DAO Oce Note

Oce Note Series on

Global Mo deling and Data Assimilation

Richard B Ro o d Head

Data Assimilation Oce

Goddard Space Flight Center

Greenbelt Maryland

New Data Typ es Working Do cument

Joanna Joiner

Rob ert Atlas

Steve Blo om

Genia Brin

Stephen Cohn

Arthur Hou

Dave Lamich

Dave Ledvina

Richard Menard

LarsPeter Riishjgarrd

Richard Ro o d

Arlindo da Silva

Meta Sienkiewicz

Jim Stobie

Runhua Yang

Data Assimilation Oce Goddard Laboratory for Atmospheres

Goddard Space Flight Center Greenbelt Maryland

This paper has not been published and should

be regarded as an Internal Report from DAO

Permission to quote from it should be

obtained from the DAO

Go ddard Space Flight Center

Greenb elt Maryland June

Abstract

This working do cument describ es activities related to the New Data Typ es Group of

the Go ddard Data Assimilation Oce DAO and how these activities are prioritized

Included in this do cument are science drivers for new data typ es brief descriptions

of assimilation metho dologies including advanced metho dologies b eing explored at the

DAO a list of new data typ es b eing considered for assimilation and validation at the

DAO and how these data typ es are roughly prioritized This do cument will b e up dated

p erio dically

Online versions of this do cument are available from

ftpdaogsfcgovpubofficenotesonpsZ postscript

Contents

List of Tables

Intro duction

Scientic drivers

Overview

Hydrological Cycle

LandSurfaceAtmosphere Interaction

SurfaceAtmosphere Interaction

Radiation Clouds Aerosols and Greenhouse Gases

Trop ospheric circulation and temp erature

Stratospheric circulation and temp erature

Constituents

Assimilation Metho dology

Statistical Analysis

Direct radiance assimilation

Traditional retrieval assimilation

Consistent Assimilation of Retrieved Data CARD

Physical Space

Phase Space

Implementation

Data ow and Computational Issues

Instrument Team Interaction

Priorities

Priorities group ed by science topic

Temp erature

Moisture Assimilation

ConvectivePrecip Retrieval Assimilation

Land Surface

Ocean Surface

Constituents

Wind prole

Aerosols

Priorities group ed by

Priorities group ed by use in GEOS

References

App endix

App endix A Acronyms

App endix B Instruments and

List of Tables

Seminar Sp eakers for New Data Typ es Group

Visitors Consultants and Collab orators

Priorities for assimilating temp erature data

Priorities for assimilating water vap or data

Priorities for assimilating convective retrievals

Priorities for assimilating landsurface data

Priorities for assimilating o ceansurface data

Priorities for assimilating ozone data

Priorities for assimilating CO data

Priorities for assimilating wind proles

Priorities for use of aerosol data

Highpriority data typ es from POES satellite

Highpriority data typ es from UARS satellite

Highpriority data typ es from TRMM satellite

Highpriority data typ es from the ADEOS satellite

Highpriority data typ es from EOS AM satellite

Highpriority data typ es for rst lo ok system

Highpriority data typ es for nal platform

Highpriority data typ es for reanalysis andor p o cket analysis

Intro duction

The mandate of the Data Assimilation Oce DAO at NASAGo ddard is to provide

highquality data sets to b e used by the science community to study a numb er of

Earth Systems problems As part of the DAO mandate a ma jor eort is b eing undertaken

to assimilate new data typ es from current and future instruments Emphasis is given to

instruments that will y as part of NASAs Mission to Planet Earth MTPE program

particularly those on b oard the Advanced Earth Observing System ADEOS the Tropical

Rainfall Measuring Mission TRMM and the Earth Observing System AM EOSAM

scheduled for launches in and resp ectively currently this do cument only

addresses satellites through EOSAM The instruments ab oard these satellites are de

signed to measure quantities related to atmospheric and surface parameters of particular

relevance to Earth Systems study More detailed information ab out comp onents of the

DAOs GEOS system can b e found in Takacs et al Pfaendtner et al

and Schub ert and Ro o d Do cumentation on the DAOs PhysicalSpace Statistical

Analysis System PSAS can b e found in da Silva and Guo

The eort required to assimilate new data typ es is nontrivial For example at other

