Modelling Noise Pollution
There’s a lot of information available when using Property Finder’s app and website to search for the perfect property location, including photographs of the property, it’s location within a map of the surrounding area, and descriptions and characteristics of the home itself (the number of bedrooms, square footage, tower facilities, and more). However, there are also several intangible features of a property that are not easy to get from a property listing and that only become apparent during a viewing. For example, how much sun the living room gets; how heavy the traffic to your office in rush hour is; what the view from the kitchen is like; how noisy it is on the balcony.
For this last example, measuring noise pollution, we can now attempt to make the intangible tangible and ultimately, quantifiable. Using GIS techniques, we have the ability to cut through the noisy data, giving users access to noise pollution information to put in a final pitch-perfect offer.
Why is noise pollution important?
Noise pollution is an important consideration when choosing a property as it can have a drastic impact on your quality of life whilst occupying it, and ultimately, it can impact the price you would expect to pay. Whilst noise hazards are usually obvious during a viewing, this is not always the case. Road traffic levels can fluctuate during the day, mosques only sound prayers intermittently and construction sites spring up constantly in the UAE, often without warning.
How noise pollution is calculated:
3D model
A 3D surface has to be created in GIS in order to model the effect that physical barriers have on the direction and absorption of sound. Within our area of interest, this primarily takes the form of buildings.
Elevation data
The terrain elevation, whilst almost negligible in a desert, does have an impact. There still is around 35 metres difference between the lowest point in Dubai (0 metres, the shoreline) and the highest (35 metres, Akoyo Oxygen). Discounting this from the analysis could affect results and therefore the base of the model has to incorporate LiDAR elevation data — a grid surface of the bare-earth captured by satellite.
Noise data
Whilst we have no explicit noise data, we do have datasets from which noise-producing features and their associated volume can be inferred. For this proof-of-concept, three features have been used:
Roads — Derived from open source datasets. The amount of traffic and therefore the decibel level can be inferred from the ‘maximum speed’ values that come with the data.
Construction sites — Current developments in construction phase, taken from Property Finder database.
Mosques — Derived from open source datasets
Sound behavior tool
A tool was required to try and model the behavior of sound, mimicking the way it depreciates from source, and echoes off vertical surfaces such as walls. The analysis for this was conducted in python utilizing the open source libraries GDAL and GRASS.
No tool exists within desktop GIS designed to model the behavior of sound. Luckily however, there are some functions within the libraries designed to mimic the way light travels, that can be edited slightly and strung together to roughly mimic the behavior of sound.
The tool works by creating a line of site from all features that create noise, resulting in areas that can be seen from those features. The Euclidean distance from all features is then calculated in order to model the depreciation of sound over distance. These values are then clipped to the areas viewable from the feature, thus modelling the way that sound travels in a straight line and deprecates uniformly from source.
A feature of sound not shared by sight however is its ability to reflect off hard surfaces such as large concrete buildings (echo). This is simulated by running the line of site tool from all surfaces contacted by the original analysis. The Euclidian distance tool is run again, this time only for a distance on 20 metres, and the results added to the original surface. This secondary process imitates the true behavior of sound waves, and is important as it will model the way noise ‘bounces’ to a small extent around building corners.
The results
Figure 1 — Euclidean distance run from roads features
The results of this analysis offer a compelling visual product, and approximately model the real-world coverage and severity of noise pollution on to listings in the local area. This information could be provided to the consumer to allow them to assess the current, or even future, impact of sound on to their choice of property.
Figure 2 — Final surface visualized in 3d
Further study
Whilst this is useful for visualisation purposes, there is also further use for it. Each pixel within the surface created (the red/orange seen above) is actually indicating a value in decibels. This means that the values from the surface can be extracted and used as input variables for our data models e.g property pricing indexes. It can also be quantified against a value the consumer would be able to relate to (e.g. dark orange = 70–80 decibels: automobile interior or telephone bell).
If you liked this article and want to be part of our brilliant team of engineers that produced it, then have a look at our latest vacancies here.