Linear regression

Urban heat island

Random forest

Demography

Ecology

Biology

Statistics

Geotechnical engineering

Yangtze river

China

Archaeology

Socioeconomic status

Vegetation (pathology)

Computer science

Physical geography

Geology

Regression analysis

Mathematics

Geography

Hydrology (agriculture)

Political science

Population

Sociology

Environmental science

Land use

Medicine

Meteorology

Interdependence

Machine learning

Urban area

Pathology

Sustainable Cities and Society

Land surface temperature (LST) is influenced by a variety of factors. Relative studies have been carried out extensively in recent years. However, the intricate relationship has yet to be fully investigated. To assess the relative importance and interdependent effects of urban characteristics on LST, we chose the central part of Shanghai as the study area and used the random forest (RF) regression model to quantify 19 different variables on 3 different grid scales. Our results suggest that Building density (BD), percentage of water area (P_Water), percentage of vegetation area (P_Veg), building height (BH), distance to the Yangtze River (DIST), and mean patch area of green space (AREA_MN) are the 6 most important influencing factors, whereas socioeconomic factors such as point of interest (POI) and nighttime light data (NTL) are the least correlated to LST changes. Meanwhile, most of these 6 variables showed non-linear correlations with LST. In particular, the Yangtze River has a steady cooling effect up to 20 km, after which it declines rapidly and loses its effect at 25 km, and BH shows a fluctuated correlation with LST. This improved knowledge of the relationship between urban characteristics and LST will serve as a guide for future urban policymakers and planners.

The dominant factors and influence of urban characteristics on land surface temperature using random forest algorithm