Visualization of 40M Cell Towers

Deck.gl-raster looks really cool. I haven't seen many attempts at doing client side raster processing.

I've been playing around with this map and a similar map of population density[1]. I wouldn't be surprised if this cell tower map was almost one-to-one match to a light pollution map: the contrast between North and South Korea is just as stark in the radio as in the visible!

But generally you're right, if you zoom out on the population density map, there simply isn't that level of distributed population Central Europe as anywhere in the United States. Europe is more similar to parts of India, China, and West Africa/Nigeria. It looks like the US most closely matches parts of Eastern Europe or Western Russia, but the US cities are substantially more suburban and "fuzzy". Aside from the stark difference between Kiev and Atlanta, in terms of population density the Eastern US seems most similar to Ukraine.

[1] https://luminocity3d.org/WorldPopDen/

And as a local, how sparse most of Australia is...

Comparing Sydney to San Francisco and London at the same zoom levels is fascinating (to me) and trying to find other areas on the globe as sparse as the dark bits of Australia is also fun...

Try the Scottish highlands. I think one of the differences is with Europe most of it is habitable, but North America has loads of desert and mountains etc.

Great site, nice to see where the local towers are. I wanted to know what tower my phone was connected and this app gave me the Cell ID (CID) https://play.google.com/store/apps/details?id=com.wilysis.ce... which matched the map. Super cool!

All I see is a blue void (tried two browsers) - do I have to click something?

@dang can we update the title? It seems to be misleading...

just summing the selection area around the cursor

40 million seems way, way too high for active towers. i doubt north texas area has 200,000 active cell towers.

For now not so much gps installed on Base Stations(towers), case only at CDMA Station GPS was strictly necessary. On UMTS or LTE it’s possible but rarely could be found because it’s required for infrequent used features. Needed time synchro received via transmission.

my guess is these are cell ids, not towers. no way north texas near dallas has 200k towers. also, no mention of active or inactive. 82% of lat,longs are distinct to 5digits (about tree resolution), which implies most towers have only a single cell id. 40mil seems way to high for active.

Why would cellular lose its "cell" meaning? The radios are still cellular which means a small geographic space in served by a particular directional physical antenna(s).

Your phone only needs to be able to hit the antenna in its cell. It doesn't need to talk to other phones or more distant antennas. This is what allows phones have have relatively low power radios that reside in your pocket without big external antennas.

This has already happened in LTE--everything is packet-based and any traditional "telephone" things are provided as a service (from a system call the IP Multimedia Subsystem) on top of that. IIRC, 3G networks were the last to make some distinction between packets and voice circuits.

I've yet to see any voice chat app work as well as cellular calls. Until then, it's still a phone with a computer attached.

Yes I agree. The most important thing about this data visualisation has surely got to be what actual data it’s visualising.

Looking at my local area. It definitely does NOT seem to be physical cell towers.

Some data points are plotted in the sea for one thing.

So it kind of doesn’t matter how fancy or well done it is if we can’t all agree on what it’s actually showing.

correct, data is based on sampling via devices.

I don't have the exact size comparison between this approach vs PNG compression, but i guess the difference would be similar to the difference between DEFLATE vs brotli.

I'm curious about what you just asked of me though, i will make the actual measurements, and will update this page with the results when i got time.

You are right, it's my mistake

Most of them are duplicate entries but with different levels of accuracy. So the same tower may have 10+ cell sites on it. So its represented 10 times. Multiple carriers in multiple technologies. Whenever they were mapped, possibly over years, the granularity level of GPS coordinates were different, so a lot of them are just in the general area.

it's based on where the single was picked up the final coordinates are calculated. Signal travels much further in open water therefore you can pickup very distant cells, skewing results.

You're correct! Nice find.

I experienced the same when I moved from China to US. In China the first time I ever experienced lost of signal was somewhere in the mountains in Tibet. Signal is just everywhere there.

US is incredibly bad on that. I live in the bay area and I can make an easy 30min drive to a place with no reception from any carriers.

I'm wondering why...

The national parks generally have no coverage - which is kind of perfect given what you go there for. Once you get off the East coast, there are even more no reception areas. In Wyoming and Idaho for example I had to drive my car some miles to the nearest town to get any reception since it was so valley-y and sparsely populated.

While the US does have a lot of no cell zones, those areas do have coverage. I wonder if your phone operated on GSM bands that didn't cover those areas. For a while in it was pretty common that if you wanted coverage in rural areas you opted for Verizon's CDMA network.

I think technology differences play a factor.

CDMA gets longer range than 3/4G, so it's useable in a sparse network so long as it's not trying to service too many people at the same time. I'm guessing the US launched lots of CDMA infrastructure back when it wasn't considered obsolete by most of the world, and hasn't upgraded it's rural cellular networks.

Australia set down all it's "legacy" CDMA stuff over a decade ago, which makes the large dark patches that cover lots of the continent very under-served.

