Influence of non-urban reference delineation on trend estimate of surface urban heat island intensity: A comparison of seven methods

Increasing concern over urban heat mitigation creates an emerging necessity to track the evolution of urban heat island intensity (UHII) to assess the effectiveness of existing plans in combating urban heat. However, complex human-environment interactions make it difficult to accurately isolate urban effects in UHII evolutions due to unstable non-urban references. This study provided the first test on the sensitivity of surface UHII (SUHII) trend estimate to seven methods of non-urban reference delineation across 281 Chinese cities. The corresponding seven non-urban references include administrative non-urban areas, administrative non-urban areas excluding extreme elevations, administrative low plant areas, suburb, outer suburb, outer suburbs progressively excluding unstable areas interfered by evident greening/desertification trends and the SUHI footprint of neighboring cities, respectively. Results revealed that the selection of different non-urban references significantly altered SUHII and the nature of observed surface thermal island (heat or cold) in 74% and 8% of cities, respectively. Besides, due to diverse and inconsistent land use and land cover changes across and nearby non-urban areas, the way of non-urban reference delineation also affected the amplitude and even direction of statistically significant SUHII trends in 41% and 21% of cities, respectively. Among the seven methods, the use of suburb which was collectively interfered by urban and non-urban changing factors, produced a large number of misestimated significant downtrends (2.7-4.5 times more cities than other methods) and was therefore in general not recommended. Administrative non-urban areas were not suitable to cities with evident greening/desertification trends, yet the applicability can be largely enhanced by removing areas with extreme elevations. Low plant areas were in general feasible, except for the cities with suburban low plants additionally warmed up by urban areas. Outer suburb was relatively the optimal, particularly the areas with all concerned interferences excluded. Overall, this study provided empirical evidence supporting the need to use stable non-urban references for reliable trend estimate of urban-induced SUHII variations. The results may benefit more accurate assessments of the causes, progress, and consequences of SUHI evolution, thereby minimizing the chances of policy failure in SUHI mitigation.