Why have Koonin and Klotzbach reached contradictory conclusions about the change in hurricanes?
1) The technology for assessing hurricanes has improved over time. Satellite images from space are processed to extract information about wind speed. If I understand correctly Klotzbach is using "best tracks" data and Koonin has reduced the resolution of satellite images of recent satellite images to the lowest resolution of match the resolution of the oldest images before processing them. In theory, the homogenized ADT HURSAT data should be superior to the apparently non-homogenized data Klotzbach is using. This leads to a big decrease in the total number of hurricanes being counted: 26% in the early period, 36% in the late period. For major hurricanes, an 8% decrease early major hurricanes and at 31% decrease in late major hurricanes. One might expect that homogenization to have caused Koonin to report a smaller increase than Klotzbach, but other factors must be more important.
2) Klotzbach studied the period 1990-2021, which isn't highly perturbed by the active phase of the AMO from about 1997-2012, while Koonin used 1979-2017. In an earlier paper Koonin started in 1982, but this new paper creates missing data for one ocean basin that wasn't monitored by satellite form 1979-1982. Koonin's early period is mostly from when the AMO Index was negative and his late period is mostly from when the AMO is positive. As I noted early, there is no global trend in Koonin's data outside the North Atlantic.
3) If you check out Koonin's Supplementary Figure 3, Best Tracks and ADT HURSAT differ substantially (often 20 kt in wind speed? more than enough to convert category 2 to 1 or 3). Quantifying hurricanes "activity" by the peak wind velocity in one 6-hour period of several dozen periods is not a very reliable process. I intend to pay more attention in the future to accumulated cyclone energy, which takes into account several dozen processed images per hurricane. It varies with v^2, as does the drag force in physics.
It's a subject I hope to dig into some more at a later stage, but my brief answer at this stage is that they probably don't reach contradictory conclusions.
Phil Klotzback's email explained that if you consider the metric used by James Koonin, they find the same trend. Slightly different results due to timescale and other factors.
Steve: You've usually read things more carefully and not jumped to conclusions as quickly as I, but it is my impression that Kossin and Klotzbach are looking at such different datasets that they can and likely are inconsistent with each other. Homogenization is big, periods studied, even the definition of a major hurricane. The supplementary data below show inconsistent Best Tracks is from ADT-HURSAT.
"Several global TC metrics have decreased during this period, with significant decreases in hurricane numbers and Accumulated Cyclone Energy (ACE). Most of this decrease has been driven by significant downward trends in the western North Pacific."
In Kossin's data, the total number of TCs in the Eastern Pacific decreased by 322 (13%) while increasing by 805 (102%). Major TC's decreased by 41 (7%) in the EP and increased by 393 (289%) in the NA. Klotzbach thinks the major driver of the trend is the EP, but changes in the NA dominate in Kossin's data.
Did you receive an email from me with my spreadsheet of Kossin's data at your old SOD address?
Re-sent (forwarded) email with spreadsheet that was originally sent on March 9, using an scienceofdoom address that worked many years ago. Hope it works this time.
SOD: When AR6 appeared, I re-analyzed the data in Table 1 of Kossin (2020), AR6's primary source. Suspiciously, the data was not organized in a way that made it easy to subtract the change in the North Atlantic – where I expected the trend analyzed by Kossin to be biased by the AMO – from the global data. In the global homogenized data (ADT-HURSAT). After simply rearranging the data, I calculated:
1) There was a 5% increase in the total number of global TCs between Kossin’s early (1979-1997) and late (1998-2017) periods and a 5% DECREASE in global TCs outside the North Atlantic. The increase in the North Atlantic was 102%! (The next biggest change was a 24% DECREASE in the South Pacific.)
2) There was a 20% increase in the total number of major global TCs and only a 4% increase in major TCs outside the North Atlantic. The increase in major TCs in the North Atlantic was 289%! (The next biggest change was a 38% increase in the Southern Indian Ocean.)
3) There was a 4% increase in the percentage of major TC’s globally and only a 2% increase outside the North Atlantic. This increase in major TCs in the North Atlantic was 16% (18% to 34%).
Based on the data in Kossin (2020) and contrary to its conclusions, THERE HAS BEEN NO GLOBAL TREND IN TCs. There is merely one basin, the North Atlantic, which – during the period of this analysis – has behaved radically differently from the other six basins for well-understood reasons. The trend outside the NA appears negligible.
At this point in my research, I got bogged down in trying to understand what is known about long-term changes in TCs in the North Atlantic. In Part 1, SOD shows the lack of a long-term trend in the number of landfalling hurricanes in the NA. For all NA hurricanes, the simplest information I located was from Klotzbach (2015), which shows three active periods in the NA (1880-1900, 1926-1969, and 1995-2012) and two inactive periods (1901-1925 and 1970-1994) with 61% correlation between ACE and the AMO Index in any year. The active periods had ACE (accumulated cyclone energy) 2-3 fold higher (my estimate) than the inactive periods. This is crudely comparable to the 102% increase in NA TCs and 289% increase in NA major TCs in Kossin (2020). Klotzbach summarized the data differently: The 25 years with the highest AMO index mostly from Kossin’s late period) had 3.2-fold more major hurricane days than the 25 years with the lowest AMO (mostly from the early period). There is no reason to believe that the enormous change in the NA between Kossin’s early and late periods has anything to do with global warming.
Why have Koonin and Klotzbach reached contradictory conclusions about the change in hurricanes?