NWP centers the eort to incorp orate new data typ es has required approximately

p ersonyears p er new data typ e and is an ongoing eort New data typ es currently b eing

considered for assimilation at the DAO may b e classied into two categories Data

typ es from instruments with an assimilation heritage at the DAO and other centers eg

TOVS radiances Scatterometer measurements etc Data typ es that do not have a

longterm op erational heritage eg precipitation surface wetness etc Data typ es in

are considered to b e a higher risk than those in The DAO eort to assimilate new

data typ es includes developing advanced assimilation metho dologies that will b e required

in order to handle the large volume of data from future instruments Monitoring new data

typ es b efore during and after assimilation will b e an imp ortant comp onent of the new

data typ es eort The eort will also involve interacting with instrument teams as well as

interfacing with other DAO groups such as the those concerned with covariance tuning

quality control op erations and development

This do cument attempts to dene and prioritize the activities of the DAOs New Data

Types group This is a working do cument meaning that it will evolve as exp erience is

gained with new data typ es Parts of the do cument are incomplete at this time and will

b e revised in the future This do cument will b e used by DAO sta to guide decisions

in how resources related to New Data Types will b e allo cated in the present and future

It will also provide instrument teams with information ab out what will b e required from

them in order to eectively utilize observations in the Go ddard Earth Observing System

Data Assimilation System GEOSDAS Details on related topics such as the Kalman

ltering eort Observation System Simulation Exp eriments OSSE Observation System

Exp eriments OSE and bias correction will app ear elsewhere or will b e incorp orated into

the do cument at a later time Do cumentation on related topics can b e found elsewhere

monitoring can b e found in da Silva et al b quality control in Dee and Trenholme

quality assurance in Schub ert and op erations in Stobie

The outline of the do cument is as follows Section describ es the scientic goals that

drive the eort to incorp orate new data typ es in the DAOs Data Assimilation System This

section b egins with an overview and follows with subsections devoted the hydrological cy

cle landsurfaceatmosphere interaction o ceansurfaceatmosphere interaction radiation

clouds aerosols greenhouse gases atmospheric circulation and constituents For each

scientic driver examples of relevant data typ es are given Section discusses traditional

and advanced metho dologies for assimilating data and provides examples of relevant data

typ es for each metho d Section describ es several topics related to implementation com

putational issues data ow and instrument team interaction Finally new and existing

data typ es are prioritized in section Crossreferences of instrument to data typ e scientic

driver use in GEOS etc are provided

Scientic drivers

Overview

The NASA Mission to Planet Earth MTPE is a program designed to make use of ground

aircraft and satellitebased measurements in order to b etter understand the systems that

govern the Earths climate their interactions and their variations NASAs Earth Observ

ing System EOS and several other satellite systems are key comp onents of this program

Driving the selection of new data typ es in the GEOSDAS is a subset of scientic ob jec

tives outlined by the MTPE program the United States Global Change Research Program

USGCRP and the Intergovernmental Panel on IPCC These ob jectives

include improving our understanding of the hydrological cycle interaction b etween the

land and o cean surfaces with the atmosphere and transp ort of heat moisture and momen

tum the interaction of clouds and aerosols with radiation and its impact on climate

chemistry and transp ort of atmospheric constituents such as ozone in b oth the trop osphere

and stratosphere These and other Earth systems topics are briey discussed b elow in the

context of data assimilation Examples of relevant data typ es are provided

Hydrological Cycle

Water substance plays an integral role in many of the pro cesses op erating within the Earth

System In the atmosphere latent heat release through condensation of water vap or pro

ducing clouds is a ma jor source of energy that regulates atmospheric circulation A large

fraction of the energy transferred from the surface to the atmosphere is in the form of latent

heat due to evap oration which itself dep ends on the moisture gradient b etween the surface

and atmosphere In addition the presence of clouds and precipitation determines the avail

ability of solar radiation and water at the surface Water vap or is also the predominant

greenhouse gas and plays a crucial radiative role in the global climate system

Clearly an accurate depiction of the global atmospheric moisture eld its vertical and

horizontal transp ort and the transfer of water across its b oundaries is critical to under

standing the hydrologic cycle and its impact on climate In the past moisture information

used in the GEOSDAS has come primarily from rawinsonde ascents Satellite moisture

estimates oer a signicant improvement over the spatial and temp oral sampling prob

lems of the existing rawinsonde network but they also have limitations Polar orbiting

satellites such as those in the Defense Meteorological Satellite Program DMSP carry

microwave instruments which are sensitive to atmospheric moisture The Sp ecial Sensor

