New paper asks: 'Would the 'real' temperature dataset please stand up?'; finds 'We have no ability to know' the true temperature data

A new paper published in Environmental Research Letters examines 8 datasets of temperature & precipitation observations in China over the past 105 years and finds significant unexplained differences between datasets, concluding,

"Because observations coming from different datasets do have differences, which one we can believe among the various so-called ‘observed climate datasets’? Indeed, we have no ability to know the ‘truth value’"

Indeed, two of the Chinese datasets in the paper show there was essentially no warming in China from 1950-2005.

Furthermore, the paper notes that because we don't know which temperature datasets are correct [if any], climate models based upon flawed temperature data will result in erroneous "tuning" of climate models and therefore biased projections i.e. GIGO]. Excerpt from the conclusion:

All of these results bring about a new challenge in the field of climate change. So-called ‘observed climate datasets’ play important roles in driving hydrologic models, evaluating global circulation models (GCMs) and regional climate models (RCM). Because observations coming from different datasets do have differences, which one we can believe among the various so-called ‘observed climate datasets’? Indeed, we have no ability to know the ‘truth value’; what we need to do is reduce the disagreement among the ‘observed datasets’ and depress their uncertainty.

As lamented in Climategate emails by Phil Jones to James Hansen, if the Chinese stopped their commendable practice of moving thermometers away from contamination by the Urban Heat Island [UHI] effect and instead taken temperature readings contaminated by airport jet exhaust [as the rest of the world did], the Chinese temperature data would have shown more artificial warming.  

One can't deny that temperature data contradictions, contamination, errors and tampering are indeed a worldwide climate phenomenon. 

Would the 'real' observed dataset stand up? A critical examination of eight observed gridded climate datasets for China

OPEN ACCESS

Qiaohong Sun, Chiyuan Miao, Qingyun Duan, Dongxian Kong, Aizhong Ye, Zhenhua Di and Wei Gong
PDF (5.39 MB) View article

Abstract

This research compared and evaluated the spatio-temporal similarities and differences of eight widely used gridded datasets. The datasets include daily precipitation over East Asia (EA), the Climate Research Unit (CRU) product, the Global Precipitation Climatology Centre (GPCC) product, the University of Delaware (UDEL) product, Precipitation Reconstruction over Land (PREC/L), the Asian Precipitation Highly Resolved Observational (APHRO) product, the Institute of Atmospheric Physics (IAP) dataset from the Chinese Academy of Sciences, and the National Meteorological Information Center dataset from the China Meteorological Administration (CN05). The meteorological variables focus on surface air temperature (SAT) or precipitation (PR) in China. All datasets presented general agreement on the whole spatio-temporal scale, but some differences appeared for specific periods and regions. On a temporal scale, EA shows the highest amount of PR, while APHRO shows the lowest. CRU and UDEL show higher SAT than IAP or CN05. On a spatial scale, the most significant differences occur in western China for PR and SAT. For PR, the difference between EA and CRU is the largest. When compared with CN05, CRU shows higher SAT in the central and southern Northwest river drainage basin, UDEL exhibits higher SAT over the Southwest river drainage system, and IAP has lower SAT in the Tibetan Plateau. The differences in annual mean PR and SAT primarily come from summer and winter, respectively. Finally, potential factors impacting agreement among gridded climate datasets are discussed, including raw data sources, quality control (QC) schemes, orographic correction, and interpolation techniques. The implications and challenges of these results for climate research are also briefly addressed.