John Abatzoglou
Currently a senior studying atmospheric science at the University of
California at Davis. He is interested
in weather modeling (both synoptic and mesoscale), atmospheric phenomena, as
well as climatic patterns (e.g. El Nino).
His project examined the effect certain atmospheric parameters (clouds,
moisture, temperature) have on the quantity of longwave radiation reaching the
surface of the Earth.
University of California Davis
Mentor: Dr. Catherine Gautier
ESTIMATING DOWNWARD LONGWAVE RADIATION BASED
ON
CLOUD, MOISTURE, AND TEMPERATURE PARAMETERS.
ABSTRACT
Based
on a compulsive series of model runs, an algorithm was derived which estimates
the downward longwave irradiance at the surface of the Earth given appropriate
input from surface and satellite data surrounding cloud, moisture, and
temperature. A cloud-base model was also implemented to approximate a
cloud-base altitude given a certain cloud top altitude and cloud optical depth.
The algorithm was then tested with TOGA COARE surface and ISCCP satellite data
taken in the western-equatorial Pacific from November 1992 through February
1993, and further compared with downward longwave radiation estimates predicted
by ISCCP. Data was obtained from two separate semi-stationary ships and
assessed on a weekly and hourly basis. The results indicate that the
estimations made were of higher accuracy than those made by ISCCP; however, the
estimated values did not appear to have a strong correlation to the measured
values. Our errors were on the order of 2% - compared to 10% errors for ISCCP.
Further, it was conveyed that more accurate data on a reduced spatial scale is
needed in coupling satellite and surface measurements in order to accurately
estimate downward longwave radiation.
INTRODUCTION
Meteorologists
may have a hard enough time forecasting what the weather will be like tomorrow,
let alone forecasting the climate of our planet years from now. In terms of a
long-term climatic prediction, you might assume climate modelers throw darts.
But through a thorough understanding of the Earth's climate forecasters can
better model our future climates. One of the biggest research topics concerning
climatologists today is the interaction between land, ocean, and atmosphere on
a global scale. The close bind between the energy exchanges at the interfaces
of these media account for much of the circulation patterns our Earth endures.
Moisture, temperature, and the role of clouds are some of the atmospheric
variables that are of particular interest in climate modeling. In the following
research, I set out to observe how these three variables contribute to
downwelling longwave radiation (DLR) reaching the Earth's surface.
EARTH RADIATION BALANCE
If
our planet is constantly being heated by the sun why isn't it getting hotter
and hotter? The answer lies in the equation of Earth Radiation Energy Balance. This formula accounts for the shortwave
radiation from the sun, the longwave radiation emitted by the Earth and its
atmosphere, and the latent and sensible heat fluxes. Overall our planet is able
to maintain a constant temperature through a complex interaction between the
sun, our atmosphere, and the surface of our Earth. The distribution of the net
radiation dictates climate
patterns on a global scale.
Let
us specifically look at downward longwave radiation (DLR). DLR is the component
of longwave radiation that is emitted by the atmosphere, which reaches the
Earth's surface. Note that longwave radiation is either absorbed by or
transmitted through a layer. That means that the further a distance an emitted
energy impulse must traverse, the larger amount of this energy will be lost in
absorption or attenuation by the atmosphere. Hence, a majority of the DLR
reaching the Earth's surface will come from levels (with atmospheric emitters)
close to the surface. Since moisture and clouds are two highly variable and
important contributors to DLR, knowledge of these values, and their respective
emitting temperatures, is of utmost importance. Therefore our three variables,
clouds, moisture, and temperature, will prove to be highly valuable in
estimating DLR.
THE PERPLEXING PARAMETERS
CLOUDS
Clouds
affect climate, and are in turn clouds are affected by climate. This complex
two-way interaction is what makes the impact of clouds so mystifying in
determining the Earth's radiation balance. What role do clouds play in the
energy budget? Do they cool our planet by reflecting incoming oslar shortwave
radiation? Or do they in fact warm the Earth by trapping in much of the
outgoing longwave radiation?
