In AR4 Urban Heat Island (UHI) effects were concluded to be real local phenomena with negligible impact on large-scale trends. UHI and land-use land-cover change (LULC) effects arise mainly because the modified surface affects the storage and transfer of heat, water and airflow. For single discrete locations these impacts may dominate all other factors.
Regionally, most attention has focused on China. A variety of investigations have used methods as diverse as SST comparisons (e.g., Jones et al., 2008), urban minus rural (e.g., Ren et al., 2008; Yang et al., 2011), satellite observations (Ren and Ren, 2011) and observations minus reanalysis (e.g., Hu et al., 2010; Yang et al., 2011). Interpretation is complicated because often studies have used distinct versions of station series. For example, the effect in Beijing is estimated at 80% (Ren et al., 2007) or 40% (Yan et al., 2010) of the observed trend depending on data corrections applied. A representative sample of these studies suggest the effect of UHI and LULC is approximately 20% of the trend in Eastern China as a whole and of the order 0.1°C per decade nationally (Table 1 in Yang et al., 2011) over the last 30 years, but with very substantial uncertainties. These effects have likely been partially or completely accounted for in many homogenized series (e.g., Li et al., 2010b; Yan et al., 2010). Fujibe (2009) ascribes about 25% of Japanese warming trends in 1979–2006 to UHI effects. Das et al. (2011) confirmed that many Japanese sites have experienced UHI warming but that rural stations show unaffected behaviour when compared to nearby SSTs.
There is an important distinction to be made between UHI trend effects in regions underseeing rapid development and those that have been developed for a long time. Jones and Lister (2009) and Wilby et al. (2011) using data from London (UK) concluded that some sites that have always been urban and where the UHI has not grown in magnitude will exhibit regionally indicative trends that agree with nearby rural locations and that in such cases the time series may exhibit multi- decadal trends driven primarily by synoptic variations. A lack of obvious time-varying UHI influences was also noted for Sydney, Melbourne and Hobart in Australia by Trewin (2012). The impacts of urbanization also will be dependent on the natural LULC characteristics that they replace. Zhang et al. (2010) found no evidence for urban influences in the desert North West region of China despite rapid urbanization.
Global adjusted data sets likely account for much of the UHI effect present in the raw data. For the US network, Hausfather et al. (2013) showed that the adjustments method used in GHCNv3 removed much of an apparent systematic difference between urban and rural locations, concluding that this arose from adjustment of biased urban location data. Globally, Hansen et al. (2010) used satellite-based nightlight radiances to estimate the worldwide influence on LSAT of local urban development. Adjustments reduced the global 1900–2009 temperature change (averaged over land and ocean) only from 0.71°C to 0.70°C. Wickham et al. (2013) also used satellite data and found that urban locations in the Berkeley data set exhibited even less warming than rural stations, although not statistically significantly so, over 1950 to 2010.
Studies of the broader effects of LULC since AR4 have tended to focus on the effects of irrigation on temperatures, with a large number of studies in the Californian central belt (Christy et al., 2006; Kueppers et al., 2007; Bonfils et al., 2008; Lo and Famiglietti, 2013). They find cooler average temperatures and a marked reduction in DTR in areas of active irrigation and ascribe this to increased humidity; effectively a repartitioning of moist and dry energy terms. Reanalyses have also been used to estimate the LULC signature in LSAT trends. Fall et al. (2010) found that the North American Regional Reanalysis generated overall surface air temperature trends for 1979–2003 similar to observed records. Observations-minus-reanalysis trends were most positive for barren and urban areas, in accord with the results of Lim et al. (2008) using the NCEP/NCAR and ERA-40 reanalyses, and negative in agricultural areas.
McKitrick and Michaels (2004) and de Laat and Maurellis (2006) assessed regression of trends with national socioeconomic and geographical indicators, concluding that UHI and related LULC have caused much of the observed LSAT warming. AR4 concluded that this correlation ceases to be statistically significant if one takes into account the fact that the locations of greatest socioeconomic development are also those that have been most warmed by atmospheric circulation changes but provided no explicit evidence for this overall assessment result. Subsequently McKitrick and Michaels (2007) concluded that about half the reported warming trend in global-average land surface air temperature in 1980–2002 resulted from local land surface changes and faults in the observations. Schmidt (2009) undertook a quantitative analysis that supported AR4 conclusions that much of the reported correlation largely arose due to naturally occurring climate variability and model over-fitting and was not robust. Taking these factors into account, modified analyses by McKitrick (2010) and McKitrick and Nierenberg (2010) still yielded significant evidence for such contamination of the record.
In marked contrast to regression based studies, several studies have shown the methodologically diverse set of modern reanalysis products and the various LSAT records at global and regional levels to be similar since at least the mid-20th century (Simmons et al., 2010; Parker, 2011; Ferguson and Villarini, 2012; Jones et al., 2012; Vose et al., 2012a). These reanalyses do not directly assimilate the LSAT measurements but rather infer LSAT estimates from an observational constraint provided by much of the rest of the global observing system, thus representing an independent estimate. A hypothesized residual significant warming artefact argued for by regression-based analyses is therefore physically inconsistent with many other components of the global observing system according to a broad range of state-of-the-art data assimilation models (Box 2.3). Further, Efthymiadis and Jones (2010) estimated an absolute upper limit on urban influence globally of 0.02°C per decade, or about 15% of the total LSAT trends, in 1951–2009 from trends of coastal land and SST.
In summary, it is indisputable that UHI and LULC are real influences on raw temperature measurements. At question is the extent to which they remain in the global products (as residual biases in broader regionally representative change estimates). Based primarily on the range of urban minus rural adjusted data set comparisons and the degree of agreement of these products with a broad range of reanalysis products, it is unlikely that any uncorrected urban heat-island effects and LULC change effects have raised the estimated centennial globally averaged LSAT trends by more than 10% of the reported trend (high confidence, based on robust evidence and high agreement). This is an average value; in some regions with rapid development, UHI and LULC change impacts on regional trends may be substantially larger.