S. cities

The datasets of National Centers for Environmental Prediction (NCEP I and II) and Climatic Research Unit – NCEP (CRUNCEP) were evaluated by comparing against observational data from the National Oceanic and Atmospheric Administration (NOAA) at monthly and annual scales, site and regional scales. The evaluations were conducted for air temperature (maximum, mean, and minimum) and precipitation at the 90 largest cities across the United States. At annual scale, NCEP I, NCEP II and CRUNCEP were slightly yet significantly underestimated for mean temperature (1%) and maximum temperatures (1%). The NCEP I was significantly underestimated for minimum temperature (1%) while overestimated for total precipitation (16%); the NCEP II was significantly underestimated for minimum temperature (2%) and total precipitation (2%); the CRUNCEP was significantly overestimated for minimum temperature (2%) and total precipitation (6%). At monthly scale, three datasets were underestimated for a majority of months in terms of monthly mean temperature, monthly maximum temperature, monthly minimum temperature, and monthly total precipitation except an overestimation of CRUNCEP on minimum temperature (48%) and monthly total precipitation (43%). At the seasonal scale, three datasets had smaller biases in summer season and larger biases in winter season for temperature, while larger biases in summer season and smaller biases in winter seasons for precipitation. Overall, the CRUNCEP dataset had a slightly better performance than NCEP I and NCEP II datasets. This multi-scale evaluation of the three most-widely used regional climate datasets provides insightful information for atmospheric science studies, particularly urban heat impact investigation.