Evidence of Urban Blending in Homogenized Temperature Records in Japan and in the United States: Implications for the Reliability of Global Land Surface Air Temperature Data

Abstract

In order to reduce the amount of non-climatic biases of air temperature in each weather station’s record by comparing it to neighboring stations, global land surface air temperature datasets are routinely adjusted using statistical homogenization to minimize such biases. However, homogenization can unintentionally introduce new non-climatic biases due to an often-overlooked statistical problem known as “urban blending” or “aliasing of trend biases”. This issue arises when the homogenization process inadvertently mixes urbanization biases of neighboring stations into the adjustments applied to each station record. As a result, urbanization biases of the original unhomogenized temperature records are spread throughout the homogenized data. To evaluate the extent of this phenomenon, the homogenized temperature data for two countries (Japan and United States) are analyzed. Using the Japanese stations in the widely used Global Historical Climatology Network (GHCN) dataset, it is first confirmed that the unhomogenized Japanese temperature data are strongly affected by urbanization bias (possibly ~60% of the long-term warming). The United States Historical Climatology Network dataset (USHCN) contains a relatively large amount of long, rural station records and therefore is less affected by urbanization bias. Nonetheless, even for this relatively rural dataset, urbanization bias could account for ~20% of the long-term warming. It is then shown that urban blending is a major problem for the homogenized data for both countries. The IPCC’s low estimate of urbanization bias in the global temperature data based on homogenized temperature records may have been biased low due to urban blending. Recommendations on how future homogenization efforts could be modified to reduce urban blending are discussed.

Challenges in the Detection and Attribution of Northern Hemisphere Surface Temperature Trends Since 1850

“…Meanwhile, the urban and rural ST implies 60% more warming than the rural-only ST (0.885°C/100y compared to 0.554°C/100y) and even more warming than any of the other ST estimates (also already noted by C2021). This is consistent with other studies that have argued that the land component of current global temperature estimates remains significantly contaminated by urbanization biases, e.g., Soon et al (2015), Zhang et al (2021), Scafetta (2021), Katata et al (2023).…”

The Detection and Attribution of Northern Hemisphere Land Surface Warming (1850–2018) in Terms of Human and Natural Factors: Challenges of Inadequate Data

“…AR6 has explicitly argued that urbanization bias represents less than 10% of the longterm warming, but several recent studies have disputed this claim [5,7,9,10,78]. Meanwhile, AR6 argues that the Matthes et al (2017) solar forcing dataset has been confirmed to be reliable, yet this claim is challenged by several studies arguing it has not yet been satisfactorily resolved which (if any) of the many solar forcing datasets are most reliable [5,7,12,15].…”

“…However, we caution that the current approaches of using statistical homogenization techniques to correct temperature records for non-climatic biases are prone to "aliasing effects" [74,75], including "urban blending" [5,76,77]. Katata et al (2023) have offered some potential modifications to temperature homogenization to reduce or remove this problem [78].…”

“…However, C2021 [5] noted that such algorithms can inadvertently lead to "urban blending" whereby the urbanization biases of urban stations are partially "aliased" [74,75] into the homogenized records of rural neighbors [5,76,77]. This was recently confirmed by Katata et al (2023) [78].…”