An Investigation into the Impact of using Various Methods to Estimate Arctic Surface Air Temperature Anomalies

Time series of global and regional mean surface temperature anomalies are a common metric used to estimate recent climate change. Various techniques can be used to create these time series. The degree of difference from using five different techniques (Linear Interpolation, Global Ordinary Kriging, Global Simple Kriging, Not Interpolating and Regridding, and Not Interpolating) to estimate Arctic Surface Air Temperature (SAT) anomalies over land and sea ice were investigated using reanalysis data as a testbed. The differences in estimation of patterns in Arctic SAT anomalies and of Arctic area average anomalies were investigated. It was found that kriging techniques provided the smallest errors in estimates of Arctic SAT anomalies relative to the reanalysis 'truth'. Global Simple Kriging was often the best kriging method when a representative mean was chosen, especially over areas of sea ice. Linear Interpolation was the third most representative technique on average with Mean Absolute Errors (MAEs) up to 0.55 K larger than the kriging techniques. But, Linear Interpolation MAEs can be within 0.0002 K of MAEs produced by kriging techniques and Linear Interpolation is also, for some years, more likely than kriging techniques to produce estimates with the smallest errors. Not Interpolating and Regridding errors are generally slightly larger than Linear Interpolation errors. Not Interpolating provides the least representative estimates of Arctic anomalies. However, while Not Interpolating techniques provide the least representative estimates of Arctic anomalies these estimates are useful checks for confirming that the estimates from interpolating techniques are reasonable.