Fine-scale mapping of urban construction’s heat pattern
Suggested Elective course: Radar Remote Sensing (Q1)
Required skills: Python programming competence
In collaboration with the municipality of Rotterdam.
The phenomenon of urban heat islands generally refers to the heating of urban areas compared to their undeveloped surroundings. This phenomenon has several implications for the environment and human life, as higher urban temperatures can increase the amount of urban smog, increase air pollution, and affect human health. Remote sensing is an efficient tool to detect urban heat on a large scale. Many applications, however, require daily measurement of urban temperature at fine details. There is a trade-off between the temporal and spatial resolution of current thermal remote sensing images. Geostationary meteorological satellites (e.g., FengYun-2F) provide near-continuous thermal observations but are characterized by low spatial resolution (> 3 km). The spatial scale of polar satellites with daily images (e.g., MODIS Terra/Aqua or Sentinel 3) is also rather coarse (1 km). On the other hand, satellite-derived temperatures with a spatial resolution of 100 m can only be acquired by platforms with a low revisit cycle, such as the Landsat 9 satellite. In addition, optical thermal satellite imagery is weather dependent, i.e., clouds and poor weather conditions affect land surface temperature (LST). To overcome this drawback, this study aims to produce temperature maps of urban areas using SAR imagery to achieve fine spatial and temporal levels.
This study focuses on statistical downscaling, also known as thermal sharpening. This is a widely used downscaling method that establishes a statistical relationship between land surface temperature and surface parameters (predictors). In this study, surface parameters can be extracted using high-resolution, weather-independent SAR imagery and used to build a model to represent the temperature map at a fine level of detail. The development of complex statistical downscaling algorithms using sophisticated regression techniques (e.g., convolutional artificial neural networks) is considered to build and subsequently extract the urban heat map at a fine level so that the temperature of each individual built-up structure can be determined.