MicrowaveImager SSMI for instance gives an accurate estimate of the integrated wa

ter vap or in an atmospheric column referred to as Total Precipitable Water TPW but

contains little information ab out the vertical distribution of moisture which is critical to un

derstanding many of the physical pro cesses mentioned ab ove The SSMI observations are

also limited by the fact that they are only available over the o ceans in precipitationsea ice

free regions SSMT provides similar coverage but contains additional channels that pro

vide information ab out the vertical structure of moisture Other satellites such as TIROS

Op erational Vertical Sounder TOVS which carries the Highresolution Infrared Sounder

HIRS and the Microwave Sounding Unit MSU have b etter geographical coverage over

land as well as o cean and can resolve the integrated water in approximately two thick

slabs Future microwave and IR instruments such as MHS AMSUB AIRS and IASI

will have improved vertical resolution for moisture sounding Although still fairly crude

the moisture estimates that arewill b e available from these satellites are valuable b oth for

assimilation validation and bias estimation For example information ab out global mois

ture derived from SSMI and TOVS has b een assimilated at DAO and other centers eg

Ledvina and Pfaendtner Derb er private communication Andersson et al and

has signicantly impacted global analyses

Signicant discrepancies have b een found b etween precipitation estimates from analysis

systems and those derived from satellite measurements Systematic dierences in b oth

the intensity and spatial distribution of monthly mean precipitation are esp ecially large in

the tropics Current research at the DAO is directed towards assimilating precipitation

data Improving the hydrological cycle in analyses in b oth the tropics and extratropics

will ultimately aid in the understanding of climate sensitivity climate variability the role

of dynamical feedback and other physical pro cesses Improving the hydrological cycle in

data assimilation systems will also b enet studies in which precipitation estimates from a

DAS are used to drive surface hydrological mo dels

LandSurfaceAtmosphere Interaction

The landsurface is an imp ortant comp onent in the Earth System aecting energy balance

the hydrological cycle and chemical cycles Latent heat sensible heat and momentum

uxes at the landsurface mo dulate nearsurface turbulence and b oundary layer

on diurnal and longer timescales The landsurface plays an imp ortant role in the hydro

logical cycle receiving water from the atmosphere in the form of rain or snow with the soil

and vegetation acting as reservoirs Precipitation that do es not inltrate the soil forms

surface runo The biosphere over the land surface also aects the carb on cycle through

photosynthesis to pro duce the greenhouse gas CO

2

Although still in its infancy land surface data assimilation has generated interest in

b oth environmental and meteorological communities Over the past decade stateofart

land surface mo dels LSM have b een coupled with general circulation mo dels Remote

sounding instruments that infer land surface quantities will play an integral role in land

surface mo deling and data assimilation in the future Many new data typ es can b e inferred

remotely from satellite instruments that provide sup erior geographic coverage of the land

surface as compared with conventional surface station observations These satellitederived

data typ es include surface skin temp erature from infrared instruments such as TOVS

AVHRR MODIS predecessor and MODIS as well as snow water equivalent content and

soil moisture information from passive microwave instruments such as SSMI and in the fu

ture TMI In addition to measurements that can b e assimilated satellitederived quantities

such as vegetation indices from AVHRR and MODIS can also b e used to sp ecify or esti

mate mo del parameters in some LSMs Use of these data should improve b oth shortterm

forecasts and analyses of climate events such as El Nino

OceanSurfaceAtmosphere Interaction

The o cean surface including regions immediately ab ove and b elow is an interface b etween

two of the great subsystems involved in the Earth System the o cean and the atmosphere

Considerable Earth Systems Science research will b e devoted to the interplay of dynamics

and features b etween these two subsystems There is a considerable disparity in the nature

of our understanding of these subsystems and their interaction A b etter understanding

of the workings of the Earth System esp ecially for long time scales will entail a detailed

knowledge of the b ehavior of the coupled atmosphereo cean system El Nino and its global

eects is certainly an example of this Thus it is imp ortant to have improved o cean surface

data to b etter delineate the mechanisms that govern the atmosphereo cean interaction

