2cultures.net(.au)

I now have three trajectory models of the Chelyabinsk meteoroid to share in Google Earth, from the three teams I am aware of who have published detailed results. The resulting KMZ file comes with a useful new feature: I’ve added geopositioned screenshots of the most useful videos, so you can now “fly” into each vantage point to check how the computed trajectories compare to the view in each video from the location it was taken. Here’s a quick tour on YouTube of the KMZ file:

Briefly, here’s what I’ve done: Initially, Google Earth allowed us to locate and measure YouTube videos to determine the angles of shadows, which enabled trajectory calculations, which have now been visualized in Google Earth. These trajectories can in turn be inspected visually from the vantage point of any number of geopositioned videos, resulting in an interesting additional method for verifying the accuracy of diverging trajectory calculations, short of going to these locations and making measurements in situ.

Before I do a walk-through of the trajectories and the videos, an exciting piece of news: A week or two ago, Jorge Zuluaga and Ignacio Ferrín — the duo calculating the trajectory of the meteoroid at the Physics Institute at the Univeristy of Antioquia in Medellín, Colombia — got in touch to discuss my original blog post on the use of Google Earth and YouTube as an ad-hoc sensor network for raw data on the meteoroid. They had taken my method and given it a rigorous mathematical make-over, aggregating information from four especially useful videos, including those contributed by commenters on Ogle Earth in the days after the event. From this trajectory, they calculated an orbit, concluding that the Chelyabinsk meteoroid was an Apollo-class Near Earth Asteroid.

In a classy move, Jorge and Ignacio asked me to co-author the paper they were writing, for having originally come up with the method which they then greatly improved upon. In the ensuing collaboration we identified additional useful videos, worked on the included Google Earth visualizations, and honed the prose. The paper, The orbit of the Chelyabinsk event impactor as reconstructed from amateur and public footage, has just been published to arxiv.org (here’s the PDF). The KMZ file discussed in this post (download and open in Google Earth) is the one attached to the paper. Note that the paper makes a point of thanking several commenters by name for their contributions to the original blog post. Citizen science FTW!

Now for some notes on the different trajectories:

Accuracy I: The NASA trajectory (in red) is derived from coordinates courtesy of Sebastien Pauleau, who calculated them from a close study of the trajectory map NASA released here, modeling the result on a WGS84 datum globe. He reports that while the accuracy of his model is calculated to 4 decimal points (an error of within 10m around Chelyabinsk, the limit of Google Earth’s accuracy with respect to the positioning of its imagery), it’s not possible right now to know how accurately NASA plotted its map. The Colombian trajectories (pink, green, black, orange), are also plotted to 4 decimal points, while the Czech results (blue) are plotted to 3 decimal points (within 100m), though of course the real error bars are a lot larger in at least all-but-one case (or else all the calculated trajectories would all have to lie within 100m of each other). The Colombian coordinates were shared directly; the Czech coordinates were published here.

Landing sites: We know that a good-sized chunk landed in Lake Chebarkul, and in the original coarse calculation I used that information as an input, fixing the lake as an endpoint for the trajectory. None of these more accurate “pro” trajectories make this assumption, and it is clear from subsequent news articles that the landing area extends beyond and around the lake. As a result, all calculated trajectories overshoot Lake Chebarkul, intersecting Earth just past the town of Miass.

Accuracy II: Because all these calculated trajectories are straight lines, there is an important caveat: From having looked closely at many geopositioned photos and videos, it seems clear that the real trajectory of the meteoroid changes as a result of the main explosion just south of Chelyabinsk. The post-explosion path seems to aim at bit more steeply at Earth, and may even have changed its azimuth (direction). Also, as the meteoroid slows down through atmospheric friction, it will begin to “fall” in a more classic arc. No one straight line can model such a more complex path with complete accuracy.

As a result, I think all current calculated trajectories overshoot the real landing site. I suspect most of the meteor mass landed between Lake Chebarkul and Miass, not beyond Miass. There are no videos from Miass showing a path flying overhead, though the calculated trajectories do assume such a path. One Miass video in particular (“Miass, near chimney” shows the contrail almost perfectly head-on, suggesting the the main part of the meteoroid landed in front of the viewpoint (towards Chebarkul).

