The first part of this thesis, describes the methods to calculate the temperature anomalies and the errors of the different temperature data sets provided by the United States Historical Climatology Network Version 2 (USHCN V2) for the contiguous United States. For all present data sets, the anomaly and error calculations were performed. Among other results, two warmer and two colder time periods were observed and analyzed with the help of statistical methods. The second part extends the analysis of the annual Surface Air Temperature (SAT) anomalies and estimates the random error of observations that includes instrumental and data adjustment errors of the USHCN V2 stations and is based on the existing meta data for instrument and the homogenization work of Menne et al., 2009, Folland et al., 2001 and Brohan et al., 2006. Furthermore the different mean SAT anomaly time series are compared with the time series developed in Menne et al., 2009, which uses another grid box resolution. In Chapter 4 the main focus is on the so-called long-term stations, i.e. stations with at least 100 years of continuous records. Thus only these stations are used to compute the temperature anomalies of the contiguous United States. Moreover the anomalies of these long-term stations are used in the last chapter in order to determine the optimal temperature anomalies. These calculations are based on the method derived in Shen's paper about optimal regional averages, 1998, such that the Mean Square Error is minimized. Hence the optimal averages and the minimized errors are computed with the help of Empirical Orthogonal Functions for every year between 1897 and 2010.