Past observations of o cean surface variables have largely b een obtained from in situ

platforms ships buoys and island data These observing systems provide information on

sea level pressure winds humidity and sea surface temp erature Such observations have

suered from a variety of problems ranging from unrepresentative sampling to inherently

p o or quality The most serious problem with these historical data sources lies in their p o or

sampling of the Tropics and the southern o ceans

There are currently in place satellitebased sensing systems which provide information

related to sea surface temp erature and sea level winds For example one source of infor

mation concerning winds in the lower atmosphere cloudtracked winds CTW has b een

available for a numb er of years Research using data from the current spacebased o cean

surface wind sensors SSMI and Scatterometer has found p otential b enets and dicul

ties in using these data The b enets are clear the satellite sensors provide a tremendously

improved sampling in b oth space and time of the o cean surface wind The diculties with

using data from these sensors lie in drawing the prop er inferences from the information

provided by the sensors For example the SSMI provides information related to the wind

sp eed near the o cean surface provided a numb er of assumptions p ertaining to the windsea

state relation and the nature of the b oundary layer near the sea surface are valid Similar

problems arise in the consideration of Scatterometer data

Improvements in the sensing of winds near the o cean surface will b enet a numb er of

Earth Science research enterprises More accurate surface winds will lead to improved esti

mates of momentum uxes and have a p otentially p ositive impact on the uxes of sensible

heat and moisture for use in atmospheric mo deling Similarly improved estimates of momen

tum heat and fresh water uxes will b e used for o cean mo dels Such wind data will provide

imp ortant forcing or b oundary conditions for research in coupled atmosphereo cean mo d

eling Ultimately the assimilation of surface wind data into a coupled atmosphereo cean

system will b e used to correct the state of such a system Another area of interest for

these wind data is in o cean wave mo deling accurate highresolution winds will improve the

pro cess of data assimilation of surface winds into o cean wave mo dels Finally improved

surface wind datasets obtained from the global assimilation of these new surface wind data

will provide more useful forcing for oine mo del studies such as seaice transp ort studies

Radiation Clouds Aerosols and Greenhouse Gases

The transfer of electromagnetic radiation in the atmosphere ie absorption emission

scattering and refraction of energy by particles clouds and aerosols and molecules is the

most imp ortant pro cess aecting energy transfer Radiative transfer in the atmosphere

is generally separated into two regions Solar or shortwave radiation that p eaks near

m in the visible part of the electromagnetic sp ectrum Terrestrial or longwave

radiation that p eaks in the infrared part of the sp ectrum near m One consequence of

molecular absorption is the socalled greenhouse eect in which a gas is virtually transparent

to incoming shortwave radiation while absorbing and reemitting longwave radiation The

ma jor greenhouse gases are CO H O and O with CO CH N O and others playing

2 2 3 4 2

a smaller roll The CO concentration remains relatively homogeneous in b oth space and

2

time although observations show a seasonal variation and longterm increase In contrast

concentrations of H O O and some of the minor greenhouse gases exhibit relatively large

2 3

spatial and temp oral variations Climate changes resulting from variations in greenhouse

gas abundances are dicult to predict b ecause of complex radiativeconvective feedbacks

and remain a controversial issue

Cloud prop erties have a large impact on energy balance at the top of the atmosphere and

the surface In the shortwave susp ended liquid water in clouds reects incoming solar radia

tion back to space In the longwave liquid water drops absorb and reemit radiation Molo d

et al have shown that systematic errors in the GEOS general circulation mo del

GEOSGCM linked to physical parameterizations of clouds convection and rainfall re

sult in latitudinallyde p end ent systematic dierences b etween observed and GEOSderived

long and shortwave ux at b oth the upp er and lower b oundaries of the atmosphere Cur

rently satellite observations eg from the Earth Radiation Budget Exp eriment or ERBE

are used to diagnose problems with mo del parameterizations

Aerosols like cloud particles can signicantly aect the heat balance of the Earth

atmosphere system by scattering and absorbing incoming solar radiation Aerosols or

susp ended particles in the atmosphere include volcanic dust sea spray dust generated from

wind smoke from forest res and biomass burning particles pro duced during combustion

chemical reactions involving naturally o ccurring gases or gases formed during combustion

and cataclysmic impacts b etween the Earth and other solar system b o dies Temp oral and

spatial variations in aerosol distributions have not b een given much attention in current

data assimilation systems

There currently exists an abundance of radiationrelated data inferred from satellite

and groundbased instruments such as cloudtop temp erature and cloud fraction in the

