6 ms·

The island he’s singled out was flown over within days of her disappearance and the pilot did not see any signs of an airplane. It would be wonderful to finall

The island he’s singled out was flown over within days of her disappearance and the pilot did not see any signs of an airplane. It would be wonderful to finally solve this mystery, but it’s not as though this idea is some recent breakthrough. It was thought of at the time, which is why the pilot was sent to look.

secabeen 7y ago
The idea they're having, which is hidden in the article, is that she crashed in the shallow water around the island, and the wreckage slipped down the underwater slope, out of sight. At a later time, landing gear was pictured on the beach, broken loose from the rest of the debris, still underwater.
Etheryte 7y ago
The article addresses this point, in case you missed it. > According to the official report, a search pilot saw “signs of recent habitation” there. But because nobody waved them down, the search team left and the Navy dismissed the theory. What the sailors didn’t know was that the island had been uninhabited for 40 years. > Earhart landed on the very edge of the island, Dr. Ballard believes. As tides rose, her plane may have slipped down the underwater slope.
- mannykannot 7y ago
  > What the sailors didn’t know was that the island had been uninhabited for 40 years. It seems that it was occupied for several days by the 24 survivors of the shipwrecked SS Norwich City in November 1929, who were forced to abandon the wreck as it caught fire. I believe it is the ship in the photograph. https://en.wikipedia.org/wiki/Nikumaroro#SS_Norwich_City_wreck https://en.wikipedia.org/wiki/Nikumaroro#SS_Norwich_City_wre...
thaumaturgy 7y ago
I was on a SAR mission a couple of years ago for a lost aircraft outside of Truckee, CA. In that case, there was flight radar data available for most of the flight path, Civil Air Patrol was out in force with multiple overhead flyovers, CHP and Air National Guard were also out, we were receiving intermittent ELT signals, there were hundreds of ground searchers and UTVs/ATVs/snowmobiles, and all of our search activity was recorded and coordinated through a well-organized incident command with specific search assignments for each day. We still failed to locate the aircraft. It was found a month or so later by some random recreationalists. Any number of circumstances may have prevented it from being spotted the first time. It's far, far harder to locate missing aircraft than most people realize.
- chx 7y ago
  Just a thought exercise to show why this is so hard: let's say you know where the aircraft was a minute before crashing but not the direction of it. If an aircraft is flying at a modest 200 km/h (that's a Cessna 172 cruising) then in a single minute it can fly more than 3km. The area of a circle with a radius of that is close to 35km^2. That's half of Manhattan Island.
  - pacala 7y ago
    How difficult would be to fly one hd filming plane above, then scour the imagery with a fine comb, possibly computer vision aided?
    - wil421 7y ago
      You would need different angles especially if there was tree cover or mountainous terrain.
    - thaumaturgy 7y ago
      Very. A similar tactic was used when Steve Fossett disappeared (https://en.wikipedia.org/wiki/Steve_Fossett#Disappearance_and_search https://en.wikipedia.org/wiki/Steve_Fossett#Disappearance_an...), and it was entirely unsuccessful. There are a few challenges: 1. There is a direct relationship between the time and effort required to take imagery, and the image resolution. Which is to say, if you want imagery that's detailed enough that you can find likely evidence without a lot of false positives, then you have to spend a ton of time in the air at low altitude, and if you want to effectively search a large area, then you have to spend time at higher altitude where you're too high up to get useful imagery. 2. Humans aren't good at sorting out natural features from unnatural ones. My favorite example of this is the face on Mars. Same deal when trying to search through imagery for debris. People have a preconception of what a plane crash looks like, and it's usually wrong. They'll confuse light-colored rocks, trees broken by snow loading, and all kinds of other things for evidence, while completely missing actual evidence. So the problem with these systems for search managers is that once someone flags something in a picture, you're kinda obligated to check on it, which takes resources away from more likely search areas. 3. The technology isn't there (yet). Computer vision may get better than humans at this soon, but not yet, and certainly not in the kind of technology that even well-funded search organizations or adventurers has access to. NSA? Maybe. Us? Nah, best we get is grainy photographs of features 1m across from high up. IR imagery currently has similar problems, there's just too much background noise in most cases. Some kind of multi-spectrum imagery fed into a big convolutional NN might be able to do a better job of telling the difference between a rock and some twisted metal and plastic composites at 1m resolution, but I don't know if anybody's actually invented that yet. I'm using 1m resolution here as a substitute value for whatever would be appropriate to cover a large area with reasonable resources. After the Camp Fire, Alameda County brought a really cool tech team out that covered large areas of the burn with imagery from drones. They got a 1cm aerial resolution that was good enough to satisfy insurance investigators. But, it required a lot of flight time and computing power to do that, and they covered a relatively small area compared to the search area for a missing aircraft.

secabeen 7y ago

The idea they're having, which is hidden in the article, is that she crashed in the shallow water around the island, and the wreckage slipped down the underwater slope, out of sight. At a later time, landing gear was pictured on the beach, broken loose from the rest of the debris, still underwater.