Other viewpoints in Miass (“Miass, near plant” especially) suggest that some of the calculated trajectories do a better job of modeling the pre-explosion path, while others are more accurate for the latter part of the path. It’s important to note that around Miass, because the meteoroid was so close to Earth, very small differences in the calculated path can have a very large perceived effect, with changes of just a few hundred meters radically altering the perceived view.

From what I understand, it’s possible to construct even more accurate trajectory models that do not assume a single straight line, and I think this is where astronomers’ efforts will lie in the future.

Accuracy III: One word of caution about the geopositioned screenshots in the KMZ file: It’s not possible to accurately compensate for fish-eye effects and other distortions in Google Earth beyond basic field-of-view adjustments, and the videos do not always contain sufficient environmental references to precisely measure the heading, tilt and roll of the camera viewpoint. So these videos cannot be used for detail work, though they do work well when trajectories diverge greatly, as is the case near Miass.

Interact: The KMZ file is fully editable, so feel free to edit the embedded screenshots (Right-click an item in the Places sidebar, select Get Info) to see if you can get a better fit against the calculated trajectories. Although it is tempting, I tried not to align the meteoroid path in the video capture with the calculated trajectories in Google Earth, relying instead on clues from the surrounding environment. It’s possible to infer quite a lot from aligning the placement and angles of objects in the geopositioned video capture with imagery on the ground.

Interact II: Do play with the opacity slider at the bottom of the Places sidebar (click the gradient button, if you need to); this makes it much easier to make comparisons (watch the video above to see how). Finally, if you’re wondering how I managed to fly around so smoothly in Google Earth — I use this to navigate 3D space.

Like many others, I was absolutely astounded by the meteor strike over Chelyabinsk when I woke on Friday morning. One silver lining to our self-surveilling society is that an event of this magnitude is certain to get caught on the myriad of always-on dashboard- and web cams. I for one could not get enough of the videos.

Might it be possible to use today’s footage and Google Earth to have an initial go at mapping the meteorite’s trajectory? I was pondering this question some 2,500km away from Chelyabinsk when I chanced upon this video:

That place is easy to find — it’s Revolution Square at the absolute center of Chelyabinsk, looking almost directly south. It is also easy to measure — the distance between the two central light poles is 32 meters, as per a quick measurement in Google Earth, while the five lanes of traffic going right to left (west to east) measure 19 meters. From this it is easy to estimate the height of the light poles to be around 12 meters — an estimate corroborated by numerous panoramas in Google Earth showing people next to these lamp posts, giving us added data points.

Using all this information, I was able to do some image analysis in Photoshop on the lengths and angles of the shadows as the meteor streaked across the sky. Here’s an animated gif showing the result of that:

The ensuing grade-school mathematics (SOHCAHTOA!) resulted in three lines of sight at three instants a few seconds apart. (For the sake of the record, I roughly calculated them to be towards 122 degrees at an inclination of 33 degrees at 9:20:28.7, towards 187 degrees at an inclination of 40 degrees at 9:20:32.2, and towards 222 degrees at an inclination of 32 degrees at 9:20:33.4. These times are the video’s own timeline, though they appear to correlate closely with the timelines of other videos.)

That allowed me to draw a plane in Google Earth which should include the meteor’s path, though it does not allow me to know the distance of the meteor from Chelyabinsk or its speed.

However, we have more clues. We know a fragment of the meteor landed in Lake Chebarkul, roughly 70km WSW of Chelyabinsk. Gratifyingly, the plane generated from the above video roughly intersects with the crash site. Also useful was the estimate by the Russian Academy of Sciences that the meteorite hit the Earth’s atmosphere at around 50,000 km/h, shattering at an altitude of 30-50km. If that was the rough speed of the meteor as it burned up in the video, then the 4.7 seconds between the first and last shadow measurements would have seen it travel 65 km. Fitting a 65km line between these two lines of sight allows us to draw a straight line path for the meteor towards the crash site, with the first measured time yielding a height of 29km, which is the moment the meteor first brightened enough to give a clear shadow.

