Optimal Blending of Land Surface Data Sets for Global Coverage

A nearly-global surface temperature data set was produced from numerous sources. Sea surface temperatures are provided by the OISST, which is a blend of satellite and in situ observations. Land surface temperatures comes from a combination of three inputs. In situ measurements are provided by the Global Historical Climatological Network (GHCN), which contains over 1,000 observations in near real time each month. The Special Sensor Microwave Imager is used to provides coverage over land surfaces that are not snow covered. The Microwave Sounding Unit (MSU) provides observations over snow covers surfaces at latitudes above 30 degrees. The MSU makes observations in the lower troposphere. However, there is a strong correspondence in high latitudes between lower tropospheric temperature anomalies and land surface temperature anomalies. This correspondence is strong enough that the MSU observations can be used to identify spatial structures in the land surface anomalies. None-the-less the magnitude between anomalies at the surface and lower must be calculated from the differences in GHCN observations and the overlying MSU values. These differences are interpolated across areas of missing surface observations and added to the MSU anomalies to provide full coverage. Independent high resolution validation of this approach supports the procedure. Optimal blending is used to minimize the root mean errors in the final fields. The resultant global product had errors less than 1oC at one degree resolution. By blending the various data sets, we are able to provide nearly full global coverage, and gain valuable insight into the spatial structure of surface temperature anomalies. The product is based on 12 years of data, and is available free of charge in near real time.