We
know that clouds cool Earth in the shortwave spectrum by reflecting incoming
sunlight. If you have ever been outside on a hot day you would probably have
seeked out relief from the heat in the shade of a tree. In the shade much of
the direct solar radiation is reflected. In the same way clouds reflect solar
radiation and cool our planet, the shade of a tree cools you. Data reveals that
a cloudless Earth would be some 12 degrees warmer [Harrison, et. al, 1990].
However, we also know that clouds
warm our planet by absorbing outgoing heat emitted from the surface and
re-radiating it back down. So if you are cold you put on extra layers of
clothing, right? Your body is able to capture more of its emitting heat in
these layers, which can further be re-radiated back towards the body itself.
This blanketing effect, affectionately called the "greenhouse
effect", that clouds have warms our planet's’ surface by about seven
degrees[Harrison, et. al, 1990]. Estimates boast that clouds contribute over
half of the re-radiated energy the surface receives from the atmosphere [Warren, et al, 1988]. The greenhouse effect is crucial to our
planet, without our atmosphere, surface temperatures would be some 33 degrees
cooler!
So clouds cool in the shortwave and
warm in the longwave. Studies have shown that the influence of clouds results
in an average additional 44.5 W/m^2 of shortwave radiation departing the top of
the atmosphere, while their blanketing effect trapped some 31.3 W/m^2 of the
longwave [Harrison, et. al, 1990]. From this we could simply extract that
clouds cool our planet. But the equation is not so easy; what really matters is
the net effect clouds have on our planet. In other words how do other
atmospheric variables and constituents interact with clouds and with cloud
formation?
If our planet experiences an
enhanced greenhouse effect, or "global warming", will the extra heat
promote increased evaporation hence cloudiness, reflecting more shortwave
radiation and retarding the warming process? Or perhaps these extra clouds will
add to the warming process by emitting greater thermal radiation towards the
surface. These are questions that researchers need to answer if we are to solve
this climatic puzzle. The answer lies in the species of clouds residing in the
skies above. High, thin clouds imply warming conditions as they tend to trap
thermal emissions while being translucent to solar radiation. On the other
hand, low, thick clouds have a cooling effect as they efficiently reflect in
coming solar radiation while still emitting high rates of thermal radiation. A
warmer climate will include a different variety and distribution of clouds.
Enumerating the influence, these clouds have on the energy balance and properly
quantifying cloud distributions will allow climatologists to more accurately
predict what the future climates hold in store for us.
The
presence of clouds will obviously enhance DLR as water droplets, which makeup
clouds, serve as emitters of radiation. Thick clouds may act opaque or
"black", allowing no DLR radiation from higher levels to penetrate
the clouds. Hence from these "black" clouds all the DLR comes from
the cloud level and below. Thinner clouds emit radiation at weaker amplitudes
and may be somewhat transparent, thus allowing some DLR radiation from higher
levels to contribute to the total DLR.
The
magnitude of DLR from clouds is directly proportional to their cloud base
emitting temperature. Therefore, one would agree that low clouds, with warmer
bases, contribute much greater to DLR than do colder, higher clouds. However,
we can not easily measure the ever-so-important cloud base height parameter
from satellites; we can measure cloud top pressure and cloud optical depth.
Using these two parameters and a cloud classification scheme we can estimate the
geometrical cloud thickness and approximate a cloud-base altitude [Ridout and Rosmond, 1996]. Our cloud parameters become as follows:
optical depth of cloud, cloud base altitude, and fractional cloud coverage.
MOISTURE
Water vapor is the most variable of
all the trace gases in our atmosphere, and though many tout carbon dioxide to
be the leading greenhouse gas, it is in fact water vapor that leads most to the
re-heating via DLR. Therefore we would expect the amount of water vapor to be a
lead component in our computation of DLR.