Download the visualizations for this as a KMZ file to open in Google Earth. Do play with the opacity slider of the overlay to check the alignments yourself — it’s most of the fun.

How does this data square with the Meteosat 9 image that has being doing the rounds? At first glance, not well: Overlaying the image in Google Earth and aligning the border with Kazakhstan shows a 240km contrail that appears to end some 75km to the ENE of Chelyabinsk, even though the path when traced on the ground also leads directly to Lake Chebarkul.

At first, I thought the image might have been taken 5 minutes earlier, before the meteor streaked straight across Chelyabinsk proper, because the image’s metadata gives us a time of 3:15:00Z, or UTC, which is 6 hours behind Chelyabinsk time. But no meteor is going to take 5 minutes to traverse 75km, so we’ll just have to live with the time discrepancy. Webcams are not atomic clocks.

Much more interesting is the fact that if you look at the position of Meteosat 9, which is in a geostationary orbit, you see that Chelyabinsk is near the horizon of its view of Earth. This leads to extreme foreshortening in the snapshot of the meteor’s contrail:

(Notice the outline of the Sea of Azov in the foreground. Here is another version showing the thermal impact.)

The version used in the overlay is an enhanced view of this image, taken from the same angle. (The blacked-out upper right-hand corner of the overlay is behind the horizon as seen from Meteosat 9.). If you simulate this view of Chelyabinsk in Google Earth, you see that in fact, the contrail aligns quite nicely over Chelyabinsk considering that it would be 30km high and at such an extreme angle over the horizon. So the 4.7 seconds of maximal brightness (with contrail) do get to happen just south of Chelyabinsk proper, as per the above video, and without contradiction by Meteosat 9.

I feel this post would not be complete without some big caveats: I am not a trained scientist; I don’t know if meteors travel through the atmosphere in straight lines or at constant speeds (I assume they don’t, but that it doesn’t matter for back-of-the-envelope type calculations). Still, it is satisfying to know that with judicious use of Google Earth, YouTube and Photoshop you can get quite far in the meteor simulation game. I can’t wait to see what the professionals come up with.

China is not the first place to see massive building projects wipe out organic urban geographies, but the scale and pace of the current transformation is unprecedented. Some recent projects have nabbed the headlines — the Olympics in Beijing, the Shanghai World Expo, the transmogrification of Kashgar — but hundreds of less well-known cities are seeing similar levels of change.

Every so often, the master plans of developers don’t align with those of an existing home owner. The negotiations, coaxing and intimidation that follow can reach absurd levels, while surrounding buildings and utilities are removed. The remaining “nail house” (钉子户, dīngzi hù) fittingly reifies the defiance of its owner, but in the end, the lone holdout almost never proves a match against the unstoppable forces of dialectical materialism.

Nail houses captivate — and not just because of the good visuals or the underdog empathies they arouse. They are a rip in the space-time continuum to a counterfactual world, where China’s vernacular architecture is ceded a place. For as long as they hold out, nail houses are waypoints to the past, standing askance against the all-too straight lines of futurism with Chinese characteristics.

I spent some time this past weekend hunting down some of the more famous nail houses, and collecting them into a KML file for download in Google Earth. Google Earth’s database of historical imagery provides a unique time-lapse tool for peering at China’s ongoing urban reinvention, and it’s allowed me to positively identify the locations of seven famous nail houses, including the most recent one in Wenling. In some cases, Google Earth has historical imagery showing the nail house in situ; in other cases, you only see the before– or after view.

When visiting each of these nail houses in Google Earth, play with the historical imagery timeline in the top toolbar to visualize the progression of the surrounding development over the past 10 years or so. The historical imagery is not available in Google Maps or on other mapping services, so you need to use Google Earth for this.

To whet your appetite, here are the nail houses in question, sometimes with associated geospatial context:

Chongqing, 2007 (population 29 million): This was the first nail house to gain international recognition. In Google Earth you can clearly see that until 2003 at least, the location was a very centrally located “snacking” street with food stalls.

