One of the critical issues in Emanuel 2005 (and presumably for other like studies) is how measurement inhomogeneity is dealt with. Emanuel relies on Landsea 1993 for a procedure to downward adjust pre-1970 wind speeds. However, prior to Emanuel’s article, Landsea – who seems vastly more authoritative than Emanuel on hurricane data – had pointed out further inhomogeneities between 1991 and present-day measurements, requiring the upward adjustment of data from the 1980s and 1970s. Indeed, his response to Emanuel raises the issue of whether the need is to adjust 1970-1990 data upwards rather than adjust pre-1970 data downwards.
It is now understood to be physically reasonable that the intensity of hurricanes in the 1970s through to the early 1990s was underestimated, rather than the 1940s and 1960s being overestimated.
It seems clear to me that every hurricane measurement needs to have the estimation method attached to it. For Emanuel to draw lurid trend conclusions with such weak analysis of homogeneity is all too typical of climate science.
Most of Landsea’s publications are available online here) . Landsea et al (Science 2006) and Landsea et al 2004 – re-analysis of Hurricane Andrew – are especially relevant to homogeneity issues.
One of the key changes in the last 15-20 years has been the increasing availability of direct overhead satellite images(rather than oblique) images and, prior to that, the introduction of infrared imagery in the mid-1980s. These changes are reported to have the effect of frequently increasing measured wind speeds. Landsea et al 2006 observe of Hurricane Hugo in 1989:
For example (13), Atlantic Hurricane Hugo was estimated to have a maximum sustained surface wind of 59 m s–1 on 15 September 1989, based on use of the Dvorak Technique from an oblique observational angle. But in situ aircraft reconnaissance data obtained at the same time revealed that the hurricane was much stronger (72 m/s) than estimated by satellite. This type of underestimate was probably quite common in the 1970s and 1980s in all tropical cyclone basins because of application of the Dvorak Technique in an era of few satellites with low spatial resolution.
Landsea et al 2006 illustrates 5 North Indian Ocean cyclones in the 1980s which were under-estimated, saying:
The North Indian basin tropical cyclones shown here are listed in the best track data set as category 3 or weaker, but were probably category 4 or 5. Similar underestimates may have been common in all ocean basins in the 1970s and 1980s. Trend analyses for tropical cyclones intesities are therefore highly problematic….Another major tropical cyclone, the 1970 Bangladesh cyclone”¢’¬?the world’s worst tropical-cyclone disaster, with 300,000 to 500,000 people killed”¢’¬?does not even have an official intensity estimate, despite indications that it was extremely intense (22).
Image of Indian Ocean Cyclone from Landsea et al 2006.
They go on to say:
Ongoing Dvorak reanalyses of satellite images in the Eastern Hemisphere basins by the third author suggest that there are at least 70 additional, previously unrecognized category 4 and 5 cyclones during the period 1978–1990. The pre-1990 tropical cyclone data for all basins are replete with large uncertainties, gaps, and biases. Trend analyses for extreme tropical cyclones are unreliable because of operational changes that have artificially resulted in more intense tropical cyclones being recorded, casting severe doubts on any such trend linkages to global warming.
Landsea et al 2004 , published before Emanuel 2005, also pointed out these problems in the specific context of a re-analysis of Hurricane Andrew (1992), which was re-classified from a Category 4 to a Category 5 hurricane. They observed:
New global positioning system (GPS) dropwindsonde observations in hurricane eyewalls”¢’¬?first collected in the eastern North Pacific from Hurricane Guillermo in 1997″¢’¬?suggest that the intensities of all of the hurricanes in the aircraft reconnaissance era up through 1998 should be re-examined when the primary method for estimating surface winds was from flight-level wind adjustments….Research using GPS dropwindsondes in the late 1990s and early 2000s has demonstrated that stronger winds exist at the surface in the hurricane eyewall than originally had been believed.
The amount of the increases is summarized below – for PDI which rises as the cube of the wind speed, the impact of these adjstments is even more dramatic.
Excerpt from Landsea et al 2004.
