Verification of tracks made on Gert (2017) using GEFS mean fields
This is a repost from August 2017
I made 12 tracks about Hurricane Gert. Two of those tracks were accidentally made from the same run. For this analysis, I chose the more erroneous one, and discarded the other. Additionally, there was a run, where I missed the system entirely. Intensity forecast was pretty good, but position errors became huge, as Gert recurved.
My analysis begins at August 5, 06Z, which is the earliest time, at which some of my tracks had cyclogenesis. Position analysis ends at August 17, 18Z, which is the last best track entry. Intensity analysis continues until August 21, 06Z, which is the last time, at which one of my tracks has a tropical cyclone.
Since I give only SSHS categories in my tracks, I don't calculate intensity errors in units of speed, but in categories. For this purpose, I define tropical storms as "Category 0", tropical depressions as "Category -1" and non-existence of a tropical cyclone as "Category -2".
I compare intensities only in cases, when either the best track data or my track has a tropical cyclone, or both. In other words, I leave the comparison out of calculations, if both best track data and my track lack a tropical cyclone.
In the 11 tracks, formation times ranged between August 5, 06Z and August 11, 18Z. Median was August 8, 18Z and average was August 8, 04Z.
The 12 peak intensities (the one complete miss included) ranged between non-formation (Category -2) to Category 5. Median was Category 1.5 and average was Category 1.9.
Dissipation times ranged between August 10, 00Z and August 21, 12Z. Median was August 17, 12Z and average was August 17, 08Z.
The following lines will have the format:
date and time (UTC) yyyy/mm/dd/hh ... number of true detections ... number of misses ... number of false detections ... number of position comparisons ... number of intensity comparisons ... average error in intensity category ... bias in intensity category ... average error in latitude ... bias in latitude ... average error in longitude ... bias in longitude ... upper limit for scalar error in position (nautical miles)... calculated scalar bias in position (nautical miles)
2017/08/06/12 ... 0 ... 0 ... 4 ... 4 ... 4 ... 1.3 ... +1.3 ... 1.7 ... 1.7S ... 2.4 ... 0.1E ... 170 ... 100
2017/08/09/12 ... 0 ... 0 ... 6 ... 7 ... 6 ... 2.5 ... +2.5 ... 2.8 ... 1.0S ... 5.1 ... 2.8W ... 340 ... 170
actual cyclogenesis:
2017/08/13/00 ... 10 ... 2 ... 0 ... 11 ... 12 ... 2.3 ... +2.0 ... 4.6 ... 4.1S ... 8.4 ... 6.6W ... 550 ... 450
2017/08/14/18 ... 9 ... 3 ... 0 ... 10 ... 12 ... 1.8 ... +0.8 ... 7.2 ... 7.2S ... 11.9 ... 11.8W ... 780 ... 770
2017/08/17/00 ... 7 ... 5 ... 0 ... 7 ... 12 ... 2.0 ... -1.9 ... 14.9 ... 14.9S ... 27.2 ... 27.2W ... 1730 ... 1730
becomes extratropical:
2017/08/17/18 ... 0 ... 0 ... 5 ... 5 ... 5 ... 3.2 ... +3.2 ... 19.8 ... 19.8S ... 43.9 ... 43.9W
And for lead times:
168h ... 0 ... 0 ... 6 ... 6 ... 6 ... 2.5 ... +2.5 ... 3.3 ... 1.2S ... 6.1 ... 3.3W ... 410 ... 210
192h ... 0 ... 0 ... 7 ... 8 ... 7 ... 2.7 ... +2.7 ... 3.2 ... 1.9S ... 5.7 ... 3.0W ... 380 ... 210
216h ... 0 ... 0 ... 7 ... 9 ... 7 ... 3.0 ... +3.0 ... 3.8 ... 2.2S ... 6.8 ... 4.0W ... 450 ... 260
240h ... 2 ... 0 ... 6 ... 10 ... 8 ... 2.8 ... +2.8 ... 4.4 ... 3.0S ... 6.5 ... 4.6W ... 450 ... 320
264h ... 6 ... 0 ... 2 ... 10 ... 8 ... 2.6 ... +2.6 ... 5.0 ... 4.5S ... 7.6 ... 6.7W ... 520 ... 470
288h ... 8 ... 1 ... 0 ... 10 ... 9 ... 2.0 ... +1.8 ... 6.1 ... 6.1S ... 9.0 ... 9.0W ... 630 ... 630
312h ... 9 ... 1 ... 0 ... 10 ... 10 ... 1.6 ... +0.4 ... 7.7 ... 7.7S ... 12.4 ... 12.4W ... 840 ... 840
336h ... 8 ... 3 ... 0 ... 9 ... 11 ... 1.7 ... -1.4 ... 10.8 ... 10.8S ... 17.9 ... 17.9W ... 1200 ... 1200
360h ... 4 ... 4 ... 2 ... 6 ... 9 ... 2.0 ... -1.1 ... 12.9 ... 12.9S ... 29.1 ... 29.1W ... 1820 ... 1820
384h ... 3 ... 2 ... 3 ... 3 ... 8 ... 2.1 ... +0.4 ... 12.4 ... 12.4S ... 24.9 ... 24.9W ... 1590 ... 1590
Here are the tracks once again.
This is from July 25, 18Z run, although I erroneously earlier said, that this would be from July 31, 12Z run:
This is the analysed track from July 31, 18Z run:
This is the discarded track from July 31, 18Z run:
August 1, 00Z run:
August 1, 06Z run:
August 1, 18Z run:
https://uploads.disquscdn.c...