Shenzhen, 2007 (population 10 million): Turn on 3D buildings to find this nail house underneath one of the world’s tallest buildings, the Kingkey 100, completed in 2011.

Guangzhou, 2007 (population 13 million): Guangzhou has seen some amazingly rapid development along the Pearl River this past decade:

Here’s the view in 2007:

Changsha, 2008 (population 2 million):

Beijing, 2010 (population 20 million):

Here’s how that particular place slowly turned into a nail house over the best part of a decade:

Kunming, 2010 (population 3 million):

And finally, Wenling, 2012 (population 1.4 million):

In Google Earth, notice the recently built railway station to the West, but also the copious fields through which the road could conceivably have meandered. The road bisecting the nail house is not yet visible in Google Earth, as the most recent imagery there is currently from 2010.

There are other nail houses out there, but in several cases Google Earth’s imagery hasn’t yet caught up with recent news. If you find any yourself, let me know and I’ll add them to the collection.

Much fun has been had at Apple’s expense this past week over the half-baked mapping product it released with iOS 6. The main benefactor has been the Google Maps team, whose painstaking work and years of experience are no longer in danger of being taken for granted. Publishing digital maps that “just work” is proving to be a lot more difficult than most realized, with The Atlantic and TechCrunch pitching in to tell that story.

As Internet serendipity dictates, the Apple Maps blowback meme soon crossed another meme — the flare-up in the ongoing territorial dispute over the Senkaku/Diaoyu islands, with China, Taiwan and Japan all claiming sovereignty. First, there was the (incorrect) rumor doing the rounds in Beijing that Apple Maps had taken sides:

RT @niubi: W beijing friends who say they will boycott iPhone 5 as the new apple map system puts the diaoyu islands in Japanese territor …
— Howard French (@hofrench) September 16, 2012

 

Upon closer inspection, it became apparent that while there are no labels or borders depicted near the islands, the islands themselves are rendered twice in the map view of Apple Maps — with a search for “Senkaku Islands” dropping a pin on one set of islands, and a search for “Diaoyu Islands” dropping a pin on the other:

When viewed in satellite view, the island labeled by the Diaoyu search disappears:

How does Google Earth/Maps stack up against Apple’s efforts?

First, there is clear evidence that Google also grappled with raw data containing a duplicate set of islands. Zoom in close enough in the satellite view and you see that the non-satellite seafloor dataset makes room for satellite imagery in a repeated, transposed pattern along a NE/SW axis, separated by the same margin of error that separates Apple Maps’ islands:

It’s fair to conclude that Google ran the seafloor masking algorithm before any human map error correction. In this case, the result is just some extra satellite imagery of island-less — but not completely empty — ocean: Just at the edge of the mask we see what I assume is a Japanese patrol vessel plowing full-steam ahead towards the islands, in an image captured on July 30, 2009.

In Google Maps’ map view, meanwhile, there is just one set of islands represented — the southwesterly one, as this is where georeferenced satellite imagery actually places the islands. This implies that at some point, a human was involved in noticing and correcting the duplication.

In Google Earth, turning on historical imagery removes the seafloor mask, so you can see the satellite image tiles in their entirety. One nearby tile alerts us to the presence of an outlying islet some 100km east of the main group; Google’s masking algorithm doesn’t quite pick it up. It is therefore invisible in the satellite view of Google Maps, and in Apple Maps too.

How might the duplicated islands have come about? Close inspection in Apple Maps (on an iPad — Apple’s maps are mobile only) reveals that the datasets are not identical; the contours of the islands are drawn differently. They thus come from different sources. And while all mapping data sources can contain one-off errors, officially sanctioned maps of China are all offset spatially by a variable amount in a ham-fisted attempt at thwarting GPS-based georeferencing by mere civilians.

My strong suspicion, then, is that the duplicate island belongs to an intentionally inaccurate Chinese dataset. There is some irony in that.

