Skip to main content

Full text of "NASA Technical Reports Server (NTRS) 19820025063: Evaluation and comparison of tropical analyses during DST-5 and DST-6"

See other formats



UHMET 80-05 


James C. Sadi el" 4 

Department of Meteorology 
University of Hawaii 

Final Report (Hawaii Univ. ) 31 p 

HC A03/MF AC 1 CSCL 04B 





Final Report 

National Aeronautics and Space Administration 
Goddard Space Flight Centen 
Grant No. NAS 5-25376 ^ 



Data systems Tests (DST) 5 and 6 were conducted to assess the adequacy of 
the global data base for numerical analysis and forecasting and In the pi’oeess 


to determine the Impact of meteorological satellite data, A voluminous 
literature has accumulated on the evaluation of satellite data, particularly 
satellite soundings, and their Impact on analysis and forecasting^ however, 
the verdict Is not yet clear. Ghil et a1 ., 1979; Atlas et al . , 1979; and 
Atlas, 1979 reported modest improvement in analyses and forecasts using 
satellite data with the Goddard Laboratory for Atmospheric Sciences (GLAS) 
models and In selected cases the satellite soundings provided critical data 
In particular weather situations which led to significant positive impacts 
relative to either the GLAS or National Meteorological Center (NMC) models 
using no satellite data. Tracton and McPherson, 1977; Miller and Hayden, 1978; 
and Tracton et al . , 1980 reported no significant improvement with NMC models 
and even a negative impact during DST-5 tests. The results of these exhaustive 
tests indicate that the satellite data impact is probably model- and season- 
dependent but definitely dependent on the method of data assimilation and the 
numerical model used to produce the analyses or initial conditions. None of 
the tests have included an evaluation of the satellite data impact in the 
tropics for this is a much more difficult task. The pressure height gradients 
are small and there is little "ground truth" for evaluating the global sound- 
ings. The primary analysis within the tropics must be of the wind field and 
the satellite contribution is restricted to essentially two levels and only a 
portion of the globe unless there is a dedicated program of bogusing data from 
polar-orbiting satellites. The only extensive tropical forecast experiment 
was for a 5-day period in 1965 (Miyakoda et al . , 1974). The results were mixed-- 
good and bad. For the first 24-hour forecast the initial conditions of the 


height and wind fields at all 10 levels were detennined by an experienced 
’ • * 
analyst. The first 24 hour forecast of the tropical precipitation pattern, 

based on this subjective analysis, compared well with observations of deep 

convective cloudiness observed by satellite. This severe test indicated that 

the model structure and physics were capable of an adequate forecast given 

the proper initial conditions. However, subsequent forecasts, based on updated 

numerical model analysis and initialization, deteriorated rapidly. This one 
test supports the contention of Atlas (1979) that the impact of satellite data 
depends primarily on improving the initial conditions and not on differences 
in forecast models. 

This report discusses a limited exercise to evaluate the adequacy of the 
PST data base for analyses in the tropics. A few days of subjective analyses 
at two levels are compared with numerical analyses from the models of the 
Goddard Laboratory for Atmospheric Science (GLAS), National Meteorological 
Center (NMC), and NMC modi fled by the University of Hawaii under a separate 
study conducted by Murakami (Sumi, 1980). The comparisons are both subjective 
and objective. Subjective comparisons were made of the vorticity and divergence 


fields as well as the wind fields and synoptic systems. The energetics of the 
upper troposphere over selected regions determined by Sumi (1980) for the 
different analyses are also discussed here. In addition, GLAS tropical 
analyses and the utility of satellite soundings and derived winds in subjective 
tropical analyses are evaluated. 

(1 ) Data Base 

The primary data base was the level Il-b data compiled for the 
DST-5 and DST-6. The file contained (a) conventional data received at NMC 



enhanced by extending the cutoff time to VO hours} (b) VTPR soundings; (c) 
cloud motion vectors together with an enhanced tet‘ provided by NESS; (d) a 
special set of cloud motion vectors determined at University of Wisconsin; 

(3) Nimbus 6 soundings specially processed by 6ISS; (f) aircraft reports from 


specially equipped wide-bodied aircraft (AIDS); and (g) TWERLE data processed 
at NCAR. 

