The Berkeley Earth Surface Temperature project has brought together the world’s instrumental temperature data and conducted an improved analysis of land-surface climate change since 1800.  This effort began by building an expanded dataset incorporating 1.2 million station-years of observations from 36000 sites (compared to 7280 stations in the efforts led by NOAA).  To this dataset we apply advanced statistical techniques to infer the time-evolution of Earth’s temperature field in a way that preserves basic spatial relationships and avoids gridding.  At the same time, reliability assessment procedures automatically compensate for various biases and inhomogeneities in the temperature data arising from measurement problems, urbanization, and other confounding factors.  Lastly, modern statistical techniques allow us to place robust uncertainty estimates on the resulting reconstruction of Earth’s temperature history.  The ultimate result of the expanded dataset and improved methodology is an analysis of climate change over land with inferred uncertainties that are often 60-75% smaller than comparable uncertainties reported in prior work.

 

The Berkeley Earth project’s fresh look at climate change has largely confirmed prior observations of global warming during the twentieth century (e.g. 0.911 ± 0.042 C over land since the 1950s).  At the same time, we have extend the instrumental record 50 years further back in time and have substantially reduced the overall uncertainties.  This talk will discuss our analysis methods, expanded dataset, and efforts to address and understand the various biases present in instrumental temperature observations.