Open Fast Traffic Sound (OFTS)

This tool uses machine learning to rapidly estimate road traffic noise exposure indicators from open data, capturing both average levels and temporal variability of urban noise.

Purpose of Tool: 

Creating road traffic noise exposure maps is a time consuming endeavor. Gathering input data and setting up calculation models (e.g. EU-CNOSSOS) typical requires weeks to months of work. As it has been observed that the intermittent noise from residential streets is differently perceived than the noise from major roads, indicators that express the temporal variability of traffic sound beyond the equivalent level are added. Calculating such indicators requires accounting for the distribution of car passages which requires long calculation times. 

In Equal Life, a surrogate model based on machine learning was trained that estimates a variety of indicators based on local topography of buildings and streets (see figure). It is fast and only relies on open data. However, it can also be refined to take into account local traffic scenario’s and building organization. 

Classification of tool: model

Required skills: python, GIS

Required input data: 

• In standard configuration OFTS uses Open StreetMap data only 

• If available, traffic information can be added 

• If available, a buildings GIS layer can be added 

• A GIS layer containing the locations of interest 

Output:  

• If a location is inside a building, levels on façade in  

• For each building a set of noise indicators is calculated and stored as a GIS layer: 

• Lnight (most and least exposed façade) 

• Sleep disturbance index, SDI, for several age groups (most and least exposed façade) 

• L50, L95 during evening, and indicator for restorative moments (most and least exposed façade) 

• Len, evening-night level for different age groups (most and least exposed façade) 

Useful links: 

The code is open source and is available on Github: https://github.com/dbotteld/Equal-Life_noise; Equal-Life_noise: Horizon 2020 Equal-Life project data-driven and hybrid models for new metrics for noise exposure related to mental health (D3.6) 

Tutorial: You can find a tutorial here