TMI/VIRS Derived Latent and Sensible Heat Fluxes over the Tropical Pacific Ocean
Balla Maggero
Marine & Oceanography, Meteorological Services, 30259, Ngong’ Road, GPO 00100 Nairobi, Kenya.
e-mail: nambuye@yahoo.com
ABSTRACT
This investigation is presented for the TMI/VIRS application to the problem of estimating monthly averages of air
temperature (ta), specific humidity (qs) and turbulent heat fluxes at the air-surface interface over the planet ocean.
The objective is directed at adaptation of retrieval mechanisms that have been used elsewhere with SSM/IAVHRR combination. In order to determine the impact, TMI/VIRS derived products are verified against
TRITON/TAO buoy array observations in the warm pool region. The results reveal extremely small biases in the
TMI-VIRS’s ta and qs. The root-mean-square errors are 0.61K and 0.92 gKg-1 respectively. This shows that the
sampling is sufficient for calculation of latent and sensible heat fluxes.
1. Introduction
The turbulent fluxes of sensible and latent can only
be measured directly by exacting in-situ
instrumentation such as on ship or buoys, but
various parameterizations allow them to be derived
from quantities observed globally from space-borne
passive microwave and passive infra-red
radiometers. They are commonly estimated with
bulk aerodynamic formulation or more sophisticated
boundary layer models; both approaches require
surface winds and air-sea differences of
temperature and humidity. As a rule, satellites
measure surface properties and atmospheric
properties aloft, but cannot provide air temperature
and humidity at the desired height above the sea
surface (eg. Chou et al., 1995, Curry et al., 1999).
So the required air-sea differences and the resulting
surface fluxes must be inferred from available data
and models. Remote sensed sea surface
temperature (SST), surface wind speed (Vs) and
total precipitable water (TPW) are primarily needed
as diagnostic variables for determining surface air
temperature, surface specific humidity and
subsequently turbulent heat fluxes. The air-sea
temperature difference is a controlling factor
because it determines the atmospheric boundary
layer stability, which in turn controls the efficiency of
the turbulent fluxes. The accuracy of air-sea flux
estimates therefore depend jointly on the
accuracies of the basic state variables; and the
validity of the parameterizations used to calculate
the fluxes from the measured quantities (Donelan,
1990). In the ensuing study, this paper aims to
demonstrate how the packages flown on TRMM
mission can contribute to
the Global Ocean Observation Systems. The
underlying motivation being estimation of turbulent
heat fluxes from sensors flying on the same
platform. Monthly averages of surface air
temperature (ta), surface specific humidity (qs) and
turbulent heat fluxes are derived from TRMM
Microwave Imager (TMI) and Visible infrared
Scanner (VIRS) measured brightness temperatures
and radiances over the warm pool region of the
Pacific Ocean (the loci and source of major
meteorological events of global concern).
Comparisons are made with the corresponding
measurements from the Triangle Trans-Ocean buoy
Network / Tropical Ocean-Atmosphere moored buoy
array reanalysis for the period 1999-2002 within a
120E180E, 25N30S domain.
2. Retrieval Techniques
Sea Surface Temperature (SST) and the Surface
saturation Humidity (qsst)
One of the best-known methods to estimate SST
from AVHRR measurement is the radiative transfer
model. The retrieval scheme employed for SST
employs a classical quadratic split-window approach
during daytime and a linear tripple-channel approach
during nighttime, both with a theoretical accuracy of
 0.4K on the instantaneous pixel level. The
advantage here is that the effect of cool skin of
ocean is inco-orperated. The missing values from
VIRS due to clouds or convective systems is
corrected by the TMI’s coarse resolution. From the
SST values, qsst can easily be determined using the
Magnus formula. The saturation humidity is then
reduced by 2% to account for the salinity of the
ocean.
Wind Speed (Vs)
The algorithm by Schlussel et al., (1995) is used to
estimate near-surface wind-speed from TMI
measurements. Vs is derived from the brightness
temperature difference between horizontally and
vertically polarized components at the same
frequency. The accuracy for the globally valid
passive wind speed retrieval is 1.4ms-1 under
conditions where no heavy rain hampers the
surface-leaving radiation in reaching the satellite. In
light rain cases, the accuracy decreases to 1.6ms-1
.
Air Specific Humidity (qs)
The water vapor content of the planetary boundary
layer is a direct function of the near surface
humidity over the ocean, (Schulz et al., 1993).
Hence, qs can inferred from TMI level 1B-11 data.
The retrieval algorithm is a linear scheme
employing 10, 19, 21, and 37 GHz brightness
temperature. This model was proved to provide
better agree-ment between SSM/I and in-situ qs
estimates (Schulz et al., 1997).
Surface Air Temperature (ta)
The ta is derived by the artificial neural network
(ANN) - approach described in Jones at al. (2001).
The ANN algorithm is a computer model of
individual elements commonly referred to as
neurons. The method and software on ANN is
available online at www.mcs.com/drt/. The ANN
algorithm used is a multi-layer perceptron (MLP)
with back propagation made up of neurons
arranged as input, output and intermediate (hidden)
layers connected by weighted links and estimates a
mapping function from one vector space X (inputs)
to another Y (outputs). The number of neurons in
the input and output layers is determined by the
dimensionality of X and Y. In this case, a four-layer
MLP architecture for ta as output is 3-10-3-1-X (Fig.
1). The inputs are total precipitable water (TPW),
sea surface temperature (SST) and wind speed, the
output is the target surface air temperature ta.
Figure 1. Schematic representation of the ANNmulti-layer perceptron (MLP) architecture used to
estimate ta.