The 1993 Adjustment
Emanuel 2005 spends no time whatever discussing inhomogeneities in the record after 1991. Instead he relies entirely on an inhomogeneity between pre-1970 measurements and 1970-1990 measurements identified in Landsea 1993 to substantially downward adjust pre-1970 measurements, which resulted in much of the rhetorical impact of his graphics (to the extent that they could be replicated – and I’m still in the process of implementing this adjustment. ) Landsea 1993 observed an empirical difference between wind speed estimation before 1970 and from 1970-1991, which Emanuel 2005 (SI) described as follows:
For the first decade or so of airborne reconnaissance, surface winds were estimated mostly by visual inspection of the sea surface. Beginning in the early 1950s, radar altimeters aboard the aircraft made possible an accurate determination of the aircraft’s absolute altitude. When combined with direct pressure measurements, this gives a good estimate of geopotential height at flight level. Surface pressure can then be estimated using empirical relationships between surface and flight level pressure. This technique, developed during the 1950s, was used without significant modification through the end of aircraft reconnaissance in the western North Pacific and until the advent of accurate dropwindsondes in the North Atlantic. Minimum surface pressure estimates were converted to maximum sustained surface wind using semi-empirical wind-pressure relations which, however, have evolved with time. For the North Atlantic, Landsea 3 has documented a change in that took place in 1970, leading to lower wind speed estimates.
Landsea 1993 provided the following table, which he used only to improve homogeneity in hurricane counts by defining 54 m sec-1 as the hurdle for counting in the early part of the series and 51 m sec-1 in the later part of the series.
From Landsea 1993
I’ve plotted the information from this table is plotted below with the earlier measurement method (vertical) and later measurement method (horizontal). To these 7 points, Emanuel 2000 fitted a slightly convex curve through the formula . The dashed red line shows a simple linear fit. The dashed black line shows the 45 degree line. If one takes Emanuel’s most recent comments at face value, this graph could equally stand for an increase in estimates for the period 1970-1990 rather than a reduction for the pre-1970 period. In his Comment on Emanuel, Landsea observed that this adjustment reached as high as 12 m sec-1 at higher speeds – a result not contemplated in the original article.
Plot of Table from Landsea 1993, showing Emanuel polynomial fit.
While Landsea’s Comment on Emanuel refers to the problem of post-1990 inhomogeneity in clear terms, he spends more time on the issue of how Emanuel implemented the pre-1970 adjustment, saying – and the above figure is helpful in understanding this:
He attempts to use a bias-removal scheme (Landsea, 1993) that recommends reduction of the tropical-cyclone winds by 2.5–5.0 m s^-1 for the 1940s–60s because of an inconsistency in the pressure–wind relationship during those years compared with subsequent (and presumably more accurate) data. However, the function used by Emanuel to reduce the winds in the earlier period goes well beyond this recommendation, as the bias removal used continued to increase with increasing wind intensity and reached a reduction of as much as 12.2ms^-1 for the strongest hurricane in the 1949–69 original data set.
Landsea then presented an alternative implementation of the pre-1970 adjustment (see below) – which is entirely inconsistent with his caveats about post-1990 inhomogeneity. To implement these comments, one should upward adjust the 1970-1990 values using the Landsea 1993 formula (for example) as means of achieving homogeneity in this period.
LAndsea 2005 Figure 1 | Derivation of Atlantic power-dissipation index (PDI). b, Three versions of the smoothed PDI for the North Atlantic using: dashed line, Emanuel’s applied bias-removal scheme; dotted line, 1993 version3 of the bias-removal scheme; solid line, original hurricane database. All three versions are identical from 1970 onwards.
This implementation is inconsistent with the following observation in Landsea’s Comment:
In major hurricanes, winds are substantially stronger at the ocean’s surface [Franklin et al Weath. Forecast.2003; Dunion et al Mon. Weath. Rev. 2003; Kepert & Wang, J. Atmos. Sci. 2001; Kepert. J. Atmos. Sci. 2001] than was previously realized, so it is no longer clear that Atlantic tropical cyclones of the 1940s–60s call for a sizeable systematic reduction in their wind speeds. It is now understood to be physically reasonable that the intensity of hurricanes in the 1970s through to the early 1990s was underestimated, rather than the 1940s and 1960s being overestimated. To examine changes in intensity over time, it is therefore better to use the original hurricane database than to apply a general adjustment to the data in an attempt to make it homogenous.
Accordingly, as a replication exercise, the calculations that I’ve done so far are reasonable implementations of Landsea’s most recent position on adjustments for the pre-1970 period – but should be adjusted upwards in the 1970-1990 period. I’ll re-do these calculations both ways now that I’ve more or less figured out how to implement the 1993 adjustment.
All authors seem to agree that the data set is in need of repair. Because the post-1970 hurricane period was quiet in the Atlantic, it doesn’t make any sense to discard prior data. However it does make sense to use it sensibly. In my opinion, the most urgent requirement is meta-data – how each particular track was calculated: documentary, airplane, oblique Dvorak, overhead Dvorak, … whatever the relevant information is. Without replicable meta-data, these hurricane calculations may be, in Judith’s terminology, a pissing contest, but the contestants seem to be pissing into the wind. If the wind speeds exceed 18 m sec-1, it’s not clear how the contest is won, but the impact on the contestants is easily predicted.