Thinking
of water vapor as an atmospheric constituent and emitter of longwave radiation
we would expect that an increase in water vapor content would lead to an
increase in DLR. We would be correct in making this assertion, as a moister
atmosphere is conducive to increased DLR. Knowing the total moisture content of
an atmospheric column is good, but knowledge of the profile of that moisture is
great. Since a majority of DLR comes from layers close to the surface of the
Earth we would expect that very moist lower layers would contribute more to DLR
than relatively moister upper layers. Henceforth, both the total moisture and
surface moisture parameters are beneficial in estimating DLR.
TEMPERATURE
Back
to the basics. Electromagnetic radiation ultimatley is derived from a body
emitting at a given temperature. All bodies whose temperatures are above
absolute zero emit radiation. Emission is directly proportional to temperature
by the Stephan-Boltzman Law. In simple terms: warmer objects emit more
radiation. Therefore it is only reasonable to expect that the temperature
parameter will have a large influence on DLR. Again, since most of the DLR
received at the surface comes from lower layers, the temperature at these
layers is of utmost importance to DLR. Therefore, knowledge of the surface
temperature stands king in terms of such variables.
The rate of change of temperature
with height, or lapse rate, does play a slight role in DLR. A very unstable
atmosphere will cool quickly with height; thus higher layers will emit lower
levels of radiation to the surface and lead to a decrease in DLR. Interestingly
such conditions are conducive to thick convective cloud formation, which then
counteract the decrease in DLR. Although small in magnitude, the atmospheric
lapse rate along with the more important surface temperature parameter are both
needed in our algorithm. Note: In the data collected we were not able to
detail a lapse rate. In each case the standard lapse rate was assumed. Errors
by assuming such a lapse rate are on the order of 2 W/m^2.
METHODS / MODEL
DERIVATION
I
derived the gist of my algorithm to model DLR through an ingenious software
tool called SBDART. Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART) is a
software package written in FORTRAN designed to calculate various radiative
fluxes in solving the Earth'’s radiative energy equation [Ricchiazzi et al,
1998]. Through a compulsive
series of crunching the variables of interest into the model I was able to
derive high order polynomials which interpolate DLR as a function of each
separate variable. The cloud (optical depth and cloud base), temperature
(surface temperature and lapse rate), and moisture (total precipitable water
and 900mb water) parameters were all assessed against a set "standard" tropical oceanic atmosphere. Other variables
including the cloud parameters, net effective radius and cloud relative
humidity, as well as concentrations of other atmospheric particulates,
including carbon dioxide and ozone, were found to have negligible effect on
DLR, and were thrown out of the model.
DLR = F1(cloud parameters) + F2(surface parameters) + F3(atmospheric
parameters)
As
expected [Frouin
et al, 1988], DLR was
positively correlated with temperature and moisture parameters. My model gave
especially strong weights to surface temperatures and total precipitable water
parameters, as these in fact dictate the lapse rate and surface moisture
parameters. We assumed, as had been done so [Frouin, 1988], that total DLR was
merely a linear combination of clear sky DLR and cloudy sky DLR.
DLRtotal = N*(DLRcloudy) + (1-N)*(DLRclear)
Special attention was given to
different types of clouds in terms of classifying geometric cloud thickness via
a polynomial involving cloud base altitude and optical depth. The model was
strictly based on the results obtained through meticulous calculations and
iterations of SBDART and the number crunching power of MATLAB.
DETAILED DERIVATION
Problem: Rectifying cloud base
height (ZCLOUD) through the known and readily available satellite products
cloud top height (ZTOP) and cloud optical depth (TCLOUD).
Motivation: Clouds (assuming
opaqueness) act as emitters of longwave energy (including DLR) - emitting
radiation at their given temperatures. The remaining DLR is derived from other
atmospheric emitters (mainly water vapor) in layers both above (though much is
lost in "black" cloud) and below the cloud base. Hence, cloud base
temperature (calculated from altitude) is critical in our estimation of DLR.
