RUDN Scientists Propose Optimal Approach to Map Urban Density by Satellite Data

RUDN University scientists have found the most accurate way to determine the spatial structure of urban confinement using satellite images. It will help create precise maps and monitor the microclimate of the urban environment, support sustainable development and even make weather forecasts. The results were published in the International Journal of Applied Earth Observation and Geoinformation.

Urban soil sealing is the covering with materials impermeable to moisture and plants (asphalt, concrete, etc.). It influence the microclimate of cities and natural areas. To monitor its dynamics and spatial heterogeneity, they use Sentinel-2 satellite images. These satellites receive images in different spectral ranges with a spatial resolution from 10 to 60 meters. Scientists interpret and analyze these images using special spectral indices. It is not always known which indexes are best to use in each specific task. For example, in a city where the soil is covered by impervious asphalt or buildings, there are no generally accepted indices. At the same time, accurate maps with information on soil sealing are crucial for sustainable urban planning, and for studying and predicting urban microclimates. RUDN soil scientists have identified two indices that give the most accurate results.

“Several global datasets are describing urban impervious cover that claims to be highly accurate. However, they all give different estimates if we consider many cities at the same time. At the same time, truly high precision is required to solve problems of urban planning and sustainable development. We analyzed several commonly used spectral indices and their thresholds to find out which ones provide the most accurate estimates. We also checked some global datasets for their applicability for studying the spatial heterogeneity of urban sealing,” said Yury Dvornikov, Ph.D., researcher at the Smart Urban Nature Center of RUDN University.

For the experiment, RUDN soil scientists selected 10 Russian cities from different climatic zones and studied Sentinel-2 data obtained at the beginning, middle, and end of the warm season 2019/2020. Having calculated possible indices, the authors compared them with real data on the condition and type of surfaces.

Soil scientists named two indices as the most suitable: mNDVI and UCI, and also indicated the optimal threshold values for them. The first is the modified normalized vegetation index, which depends on how vegetation reflects and absorbs different wavelengths of light. The second is the composite building index, which allows you to improve the spectral signal from artificial objects. The coefficient of determination (or the proportion of “explained” variance) for these two indices compared to real data was 82% or more. Only one of the existing global datasets showed a comparable result – the rest worked worse by several tens of percent.

“Our results can be used to create accurate maps of annual and even seasonal dynamics of urban sealing. This is important in situations where global data sets provide a large spread of tens of percent between estimates,”  said Yury Dvornikov, Ph.D., researcher at the Smart Urban Nature Center of RUDN University.