Europe is much more "dense" and many obstacles therefore requiring more towers to cover the same surface area.

Population density is significantly higher in Europe than in the US.

You can serve only so many phones and handle only so much bandwidth from a single cell tower. So they reduce cell sizes and install more towers to keep devices per cell down.

The points you see are wrong, but they do have cell towers in the sea (often supplied by underwater cables). In this part of the world they can be found usually around the North Sea on or for the oil rigs mainly. However the majority of these points even on land are quite inaccurate. They are captured by simple triangulation - wardriving.

This is no cell towers exactly, this is the measurements related to dedicated cellID(sector if BaseStation(tower)). So such kind if issues is possible especially at water(rf signal propagation feature) and gps mapping if present.

I found something similar, finding a number of cell towers in a lake I'm pretty sure no one would ever put a tower in (it's about 10 feed deep).

I saw that too in, for example, Lower New York Harbor.

total generated data set is 144mb,

number of tiles that has smaller size than 100bytes is : 321796

number of tiles that has size between 100b, 1kb is : 13693

number of tiles that has size between 1kb, 5kb is : 8486

number of tiles that has size between 5kb, 50kb is : 5275

number of tiles that has bigger size than 50kb is : 279

empty tiles occupies 32 byte each and avg size for non empty tiles is ~4kb

OpenCellID shows an LTE point in an agricultural field near me. There definitely isn't a cell tower there in the traditional sense of cell tower. There is an Emergency Call Box located there so maybe that is what is getting picked up?

Added the missing OSM attribution

This is just a map of cell towers. Many people with cell service in their area might not have access to a broadband internet service. Internet access via cellular doesn’t really compare in terms of costs and speed/latency. (At least in the US, cost per GB over cellular is way too high)

The potential user base is massive once you move away from the urban centers in the developed world.

Take a look at Lagos, Nigeria. Population is ~15m for ~80k towers. Only 1.4k (1.7%) of them are 4G LTE with the remaining either 3G or CDMA.

Or the State of Sao Paulo, in Brazil (home of 22% of Brazil's population and 33% of the country's GDP). Approximately 573k towers, 76k of which are 4G

For comparison, the greater Boston area has ~107k towers and 58k are 4G.

I think they're suggesting something much more simple than that:

SELECT lat, lon FROM tower_list WHERE tower_id=${The one you're connected to};

I think the is rather that there's no pre rendering of the tiles serverside, even dynamically. Its all client side. Ie, updates should just require a dataset update, not a regeneration of map tiles for a tile store in the case of pre rendering. And no need for a spatial server.

There's pockets in more places. A handful of countries had fixed wireless CDMA for home phones where it was difficult to install wires, and many of them have opened up regulations for those carriers to do mobile CDMA as well.

Mexico (e.g., Guadalajara), the Caribbean (particularly adjacent to Puerto Rico), and northern Colombia and Venezuela seem to have pockets of CDMA as well.

On the map, there appear to be small handfuls in Canada, even though Bell and Telus shutdown those networks in 2019 and 2017, respectively

Indonesia still have a bunch of CDMA towers listed, but IIRC the spectrum was auctioned off a while ago.

We shut down the last of the CDMA here in Australia back in about 2008.

Qualcomm's CDMA (CDMA2000).

Almost all 5G deployals around the world are 5G non-standalone; they have 4G LTE basestations doing handset control and 5G modulations (on their own frequency span) for data transfer.

Unrelated, this can't be a map of cell phone towers. It's probably a map of cell phone basestation locations.

no, they're just very few of them currently in the database.

I was browsing the forum of the data source -- looks like there are some users expressing that the China data is significantly out of date. I would bet that the applications which collect this data are not common on the other side of the firewall.

I think it may be related to data source. As a Chinese I can say signal is basically everywhere, far better than US.

Reminds me of the inverse of this map:

https://news.ycombinator.com/item?id=25825390

This probably counts cells (i.e. a cell will be a service from an antenna facing in a specific direction broadcasting on a specific frequency) rather than towers. One tower can operate several cells--especially true for 3G and 4G where many carriers now do carrier aggregation across more than one frequency (and thus more than one cell). My local tower for one operator alone has 12x 3G cells (3 antennas facing N/SE/SW which each carries 2x 900 MHz carriers and 2x 2100 MHz carriers). For 4G the same 3 antennas carries 700 MHz, 1800 MHz, 2100 MHz and 2600 MHz services (for a total of 12 cells). So it adds up pretty quickly.

"While 98 percent of households had a television and 69 percent had a mobile phone, only 19 percent had a computer."

https://www.reuters.com/article/us-northkorea-unicef/tacklin...

It makes sense that the towers are installed where people are, but note that this only holds true for first world nations. Africa has plenty of underserved areas. Or go and see if you can find the border between North and South Korea. You can also see the border between India and China quite clearly.