Chapter 5 - The Oceanic Heat Budget
5.5 Global Data Sets for Fluxes
Ship and satellite data have been processed to produce global maps of fluxes. Observations from ship measurements made over the past 150 years yield maps
of the long-term mean values of the fluxes, especially in the northern hemisphere. Ship data, however, are sparse in time and space, and they are being replaced
more and more by satellite data.
The most useful maps are those made by combining level 3 and 4 satellite data
sets with observations from ships, using numerical weather models. Let's look
first at the sources of data, then at a few of the more widely used data sets.
Comprehensive Ocean-Atmosphere Data Set
Data collected from observers on ships are the richest source of marine information.
et al., (1985) describing their efforts to collect,
edit, summarize, and
publish all marine observations write:
Since 1854, ships of many countries have been taking regular observations
of local weather, sea surface temperature, and many other characteristics
near the boundary between the ocean and the atmosphere. The observations
by one such ship-of-opportunity at one time and place, usually incidental
to its voyage, make up a marine report. In later years fixed research vessels,
buoys, and other devices
have contributed data. Marine reports have been collected, often in machine-readable form, by various
agencies and countries. That vast collection of data, spanning the global oceans from the mid-nineteenth
century to date, is the historical ocean-atmosphere record.
These marine reports have been edited and published as the Comprehensive
Ocean-Atmosphere Data Set COADS (Woodruff et al. 1987) available
through the National Oceanic and Atmospheric Administration.
The ICOADS release 2.4 includes 238 million reports of marine surface
conditions collected from 1784–2007 by buoys, other platform types, and
by observers on merchant ships. The data set include fully quality-controlled
(trimmed) reports and summaries. Each unique report contains 22 observed and
derived variables, as well as flags indicating which observations were statistically
trimmed or subjected to adaptive quality control. Here, statistically trimmed
means outliers were removed from the data set. The summaries included in the
data set give 14 statistics, such as the median and mean, for each of eight
observed variables: air and sea surface temperatures, wind velocity, sea-level
pressure, humidity, and cloudiness, plus 11 derived variables.
The data set consists of an easily-used data base at three principal resolutions:
- individual reports,
- year-month summaries of the individual reports
in 2° latitude by 2° longitude boxes from 1800 to 2005
and 1° latitude by 1° longitude boxes from 1960 to
- decade-month summaries.
Note that data from 1784 through the
early 1800s are extremely sparse–based on scattered ship voyages. See
Overview and an example of temperature
Duplicate reports judged inferior by a first quality control process designed by the National Climatic
Data Center NCDC were eliminated or flagged, and "untrimmed" monthly and decadal summaries were
computed for acceptable data within each 2° latitude by 2° longitude grid. Tighter, median-smoothed
limits were used as criteria for statistical rejection of apparent outliers from the data used for separate
sets of trimmed monthly and decadal summaries. Individual observations were retained in report form but
flagged during this second quality control process if they fell outside 2.8 or 3.5 estimated
standard-deviations about the smoothed median applicable to their 2° latitude by 2° longitude box,
month, and 56-, 40-, or 30-year period (i.e., 1854-1990, 1910-1949, or
The data are most useful in the northern hemisphere, especially the North
Atlantic. Data are sparse in the southern hemisphere and they are not reliable
south of 30°S. Gleckler and Weare (1997) analyzed the
accuracy of the ICOADS data for calculating global maps and zonal averages of
the fluxes from 55°N to
40°S. They found that systematic errors dominated the zonal means. Zonal
averages of insolation were
uncertain by about 10%, ranging from ±10 W/m2 in high latitudes
to ± 25W/m2 in
tropics. Long wave fluxes were uncertain by about ± 7W/m2. Latent
heat flux uncertainties
ranged from ±10 W/m2 in some areas of the northern oceans to ±30
tropical oceans to ±50 W/m2 in western boundary currents. Sensible
heat flux uncertainties tend
to be around ±5-10 W/m2.
Josey et al., (1999) compared averaged fluxes
calculated from ICOADS with fluxes calculated from observations made by carefully
calibrated instruments on some ships and buoys. They
found that mean flux into the oceans, when averaged over all the seas surface
had errors of ± 30W/m2. Errors vary season-ally and by region, and
global maps of fluxes require corrections such as those proposed by DaSilva,
Young, and Levitus (1995)
shown in Figure 5.7.
