Urban Canopy Cover with AI
Understanding the Green in Cities Better with Artificial Intelligence
Looking across a city it is a great different whether you see green splashes, dots or just grey. Both to mitigate the impact of climate change in cities, yet also for wellbeing. Urban canopy cover can help mitigating the impact of increasing daytime summer temperatures. Reading up on the Climate Change AI workshop held at NeurIPS 2019 there was a paper called Quantifying Urban Canopy Cover with Deep Convolutional Neural Networks that I wanted to read.
Urban Tree Canopy (UTC) refers to the layer of tree leaves, branches, and stems that provide tree coverage of the ground when viewed from above.
An example of existing ways to measure this is usually with a degree of segmentation and satellite images (or data from areas). It could look something like this:
Physical models show that urban trees can significantly reduce the diurnal temperature range.
In meteorology, diurnal temperature variation is the variation between a high temperature and a low temperature that occurs during the same day.
So there you have it, a short explanation of some benefits, however what does the paper in question say?
Urban Canopy Cover with AI
The abstract of the paper reads as follow.
Urban canopy cover is important to mitigate the impact of climate change. Yet, existing quantification of urban greenery is either manual and not scalable, or use traditional computer vision methods that are inaccurate. We train deep convolutional neural networks (DCNNs) on datasets used for self-driving cars to estimate urban greenery instead, and find that our semantic segmentation and direct end-to-end estimation method are more accurate and scalable, reducing mean absolute error of estimating the Green View Index (GVI) metric from 10.1% to 4.67%. With the revised DCNN methods, the Treepedia project was able to scale and analyze canopy cover in 22 cities internationally, sparking interest and action in public policy and research fields.
As mentioned above tree canopy has a lot of benefits.
Their argument is that it can reduce peak temperatures on the hot days.
Other benefits they mention in the introduction is:
- Removal of air pollution.
- Increased perceived neighbourhood safety.
- Better visual and aesthetic appeal for residents.
The writers of the research paper argues that current methods to measure canopy cover is inadequate.
Traditional vs. New
They contrast traditional measurement.
Traditional methods: rely on either overhead imagery or in-person fieldwork.
As opposed to their method that takes images on-the-ground.
Dataset: We choose Cambridge (USA), Johannesburg (South Africa), Oslo (Norway), Seao Paulo (Brazil), and Singapore (Singapore) as cities included in our training and test sets. From each of the 5 cities, we randomly select 100 available Google Street View (GSV) images along street networks to form a training-validation-test set. We then divide the 500 image dataset into a 100 image test set, 320 image training set and a 80 image validation set. We produce manual labels by carefully tracing all vertical vegetation in the images for all 500 images.
In simple English: they find the trees, bushes, etc.
I found an interesting visualisation in their appendix too representing Boston and London.
I thought it was interesting to think about the potential of combining these different datasets for moderation to get a clearer picture of whether it is true or perhaps finding a way to more accurately map how green a city is. Each method has its limitations, but together and combined it may create a realisation of the true representation of the vegetation.
This is #500daysofAI and you are reading article 285. I am writing one new article about or related to artificial intelligence every day for 500 days. My current focus for 100 days 200–300 is national and international strategies for artificial intelligence. I have decided to spend the last 25 days of my AI strategy writing to focus on the climate crisis.