Detection of non‐climatic biases in land surface temperature records by comparing climatic data and their model simulations

Abstract The 0.6 °C warming observed in global temperature datasets from 1940 to 1960 2000–2020 can be partially due urban heat island (UHI) and other non-climatic biases the underlying data, although several previous studies have argued contrary. Here we identify land regions where such could present by locally evaluating their diurnal range (DTR = T Max − Min trends between decades 1945–1954 2005–2014 1951–1960 1991–2000 versus synthetic hindcasts produced CMIP5 models. Vast of Asia (in particular Russia China) North America, a significant part Europe, Oceania, relatively small parts South America Colombia Venezuela) Africa show DTR reductions up 0.5–1.5 larger than hindcasted ones, mostly fast urbanization has occurred, as central-east China. Besides, it is found: (1) May October, globally warmed 40% less hindcast; (2) Greenland, which appears nearly free any contamination, Mean about 50% (3) world macro-regions with, on average, lowest with low (60S-30N:120 W–90 E 60 S–10 N:90 E–180 E: Central Africa, Oceania) 20–30% models’ hindcast. Yet, macro-region largest high (30 N–80 N:180 W–180 most Asia) just little bit more (5%) hindcast, indicates that models well agree only potentially problematic records. Indeed, also tree-based proxy reconstructions covering 30°N–70°N area produce significantly correspondent instrumentally-based record since 1980. Finally, compare sea surface data simulations find 25–45% 1 1940–1960 biases. By merging (assumed correct) an adjusted based model prediction, during same period found 15–25% lower reported. corrected compatible shown satellite UAH MSU v6.0 troposphere 1979. Implications for climate evaluation future estimates are briefly addressed.