Dear Amit,

Good morning. All excellent suggestions and glad you summarized them here.
These are in default use -- built-up area, LULC fields, road network,
population density, commercial activity density, and road activity points
density (like stops).
https://urbanemissions.info/india-air-quality/india-ncap-cities/

We have a template tool for playing with these combinations.
https://urbanemissions.info/tools -- look for the gridding tool.
This has been the default setup for a long time and the process works (not
just in India, but anywhere).

Question is not if we can grid the emissions. We want a new layer --
vehicle density from the images at the grid level -- a new proxy layer.
Given the buzz around AI and Satellites, this is an open question to the
community -- is this something we can make for Indian cities? If yes, how.

With best wishes,
Sarath

--
*Dr. Sarath Guttikunda*

*https://urbanemissions.info <https://urbanemissions.info/about-ueinfo/>*


On Mon, Jul 7, 2025 at 11:01 PM Amit Shirsath <[email protected]>
wrote:

> Hi Dr. Sarath,
>
> There is another approach, though not very accurate and too complex.
>
>    - There is Vahan data available at RTO level. These RTO offices
>    usually serve the whole district or municipal corporation or a few taluka,
>    etc. GIS boundaries for these entities are generally available online. So,
>    RTO can be mapped to a geographical boundary.
>    - Also, there are .tiff images available for all India on Bhuvan.
>    Using Zonal Histogram with these images, one can derive Land Use Land Cover
>    (LULC) for all India. I had made LULC at 100m x 100m grids for all India a
>    year back. So, 1 Sqkm is completely feasible.
>    - Once LULC is available, we have to identify areas where high density
>    of "Builtup Cover" is available. In theory, it is the area where some kind
>    of construction is there like buildings, bungalows, informal settlements,
>    etc. There can be different approaches to identify this, as classes
>    available with Bhuvan files may not directly correspond to what we want.
>    - I am making the assumption that the "Builtup Cover" areas will have
>    vehicle ownership or usage. This assumption has to be refined further based
>    on your requirement.
>    - Once RTO boundary and "Builtup Cover" is available, the total
>    vehicle sales can be distributed to "Builtup Cover" area proportionally.
>    - So, finally you get the number of vehicles per grid.
>    - One advantage of using Vahan data, is that the vehicles are
>    categorised by Vehicle Class, category, Fuel, Norm (BS IV, BS III, etc.)
>    and Maker. So, you can assign weight
>
>
> *Notes:*
>
>    - The vehicle classifications on Vahan portal may not be available for
>    all the years.
>    - Here, we are assuming that the vehicle purchased within an RTO
>    boundary is operating within the same RTO, which is not true in many cases.
>    For example, Nowadays, I am observing many high end consumer vehicles with
>    "DD" registration numbers operating in Mumbai. Also, some large commercial
>    vehicles like tourist buses with "NL" registration numbers operate on
>    Mumbai, Pune, Bangalore routes.
>    - However, If this type of migration percentage is lower compared to
>    the total number of vehicles sold in the last 15 years, then this method
>    can be useful.
>
>
>
>
>
>
> On Fri, Jul 4, 2025 at 5:00 PM Sarath Guttikunda <[email protected]>
> wrote:
>
>> Dear Anupam,
>>
>> Thanks for the response. The main bottleneck in this process is having
>> access to high-resolution imagery. The codes are all open and many groups
>> seem to be using them in places other than India. If you come across any
>> useful applications, we would be open to exploring those opportunities.
>>
>> On the second point, yes, GPS-based data would be much more useful, but
>> again, that's not a free dataset. The Google Distance API actually uses
>> cell signals to gather congestion-related information, but it's not a free
>> dataset, and over the last three or four years, they've really increased