3. Turbulent Heat Fluxes
High Qs and Ql mean bias, RMS and correlation
coefficient values are observed. The reason may be
attributed to wind or sampling error. The buoy
anemometers are not sensitive to very low winds.
The spatio-temporal evolution of surface specific
humidity, surface air temperature and the latent heat
are displayed in figure 2 obtained from model results.
Strong values are depicted north of Australia.
Because of a strong relation between SST and ta,
remarkable features are noticeable; representation
of the warm pool region north of Australia.
The latent (Ql) and Sensible (Qs) heat fluxes are
then estimated by the following bulk formulation;
Ql = Cd Vs  qs - qsst ---------------- (2)
Qs = Cd Vs  ta - SST -------------- (3)
where  and Cd are air density and drag coefficient
respectively. Cd is estimated according to the Louis
et al. (1982). Note that the model SST is used in
estimating the sensible heat flux, implying that the
air-sea feedback process are considered to some
extent.
4. Preliminary Findings
Several measures of the closeness of the TMI/VIRS
estimates to the ground truth, i.e. TRITON/TAO
values are the bias, root-mean-square (RMS.), and
correlation coefficient.
The results are presented in table 1. The biases for
the qs and ta of -0.077gKg-1 and 0.06K respectively.
The RMS were 0.921gKg-1 and 0.611K, and
correlation coefficients of 0.844gKg-1 and 0.860K
respectively for qs and ta.
Element
ta (K)
qs (gKg-1)
Qs (Wm2)
Ql (Wm-2)
Mean
Bias
0.06
-0.077
17.15
14.81
RMS
0.611
0.921
55.6
52.23
Corr.
Coeff.
0.860
0.844
0.582
0.616
Table 1. Intercomparison results between TMI-VIRS
derived ta, qs Ql Qs with TRITON data.
5. Summery
This paper has described the SSM/I-AVHRR
retrieval techniques for estimation of surface air
humidity (qs), surface air temperature (ta) and applied
them to TMI-VIRS. The results for the flux fields
have shown a reasonable agreement and more or
less converging to the same patterns with the
TRITON buoy array datasets.
Precipitation Measurement (GPM) should be
successiful. finer spatial/temporal time scale desired
for most application could be achieved. The relatively
dense sampling would be tremendiously useful in the
study of these fluxes: more uniform error
characteristics on the
Acknowledgement:
The Author is thankful to Dr. Bentamy and two
anonymous reviewers for their constructive comments
and suggestions for this manuscript. Data used in this
study include data provided by the NASA-funded EOS
data and information system archive at Goddard Space
Flight Centre, Greenbelt, MD, and the Marshall Space
Flight Centre, Huntsville, Alabama, USA.
Sea-truth data is by the TRITON/TAO Project Office at the
(JAMSTEC), and the PMEL, NOAA for the buoy data
used. Don Tveter developed the ANN program.
References:
Figure 2. mean evolution of top: ta – contour interval 1K,
centre: qs – contour interval 2gkg –1 and bottom: Ql contour interval 20 Wm-2. Large values show heat
transfer from atmosphere to the ocean. The loss of latent
heat is not very significant.
The statistics reveal extremely small biases in the
TMI-VIRS derived qs and ta. The RMS errors are
0.611K and 0.921 gKg-1, and large correlation
coefficients of 0.860 and 0.844, respectively. Qs and
Ql gave large values for mean bias and RMS errors.
It may be concluded that the errors lie with
insensitive of the buoy anemometers to very low
winds and or the sampling error. The TMI-VIRS
observations may provide an opportunity to make
contribution in turbulence estimation and it is the
author’s hope that the proposed Global
Chou, S.H., R.M. Atlas, C.L. Shie, J. Ardizzone,:
Estimates of surface humidity and latent heat fluxes over
the oceans from SSM/I data. Mon. Wea. Rev., 123, 24052425, 1995.
Curry, J.A., C.A. Clayson, W.B. Rossow, R. Reader, Y-G,
Zhang, P.J. Webster, G.Liu, and R.S. Sheu: Highresolution satellite-derived dataset of the surface fluxes of
heat, freshwater and momentum for the TOGA COARE
IOP. Bull. Amer. Meteor. Soc., 80, 2059-2080, 1999.
Donelan, M.A.: Air-sea interaction. The sea: Part B Ocean Engineering Science, 9, (B. Lenchaute, and D.
Hanes, Eds.), 239-292, 1990.
Jones, C., P. Peterson, C. Gautier and W.T. Liu, Satellite
observations of Latent heat fluxes in the Tropical Pacific
Ocean. WCRP/SCOR workshop on Intercomparison and
Validation of Ocean-Atmos. flux fields, Potomac, MD,
USA. WMO/TD-NO. 1083, pp 209-214 2001.
Louis, J., M. Tiedtke and J. F. Geleyn, A short history of
PBL parameterization at ECMWF. Workshop on Planetory
Boundary Layer parameterization, ECMWF, 59-80 1982.
Schulz, J., P. Schlussel and H. Grassi, Water Vapor in the
atmos. boundary layer over oceans from SSM/I
measurements. Int. J. Remote Sens., 14, 2773-27891993.
Schulz, J., J. Meywerk, and S. Ewald: Evaluation of
satellite-derived latent heat fluxes. J. Climate, 10, 27822795, 1997.
Schlussel, P., L. Schanz, and G. Englisch: Retrieval of
latent heat flux and longwave irradiance at the sea surface
from SSM/I and AVHRR measurements. Adv. Space.
Res., 16, 107-116, 1995.
Shlussel, J. and A. Albert: Latent heat flux at the sea
surface retrieved from combined TMI and VIRS
measurements of TRMM. Int. J. Remote Sens., (in press),
2001.