While the data set was Impressive, It must be emphasized that no 
special effort, beyond extending the cutoff time to 10 hours, was made to 
enhance the conventional data base and regular aircraft reports. The greatest 
effect of this Is reflected In the tropical regions where the normally scanty 
data stem from a combination of factors such as (1) relatively fewer stations; 
(2) missed observations; (3) poor communications; and (4) lack of a program 
for collecting AIREPs by most tropical weather services (Figs. 1 and 2). In 
Fig. 1 our analysis over Africa at 250 mb for 00 GMT, 3 September Is based on 
only five rawin observations supplemented by time and space continuity, 
climatology, bogused winds estimated from satellite cloud photographs and a 
few satellite winds in the Atlantic. Figure 2 shows the aircraft wind 
observations available for 12 GMT, 15 August 1974 during the GATE. No special 
AIDS data are Included. It Is unlikely that air traffic decreased between 
1974 and 1975. 

Compared with Africa the data over South America were as sparse; 

however, the satellite winds from NESS and Wisconsin permitted a reasonable 


analysis at 250 mb. At 850 nib adequate analysis was impossible over either 
South America or Africa. 

The primary data base was supplemented by satellite photographs 
from three sources: (1) the global band tropical mercator projection mosaics 
of NOAA-4 from NESS; (2) high resolution DMSP orbital strips borrowed from 

the archives of the World Data Center for Glaciology; and (3) |^]es West ‘ 1 

pictures front the NESS station In Honolulu. These were used to tbcate and I 

estimate the Intensity of synoptic systems such as tropical cyclones, tropical | 

Upper tropospheric troughs (TUTT) and embedded upper tropospheric cyclones, ^ 

“ I 

troughs, ridges, jet streams, lower tropospheric convergence zones, mid- c 

laH^’ide lows and fronts. Through established subjective models, wind fields 
were inferred from the cloud patterns (Anderson et al . . 1969; Sadler, 1963, 

1976a, 1976b, 1978). As an example, the direction and speed of the easterly 
jet entering the east side of Fig. 1 In the equatorial zone was Inferred from 
the satellite -observed cirrus clouds. This jet, extending from south of India, 
persisted arid on the next day (4 September) penetrated Into equatorial Africa; 

55 kt winds were observed at 250 mb at Nairobi. The increase from less than 
10 kts on 3 September to greater than 50 kts on 4 September could not have , 

been adequately interpreted without the history of the jet obtained from the 
satellite-observed cirrus. This anomalous jet, which deviated considerably 1 

from climatology, will be referred to later in a discussion of the energetics. j 

(2) Selection of Period and Levels for Analysis 

The days for analysis within each DST period were selected mostly 
on the basis of the tropical synoptic activity as revealed by the global 
satellite pictures. The selected DST-5 period of 30 August - 4 September 1975 
contained seven tropical cyclones in various stages of develd^ent and a well- 
developed TUTt in the North Atlantic and the North Pacific. In\\the eastern 
Pacific, Hurricane Katrina attained an intensity of 115 kts, Jewell became a 
minimum strength hurricane and the remnants of lisa passed north of Hawaii. 

Hurricane Doris foniied north of 30N In the Central Atlantic and Hurricane 
Caroline traversed the Gulf of Mexico and entered Mexico. Typhoon Tess formed 
near Guam and tropical storm Susan formed southeast of Japan. 



The DST-6 period of 29 February - 3 March 197v contained a typhoon 
off northeast Australia and a well-developed, TUTT 'in the southeast Pacific. 

In addition a deep trough In the westerlies with a strong frontal system 


passed through Hawaii. 

The standard levels of 850 mb and 250 mb were chosen for analysis 
to maximize the data by utilizing low- and high-level satellite cloud motion 

vectors and aircraft reports which are concentrated near the 250 mb level. 

* * 

(3) Data Plotting. Analysis and Grid Point Data Extraction 

A large scale (1:15,000,000) map, even though unwieldy (8 ft long) 
and time consuming In line drawing, was necessary to accommodate the large 
volume of satellite cloud motion vectors procesi-ed from GOES East and West 
over the Western Hemisphere. The Il-b data were plotted with varying plotting 
models to distinguish the different data types and sources. 