When it comes to labels, Google approaches the dispute in a typically Googley way: In Google Maps (though not in Google Earth, curiously), we get more information, not less:

The islands are named in both Japanese and Chinese. In Japanese, the islands are labeled 尖閣諸島 (Senkaku-shotō, or “Senkaku group of islands”); in Chinese, they are called 钓鱼诸岛 (Diàoyúqúndǎo, or “Diaoyu group of islands”). Individual island names are also labeled in both languages.

This approach to disputes is similar to Google’s dual labeling of the Persian/Arabian Gulf and the East Sea/Sea of Japan, though it is a policy that is at times applied haphazardly: For a similar dispute between Korea and Japan over the Liancourt rocks, Google Maps stays mute while Google Earth gets an expository popup label. And Google doesn’t currently label the Senkaku/Diaoyu islands with a Roman alphabet, which makes it difficult for most Google Maps users to comprehend, unless they have the third-party Panoramio photo layer turned on.)

Finally, a few more observations re Apple Maps that don’t quite fit anywhere else:

• For the user interface fanatics that Apple engineers clearly are, the two-fingered navigation of Apple Maps in 3D mode is far less intuitive that for the iPhone’s Google Earth app. The latter is far better at guessing whether I want to tilt the view, rotate or zoom.
• When using iOS 6 within China’s great firewall, Apple Maps’ satellite view only shows imagery of China, with the rest of the world blacked out:

  

  Image courtesy of The Next Web

  I’m intrigued by the technical “solution” behind this behavior: One likelihood is that Apple’s servers check the iOS device’s geolocated IP address before deciding which imagery dataset to send it. Google, in contrast, quarantines its country-specific datasets behind separate URLs (maps.google.co.in, maps.google.kr…), leaving the choice of dataset up to consumers, with governments free to block URLs from their citizens if they dare.

  As a corollary, the Chinese Maps dataset provided to Apple by Autonavi is not available globally, but only to users inside China. If you repeat Anthony Drendel’s search for “Lijiang Teachers College” outside China, you get this view:

  
  From left to right: Drendel’s map view of the teachers’ college inside China, my map view from outside China and my satellite view from outside China.

  it is possible that Apple’s solution does not require a conspiratorial collaboration with the Chinese authorities. As the logo in the bottom right of the screen shows, Drendel’s dataset is provided by Autonavi, which is a company that specializes in Chinese mapping data only. Maybe Apple simply hasn’t had the time or the legal license to properly mash together Autonavi’s dataset with Apple’s global dataset, including Tomtom’s. A lack of time would also explain the duplication of the Diaoyu/Senkaku islands.

If you are a web-based service provider and don’t have operations in a specific country, then that country’s laws cannot constrain the services you provide to users there. If Bahrain were to decide it doesn’t like Google’s maps, it could try to block the relevant URLs wholesale, but it has no legal recourse to compel Google to censor its maps. That’s because there are no Googlers, Google servers, or Google Street View cars in Bahrain.

But with Google’s global operations now having expanded to over 40 countries, far more countries are able to regulate the services Google provides locally. And when it comes to mapping and location-based services, some of them most definitely do.

Over the years, China and India have been the two main countries tracked on this blog for authoritarian tendencies when it comes to borders, names and third-party content. Because of the size of these two markets, Google has generally acquiesced to the legal restrictions governing locally published web maps, but has always made sure to contain the damage to the datasets aimed at these countries. Local users have always had recourse to the unmolested content served from outside the country — the reference content of maps.google.com and Google Earth.

Recently, it has become apparent that South Korea has also been placing constraints on what maps.google.co.kr is allowed to show. Because South Korea’s laws are easily accessible, this makes for an interesting case study in how Google tries to accommodate local laws in its local services without compromising the integrity of its global dataset.

The main way in which maps.google.co.kr differs from maps.google.com is the maximum zoom level available for the satellite view of Korea, which maxes out at a far lower resolution than for the global map. On maps.google.co.kr, the furthest you can zoom into Korea is this: (Click to go to these locations on the relevant maps)

On maps.google.com, you can zoom in a further three steps.

On Google Earth, of course, you can zoom in as far as the DigitalGlobe imagery will bear.