Radiative transfer models assume clouds to be at a certain altitude, existing
as a homogenous layer; however studies have shown that there are some serious
errors with this simplification. Further it is important to realize that there
exist large variabilities amongst these distinct cloud cases. Time permitting,
the algorithm used to determine cloud-base altitude should take into account
these cloud pattern schemes such as developed in the Intercomparison of
Radiation Codes in Climate Models Phase III (ICRCCM) project.
Idea: Although there is no simple
way to derive a cloud base height given the two input parameters, we can assume
a model which allows us to specify a geometric cloud thickness, hence base
height, for the parameters involved.
ISCCP Cloud Scheme
The particular cloud scheme used in
the approximation is taken from the ISCCP stage C2 data analysis [Rossow, 1991]. Cloud types are classified according to
ZTOP and TCLOUD. We specify seven different cloud types according to cloud-top
altitude: high-top clouds (cirrus, cirrostratus, and deep convective),
medium-top clouds (altocumulus, nimbostratus), and low-top clouds (cumulus,
stratus). Each cloud is assigned a given cloud thickness in millibars. As for
clouds with large optical thicknesses (see stratus, nimbostratus, and deep
convective), we artificially model increased geometrical thickness of a cloud
layer with increasing optical depth through an empirical formula that allows
the base of these clouds to slope down to 1 km (for nimbostratus and deep convective)
and to the surface (for stratus). This scheme has possible difficulties and
should further be improved upon.
Quick Hydrostatic
Given a certain cloud thickness in
pressure units we can calculate the thickness in meters by using a quick
hydrostatic approximation (dp= - rho*g*dz). Using the equation of state and
differentiating we can see that this requires a temperature profile of the
atmosphere. This profile is derived assuming that of the "standard"
tropical atmosphere. Basically we assume a constant lapse rate of 6.5 K/km plus
or minus the stability parameter LRD (lapse rate deviation). We then compute a
first approximation for the cloud depth based on the mean atmospheric
temperature (T(ZTOP)+T(0))/2. Having computed this value we get a first approximation
to our cloud base altitude; however we can further improve upon this by
reiterating using a mean atmospheric temperature between ZTOP and our initial
estimation of ZCLOUD. This gives us an improved estimation of our cloud base
altitude.
COMPUTER PROGRAM:
zbottom.m
This is the main program that computes the cloud-base altitude given the input parameters ZTOP, TCLOUD, LRD. The cloud-parameterization here assigns each cloud-type a given pressure depth. The algorithm uses a reiteration of an initial estimate to get a new and improved (hopefully) hydrostatically sound estimate for ZCLOUD. A bottom limit is given for thick clouds. Stratus may extend to the surface, whereas nimbostratus and deep convective clouds bottom out at the 1-km level. For a graphical example for a typical cloud base scheme assuming "standard" tropical atmospheric conditions click here.
T.m
This subprogram infers a temperature
for a given altitude and temperature profile. Here we assume a constant lapse
rate of 6.5 K/km plus or minus LRD. The program goes on to compute the
temperature at the given altitude computed from the surface temperature
parameter.
P.m
This subprogram infers a pressure
for a given altitude and thermostructure of the atmosphere, assuming an
atmosphere in hydrostatic balance with a surface pressure of 1013.25 mb. We
make use of the average temperature of the atmosphere by taking the average
between the top and bottom of the layer at hand.
Motivation: Clouds both warm and
cool the Earth: by re-radiating outgoing terrestrial radiation and by
reflecting solar radiation, respectively. The net effect of clouds on the
Earth's radiation budget requires knowledge of the type of cloud and its
radiative properties. Knowledge of the effect certain cloud types have on the
Earth'’s energy flux can lead to improvde estimates on climates of the future
and the role clouds play in global climate.