The analyst then noted on the charts Information which was obtained 
from other sources- -the position and intensity of tropical cyclones from 
published sources and the global satellite pictures; the position and intensity 
of other satellite-observed significant cloud masses associated with tropical 
disturbances, convergence zones, TUTTs, monsoons, fronts, trade winds, etc.; 
estimated winds from the polar-orbiting satellites outside the area covered 
by the geostationary satellites; the position of other surface features such 
as subtropical highs and ridges from the NMC global band surface analysis. 

The wind direction and speed were then analyzed using overlying 
acetate and various-colored grease pencils. This method facilitates the over- 
laying and ajustment of analyses to assure time, space and vertical continuity. 

After completion of all analyses, they were traced onto the plotted 
charts and grid point data were manually extracted at 2.5° grid spacing for 
subsequent use in comparison with other analyses and derived fields. 



The satellite sounding data were plotted on separate charts of the 
same scale and analyzed for thickness and temperature patterns. No analyses 
were available for the reference level (1000 mb); therefore, height analyses 
of the constant pressure surfaces were not accomplished. 


Prior to evaluating and comparing various tropical analysis schemes, 
suitability of the data for subjective analysis should be commented on. 
Subjective tropical analysis by an experienced analyst working In a research 
mode should be superior to an objective tropical analysis for many obvious 
reasons; however, the quality of both Is dependent on data; the greater the 
quantity and quality of data the more alike the analyses become and converge 
on the true solution. Even though I selected the levels of maximum data 
(excluding the surface) for this exercise, on the global scale there remained 
large areas with very few data; as discussed In the Introduction, analysis 
over these areas by any method Is little better than climatology. 

(1 ) Winds 

Within most of the Western Hemisphere, due to the GOES satellites 
and at least some effort In collecting aircraft observations, the DST data 
base approaches that needed for an adequate tropical wind analysis at two 
levels after making the reasonable assumption that the satellite and aircraft 
winds are representative of the flow at the selected levels of 250 mb and 850 
mb. To Illustrate the impact of the satellite winds. Eastern Pacific sections 
of the analyses are shown In Figs. 3 and 4. This Is a region of very few 
conventional observations and prior to the satellite era its meteorology was 
relatively unknown. The 250 mb analysis for 3 March (Fig, 3) obtained by 
meshing the few rawinsonde observations (circled), the few aircraft reports 


west of 180**' and a good distribution of satellite winds Is probably close to 
the true solution of this very complex circulation pattern caused by the 
extensive and strong tropical upper tropospheric trough (TUTT) and two 
Intense mldlatltude cyclones. The addition of data would not change the 
analysis significantly. The 850 mb analysis for 2 Septetrt>er (Fig. 4) with 
only satellite data over the ocean depicts a simple trade wind flow In the 
South Pacific and a rather complex monsoon type flow In the North Pacific. 

The trade wind flow Is well sampled but even In the north the few well placed 
satellite winds In combination with the satellite-observed cloudiness and 
synoptic modeling of Hurricane Katrina and decayed Hurricane Jewell permits a 
reasonable analysis. 

Adequate analyses of the pressure height and temperature fields 
have never been possible in the tropics due to the sparsity of rawinsonde 
stations, small gradients, and the mixture of rawinsonde types whose differences 
or biases are comparable to the synoptic gradients. 

In such a situation satellite soundings should have maximum impact. 

The spatial resolution of the data is good and from the same instrument the 
gradients should be accurately determined even though the absolute values may 
be biased. The problem of specifying the height of the reference level remains 
but even here, due to small pressure gradients, the error should be less than 
In higher latitudes (except in the vicinity of tropical cyclones). Tropical 
cyclones must be man-"bogused" into the data base whether it be wind or height 
analysis. Assuming the soundings are useful in the tropics, the added 
advantage is their availability for all standard levels. 

Our subjective analyses of the thickness field indicate that the 
satellite soundings can reasonably depict the height field at 250 mb over the 

8 . 

tropical oceans. Figure 5 shows an analysis of the 1000-250 mb thickness 
field from the Nimbus 6 soundings In the Central and Western Pacific on 
4 September. Added to the chart are the ridge and trough positions taken 
from the wind analysis. The positions of the N.H. subtropical ridge (STR) 
extending from China across Japan and northeastward, the N.H. subequatorlal 
ridge (SER) oriented east-west near ION, the S.H. ridge from northern Australia 
to the equator near 175W; the N.H. tUTT system extending across the West 
Pacific and separating the STR from the SER In eastern China and the omega 
pattern north of Hawaii are all reasonably well located by the satellite 
soundings; certainly better than could be obtained by determining the height 
field only from the few radiosonde stations. How much the pattern would be 
altered by the addition of the 1000 mb height field is unknown. 