Why is Google’s Korean map behaving this way? In short, because of Korea’s Spatial Data Industry Promotion Act from 2009, specifically Article 7, which states that:

Spatial data business operators may produce and distribute any processed spatial data. In such cases, processed spatial data shall not include any spatial data on any military base provided for in subparagraph 1 of Article 2 of the Protection of Military Bases and Installations Act nor on any military installation provided for in subparagraph 2 of the said Article.

Article 2 of the Protection of Military Bases and Installations Act, in turn, explains that:

And considering the existence of the most heavily militarized border on the planet between North Korea and South Korea, this means a substantial part of South Korea is riddled with military installations.

By limiting the maximum resolution of its Korean imagery on maps.google.co.kr, Google appears to have satisfied Korean regulators that it is obeying the relevant Korean laws. Thus, Google avoids having to blur or otherwise censor the satellite imagery base layer for Korea — something which it has successfully managed to avoid in China, India and elsewhere.

When it comes to road and point-of-interest data, the datasets for maps.google.com and maps.google.co.kr appear to be identical translations, probably because the data is locally sourced. This would explain why large areas near the border remain unmapped, and military installations remain unlabeled. Many of these installations on both sides of the border are labeled in Google Earth by third-party contributors, however.

There is one final quirk to how Google delivers map information about Korea. When viewing satellite imagery in maps.google.com, you can zoom in on a highly accurate rendition of the Military Demarcation Line (MDL), which indicates exactly where the front was when the armistice agreement was signed in 1953.

A far more vague MDL is depicted on maps.google.co.kr and in the map view of maps.google.com, which is strange, considering that we can assume the North Koreans know exactly where the line is:

Google Earth also carries the vague version of the line, but additionally displaces it by around 500-1000 meters to the east.

I’m putting this down to a glitch, however. You can always load up on Korean War maps as a KMZ file for Google Earth.

How does Google’s Korean map compare to the local competitor, local.daum.net? When zooming in on satellite imagery with Daum, at a certain point the border region becomes completely blacked out, putting Google Maps at a relative advantage there.

For satellite imagery outside of the immediate border region, however, Daum lets you zoom in far closer than on Google’s Korean map, without any further apparent limitations, though the imagery is in all likelihood censored to obfuscate military installations, something which Google has always said it is unwilling to do. It’s depiction of the DML is equally vague as on  maps.google.co.kr

Daum’s web-based maps also offers historical satellite imagery of South Korea in its web-based map — something which Google only offers in Google Earth — and even historical Street View imagery of Korea, which is something Google doesn’t offer anywhere (yet?).

The big picture
It is apparent (to this non-lawyer) that in containing online censorship to within the legal jurisdiction of the country making the demand, Google (and Twitter and Tumblr) are trying to build a legal norm. Their solution may well satisfy the letter of the law, but it does not really satisfy the intent, as the censored content is always available uncensored on another domain controlled by Google. It would be ridiculous, for example, for sensitive sites in India to be redacted if they remain readily visible on Google Maps in Pakistan. In the same way, it is pointless for a British court to enforce a super-injunction by only demanding that Twitter remove an offending Tweet from the UK version of the Twitter stream, considering how easy Twitter makes it to circumvent the block. (I approve, of course.)

The risk is always there, though, that a country might decide to innovate its legal approach so as to punish Google or Twitter locally for a refusal by global HQ to resort to censorship globally. It’s not unthinkable, as this recent report on Extraterritorial Jurisdiction from Harvard’s Kennedy School of Government makes clear:

While states tend not to assert criminal jurisdiction over foreign companies directly, they do make use of a range of parent-based methods of regulation (i.e. domestic measures with extraterritorial implications) under which a parent company may be held criminally responsible for contributions to the crimes of others, including the foreign conduct of foreign actors, or failures to ensure compliance by those over which it is able to exercise control. [p.142]

There are many subtle ways a state can retaliate against local operations. Just ask Google’s Chinese operations. The one big deterrent would be the promise of a robust retaliation from the US government in terms of trade policy.