Idea: Through numerous iterations of
the SBDART model we varied the optical depth (TCLOUD) and cloud base altitude
(ZCLOUD) to test the sensitivity of these parameters with DLR. It was
discovered that increased DLR was associated with increased TCLOUD and
decreased ZCLOUD, as expected. Further, it appeared as though a plateau was reached
for clouds where TCLOUD>11, with the effects of increased thickness being
most pronounced at smaller levels of thickness. As for ZCLOUD, not
surprisingly, the lower the cloud bases height, the warmer the temperature from
which the emission will come from. The model seemed to have some difficulties
with high, thick clouds; however these problems were dismissed due to the
physical improbability of such clouds. Best-fit polynomials were assigned for
given ZCLOUD levels (as a means of improving the fit), as were clouds with
TCLOUD values in excess of 11 tau. As for clear skies, repetitive SBDART
revealed that clear skies under "standard" conditions yielded a DLR
value of 400 W/m^2.
COMPUTER PROGRAM:
dlwrad.m
This is the program that ties all
the strings together. All moisture, temperature, and cloud parameters are sent
in and we magically extract our DLR value. Special attention is given to clear
sky conditions and those conditions where TCLOUD>11. A best-fit polynomial
is assigned for both TCLOUD and ZCLOUD values. Having computed the baseline DLR
for the specified cloud scheme - we now modify this number due to the effect of
both the temperature and moisture profiles of both the atmosphere and the
surface. Again, recall that these are all modifications against the
"standard" tropical oceanic atmosphere.
Motivation: Since water vapor is one
of the prime emitters of longwave radiation, knowledge of its concentration in
the vertical is necessary in accurately solving the radiative transfer
equation. A moist atmosphere containing a high water concentration will have a
larger component of DLR than a dry atmosphere. The temperature structure of the
atmosphere also plays a role as the temperature at the various layers is
directly used in the Stephan's Boltzmans equation in calculating the emission
from that layer. By using the temperature and moisture profiles of the given
atmosphere we can improve upon our algorithm by taking these factors into
account.
Idea: Through numerous iterations of
the SBDART model we varied the temperature and moisture profiles of the
atmosphere to arrive upon a range of DLR values. Assuming in the tropics that
the moisture and temperature profiles are fairly consistent in shape we
quantify differences from the "standard" in terms of a ratio
(moisture) and deviation (temperature). Moisture Profile Difference (MPD) is
defined as the ratio of total precipitable water in a column of air versus that
of the "standard" tropical atmosphere. A MPD of 1 would match that of
the "standard", whereas a MPD greater than 1 would indicate a moister
atmosphere. Lapse Rate Deviation (LRD) is defined as the average lapse rate in
the troposphere minus the average lapse rate in the "standard"
troposphere. The "standard" troposphere has an average lapse rate of
0.0065 K/m. Hence an unstable atmosphere leads to a positive LRD value, for
example. The model was run for both clear and cloudy sky cases, also accounting
for the influence of clouds. The data set was then processed through a Taylor
polynomial to get a best-fit approximation.
COMPUTER PROGRAM:
atmos.m
This program alters the computed DLR
by taking into account the moisture and temperature profile of the atmosphere
at study. The LRD and MPD parameters are passed in, along with TCLOUD (used to
differentiate between clear and cloudy). Through a rigorous data run we have
computed polynomials to account for the effects of both moisture and
temperature in cloudy and clear conditions. Upon evaluation of these
polynomials with the appropriate values (note: MPD-1) we make an adjustment to
our computed DLR.
Motivation: As noted above, the
profile of the atmosphere is important, but of special importance is that of
the bottom of the atmosphere, or the surface. Knowing the surface temperature
can lead to deriving the emitting radiative temperature at any layer using the
temperature profile, or lapse rate. Of lesser importance is the moisture
content at the surface. Given a certain total precipitable water as specified
in the atmospheric adjustment segment we can approximate the moisture at the
surface, assuming a typical moisture profile. Upon doing so we can observe the
effects of having relatively moist (increased DLR) or relatively dry (decreased
DLR) surface conditions.
