Abstract
In the present work a recently developed product, the South American Mapping of Temperature (SAMeT), is presented. SAMeT minimizes certain deficiencies associated with both the eventual lack of observations and incorporates temperature corrections due to elevation or lapse rate of temperature (LRT) considerations. SAMeT is based on combinations of observed 2‐m temperatures and ERA5 reanalysis, as well as LRTs values computed from maximum (Tmax) and minimum (Tmin) temperatures. The LRT were computed using a simple linear regression on a 40‐year ERA5 reanalysis dataset and a digital elevation model dataset from Global Digital Elevation (GTOPO30). It was obtained that the computed LRT is consistently smaller than the standard LRT for all the seasons and studied regions. The South American territory was divided into four subdomains and the evaluation performed via a cross validation. This methodology consists in removing 90% of the observations to form OBS90 and generate the SAMeT fields. The remaining 10% (OBS10) is used for validation purposes. An additional dataset, hereafter referred to as OBS90i, is obtained using Kriging interpolation of OBS90 along with GTOPO30 and temperature adjustment by a standard LRT (−6.5°C·km −1 ). The evaluation period was from January 2011 to December 2013. Thus by using the standard LRT, a systematic negative bias for the temperature is obtained. On the other hand, ERA5 displays underestimation for Tmax and overestimation for Tmin. Furthermore, in comparative terms ERA5 displays the larger errors; thus, it is suggested that a bias remotion is important for any application involving these data. The systematic application of computed values for LRT in addition to a combination between observations and ERA5 data allows SAMeT to generate fields with smaller errors compared to observations than ERA5 while keeping spatial correlations. Thus, the present product brings an innovation to the temperature space fields for the South American region.