The quality of the sound environment has become a key indicator for assessing urban livability and overall ecological conditions. Traditional monitoring methods based on sound pressure levels (SPL) fall short in fully reflecting residents' subjective auditory experiences. Moreover, these methods rely on costly networks, multisource data, or complex attenuation models, yet remain limited to street- or point-scale observations. To address these limitations, this study proposes a modeling framework that integrates multiscale and multispectral remote sensing imagery with deep learning to estimate the Sound Environment Discomfort Level (SEDL) at the residential-block scale and validates the approach across mainland China. To construct the estimation model for the SEDL, the study proceeded in the following steps: 1. SEDL data reflecting subjective perception were acquired from residents through a large-scale survey that yielded 516,804 valid questionnaires; 2. multispectral imagery derived from satellite remote sensing was employed to construct multiscale visible and invisible spectral features; 3. a deep learning model based on Vision Transformer (ViT) was designed and applied. By learning the complex, nonlinear relationship between multispectral, multiscale remote sensing data and survey-derived SEDL, this model effectively estimates SEDL in an end-to-end manner using remote sensing data. Experimental results demonstrate that the proposed model achieves strong performance on the test set (MAE = 0.28, R² = 0.63), effectively capturing the coupling between residents' auditory perceptions and remote sensing features, enabling the estimation of urban SEDL to be upscaled to all major cities across mainland China while maintaining a high spatial resolution. Analysis revealed the complementary importance of visible and nonvisible spectral bands in modeling sound environment perception, providing novel evidence at a national scale that remote sensing information can effectively characterize urban residents' auditory discomfort and uncovering inter-provincial differences in spectral contributions to SEDL. Leveraging these findings, we generated China's first residential-block-scale SEDL map covering all cities, offering critical guidance for urban planning, environmental management, and public health policy.
Chen et al. (Tue,) studied this question.