Idea: Through numerous iterations of
the SBDART model we varied the temperature and moisture levels at the surface
to arrive upon a range of DLR values. Assuming the "standard"
tropical atmosphere we pass in a surface temperature deviation (STD) which is
the actual surface temperature minus this "standard" temperature.
Likewise the surface moisture deviation (SMD) is computed as the moisture at
the surface minus the "standard" surface moisture taking into account
the MPD ratio. Upon running the model we noted a distinct variation regarding
the surface parameters with total atmospheric moisture. Hence we examined the
surface effects under three atmospheres: 1) dry (MPD<0.75); 2) normal
(0.75<MPD<1.5); 3) moist (MPD>1.5). The data set was then processed
through a Taylor polynomial to get a best-fit approximation for STD and SMD for
each case.
COMPUTER PROGRAM:
surf_adj.m
This program alters the computed DLR
taking into account the moisture and temperature at the surface. STD and SMD
are the parameters passed in, along with MPD (used to differentiate between the
three moisture schemes). Through a rigorous data table evaluation we have
computed polynomials to account for the effects of both moisture and
temperature in three cases. Upon evaluation of these polynomials with the
appropriate values we make an adjustment to our computed DLR.
DATA
As
a means of testing my data I dug through the Tropical Ocean Global Atmosphere -
Coupled Ocean-Atmosphere Response Experiment (TOGA COARE ) datasets in search of longwave radiation data. The TOGA COARE program
was devised to look at the relationship between the heat and energy transfer
from the tropical oceans (big heat source) poleward and global circulation
patterns. In short, in our global energy budget the equatorial
regions receive excess heat, which they then transport to higher latitudes
driving many of the Earth's circulation patterns.
The
R/V Xiangyanghong #5 (RX) data was taken from November 1992
through February 1993 on a ship stationed at 2 05'S, 154 30'E. Surface meteorological data as well as radiation data was recorded
aboard the ship. Radiation measurements were taken with an Eppley radiometer.
Over the same time period International Satellite Cloud
Climatology Project (ISCCP)
data was obtained from satellites including an estimate of DLR for the entire
TOGA COARE campaign region (20S - 20N, 120E - 170 W) on a 2.5 x 2.5 degree
spatial scale. During the same time period data was collected aboard a ship, Franklin, which meandered around a similar locale. Surface meteorological measurements as well as DLR readings were
taken every 15 minutes over the time period. Similarly, ISCCP data was paired
up with the surface data obtained from the ship.
I
proceeded to compute weekly averages of the variables of interest for our
specific locale, such to use as input in our algorithm. Unfortunately there
were many gaps in the data, including a lack of surface pressure data for the
Franklin set (assumed pressure = 1010 mb), as well as numerous hourly gaps and
inconsistency problems. This reduced the size of our databank vastly; however
we did have enough continuity to compute our algorithm for nine separate weeks
for the RX, and for six weeks over the Franklin.
Unsatisfied
with the results via weekly averages, I reduced the temporal scale of
measurement down to hourly intervals. As a means of comparing data I looked at
the DLR for a 24-hour period (Nov 23) on an hourly basis using the Franklin and
ISCCP data. On a whole there were 21 hourly paired ISCCP/surface observations
and 15 separate paired weekly averages used in our validation analysis.
RESULTS/DISCUSSION
My
(JOHN) estimated
downward longwave irradiance
varied only between 407 and 436 W/m^2. However, this range may not be
unreasonable given the stability of atmospheric conditions in the tropics. The
actual measured irradiance varied from 406 to 446 W/m^2, while the irradiance
estimated by ISCCP varied from 433 to 492 W/m^2. Atmospheric conditions
remained fairly constant throughout the period. Temperature and moisture
parameters remained nearly homogenous (hence these parameters could not be
soundly validated), while ISCCP cloud parameters varied the most.