>> the price tag for using that API.
>>
>> In our emissions distribution step, having granular information is much
>> more useful -- granular in space and time. Using an image provides a static
>> dataset on where vehicles are on the road and also where parking lots are
>> (space dimension). This, again, is a proxy that could align with just road
>> density, which is the theory we've been using for the longest time. If we
>> can add another layer of vehicle density to it, the spatial analysis of
>> emissions at higher resolutions, especially in big cities, will improve,
>> along with some of our other analyses.
>>
>> Hoping others on the list can add to this discussion.
>>
>> With best wishes,
>> Sarath
>>
>> --
>> *Dr. Sarath Guttikunda*
>>
>> *https://urbanemissions.info <https://urbanemissions.info/about-ueinfo/>*
>>
>>
>> On Tue, Jul 1, 2025 at 9:35 PM Anupam Sobti <[email protected]>
>> wrote:
>>
>>> The constraint here might be availability of usable high resolution
>>> imagery. I believe Google prohibits use of its imagery and the imagery that
>>> these tutorials use are not available in India. Some people use night time
>>> lights overlayed on street networks as a proxy for vehicle density but I
>>> think GPS based google data would be more reliable.
>>>
>>> Regards
>>> Anupam
>>>
>>> On Wednesday, June 18, 2025 at 12:18:08 PM UTC+5:30 Sarath Guttikunda
>>> wrote:
>>>
>>>> In our group, we frequently perform extensive numerical analysis,
>>>> particularly to understand emission intensities, both spatially and
>>>> temporally. The spatial understanding of emissions is a significant
>>>> component of our work, as it is crucial for accurately placing emissions
>>>> before they are modeled and concentration maps are generated.
>>>>
>>>> This process is technically known as the "gridding of emissions." For
>>>> example, if we know there are one thousand trucks operating in a city, each
>>>> traveling one hundred kilometers a day, we can multiply these figures by an
>>>> emission factor for a specific pollutant to determine the total emission
>>>> intensity of trucks moving within the city's airshed. The question then
>>>> becomes: how do we distribute these emissions into various grids for a
>>>> city? We typically work with one-square-kilometer grids on average, and you
>>>> can see some examples below.
>>>> https://urbanemissions.info/india-air-quality/india-ncap-city-airsheds/
>>>>
>>>> One of the proxies we use for trucks is highways. The assumption is
>>>> that most trucks will travel on highways and spend the majority of their
>>>> time there. Therefore, we assign a higher weight to the grids that
>>>> intersect with highways. We also incorporate other layers of information
>>>> with additional weights. For instance, industrial hubs, commercial hubs,
>>>> malls, and markets are places where these vehicles are likely to go and
>>>> spend some time. This methodical approach generates various weighting
>>>> functions, and once we have the emission intensities, it produces a gridded
>>>> emission file. So far this method of madness works and we have a good
>>>> understanding of how the layers are behaving with some plus minus. We have
>>>> an example tool to play with this method --
>>>> https://urbanemissions.info/tools/
>>>>
>>>> We aim to improve this process. One of the layers we introduced in the
>>>> past was speed information from the Google Maps API. We can download speed
>>>> data, which also indicates congestion times. We utilized this as another
>>>> proxy to understand where and for how long vehicles spend time, and
>>>> accordingly, assign weights. See example image for Mumbai here -
>>>> https://urbanemissions.info/india-apna/mumbai-india/
>>>>
>>>> A new approach we want to explore, given some recently available
>>>> information (and algorithms), is vehicle density. This would again be
>>>> a static input. For example, if you take a satellite image and apply an
>>>> algorithm, you could determine how many vehicles are visible within each
