Temporally Weighted Land Surface Temperature: From Multiyear Observations to a Target-Year-Referenced Thermal Surface

Abstract

Abstract

Accurate characterization of land surface temperature (LST) is essential for urban thermal environment analysis. However, commonly used LST representations have their inherent limitations. For example, single-year LST may be influenced by year-specific anomalies, whereas multiyear means may obscure recent changes. Neither representation addresses how relevance to the target year and persistent historical information can be balanced within a single spatial LST surface. This study therefore proposes a temporally weighted land surface temperature (TWLST) method, which uses temporal decay weights and a half-life parameter to achieve the above balance and produce a continuous LST surface for spatial analysis. The framework was applied at regional and city scales in the Beijing–Tianjin–Hebei region and Beijing city, respectively, to examine its applicability. Spatial overlap between TWLST high-temperature (HT) areas and reference LST HT areas ranged from 0.577 to 0.733 and from 0.666 to 0.863 in the two applications. Historical annual LST anomalies were higher in TWLST-specific HT areas than in target-year-specific HT areas, with median paired differences of 1.16 °C and 2.49 °C, respectively. These results indicated that TWLST preferentially identified persistently warm locations while preserving major high-temperature patterns. The modeling and interpretation analyses further indicated that changes in LST representation could influence model performance and the relative importance of explanatory variables.