Overall
my algorithm boosted an error of only 2.1 %, while ISCCP's error was a whopping
10.0 %. Specifically the RX and Franklin
weekly averages yielded errors of 1.32% and 1.48%, while the hourly Franklin resulted in an error of 2.9%. Looking at the consistency of the measured
variables one can only expect to have a fine outcomming range of irradiance
values. ISCCP data seemed to overestimate on many behalves. Not only were their
estimated DLR measurements in excess, but so (one may speculate) might have
been their moisture content and cloud parameters. When comparing my DLR estimates to the estimates
generated by ISCCP there is
a very strong correlation. This points towards the source of error being the
data itself!
Though
my estimated values did not differ significantly from the measured values, the
positive correlation was not strong. Upon further investigation my estimations
apparently gave too much weight to both the total atmospheric moisture, as well
as the optical depth of the cloud. Further refinement on behalf of these parameters
may lead to better results for this particular dataset.
Moisture
errors seemed to be due to an issue of overestimation. The presence of water
vapor may be the largest contributor to DLR, however we may overestimate its
radiative potential in our model. But due to the extremely fine variability of
values we can not show any significant evidence that this alone is the problem.
Total precipitable water estimation was particularly large relative to our
locale, hence another possible source of error. Also, ISCCP moisture parameters
do not seem to vary significantly over a temporal scale. Total precipitable
water and 900 mb layer moisture data seemed to be represented in a step-like
fashion, only being updated every 12-16 hours. This is not sufficient data to
make finite DLR estimates, especially on an hourly basis. Hence knowledge of
precise moisture content (especially total precipitable water) is crucial in
our radiative flux calculation.
The
errors associated with the optical depth may be due to the cloud classification
algorithm and scheme. Cloud classification is very difficult to define, and
becomes even fuzzier when taken over such a large time scale. The cloud
classification scheme may need to incorporate a larger variety of clouds,
possibly specific to climate of the particular locale. Tropical clouds are
relatively consistent, especially in terms of base height, hence it might be
wiser to assume cloud-base altitudes from climatalogical data. Another possible
idea is to include an algorithm which will calculate the flux data without
assuming that there is just one homogenous cloud structure (something more
complex than the plane-parallel radiative transfer scheme considered). Allowing
and accounting for various types of clouds at different levels will result in
very different outcomes, though obtaining adequate measurements from satellites
may be trying. Cloud fractional coverage at numerous layers would be needed, as
would the respective optical depths and pressure levels of these clouds. Furthermore,
the problem of cloud overlap would need to be taken into account.
An
interesting observation was that of cloud fraction. It was assumed that DLR
would be directly proportional to fractional cloud cover; however error plots
reveal a tendency to underestimate DLR in decreased cloud cover, while
overestimating it in increased cloud coverage. Hence the once thought to be
linear relation between cloud fraction and DLR may be proven otherwise [e.g.,
Frouin et al., 1988]. Both JOHN and ISCCP estimates show a strong
positive correlation between DLR and fractional cloud coverage; yet the correlation between the actual measured
values and cloud fractional coverage show no such correlation. Looking at the ISCCP data we can see a very
strong correlation between estimated cloud-base heights and cloud coverage. There appears significantly larger cloud
coverage associated with higher cloud bases. Whether it is operational or
mechanical, it appears the ISCCP data may in fact miss much of the low
cloudiness, which is of vast importance in calculating DLR.
One
source of data error is the spatial scale of the data. ISCCP data was taken and
averaged over a fairly large area (2.5 x 2.5 degrees) and may not be indicative
of the conditions measured directly over the RX or Franklin sites. Although
variability over the tropical oceans may be small, site specific data may lead
to more accurate results. Variable and inhomogeneous cloudiness at in situ
sites is not captured through ISCCP data and can contribute to large errors.
Upon further investigation, ISCCP cloud data shows significant spatial
fluctuations for both the optical depth and cloud top pressure parameters.
Hence the stark variability and inaccuracy of the cloud parameters may explain
our errors in estimating DLR.
