TCs within Reanalyses: Evolving representation, trends, potential misuse, and intriguing questions Robert Hart (rhart@met.fsu.edu) Danielle Manning, Ryan Maue Florida State University Mike Fiorino National Hurricane Center 3 rd WCRP Intl Conference on Reanalysis: 30 January 2008
Strengths It has been noted several times here that the strengths of reanalyses lie in largescale This is even true in the tropics, where the reanalyses have the greatest challenge, as we have learned this week.
Gray Genesis Potential ERA-40 v OBS YGP = yearly genesis potential = low-level vort * vertical wind shear * convection Source: Dr. Michael Fiorino (NHC)
However What about the year-to-year variability Details, such as TCs themselves? Data changes: Satellites, ships, etc This raises a very fundamental question in reanalyses: How much of what we see is synthetic and how much of what we see is real? And this is particularly key to the TC question, in light of recent controversy.
Evidence of source data shift in analog pattern matching: If you match smoothed global 1000-500mb thickness patterns, this is the DNA of matching. Source: Hart et al. 2008 How much of this is data, and how much of this is actual climate shift?
Bifurcations in the 1970s exists in both ERA40, NCEP reanalyses Source: Hart et al. (2008), BAMS, in revision.
Motivation These issues naturally feed back upon the question of the representation of tropical cyclones (TCs) in the reanalyses. This is further begged by recent studies (e.g., Sriver and Huber 2006) which claimed that the reanalyses TC winds could be used as robust and independent measures for use in comparison to other TC trend studies (e.g. Emanuel 2005; Webster et al. 2005) This talk will first examine the evolving representation of TCs in reanalyses First step is to compare best-track (BT) TC representation to ERA40
ERA-40 TC detection v tropical wind score SHEM v NHEM ; tau=0 v tau=72 h Source: Dr. Michael Fiorino (NHC)
Data and Methodology Hurricane Floyd MSLP (contour) and 10m Wind (shaded)
Data and Methodology Hurricane Floyd Temperature anomaly from zonal mean ( C) 980mb +6 960mb +12 1002mb 990mb +6 +4 Varying degrees of muted representation of TCs, depending on resolution, data input, assimilation type, physics.
Data and Methodology Using NHC BT as a guide, track all North Atlantic storms manually within the 1.125 ERA40 data ERA40 MSLP minimum or 925mb vorticity maximum when MSLP minimum not present Could not use autotracker throughout b/c many storms were quite poorly resolved and had significant track differences from their BT counterparts Could not use BT tracks since ERA40 TC structure is subject of study
ERA40 TCs Position Difference from BT Not surprisingly, position comparison between ERA40 TC and BT TC is a strong function of distance from densest observational net.
ERA40 TCs Intensity Mean MSLP for Each SS Category 1020 1000 980 MSLP 960 940 920 900 ERA40 HURDAT 1 2 3 4 5 SS Category Over a basin-scale perspective, and across the entire ERA40, the ERA40 TC intensity is uncorrelated to the BT intensity in the NATL.
ERA40 TCs Intensity Difference In the BT database, the mean TC intensity is shown to the left. In striking contrast, the ERA40 TC intensity is not well correlated, and in fact appears to represent a function more of age of the TC (existence within the reanalysis) than the actual intensity.
ERA40 TCs Intensity Correlation Intercomparison of the prior two figures leads to this figure, which illustrates that really only land is the ERA40 TC intensity well-correlated to the BT intensity.
Structural Comparison Need to move beyond just intensity and position, and analyze structural representation Will use the cyclone phase space (Hart 2003) to quantify the structural characteristics Specifically, how tropical are ERA40 TCs? How does this representation vary as a function of BT intensity, location, size, etc? Where and when is the ERA40 TC most correct?
