Part A | 3.3
56 Part A Fundamentals
to remove bias and reduce scatter. These wind measurements also have significant error because a ship’s
superstructure distorts wind flow patterns. Thomas
et al. [3.36] discuss this issue in detail. Of particular
concern with all ship observations is the so-called fair
weather bias – the tendency for ships to avoid storms
and thus underestimate the true probability of larger
waves and winds. A number of attempts were made to
remove bias and scatter [3.37] and these efforts eventually culminated in the work of Hogben et al. [3.38]
who used observations from ships passing close to instrumented buoys to develop corrections that largely
removed bias, at least in the North Atlantic. However,
as Hogben et al. [3.38] point out, they were not nearly
as successful in the southern hemisphere where there
are far fewer offshore instruments. Nor could they do
much about the scatter inherent in subjective human
observations.
Another important historical dataset is the socalled HURDAT (National Hurricane Centers HURricane DATabases) best-track data maintained and distributed by NOAA (National Oceanic and Atmospheric
Administration) [3.39] and documented by Jarvinen
et al. [3.40]. It contains the time histories of tracks,
peak winds, central pressure, and radius for historical North Atlantic tropical storms from 1851 to the
present. NOAA provides similar information for the
eastern North Pacific but only from 1949 to the present.
The NOAA GTECCA (global tropical and extratropical cyclone climatic atlas) database contains historical
track data for global storms up to 1995. For cyclones,
the coverage starts as early as 1870 in the North Atlantic but not until 1945 for the other major basins of
the world [3.41]. Coverage of Northern Hemisphere
extratropical cyclones (winter storms) starts in 1965.
Note that these datasets do not include detailed wind
fields, just the basic intensity information that can
be used to reconstruct the detailed wind fields using various methods such as parametric models [3.42].
Others have used the historical tracks to assimilate
into much more sophisticated numerical models providing high resolution gridded wind velocity and pressure [3.43].
The accuracy of early storms in HURDAT have
been questioned especially as climate scientists have
tried to detect trends in historical storm severity. Karl
et al. [3.44] provide a fairly recent summary of these
findings. In part because of these questions, NOAA
recently undertook a re-analysis of the data underlying HURDAT. These efforts have been documented in
a series of publications, which are referenced in Hagen and Landsea [3.45]. Even after their reexamination,
the researchers in this effort readily admit that large
uncertainties remain in the storms prior to routine airborne observations which started in the early 1950s.
Emanuel [3.46] and others have noted that there is
even more uncertainty in basins outside the North Atlantic, and this uncertainty extends into the post-1950
era because of the lack of reconnaissance flights in most
basins.
3.3.2 Satellite Databases
The study of the oceans and winds from space started
in the 1970s with the launch of Skylab and Geos-3,
which were equipped with a radar-altimeter, windscatterometer, radiometer, and infrared scanner. Le
Traon [3.47] gives an overview of the state of operational satellites used in oceanography and to a lesser
extent, meteorology.
One of the most useful sensors for ocean engineering has proven to be the altimeter. The first of many
operational altimeters began with TOPEX (Ocean Topography Experiment)/Poseiden and ERS1 (Earth Resources Satellite) in 1991. Since 1998, there have been
as many as four altimeters flying simultaneously because several altimeters are needed to properly resolve
the length and time scales of energetic oceanographic
phenomena such as mesoscale eddies, storm-driven
waves, etc. The most used channels from the altimeter are wind velocity (speed and direction), wave height
and period, and sea surface height.
Sea surface height from the operational altimeters
is available in near real time and in historical archives
from various sites [3.48–50]. Though accuracy varies
by satellite, typical RMS (root mean square) errors
are less than 3 cm [3.51]. These heights are useful
in tracking geostrophic ocean currents and developing
comprehensive maps of astronomical tides in deeper
water, the latter being prohibitively expensive before
the advent of satellite altimeters. Shum et al. [3.52] assessed 20 of these tidal models and found many to be
accurate to better than 2 cm.
Wind speed and wave height and period measurements from the operational satellite altimeters are
available over the web [3.53–55] but these are typically organized by individual tracks for each satellite
or statistics from several satellites averaged over large
areal blocks. The track data must typically be filtered to
eliminate periods with heavy rainfall or close passage
to land. Several companies offer commercial products
with fully analyzed databases that can be accessed with
their proprietary software [3.56, 57].
