Desenvolvimento de modelos preditivos para avaliação de paisagem sonora com base em levantamento na orla marítima da cidade de Maceió-AL

Abstract

In recent years, the field of soundscape studies has gained increasing prominence, driven by the need to transcend traditional approaches to urban environmental management that focus exclusively on noise control policies. Adopting a multidisciplinary perspective, the assessment of urban sound environments integrates both subjective variables, related to user perception and objective physical parameters inherent to the specific context and location. In this regard, the development of analytical tools capable of understanding and predicting perceptual responses to planning decisions through predictive modeling is essential. This doctoral thesis proposes a methodology for the evaluation of urban soundscapes based on the development of predictive models and their application in public spaces in the city of Maceió, Alagoas, Brazil, specifically along the seafront of the Ponta Verde and Pajuçara neighborhoods. The study adopts an applied research design, encompassing field data collection in accordance with the ISO 12913 normative series, including soundwalks, questionnaires, and acoustic measurements. A novel methodological approach is presented, involving the construction of a mixed questionnaire grounded in methods A and B of ISO/TS 12913-2, enriched with contextual and visual information. Furthermore, a procedure for acoustic environment characterization is proposed, aligned with ISO/TS 12913-3, and complemented by in-depth analyses of interrelations among perceptual and contextual factors. This is followed by correlation analyses and the identification of relevant indicators (empirical data) and descriptors (measures of perceptual response) for model development. The modeling framework comprises both linear predictive models (multiple linear regression) and nonlinear models (artificial neural networks – ANN). The ANN-based models demonstrated up to 35% greater predictive performance compared to linear models, with one model achieving a coefficient of determination (R²) of 0.98. The methodological advancements presented herein offer a significant contribution to the technological field of urban soundscape assessment, providing support for evidence-based decision-making by researchers and urban planners. This research expands the theoretical and practical horizons of soundscape studies by proposing innovative approaches for the characterization and predictive modeling of urban acoustic environments.