Raw data are available from satellite projects, but we need processed data. Various
processed data from satellite projects are produced (Table 5.3):
Table 5.3 Levels of Processed Satellite Data
Level of Processing
Unprocessed data from the satellite in engineering units (volts)
Data processed into geophysical units (wind speed) at the time and place the satellite instrument made the observation
Level 2 data interpolated to fixed coordinates in time and space
Level 3 data averaged in time and space or further processed
The operational meteorological satellites that observe the ocean include:
- NOAA series of polar-orbiting, meteorological satellites;
- US Defense Meteorological Satellite Program DMSP polar-orbiting satellites,
which carry the Special Sensor Microwave/Imager (SSM/I);
- Geostationary meteorological satellites operated by NOAA (GOES), Japan
(GMS) and the European Space Agency (METEOSTATS).
Data are also available from instruments on experimental satellites such as:
- Nimbus-7, Earth Radiation Budget Instruments;
- Earth Radiation Budget Satellite, Earth Radiation Budget Experiment;
- The European Space Agency's ERS-1 & 2;
- The Japanese Advanced Earth Observing System (ADEOS) and Midori;
- The Eaeth-Observing System satellites Terra, Aqua, and Envisat,
- The Tropical Rainfall Measuring Satellite (TRMM); and
- Topex/Poseidon and its replacement Jason-1.
Satellite data are collected, processed, and archived by government organizations.
Archived data are
further processed to produce useful flux data sets.
International Satellite Cloud Climatology Project
The International Satellite Cloud Climatology Project is an ambitious project to collect observations
of clouds made by dozens of meteorological satellites from 1985 to 1995, to calibrate the the satellite
data, to calculate cloud cover using carefully verified techniques, and to calculate surface insolation
(Rossow and Schiffer, 1991). The clouds were observed with visible-light instruments on polar-orbiting and
Global Precipitation Climatology Project
This project uses three sources of data to calculate rain rate (Huffman, et
al., 1995, 1997):
- Infrared observations of the height of cumulus clouds from goes satellites. The basic idea is that
the more rain produced by cumulus clouds, the higher the cloud top, and the colder the top appears in the
infrared. Thus rain rate at the base of the clouds is related to infrared temperature.
- Measurements by rain gauges on islands and land.
- Radio-frequency emissions from from water droplets in the atmosphere observed by the SSM/I.
Accuracy is about 1 mm/day. Data from the project are available on a 2.5° latitude by 2.5°
longitude grid from July 1987 to December 1995 from the Global Land Ocean Precipitation Analysis at the
NASA Goddard Space Flight Center.
Xie and Arkin (1997) produced a 17-year data set based on seven types of satellite and rain-gauge data
combined with the output from the NCEP/NCAR reanalyzed data from numerical weather models. The data set has
the same spatial and temporal resolution as the Huffman data set.
Data From Numerical Weather Models
fluxes have been calculated from weather
numerical weather models by various reanalysis projects described in §4.4.
The fluxes are
consistent with atmospheric dynamics, they are global, they are calculated every
and they are available for many years on a uniform grid. For example, the NCEP/NCAR
reanalysis, available on a CD-ROM, include daily averages of
stress, sensible and latent heat fluxes,
net long and
short wave fluxes, near-surface temperature, and precipitation.
Accuracy of Calculated Fluxes
Recent studies of the accuracy of fluxes computed
by numerical weather models and reanalysis project suggest:
- Heat fluxes from the NCEP and ECMWF reanalyses have similar global
values, but the fluxes have important regional differences. Fluxes from the
Earth Observing System reanalysis are much less accurate (Taylor, 2000: 258).
- The fluxes are biased because they were calculated using numerical
optimized to produce accurate weather forecasts. The time-mean values of
fluxes may not be as accurate as the time-mean values calculated directly
- The simulation of boundary-layer clouds is a significant source of
error in calculated
fluxes. The poor vertical resolution of the numerical models does not adequately
low-level cloud structure (Taylor, 2001).
- The fluxes have zonal means that differ significantly from the same
zonal means calculated from ICOADS
(International Comprehensive Ocean-Atmosphere Data Set) data. The differences
can exceed 40
- The atmospheric models do not require that the net heat flux
time and Earth's surface be zero. The ECMWF data set averaged
years gives a net flux of 3.7 W/m2 into the ocean. The NCEP
reanalysis gives a net
flux of 5.8 W/m2 out of the ocean (Taylor, 2000: 206). ICOADS
data give a net flux
of 16 W/m2 into the ocean (figure 5.7).
Thus reanalyzed fluxes are most useful for forcing climate models needing
actual heat fluxes
and wind stress.
data are most useful
for calculating time-mean fluxes except
perhaps in the southern hemisphere. Overall, Taylor (2000) notes that there
are no ideal data sets, all
have significant and unknown errors.
Output From Numerical Weather Models
Some projects require flux data a few hours after after observations are collected. The surface
analysis from numerical weather models is a good source for this type of data.