Assessing homogeneity of land surface air temperature observations using sparse‐input reanalyses

• Main Authors: Gillespie, Ian, Haimberger, Leo, Compo, Gilbert P., Thorne, Peter
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wiley 2023
• Subjects: Sea ice
• Online Access: https://mural.maynoothuniversity.ie/16947/; https://mural.maynoothuniversity.ie/16947/1/joc.7822.pdf

State-of-the-art homogenisation approaches for any test site rely upon the avail-ability of a sufficient number of neighbouring sites with similar climatic condi-tions and a sufficient quantity of overlapping measurements. These conditionsare not always met, particularly in poorly sampled regions and epochs. Modernsparse-input reanalysis products which are constrained by observed sea surfacetemperatures, sea-ice and surface pressure observations, continue to improve,offering independently produced surface temperature estimates back to the early19th century. This study undertakes an exploratory analysis on the applicabilityof sparse-input reanalysis to identify breakpoints in available basic station data.Adjustments are then applied using a variety of reanalysis and neighbour-basedapproaches to produce four distinct estimates. The methodological indepen-dence of the approach may offer valuable insights into historical data qualityissues. The resulting estimates are compared to Global Historical ClimatologyNetwork version 4 (GHCNMv4) at various aggregations. Comparisons are alsomade with five existing global land surface monthly time series. We find a lowerrate of long-term warming which principally arises in differences in estimatedbehaviour prior to the early 20th century. Differences depend upon the exactpair of estimates, varying between 15 and 40% for changes from 1850–1900 to2005–2014. Differences are much smaller for metrics starting after 1900 and neg-ligible after 1950. Initial efforts at quantifying parametric uncertainty suggestthis would be substantial and may lead to overlap between these new estimatesand existing estimates. Further work would be required to use these data prod-ucts in an operational context. This would include better understanding the rea-sons for apparent early period divergence including the impact of spatialinfilling choices, quantification of parametric uncertainty, and a means toupdate the product post-2015 when the NOAA-CIRES-DOE 20CRv3 sparseinput reanalysis product, upon ...