August 2, 00Z run:
August 2, 06Z run:
August 2, 12Z run:
August 2, 18Z run:
August 3, 00Z run:
August 3, 06Z run:
The total miss was July 31, 12Z run.
I made 12 tracks about Hurricane Gert. Two of those tracks were accidentally made from the same run. For this analysis, I chose the more erroneous one, and discarded the other. Additionally, there was a run, where I missed the system entirely. Intensity forecast was pretty good, but position errors became huge, as Gert recurved.
My analysis begins at August 5, 06Z, which is the earliest time, at which some of my tracks had cyclogenesis. Position analysis ends at August 17, 18Z, which is the last best track entry. Intensity analysis continues until August 21, 06Z, which is the last time, at which one of my tracks has a tropical cyclone.
Since I give only SSHS categories in my tracks, I don't calculate intensity errors in units of speed, but in categories. For this purpose, I define tropical storms as "Category 0", tropical depressions as "Category -1" and non-existence of a tropical cyclone as "Category -2".
I compare intensities only in cases, when either the best track data or my track has a tropical cyclone, or both. In other words, I leave the comparison out of calculations, if both best track data and my track lack a tropical cyclone.
In the 11 tracks, formation times ranged between August 5, 06Z and August 11, 18Z. Median was August 8, 18Z and average was August 8, 04Z.
The 12 peak intensities (the one complete miss included) ranged between non-formation (Category -2) to Category 5. Median was Category 1.5 and average was Category 1.9.
Dissipation times ranged between August 10, 00Z and August 21, 12Z. Median was August 17, 12Z and average was August 17, 08Z.
The following lines will have the format:
date and time (UTC) yyyy/mm/dd/hh ... number of true detections ... number of misses ... number of false detections ... number of position comparisons ... number of intensity comparisons ... average error in intensity category ... bias in intensity category ... average error in latitude ... bias in latitude ... average error in longitude ... bias in longitude ... upper limit for scalar error in position (nautical miles)... calculated scalar bias in position (nautical miles)
2017/08/06/12 ... 0 ... 0 ... 4 ... 4 ... 4 ... 1.3 ... +1.3 ... 1.7 ... 1.7S ... 2.4 ... 0.1E ... 170 ... 100
2017/08/09/12 ... 0 ... 0 ... 6 ... 7 ... 6 ... 2.5 ... +2.5 ... 2.8 ... 1.0S ... 5.1 ... 2.8W ... 340 ... 170
actual cyclogenesis:
2017/08/13/00 ... 10 ... 2 ... 0 ... 11 ... 12 ... 2.3 ... +2.0 ... 4.6 ... 4.1S ... 8.4 ... 6.6W ... 550 ... 450
2017/08/14/18 ... 9 ... 3 ... 0 ... 10 ... 12 ... 1.8 ... +0.8 ... 7.2 ... 7.2S ... 11.9 ... 11.8W ... 780 ... 770
2017/08/17/00 ... 7 ... 5 ... 0 ... 7 ... 12 ... 2.0 ... -1.9 ... 14.9 ... 14.9S ... 27.2 ... 27.2W ... 1730 ... 1730
becomes extratropical:
2017/08/17/18 ... 0 ... 0 ... 5 ... 5 ... 5 ... 3.2 ... +3.2 ... 19.8 ... 19.8S ... 43.9 ... 43.9W
And for lead times:
168h ... 0 ... 0 ... 6 ... 6 ... 6 ... 2.5 ... +2.5 ... 3.3 ... 1.2S ... 6.1 ... 3.3W ... 410 ... 210
192h ... 0 ... 0 ... 7 ... 8 ... 7 ... 2.7 ... +2.7 ... 3.2 ... 1.9S ... 5.7 ... 3.0W ... 380 ... 210
216h ... 0 ... 0 ... 7 ... 9 ... 7 ... 3.0 ... +3.0 ... 3.8 ... 2.2S ... 6.8 ... 4.0W ... 450 ... 260
240h ... 2 ... 0 ... 6 ... 10 ... 8 ... 2.8 ... +2.8 ... 4.4 ... 3.0S ... 6.5 ... 4.6W ... 450 ... 320
264h ... 6 ... 0 ... 2 ... 10 ... 8 ... 2.6 ... +2.6 ... 5.0 ... 4.5S ... 7.6 ... 6.7W ... 520 ... 470
288h ... 8 ... 1 ... 0 ... 10 ... 9 ... 2.0 ... +1.8 ... 6.1 ... 6.1S ... 9.0 ... 9.0W ... 630 ... 630
312h ... 9 ... 1 ... 0 ... 10 ... 10 ... 1.6 ... +0.4 ... 7.7 ... 7.7S ... 12.4 ... 12.4W ... 840 ... 840
336h ... 8 ... 3 ... 0 ... 9 ... 11 ... 1.7 ... -1.4 ... 10.8 ... 10.8S ... 17.9 ... 17.9W ... 1200 ... 1200
360h ... 4 ... 4 ... 2 ... 6 ... 9 ... 2.0 ... -1.1 ... 12.9 ... 12.9S ... 29.1 ... 29.1W ... 1820 ... 1820
384h ... 3 ... 2 ... 3 ... 3 ... 8 ... 2.1 ... +0.4 ... 12.4 ... 12.4S ... 24.9 ... 24.9W ... 1590 ... 1590
Here are the tracks once again.
This is from July 25, 18Z run, although I erroneously earlier said, that this would be from July 31, 12Z run:
https://uploads.disquscdn.c...
August 2, 00Z run:
Comments
Post a Comment