Etheryte 7y ago

The article addresses this point, in case you missed it. > According to the official report, a search pilot saw “signs of recent habitation” there. But because nobody waved them down, the search team left and the Navy dismissed the theory. What the sailors didn’t know was that the island had been uninhabited for 40 years. > Earhart landed on the very edge of the island, Dr. Ballard believes. As tides rose, her plane may have slipped down the underwater slope.

mannykannot 7y ago
> What the sailors didn’t know was that the island had been uninhabited for 40 years. It seems that it was occupied for several days by the 24 survivors of the shipwrecked SS Norwich City in November 1929, who were forced to abandon the wreck as it caught fire. I believe it is the ship in the photograph. https://en.wikipedia.org/wiki/Nikumaroro#SS_Norwich_City_wreck https://en.wikipedia.org/wiki/Nikumaroro#SS_Norwich_City_wre...

thaumaturgy 7y ago

I was on a SAR mission a couple of years ago for a lost aircraft outside of Truckee, CA. In that case, there was flight radar data available for most of the flight path, Civil Air Patrol was out in force with multiple overhead flyovers, CHP and Air National Guard were also out, we were receiving intermittent ELT signals, there were hundreds of ground searchers and UTVs/ATVs/snowmobiles, and all of our search activity was recorded and coordinated through a well-organized incident command with specific search assignments for each day. We still failed to locate the aircraft. It was found a month or so later by some random recreationalists. Any number of circumstances may have prevented it from being spotted the first time. It's far, far harder to locate missing aircraft than most people realize.

chx 7y ago
Just a thought exercise to show why this is so hard: let's say you know where the aircraft was a minute before crashing but not the direction of it. If an aircraft is flying at a modest 200 km/h (that's a Cessna 172 cruising) then in a single minute it can fly more than 3km. The area of a circle with a radius of that is close to 35km^2. That's half of Manhattan Island.
- pacala 7y ago
  How difficult would be to fly one hd filming plane above, then scour the imagery with a fine comb, possibly computer vision aided?
  - wil421 7y ago
    You would need different angles especially if there was tree cover or mountainous terrain.
  - thaumaturgy 7y ago
    Very. A similar tactic was used when Steve Fossett disappeared (https://en.wikipedia.org/wiki/Steve_Fossett#Disappearance_and_search https://en.wikipedia.org/wiki/Steve_Fossett#Disappearance_an...), and it was entirely unsuccessful. There are a few challenges: 1. There is a direct relationship between the time and effort required to take imagery, and the image resolution. Which is to say, if you want imagery that's detailed enough that you can find likely evidence without a lot of false positives, then you have to spend a ton of time in the air at low altitude, and if you want to effectively search a large area, then you have to spend time at higher altitude where you're too high up to get useful imagery. 2. Humans aren't good at sorting out natural features from unnatural ones. My favorite example of this is the face on Mars. Same deal when trying to search through imagery for debris. People have a preconception of what a plane crash looks like, and it's usually wrong. They'll confuse light-colored rocks, trees broken by snow loading, and all kinds of other things for evidence, while completely missing actual evidence. So the problem with these systems for search managers is that once someone flags something in a picture, you're kinda obligated to check on it, which takes resources away from more likely search areas. 3. The technology isn't there (yet). Computer vision may get better than humans at this soon, but not yet, and certainly not in the kind of technology that even well-funded search organizations or adventurers has access to. NSA? Maybe. Us? Nah, best we get is grainy photographs of features 1m across from high up. IR imagery currently has similar problems, there's just too much background noise in most cases. Some kind of multi-spectrum imagery fed into a big convolutional NN might be able to do a better job of telling the difference between a rock and some twisted metal and plastic composites at 1m resolution, but I don't know if anybody's actually invented that yet. I'm using 1m resolution here as a substitute value for whatever would be appropriate to cover a large area with reasonable resources. After the Camp Fire, Alameda County brought a really cool tech team out that covered large areas of the burn with imagery from drones. They got a 1cm aerial resolution that was good enough to satisfy insurance investigators. But, it required a lot of flight time and computing power to do that, and they covered a relatively small area compared to the search area for a missing aircraft.