Numerous researchers have investigated the accuracy of altimeter-derived winds and waves, e.g., [3.58,
59]. These efforts show that there is a systemic bias
unique to each satellite but it is easily corrected leaving
an RMS difference with buoy current meter measure-
56 Part A Fundamentals
to remove bias and reduce scatter. These wind measurements also have significant error because a ship’s
superstructure distorts wind flow patterns. Thomas
et al. [3.36] discuss this issue in detail. Of particular
concern with all ship observations is the so-called fair
weather bias – the tendency for ships to avoid storms
and thus underestimate the true probability of larger
waves and winds. A number of attempts were made to
remove bias and scatter [3.37] and these efforts eventually culminated in the work of Hogben et al. [3.38]
who used observations from ships passing close to instrumented buoys to develop corrections that largely
removed bias, at least in the North Atlantic. However,
as Hogben et al. [3.38] point out, they were not nearly
as successful in the southern hemisphere where there
are far fewer offshore instruments. Nor could they do
much about the scatter inherent in subjective human
observations.
Another important historical dataset is the socalled HURDAT (National Hurricane Centers HURricane DATabases) best-track data maintained and distributed by NOAA (National Oceanic and Atmospheric
Administration) [3.39] and documented by Jarvinen
et al. [3.40]. It contains the time histories of tracks,
peak winds, central pressure, and radius for historical North Atlantic tropical storms from 1851 to the
present. NOAA provides similar information for the
eastern North Pacific but only from 1949 to the present.
The NOAA GTECCA (global tropical and extratropical cyclone climatic atlas) database contains historical
track data for global storms up to 1995. For cyclones,
the coverage starts as early as 1870 in the North Atlantic but not until 1945 for the other major basins of
the world [3.41]. Coverage of Northern Hemisphere
extratropical cyclones (winter storms) starts in 1965.
Note that these datasets do not include detailed wind
fields, just the basic intensity information that can
be used to reconstruct the detailed wind fields using various methods such as parametric models [3.42].
Others have used the historical tracks to assimilate
into much more sophisticated numerical models providing high resolution gridded wind velocity and pressure [3.43].
The accuracy of early storms in HURDAT have
been questioned especially as climate scientists have
tried to detect trends in historical storm severity. Karl
et al. [3.44] provide a fairly recent summary of these
findings. In part because of these questions, NOAA
recently undertook a re-analysis of the data underlying HURDAT. These efforts have been documented in
a series of publications, which are referenced in Hagen and Landsea [3.45]. Even after their reexamination,
the researchers in this effort readily admit that large
uncertainties remain in the storms prior to routine airborne observations which started in the early 1950s.
Emanuel [3.46] and others have noted that there is
even more uncertainty in basins outside the North Atlantic, and this uncertainty extends into the post-1950
era because of the lack of reconnaissance flights in most
basins.
3.3.2 Satellite Databases
The study of the oceans and winds from space started
in the 1970s with the launch of Skylab and Geos-3,
which were equipped with a radar-altimeter, windscatterometer, radiometer, and infrared scanner. Le
Traon [3.47] gives an overview of the state of operational satellites used in oceanography and to a lesser
extent, meteorology.
One of the most useful sensors for ocean engineering has proven to be the altimeter. The first of many
operational altimeters began with TOPEX (Ocean Topography Experiment)/Poseiden and ERS1 (Earth Resources Satellite) in 1991. Since 1998, there have been
as many as four altimeters flying simultaneously because several altimeters are needed to properly resolve
the length and time scales of energetic oceanographic
phenomena such as mesoscale eddies, storm-driven
waves, etc. The most used channels from the altimeter are wind velocity (speed and direction), wave height
and period, and sea surface height.
Sea surface height from the operational altimeters
is available in near real time and in historical archives
from various sites [3.48–50]. Though accuracy varies
by satellite, typical RMS (root mean square) errors
are less than 3 cm [3.51]. These heights are useful
in tracking geostrophic ocean currents and developing
comprehensive maps of astronomical tides in deeper
water, the latter being prohibitively expensive before
the advent of satellite altimeters. Shum et al. [3.52] assessed 20 of these tidal models and found many to be
accurate to better than 2 cm.
Wind speed and wave height and period measurements from the operational satellite altimeters are
available over the web [3.53–55] but these are typically organized by individual tracks for each satellite
or statistics from several satellites averaged over large
areal blocks. The track data must typically be filtered to
eliminate periods with heavy rainfall or close passage
to land. Several companies offer commercial products
with fully analyzed databases that can be accessed with
their proprietary software [3.56, 57].
Numerous researchers have investigated the accuracy of altimeter-derived winds and waves, e.g., [3.58,
59]. These efforts show that there is a systemic bias
unique to each satellite but it is easily corrected leaving
an RMS difference with buoy current meter measure-