It was noted during analyses of the thickness fields that certain 
areas have repetitive patterns which appear to be caused by strong horizontal 
near-surface thermal gradients anchored by topography. The best examples are 
over the mountains of eastern Africa, particularly in Ethiopia. The thickness 
pattern from 1000 to 850 mb (Fig. 6) is essentially the same as from 1000 to 
250 mb (Fig. 7) and the addition of the 1000 mb height would not appreciably 
alter this relationship over this area. It is highly unlikely that such a 
strong cold low exists through the troposphere and experience with upper 
tropospheric winds during GATE (see Fig. 2) indicate that these features are 
not real but must be due to some flaw in the method or assumptions in produc- 
ing the soundings. 





The plan to use subjective analyses to evaluate data and serve as a 
comnion base for comparing objective analyses in the tropics included both 


10 . 

tropical systems were not adequately depicted and no time continuity could 
be established. However* In the couse of attempting the comparison some 
general features of the GLAS analyses In the tropics were noted and will be 
discussed later. 

In conjunction with a parallel project under the direction of Prof. 
Murakami. University of Hawaii, comparisons were made between parameters 
derived from the Level III grid point data of the NMC and University of Hawaii 
objective analyses and our subjective analyses. The University of Hawaii 
analysis modified the NMC analysis over data sparse areas by applying a 
divergence term proportional to the satellite-observed outgoing longwave 
radiation (Sumi , 1980). 

(1) ReUtIve Vorticitv 

Figure 8 shows the 250 mb subjective analysis at 00 GMT on 1 Sep- 
tembar 1974 for the section from India eastward to 70W illustrating the 
typical data distribution and the contrast between the simple flow pattern of 
the'winter hemisphere and the complex pattern of the summer hemisphere. The 
wild data supplemented by satellite photographs and time continuity were 
sulficlent on which to base reasonable analyses for this section. The lack 
of satellite winds in the eastern South Pacific on 1 September was not typical 
of most days. The grid point winds between the equator and 20N on Fig. 8a 
ar‘:3 from the GLAS 300 mb analysis and will be discussed in a later section. 

Figure 9 shows the vorticity field derived from the analysis of 
Fig. 8 and for comparison Fig, 10 shows the vorticity derived from the Level 
III NMC 200 mb analysis. In general the vorticity patterns are alike. The 
main exceptions are in the southern Bay of Bengal, the western North Pacific 
and the South Pacific in the region centered near 12S and 173E. In general 
the subjective analysis vorticity is greater than the NMC vorticity. Notable 



examples are ovef the southern United States, along the TUTT across the 
North Pacific and from Japan southwestward to Indie. 


(2) Five-day Averaged Wind Fields and Energetics 

Under the Murakami project grid point data from the analyzed wind 
• fields were used to compute divergence, vorticity, streamfunctlon, velocity 
potential and energetics over the area from 40E eastward to lOOM. The results 
were reported In Sumi (1980) and selected figures from that report are used 
herein for additional discussion and to note differences In Interpretation of 
the results. The averaged wind fields are shown In Fig. 11, There Is little 
perceptible difference between the UHM and NMC objective analyses. The 
subjective anflysis, except for being somewhat smoother, is in general agree- 
ment with the objective analyses since data are plentiful over most of the 
area, Major differences are found in the equatorial Indian Ocean and the 
equatorial Pacific east of 170E. Sumi (1980) attributed the differences in 
the Indian Ocean to the use of climatology by the subjective analyst; however, 
this is not correct for the significantly greater wind speeds of the subjective 
analyses differ from climatology and are due to the use of "bogused” winds 
obtained from extensive and persistent cirrus streamers observed by satellite 
(discussed earlier). In the Pacific the subjective analyses maintained the 
buffer system ju^lt north of the equator with resulting west winds along the 
equator in contrast to east winds in the objective analyses. 