ERA40 TCs Structure Source: Manning and Hart 2007
ERA40 TCs Structure
ERA40 TCs Structure: Zoom Source: Manning and Hart 2007
Impact of TC Size: Example Note: These are both Category 5 at these times in the best-track. 1961: Carla 1992: Andrew 1013.5 mb 981 mb
Impact of TC size With relatively coarse grid space, just how important is TC size? Use only TCs during the late period Data quality and density homogeneous Need size data from extended best track (1988 onward): DeMaria et al. 2007 Bin by extended best-track size for the mean stormforce wind radius Very small (<83nm) Small (84-133nm) Large (133-182nm) Very large (>182nm)
Source: Manning and Hart 2007 Impact of TC size: Zoom Only when the storm becomes a sufficient size is intensity variability correctly captured. Beyond a certain size, additional size increase doesn t influence
Impact of TC Year MSLP Entire Period (1957-2001) All storms SS Cat 2 3 4 5 1 2 3 4 ERA40 cannot distinguish MSLP between Late some Period SS (1988-2001) categories All storms In order to be used for SS trend Cat detection 2 it should be 3 able to distinguish 4 between 5 all SS categories 1 SS categories are large 2bins (15-20kt) so this criterion is rather generous 3 4 Source: Manning and Hart 2007 Not sig. Sig. to 95% conf. level Sig. to 99% conf. level
These results argue for an approach that JMA took with the JRA25 through the TC BT wind synthesis of Dr. M. Fiorino. However, this raises the question of ERA40 TC Representation Summary ERA40 TCs are nearly all tropical storm intensity at best Their quality of representation in the ERA40 is a far stronger function of TC size and location than actual intensity While ERA40 representation has improved markedly in the satellite era, the ERA40 is still unable to distinguish SS 1,2; SS2,4; SS3,4; SS3,5 TCs.
Potential red herrings? TC winds are largely a red herring in timeaccumulated measures where the time integral dominates the wind. Source: Maue and Hart 2007
global TC activity scaled ACE days global TC crash in 1999 Source: Dr. Michael Fiorino (NHC)
Caution and Care With the wealth of reanalysis data out there, it is easier than ever before to produce results Often these results can at first be thought to be independent confirmation of earlier results The great responsibility lies in looking at the details of the data and considering how much independence there really is, and how much of the signal is synthetic vs. real
Let s end with an intriguing question Dr. Kevin Trenberth on Monday (V1-221) showed improved estimates of the energy transport by the atmosphere and ocean using state of the art estimates from reanalyses and other sources Among these transports lies the role of tropical cyclones Despite the greatly muted TC representation, can reanalyses be used to indirectly extract the potential magnitude of the TC role in climate?
Active vs. Inactive Hemispheric TCs How does the winter following these anomalous TC seasons differ? Can the TC role in heat transport be implied through reanalyses?
Stationary Eddies Only [Long-term mean winter long-wave pattern] Following inactive recurving TC season Following active recurving TC season Source: Hart et al. 2007
Need to keep alert for other red Removing anomalous AO years herrings Removing anomalous ENSO years Removing anomalous NAO years Removing anomalous PNA years
Concluding Summary Reanalyses are clearly a powerful tool, and all powerful tools require both caution and responsibility to avoid misuse and red-herring attribution Reanalysis representation of NATL TCs over the past decades seemingly has been greatly effected by data availability : The size of an actual TC and its location is far, far more influential in reanalysis quality than its actual intensity (in the NATL)! Current reanalyses are raising many key climate questions Why has global TC activity been declining since the mid 1990s, and crashed in 1999 and 2007? Is the winter climate change following aggregate NH TC activity causation or correlation? Hopefully, future reanalyses will help answer these questions.
Concluding Summary References: Hart, R., R. Maue, and M. Watson, 2007: Estimating the atmospheric and SST memory of tropical cyclones through MPI anomaly evolution. Mon. Wea. Rev., 135, 3990-4005. Manning, D, 2007: The utility of the ERA40 CPS in Trend Diagnosis and North Atlantic Tropical Cyclone Reanalysis. M.S. Thesis, Florida State University. Maue, R. and R. Hart, 2007: Comment on: "Low frequency variability in globally integrated tropical cyclone power dissipation." Geo. Res. Letters, 34, L11703, doi:1029/2006gl028283 Manning, D. and R. E. Hart, 2007: Evolution of North Atlantic ERA40 Tropical Cyclone Representation. Geo. Res. Letters, 34, L05705. doi:10.1029/2006gl028266 Hart, R.,2003: A cyclone phase space derived from thermal wind and thermal asymmetry. Mon. Wea. Rev., 131, 585-616.