Upon
evaluating our particular ISCCP parcel spatially in comparison to other parcels
in the dataset along lines of similar latitude we saw some interesting
observations. Whether the findings are that of a climatological phenomena or of
an error in the ISCCP dataset is unknown. However, cloud top pressures and
optical depths were consistently higher for our locale when compared to
estimates for similar latitudinal locales.
CONCLUSIONS
ISCCP
and JOHN DLR estimates show a distinct pattern with one another (weekly and
hourly). Yet both seem to miss the mark when it comes to a correlation with in
situ measurements of DLR. This only leads us to believe that errors surrounding
the ISCCP data are to blame. Researchers have complained over the years on the
difficulty of obtaining accurate DLR data; however, retrieving precise cloud
data on a sufficiently small enough spatial scale may in fact prove more
difficult. ISCCP data and in situ measurements are not suitably paired together
to run our algorithms through if we expect to obtain accurate results. My
algorithm may not be a complete failure, as it still beats the ISCCP model; it
does capture in essence the effect cloud, moisture, and temperature parameters
have in the modeled atmosphere. The verification of such model output with
actual atmospheric measurements is needed to validate and further improve upon
our model and the core physics which drives our model.
ACKNOWLEDGMENTS
I would like to thank ICESS in conjunction with NASA for providing us students with insight into
the many domains of Earth Systems Science. I wish to thank my mentor, Prof. Catherine
Gautier, for her encouraging ideas and support on this project. I would like to
thank Allison Payton for help in obtaining background information and with the
SBDART system. I also would like to thank Pete Peterson for the ISCCP data and
a crash course lesson in IDL (helmet not included). Finally, I would like to
thank the TOGA COARE researchers for devoting countless hours in the tropics,
meticulously taking careful measurements and working hard on their tans.
REFERENCES
FROUIN R; GAUTIER C;
MORCRETTE J. DOWNWARD LONGWAVE IRRADIANCE AT THE OCEAN SURFACE FROM SATELLITE
DATA: METHODOLOGY AND IN SITU VALIDATION. JOURNAL OF GEOPHYSICAL REASEARCH,
JAN, 1988, V93(NC1):597-619.
GUPTA SK. A
PARAMETERIZATION FOR LONGWAVE SURFACE RADIATION FROM SATELLITE DATA - RECENT
IMPROVEMENTS. JOURNAL OF APPLIED METEOROLOGY, DEC, 1992, V31(N12):1361-1367.
HARRISON EF;
RAMANATHAN V. SEASONAL VARIATIONS OF CLOUD RADIATIVE FORCING DERIVED FROM EARTH
RADIATION BUDGET EXPERIMENT; JOURNAL OF GEOGRAPHY, 1990:18687-18703.
RICCHIAZZI P; YANG
S; GAUTIER C; SOWLE D. SBDART: A RESEARCH AND TEACHING SOFTWARE TOOL FOR
PLANE-PARALLEL RADIATIVE TRANSFER IN THE EARTH'S ATMOSPHERE. BULLETIN OF THE
AMERICAN METEOROLOGICAL SOCIETY, MAY, 1998 V79(N10):2101-2 114.
RIDOUT JA; ROSMOND
TE. GLOBAL MODELING OF CLOUD RADIATIVE EFFECTS USING ISCCP CLOUD DATA. JOURNAL
OF CLIMATE, JUL, 1996, V9(N7):1479-1496.
ROSSOW WB; SCHIFFER
RA. ISCCP CLOUD DATA PRODUCTS. BULLETIN OF THE AMERICAN METEOROLOGICAL SOCIETY,
JAN,1991, V72(N1):2-20.
WARREN SG, HAHN CJ,
LONDON J, CHERRIN RM, JENNE RL. GLOBAL DISTRIBUTION OF CLOUD COER AND AMOUNT
OVER THE OCEAN; NCAR TECHNICAL NOTE. NATIONAL CENTER FOR ATMOSPHERIC RESEARCH.
BOULDER, CO. 1988.