Saturday, July 30, 2016
Monday, July 4, 2016
MH370 Debris Analytics ("blue collar" version)
Elegant drift analytics have been performed by Brock McEwen, independent contributor, and Richard Godfrey of the IG.
McEwen's analytics can be found here:
Brock Study
Godrey's analytics can be found here:
Richard Study
This effort is not intended to replace or supplement the efforts linked above. It is simply another way to look at the problem. I would characterize it as being derived from common sense and complete. Complete in this context means everything is included below. Unlike the studies linked above where the heavy lifting from an analytical perspective is "under the covers", and you are presented with graphical or tabular output which cannot be checked (not meaning to imply that there is reason to believe checking is needed).
A table of the debris found as of this date (07/04/16) is presented below. Not all of these finds have been officially confirmed, but they are all compelling relative to being from 9M-MRO. No distinction is made by debris size, whether it came from the right or left side of the aircraft, or whether it was an internal or external part.
A LON - LAT scatter plot of these finds is as shown below.
Map view of debris finds below.
The finds show a significant clustering around latitude -20 to -26. Outliers are the Mossel Bay find and the most recent Pemba Island find.
Shown below is the Geomar reverse drift plot using "only" the flaperon finding on Reunion Island. Reunion Island is at -21 latitude, and is representative of the latitude of the debris clusters. I have no reason to doubt the validity of the Geomar model, and it saves the time and tedium of watching which way rubber ducks travel across the ocean.
It is clear from the Geomar graphic the preferred origin of the Reunion find is along the coast of Sumatra/Java. It can be inferred that other finds in the vicinity of latitude -20 to -26 would have a similar origin. It is reasonable to conclude, based on the Geomar model, that the 9M-MRO terminus lies between latitude -10 and -20 on or near the 7th arc.
The debris data lends absolutely no support to the current area being searched by the ATSB. The fact that no debris has been found on the coast of Western Australia further supports the notion that the current search area is not correct.
Edits July 7, 2106 - DW
Since posting the above on the JW blog several comments have been made by people with relatively little technical background or a flawed background that "beg" for additional commentary here.
First is the notion of forward and reverse drift calculation. All calculations are by definition reverse calculations since the objective of the calculation is the origin of the debris not the terminal locations of the debris which comprise the set of knowns. Now it is possible to use algorithms which reverse the forcing functions to arrive at a hypothetical starting point. It is also possible to use forcing functions that operate in the forward direction and use them act on debris at a variety of starting points and subsequently cataloging the results.
Second is the notion that additional debris finds add confusion and difficulty to drift calculations, especially in the reverse direction. The reality is that more data provides more constraining information that actually serves to "focus" possible starting points. More information is always better, that is an axiom of information theory as well as Kalman filtering.
Third is the notion that the new Geomar study should be used, and the above study to be discarded. I deliberately used the above study because it addresses the debris clusters at ~20S directly. No one, including Geomar, has indicated that the above study is flawed. Debris found at Pemba Island and Mossel Bay can only narrow the findings of the above study, not broaden them. A common metric in navigation mathematics is the notion of "dilution of precision". A narrow string of debris location latitudes significantly dilutes the precision that could be obtained if the debris were more scattered with respect to latitude. This statement applies particularly to the North-South precision of the starting point.
McEwen's analytics can be found here:
Brock Study
Godrey's analytics can be found here:
Richard Study
This effort is not intended to replace or supplement the efforts linked above. It is simply another way to look at the problem. I would characterize it as being derived from common sense and complete. Complete in this context means everything is included below. Unlike the studies linked above where the heavy lifting from an analytical perspective is "under the covers", and you are presented with graphical or tabular output which cannot be checked (not meaning to imply that there is reason to believe checking is needed).
A table of the debris found as of this date (07/04/16) is presented below. Not all of these finds have been officially confirmed, but they are all compelling relative to being from 9M-MRO. No distinction is made by debris size, whether it came from the right or left side of the aircraft, or whether it was an internal or external part.
A LON - LAT scatter plot of these finds is as shown below.
Map view of debris finds below.
The finds show a significant clustering around latitude -20 to -26. Outliers are the Mossel Bay find and the most recent Pemba Island find.
Shown below is the Geomar reverse drift plot using "only" the flaperon finding on Reunion Island. Reunion Island is at -21 latitude, and is representative of the latitude of the debris clusters. I have no reason to doubt the validity of the Geomar model, and it saves the time and tedium of watching which way rubber ducks travel across the ocean.
It is clear from the Geomar graphic the preferred origin of the Reunion find is along the coast of Sumatra/Java. It can be inferred that other finds in the vicinity of latitude -20 to -26 would have a similar origin. It is reasonable to conclude, based on the Geomar model, that the 9M-MRO terminus lies between latitude -10 and -20 on or near the 7th arc.
The debris data lends absolutely no support to the current area being searched by the ATSB. The fact that no debris has been found on the coast of Western Australia further supports the notion that the current search area is not correct.
Edits July 7, 2106 - DW
Since posting the above on the JW blog several comments have been made by people with relatively little technical background or a flawed background that "beg" for additional commentary here.
First is the notion of forward and reverse drift calculation. All calculations are by definition reverse calculations since the objective of the calculation is the origin of the debris not the terminal locations of the debris which comprise the set of knowns. Now it is possible to use algorithms which reverse the forcing functions to arrive at a hypothetical starting point. It is also possible to use forcing functions that operate in the forward direction and use them act on debris at a variety of starting points and subsequently cataloging the results.
Second is the notion that additional debris finds add confusion and difficulty to drift calculations, especially in the reverse direction. The reality is that more data provides more constraining information that actually serves to "focus" possible starting points. More information is always better, that is an axiom of information theory as well as Kalman filtering.
Third is the notion that the new Geomar study should be used, and the above study to be discarded. I deliberately used the above study because it addresses the debris clusters at ~20S directly. No one, including Geomar, has indicated that the above study is flawed. Debris found at Pemba Island and Mossel Bay can only narrow the findings of the above study, not broaden them. A common metric in navigation mathematics is the notion of "dilution of precision". A narrow string of debris location latitudes significantly dilutes the precision that could be obtained if the debris were more scattered with respect to latitude. This statement applies particularly to the North-South precision of the starting point.
Subscribe to:
Posts (Atom)