1) The technology for assessing hurricanes has improved over time. Satellite images from space are processed to extract information about wind speed. If I understand correctly Klotzbach is using "best tracks" data and Koonin has reduced the resolution of satellite images of recent satellite images to the lowest resolution of match the resolution of the oldest images before processing them. In theory, the homogenized ADT HURSAT data should be superior to the apparently non-homogenized data Klotzbach is using. This leads to a big decrease in the total number of hurricanes being counted: 26% in the early period, 36% in the late period. For major hurricanes, an 8% decrease early major hurricanes and at 31% decrease in late major hurricanes. One might expect that homogenization to have caused Koonin to report a smaller increase than Klotzbach, but other factors must be more important.
2) Klotzbach studied the period 1990-2021, which isn't highly perturbed by the active phase of the AMO from about 1997-2012, while Koonin used 1979-2017. In an earlier paper Koonin started in 1982, but this new paper creates missing data for one ocean basin that wasn't monitored by satellite form 1979-1982. Koonin's early period is mostly from when the AMO Index was negative and his late period is mostly from when the AMO is positive. As I noted early, there is no global trend in Koonin's data outside the North Atlantic.
3) If you check out Koonin's Supplementary Figure 3, Best Tracks and ADT HURSAT differ substantially (often 20 kt in wind speed? more than enough to convert category 2 to 1 or 3). Quantifying hurricanes "activity" by the peak wind velocity in one 6-hour period of several dozen periods is not a very reliable process. I intend to pay more attention in the future to accumulated cyclone energy, which takes into account several dozen processed images per hurricane. It varies with v^2, as does the drag force in physics.
2) Koonin has fi
Frank,
It's a subject I hope to dig into some more at a later stage, but my brief answer at this stage is that they probably don't reach contradictory conclusions.
Phil Klotzback's email explained that if you consider the metric used by James Koonin, they find the same trend. Slightly different results due to timescale and other factors.
I'll ask Phil's permission to publish his email.
Steve: You've usually read things more carefully and not jumped to conclusions as quickly as I, but it is my impression that Kossin and Klotzbach are looking at such different datasets that they can and likely are inconsistent with each other. Homogenization is big, periods studied, even the definition of a major hurricane. The supplementary data below show inconsistent Best Tracks is from ADT-HURSAT.
https://www.pnas.org/doi/suppl/10.1073/pnas.1920849117/suppl_file/pnas.1920849117.sapp.pdf
For example, in the abstract Klotzbach says:
"Several global TC metrics have decreased during this period, with significant decreases in hurricane numbers and Accumulated Cyclone Energy (ACE). Most of this decrease has been driven by significant downward trends in the western North Pacific."
In Kossin's data, the total number of TCs in the Eastern Pacific decreased by 322 (13%) while increasing by 805 (102%). Major TC's decreased by 41 (7%) in the EP and increased by 393 (289%) in the NA. Klotzbach thinks the major driver of the trend is the EP, but changes in the NA dominate in Kossin's data.
Did you receive an email from me with my spreadsheet of Kossin's data at your old SOD address?
No, I haven’t received it
Re-sent (forwarded) email with spreadsheet that was originally sent on March 9, using an scienceofdoom address that worked many years ago. Hope it works this time.
SOD: When AR6 appeared, I re-analyzed the data in Table 1 of Kossin (2020), AR6's primary source. Suspiciously, the data was not organized in a way that made it easy to subtract the change in the North Atlantic – where I expected the trend analyzed by Kossin to be biased by the AMO – from the global data. In the global homogenized data (ADT-HURSAT). After simply rearranging the data, I calculated:
1) There was a 5% increase in the total number of global TCs between Kossin’s early (1979-1997) and late (1998-2017) periods and a 5% DECREASE in global TCs outside the North Atlantic. The increase in the North Atlantic was 102%! (The next biggest change was a 24% DECREASE in the South Pacific.)
2) There was a 20% increase in the total number of major global TCs and only a 4% increase in major TCs outside the North Atlantic. The increase in major TCs in the North Atlantic was 289%! (The next biggest change was a 38% increase in the Southern Indian Ocean.)
3) There was a 4% increase in the percentage of major TC’s globally and only a 2% increase outside the North Atlantic. This increase in major TCs in the North Atlantic was 16% (18% to 34%).
Based on the data in Kossin (2020) and contrary to its conclusions, THERE HAS BEEN NO GLOBAL TREND IN TCs. There is merely one basin, the North Atlantic, which – during the period of this analysis – has behaved radically differently from the other six basins for well-understood reasons. The trend outside the NA appears negligible.
At this point in my research, I got bogged down in trying to understand what is known about long-term changes in TCs in the North Atlantic. In Part 1, SOD shows the lack of a long-term trend in the number of landfalling hurricanes in the NA. For all NA hurricanes, the simplest information I located was from Klotzbach (2015), which shows three active periods in the NA (1880-1900, 1926-1969, and 1995-2012) and two inactive periods (1901-1925 and 1970-1994) with 61% correlation between ACE and the AMO Index in any year. The active periods had ACE (accumulated cyclone energy) 2-3 fold higher (my estimate) than the inactive periods. This is crudely comparable to the 102% increase in NA TCs and 289% increase in NA major TCs in Kossin (2020). Klotzbach summarized the data differently: The 25 years with the highest AMO index mostly from Kossin’s late period) had 3.2-fold more major hurricane days than the 25 years with the lowest AMO (mostly from the early period). There is no reason to believe that the enormous change in the NA between Kossin’s early and late periods has anything to do with global warming.
http://www.novaweather.net/Uploads/ngeo2529.pdf