Google Straps Aclima Sensors to Street View Cars to Map Air Pollution | Hacker News Reader

Google Straps Aclima Sensors to Street View Cars to Map Air Pollution

techcrunch.com

79 pointsZweihander24 comments

This is going to be interesting for many cities, but I can speak to the case of Madrid.

The Partido Popular government fixed the pollution problem by moving the sensors out of busy streets and into big parks that happen to be within the city limits[1,2].

The historical continuity of measurements has been lost, but Google will now provide data with more continuity. Or at least I hope it is provided.

[1] http://elpais.com/diario/2009/05/18/madrid/1242645857_850215... [2] http://www.ft.com/cms/s/0/022cf59c-3091-11e0-9de3-00144feabd...

As someone spending a large amount of time in India I am very concerned about the extreme pollution level there. This was the best country-wide data I could find on pollution: http://imgur.com/sPtkwDj

I would love it if Google could produce something more accurate!

http://www.citylab.com/tech/2013/09/heres-where-youre-most-l...

This is the site I usually check:

http://aqicn.org/city/beijing/

Also, if I needed it, I'd probably build one of these:

http://smartairfilters.com/en.html

Sounds like something related to Sidewalk Labs[1]?

http://www.prnewswire.com/news-releases/dan-doctoroff-and-go...

This is great to see. We are prototyping something similar - want to deploy static sensors instead of strapping the sensors to vehicles. We initially thought about the vehicle option by partnering with Uber/Lyft. Definitely cost effective but getting relevant data from the same location at different time slices will become a function of the vehicle's presence at that location and time. A bit difficult to manage with a vehicle fleet whose primary function is something else. Worth further exploration though.

Neat but doesn't pollution vary significantly day to day and seasonally? It seems like one drive through at random time/day/season wouldn't be enough data to be meaningful.

They seem to understand that:

"In the first pilot, three Street View cars collected 150 million air quality data points over a month of driving around Denver, Colorado. They measured for chemicals that are hazardous to breathe, like nitrogen dioxide, nitric oxide, ozone, carbon monoxide, carbon dioxide, methane, black carbon, particulate matter, and Volatile Organic Compounds (VOCs)."

It seems to me that they can do repeated measures (coinciding with updating street view). I'm more interested in how this will account for the fact that the monitor is on a car which is itself a source of pollution, no?

Who that car is following and how long the spend sitting at red lights (or in traffic leaving a light) would likely have a larger effect. If you have enough data points you can reliably correct for it but there's still gonna be exception cases like the industrial area that's polluted during the day and the adjacent down that the wind blows the pollution to at night (Google doesn't drive street view cars at night AFAIK).

I don't know how much Google takes season into account when routing street view cars but areas with a tourism industry would be cleaner in the off season and colder areas will be dirtier in the winter (more fuel used for heating, less trees). Also something that could be accounted for...

Now we're getting to SimCity level of coolness.

Great. But I think Google will not show up where I live in a while so where can one buy these Aclima sensors?

Also I hope they expand measurements to other parameters like particle pollution.

Monitoring pollution is a taboo topic in our city because it would show that the port pollutes a lot and of course money is more important than public health.

The article mentioned they are measuring particulates. That is exactly what you're hoping they expand, so good news! They already will be measuring that.

I hope they find a way to add this to a portable device for end users. Here in DC I've suspected the air in the metro stations (underground train) is many times worse than the above ground sampling.

Possibly unrelated, but if you are interested in the geography of Chinese air pollution, check out aqicn.org

Why not just, I don't know, install a permanent sensor. This doesn't seem very useful unless these cars are driving the same route every day.

Because they are already driving anyway and this achieves extremely high spatial density compared with a few scattered immobile sensors (it isn't dense when viewed spatio-temporally, but there are plenty of uses and nothing stops you from also fusing data with static sensors).

Well, there are a few reasons here.

1. Installing a permanent sensor means ensuring some power source

2. Any kind of planning / permissions needed to sort it out

3. They're fixed

The first two simply aren't problems if you whack them on a car. No approval needed, no lengthy agreements about the look & impact of placing them around.

Three is really important. While the same money spent on fixed sensors gives greater time resolution, it'll give worse spatial resolution. Is it worth sticking a permanent sensor down to find out that actually the pollution levels never get high enough to be a problem? How many side-roads will get covered?

> This doesn't seem very useful

A huge amount of spatially dense readings across cities? That doesn't seem like it might be useful to you?

How often does a google car go past a location? I always assumed it was less than once a year based on how often they update street view where I live.

Plus, you have no idea what time of day the car will be passing through a location.

So you can't compare the data points over time - they're not taken the same time of day. And I'm not sure if you can even compare different regions - your data may tell you that one part of the city has cleaner air than another, but it's only because one part of the city was measured in the morning and another part of the city in the afternoon.

It's unlikely that this would be the last step in climate data collection---it feels more like pilot program, ironing out the practical issues with attaching such a sensor alongside a suite of existing sensors in the form-factor of a car.

... which might be very clever of Google, since they have a self-driving car program right around the corner that would promise to put O(10,000) of their cars on the road, driving continuously in densely-populated cities. Don't mistake the short game for the long game. ;)

> Plus, you have no idea what time of day the car will be passing through a location.

I don't, but I assume google does.

> So you can't compare the data points over time - they're not taken the same time of day

Not in a simple way, no. But to say it doesn't provide useful information seems odd when the article says

> Independent scientific analysis confirmed that the mobile sensor system worked for collecting street-by-street data, and could improve upon the regional network of sensors operated by the Environmental Protection Agency.

How do you read that quote as a positive thing? The first half merely says it was successful in collecting data and the second half uses weasel words "could improve".

I've worked on projects in the past where we collected a ton of data like this, gave it to some poor graduate student or post-doc to do a bunch of data wanking with it, only to produce throw-away papers in bad journals because even. Though it was a huge amount of data it wasn't good data.

Lots of people buy and install home weather systems, and hook up their data feeds to Weather Underground (or what have you). I'm not a climatologist/meteorologist, but my understanding is that personal weather stations provide a serious bump in data volume. Perhaps the future of pollution data looks something like that? Add a few pollution sensors to a personal weather station? Market a stand-alone pollution sensor with "health" justifications?

It would be good to see road signs displaying this information in real time. It should also be a required feature in cars to show excaclty what pollution they are emitting, along with common health effects. People need to be able to make an informed choice about driving.

Reply on news.ycombinator.com