Variability of urban surface temperatures and implications for aerodynamic energy exchange in unstable conditions

Abstract

Sensible heat flux (QH) is a critical driver of surface and boundary layer meteorological processes, especially in urban areas. Aerodynamic resistance methods (ARM) for modelling QH are promising because, in principle, all that is needed is surface temperature (T0), air temperature (TA) and an aerodynamic resistance term (rH). There are significant challenges in urban areas, however, due to uncertainties in satellite‐derived land surface temperatures (LST), logistical challenges in obtaining high‐resolution air temperatures, and a limited understanding of spatial and temporal variability of rH and associated variables (e.g. thermal roughness length). This work uses an extensive LST dataset covering 6 years (2011–2016) in central London and a long‐term in situ observation network to analyse the variability of LST and rH variables. Results show that LST is spatially correlated with building and vegetation land cover with coherent thermal structures at length scales less than 500–1,000 m. Additionally, satellite‐observed LST varies with average building height (up to 10% cooler in areas with tall buildings). The rH term and associated variables are observed to vary on daily and seasonal cycles, and findings are used to model QH using five variations of an ARM‐based approach on a 100 m pixel basis. Modelled QH is compared to observations from three scintillometer paths and an eddy covariance flux tower. We find generally good agreement between observations and models, although there is uncertainty in all methods (mean absolute error ranges from 58.1–129.3 W/m2) due to the challenges involved in determining high‐resolution meteorological and surface inputs, particularly LST and friction velocity (u*). Additional complexity in evaluating modelled QH arises from anthropogenic heat sources: long‐term tower‐based observations show that TA and the radiometer‐derived T0 are warmer during working weekdays than non‐working days (up to 0.7 °C) and that there is an observed lag (2–3 hr) between energy consumption and the observed warming and modelled QH.