These differences in the wind field produced considerable differences 
in the energy conversions and the eddy kinetic energy fluxes over these regions 
as reported by Sumi, 1980 (see his Table 1 and Figs. 8 and 9), In the Indian 
Ocean between thequator and 15S the barotropic energy conversion was negative 
(barotropically stable) for the subjective analysis and strong positive 
(barotropically unstable) for the objective analyses. Between the equator and 


1SN the kinetic energy flux at the eastern boundary of 110E was six times 
greater In the subjective analysis. In the Pacific between the equator and 
15N the kinetic energy flux at the western boundary of 140E was ten times 
greater In the subjective analysis. 

The outgoing longwave radiation (OLR) obtained from satellite data 
Is a gross measure of the area and depth of convective cloud systems In the 
tropics and Is therefore related to the divergence of the upper winds and can 
be used as an Independent check on the quality of wind analyses. The 
divergent components of the wind fields from Fig. 11 are shown in Fig. 12 
together with the five-day averaged OLR. The minimum OLR values, located In 
the Bay of Bengal and over Borneo, agree best with the divergent flows of the 
subjective analyses even though OLR was used as a correction factor In the UHM 


The GLAS objective analyses and our subjective analyses could not be 
adequately compared for reasons stated earlier; however, for those few 
tropical features whi»-n could be compared the GLAS analyses proved unsatis- 
factory. Some specific examples are; 

( ^ ) Tropical Cyclones 

None of the six tropical cyclones during the August- September period 
were adequately depicted on either the surface or 850 mb analyses. On most 
days there was no evidence of the cyclones at all. Since these systems are 
seldofu reflected in the conventional data network, they must be bogused. 

( 2 ) Surface Pressu re Analysis 

In general the surface pressure analysis was very poor over the 
tropics, apparently due to unsatisfactory data checking. There were daily 











k ; 




u ’ \ P ^ \ ' 

- •■ • H" 




examples of Isolated "bull's-eye" type circular systems* each obviously 
caused by a singular poor observation. Extreme examples Mere a 1020 Isolated 
high on the equator In the Atlantic and a 1016 mb high near a 1004 mb low on 
the equator In the eastern Pacific, The systems pop In and out of the analyses 
with, of course, no continuity In time or space. 

A good surface pressure analysis Is a prerequisite for determining 
a good reference level for the satellite soundings and It Is unfortunate if 
their ultimate utility In the tropics Is limited by the quality of the surface 

(3) 300 mb Wind Analyses 

A most discouraging feature was the 300 mb wind analysis In the 
tropical belt since it is presumably based on many observations from satellite 
wind vectors, aircraft reports and rawins. The GLAS grid point wind direction 
vectors and interpolated speeds from the 00 GMT 1 September 300 mb analyses 
are plotted on Fig. 8a. The analysis Is very "nof^^y" with dirsiCtions "flip 
flopping" between grid points and speeds varying erratically. One rawin check 
is afforded by Johnston Island which is very near a grid point. The analyzed 
and observed directions differ by some 50 degrees. The analysis for this day 
is typical of the other days and it is difficult to define the problem without 
more knowledge of the analysis scheme. 

(4) The system centers and trough and ridge lines seldom coincide in the 
pressure and wind analyses. Again, knowledge of the analysis scheme is needed 
to determine the cause since we do not know if the analyses are independent or 
if some type of dynamic balance procedure is used. 




• * 

Satellite Input to the tropical data bate has been impressive and has 
continued to increase since the DST-5 and DST-6 with the addition of improved 
soundings, better specification of the SST and more geosynchronous satellites. 
However, purely objective analysis by any current scheme, even for the few 
levels of maximum data, does not take full advantage of satellite data and can 
be considerably improved by experienced human intervention. The impact of 
satellite soundings, although promising, has yet to be tested. 

Beyond the analysis problem is the more formidable one of numerical 
tropical forecasting which has scarcely been posed much less tested. If our 
main interest lies in higher latitudes, perhaps we should continue to "wire 
around" the tropics while posing the question: "What imijact would a better 
tropical analysis have on higher latitude forecasting?" Most meteorologists 
arbitrarily assume it would have a large effect because the tropical region 
is the major energy source and the large international programs of GATE and 
FGGE have been conducted to deal mainly with the tropical problems. However, 
success in linking the tropical convection scale to the large scale circulation 
has not been reported from the GATE program. Perhaps a detailed tropical 
analysis, beyond the proper specification at the interface between the tropics 
and extratropics, is unnecessary and would have little impact on the quality 
of midlatitude forecasts. The data collected during the special observing 
periods of FGGE may be sufficient to address the question and is worthy of 


Mr. Hiroshi Nishimoto aided in the analysis, Mr. Louis Oda conducted the 

data management and plotting and Mrs. S. Arita typed the manuscript. 