>>>> grid. Because this is a static image for a specific time, we cannot use it
>>>> as a layer for all-purpose gridding. However, it would serve as an
>>>> additional layer of information that accurately reflects what is happening
>>>> on the ground. It could also be used to extract information about official
>>>> and unofficial parking lots where vehicles spend a significant amount of
>>>> time on a given day. This would allow us to extract valuable insights.
>>>>
>>>> There are many online examples of this being done using geostationary
>>>> images in Europe and the United States and most of them require an image
>>>> and rest seem to work (take it with a pinch of salt -- non-it-person
>>>> speaking).
>>>>
>>>> https://www.linkedin.com/posts/giswqs_geoai-geospatial-ai-activity-7309216621997281280-46yC/
>>>>
>>>> https://up42.com/marketplace/analytics/detection-cars-oi
>>>> A commercial portal -- but seems to do exactly what we want at a price
>>>>
>>>> So, the question to the group today is this: If there is a grid file,
>>>> let us say for Bangalore -- has anyone done anything similar to create
>>>> a vehicle density map, regardless of the vehicle type? or have any
>>>> ideas on how to approach this for Indian cities?
>>>>
>>>> Please keep in mind that the ultimate goal is not to identify
>>>> individual vehicles or count vehicles from traffic cameras. The focus
>>>> is on a static image: if we have one, can we, or has anyone, worked on
>>>> creating a vehicle density map from it?
>>>>
>>>> Any sights into making an example and hopefully scaling it up is
>>>> appreciated.
>>>>
>>>> Looking forward to the follow ups.
>>>>
>>>> With best wishes,
>>>> Sarath
>>>>
>>>> --
>>>> *Dr. Sarath Guttikunda*
>>>>
>>>> *https://urbanemissions.info
>>>> <https://urbanemissions.info/about-ueinfo/>*
>>>>
>>> --
>>> Datameet is a community of Data Science enthusiasts in India. Know more
>>> about us by visiting http://datameet.org
>>> ---
>>> You received this message because you are subscribed to the Google
>>> Groups "datameet" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to [email protected].
>>> To view this discussion visit
>>> https://groups.google.com/d/msgid/datameet/c0afb013-c9b2-4f13-8b92-570b3d79b916n%40googlegroups.com
>>> <https://groups.google.com/d/msgid/datameet/c0afb013-c9b2-4f13-8b92-570b3d79b916n%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>> --
>> Datameet is a community of Data Science enthusiasts in India. Know more
>> about us by visiting http://datameet.org
>> ---
>> You received this message because you are subscribed to the Google Groups
>> "datameet" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> To view this discussion visit
>> https://groups.google.com/d/msgid/datameet/CAAj%2BwWo%3DT5siPy4red1eLVexAMoeKPE3V3d69WrLAt8tN9%2BQow%40mail.gmail.com
>> <https://groups.google.com/d/msgid/datameet/CAAj%2BwWo%3DT5siPy4red1eLVexAMoeKPE3V3d69WrLAt8tN9%2BQow%40mail.gmail.com?utm_medium=email&utm_source=footer>
>> .
>>
>
>
> --
> Regards,
> Amit Shirsath
> +91 9970944537
> www.amitshirsath.com
>
> It's easier to invent the future than predict it. -- Alan Kay
>
> --
> Datameet is a community of Data Science enthusiasts in India. Know more
> about us by visiting http://datameet.org
> ---
> You received this message because you are subscribed to the Google Groups
> "datameet" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To view this discussion visit
> https://groups.google.com/d/msgid/datameet/CALVzC30LGr84vHm4ZU4Cz%3D3HQRWcnM_TX9z0STu%3DvWbTzzL6tg%40mail.gmail.com
> <https://groups.google.com/d/msgid/datameet/CALVzC30LGr84vHm4ZU4Cz%3D3HQRWcnM_TX9z0STu%3DvWbTzzL6tg%40mail.gmail.com?utm_medium=email&utm_source=footer>
> .
>

-- 
Datameet is a community of Data Science enthusiasts in India. Know more about 
us by visiting http://datameet.org
--- 
You received this message because you are subscribed to the Google Groups 
"datameet" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/d/msgid/datameet/CAAj%2BwWrUDJqEA7O7LrRbbj6dugTCui0z-dQLstT1o8rNwyQ5ng%40mail.gmail.com.

Reply via email to