During the summer of 2003 several NOAA Research Laboratories (Environmental Technology Laboratory, Forecast Systems Laboratory, and National Severe Storms Laboratory) collaborated with the National Weather Service including the National Centers for Environmental Prediction on a program to improve surface temperature forecasts. This program used a regional observational and modeling testbed approach, focusing on the New England region. The motivation for this project was the recognition that the annual cost of electricity on a nationwide basis could decrease by $1 billion per year if the accuracy of surface temperature forecasts improved by 0.6 degree Celsius. During this project arrays of special research grade instruments were deployed throughout New England by NOAA Research, including five 915 MHz wind profilers with RASS, cloud ceilometers, surface meteorology including solar and net radiation. Measurements from these instruments were compared, on an hourly basis in real-time during the experiment, against predictions made by operational and research forecast models, including the Eta, GFS, NMM, and RUC models, as well as experimental SREF (Short Range Ensemble Forecast) models under development at NCEP. Early results from this project will be presented, focusing on phenomena that most directly affect surface temperature, including clouds and radiation, and boundary layer depth.