Appreciation Is expressed to Or. Hurakami and Mi*. Sumi for their cooperation 
In Integrating the research efforts and for ^he’us’e of their results. 


Anderson, R. K. • J. P. Ashman, F. Bittner, 6, R.* Farr, E. W. Ferguson, V. J. 

• * 

Oliver, A. H. Smith, F. C. Parmenter, 0. Slebers, R. W. Skidmore, 

J. F. Purdom, 1969: Application of meteorological satellite data In 
analysis and forecasting. ESSA Tech. Report NECS>51 plus supplements. 

Atlas, R., M. Halem, and M. 6h11, 1979: Subjective evaluation of the combined 
Influence of satellite temperature sounding data and Increased model 
resolution on numerical weather forecasting. Preprint, Proc. of the 
Fourth Conference on Numerical Weather Prediction, 319-328. 

, 1979: A comparison of GLAS SAT’and NMC high resolution NOSAT 

.forecasts from 19 and 11 February 1976. NASA Tech. Memo. 80591, Goddard 
Space Flight Center. 

Gh11, M. , M. Halem, and R. Atlas, 1979: Time-continuous assimilation of reirote- 
soundlng data and its effect on weather forecasting. Mon. Wea. Rev . . 

107 , 140-171. 

Miller, A. J., and C. M. Hayden, 1978; The impact of satellite derived 

temperature profiles on the energetics of the NMC analyses and forecasts 
during the August 1975 DST. Mon. Wea. Rev ., 106 , 390-398. 

Miyakoda, K. , J. C. Sadler, and G. D. Hembree, 1974: An experimental prediction 
of the tropical atmosphere for the case of March 1965. Mon. Wea. Rev ., 

102 , 571-591. 

Sadler, J. C., 1963: TIROS observations of the summer circulation and v/eather 
patterns of the eastern North Pacific. Proc. of Symposium on Tropical 
Meteor., New Zealand Meteorological Service. 

, 1976a: Tropical cyclone initiation by thu tropical upper 

tropospheric trough. UHMET 75-02, Dept, of Meteor., Univ. of Hawaii. 


Sadler* J» C. . 1976b: A role of the tropical upper tropospheric trough In 
early season typhoon development. Mon, Wea. Rev ., 104, 1266*1268. 


, 1978: Mid'season typhoon development and Intensity changes and 

the tropical upper tropospheric ‘trough. Hon. Wea. Rev ., 106 . 1137-1152. 

Sumi, A., 1980: A comparison of several analysis schemes on the energetics 
over the monsoon region. Dept, of Heteor. , Univ. of Hawaii unnumbered 
report under NASA grant NAS 5-25378. 22 pp 4- 12 figures. 

Tracton, S. , and R. McPherson, 1977: On the Impact of radiometric sounding 
data upon operational numerical weather prediction at NMC. National 
Meteor. Center, Washington, D.C. , Office Note 136. 

, A. J. Desmarais, R. J. van Haaren, and R. D. McPherson, 1980: 

The impact of satellite soundings on the National Meteorological Center's 
analysis and forecast system--the data systems test results. Mon. Wea . 
Rev. , 108, 534-586. 

five rawin observations are circled. 

2. Aircraft data over Africa near 250 mb for 12 GMT, 15 August 1974 illustrating the typical distribution of the 
approximately 200 available pilot reports. 



analysis over the tropical eastern Pacific Ocean at 00 GMT, 2 Septefnber 1975. 

onQINM. PftflE ® 

at 00 QfT, 1 September 1975 from 180 to 70E. 

9a. 250 mb relative vorticity at 00 GMT, 1 September derived from analysis of Fig. 3a. Positive values (N.H.) 

are thin solid lines. 

relative vorticity at 00 GMT. 1 September derived 

vorticity at 00 GMT, 1 September derived from NMC analysis 


relative vorticity at 00 GMT, 1 September derived from NMC analysis