Estimating heat storage in urban areas using multispectral satellite data and machine learning

  • SCI-E
作者: Hrisko, Joshua;Ramamurthy, Prathap;Gonzalez, Jorge E.
作者机构: CUNY City Coll, Dept Mech Engn, New York, NY 10031 USA.
CUNY City Coll, NOAA, CESSRST Ctr, New York, NY 10031 USA.
语种: 英文
关键词: Heat storage;GOES-16;Machine learning;Radiance;Heat flux;Urban;Satellite remote sensing;GBRT
ISSN: 0034-4257
年: 2021
卷: 252
基金类别: National Oceanic and Atmospheric Administration - Cooperative Science Center for Earth System Sciences and Remote Sensing Technologies (NOAA-CESSRST) [NA16SEC4810008]; City College of New York, NOAACESSRST (aka CREST) program; NOAA Office of Education, Educational Partnership Program; Department of Defense Army Research Office Grant [W911NF-18-10371]; Federal Emergency Management Agency [FEMA-4085-DR-NY]; U.S. Department of Energy's Office of ScienceUnited States Department of Energy (DOE)
摘要: A satellite-derived hysteresis model is presented for estimate heat storage in urban areas. Storage heat flux, one of the dominant terms in the urban surface energy budget (USEB), is largely unknown despite its critical relationship to various urban environmental processes. This study introduces a novel technique for quantifying heat storage by relating multispectral satellite radiances and geophysical properties to ground-truth residual heat storage computed with flux instruments. Gradient-boosted regression trees serve as the method of maximizing the relationship between satellite data and flux measurements. Several flux networks are used to train and validate the model over varying land cover types, which strengthens the robustness of the model. The model performs well under variable weather conditions such as cloudy rainy days. In comparison with other studies, the RMSE and MAE values were found to be lower than some ground-to-ground studies, and is one of few satellite-derived methods that computes direct comparison over a range of different land cover types.

Estimating heat storage in urban areas using multispectral satellite data and machine learning