ISCCP data base derived from combined infrared and visible observations and cloudliqui d

water derived from passive microwave instruments such as SSMI Future instruments such

as MODIS and the TRMM microwave imager TMI will provide additional estimates of

these quantities Groundbased aerosol observations date back to the early s Aerosol

parameters have also b een measured with satellite instruments such as SAGE HALOE

and CLAES primarily in the stratosphere Advanced instruments such as MISR have

b een designed for improved observations of aerosol prop erties Groundbased monitoring

of greenhouse gas concentration has b een taking place over the last several decades In

the future more complete global monitoring of greenhouse gases will b e accomplished with

instruments such as MOPITT These data typ es related to radiation have not yet b een used

in op erational data assimilation systems Research at the DAO is currently ongoing to more

fully exploit these new data typ es in a DAS This should lead to a b etter understanding of

the role of radiation and dynamical feedback pro cesses in regulating the climate system

Atmospheric Circulation

Accurate consistent longp erio d measurements of global mass height and temp erature

and momentum wind elds are imp ortant for Earth System studies Atmospheric circu

lations are the metho d by which sensible and latent heat mass and momentum uxes are

transp orted The study of climate variability fo cuses in part on interannual dierences and

on anomaly elds dierences from climatological means including changes in largescale

ows such as the Hadley circulation which transp orts mass energy and moisture from

the tropics to the midlatitudes Another imp ortant long timescale circulation is the El

NinoSouthern Oscillation ENSO which has b een linked to changes in the extratropical

circulation resulting in o o ds and droughts The stratospheric quasibiennial oscillation

QBO is a ma jor comp onent of the interannual variability in the stratosphere Several

do cumented deciencies in GEOS such as a weak Hadley cell and warmbiased trop opause

temp eratures in the tropics see references in Schub ert and Ro o d can p otentially b e

addressed with new data typ es including improved use of current observations

Trop ospheric circulation and temp erature

In the trop osphere and lower stratosphere b oth winds and temp eratures are observed with

a network of conventional rawinsondes This network is dense in the Northern Hemisphere

over land However observation density and quality in the tropics Southern Hemisphere

and over o ceans is much p o orer In these areas satellite observations provide excellent

geographic coverage Advances in global mo dels and data assimilation have led to

b etter global analyses and forecasts of these elds but there is ro om for improvement b oth

in the observations themselves as well as the way in which they are used in assimilation

systems Advanced instruments for temp erature sounding such as AIRS and IASI will

provide measurements with higher information content ab out the mass eld than the current

NOAA p olarorbiting satellites Yet even with improved sounding capabilities there are

still b oth technical and scientic issues to b e addressed when considering how b est to

assimilate remote satellite observations These issues will b e addressed in more detail in

section

Stratospheric circulation and temp erature

The need for improved knowledge of stratospheric winds and temp eratures is most strongly

driven by eorts to understand stratospheric ozone variability and predict anthrop ogenic

changes in the ozone distribution Recently there has b een increased interest in role of

the stratosphere in climate change esp ecially changes asso ciated with upp er trop ospheric

and lower stratospheric water vap or and ozone In addition new initiatives in trop ospheric

chemistry require b etter quantication of stratospheretrop osphere exchange

Winds and temp erature from data assimilation systems have b een central in increasing

the quanititive level of stratospheric chemistry In the lower stratosphere middle latitudes

the winds are already of sucient quality that transp ort calculations can b e used to unify

constituent measurements from dierent observing platforms The chemical studies have

also revealed places where the winds and temp eratures remain of low enough quality that

uncertainties in chemical assessments are still strongly tied to meteorological conditions

Within the current GEOS system the most notable problems with the winds lie in the

tropics and the subtropical middle latitude b oundary The transp ort b etween the subtrop

ics and middle latitudes is an esp ecially imp ortant topic b ecause mixing of middle latitude

p ollutants into the tropics can strongly p erturb the ozone sources With regard to temp era

ture eld scientists in the aircraft missions have noted a warm bias of GEOS temp eratures

in the lower stratosphere when the extreme cold events o ccur Improvements in the tem

p erature representation are needed b ecause heterogeneous chemical pro cesses b ecome more

imp ortant at extremely low temp ertures Longterm transp ort exp eriments show that re

cent versions of GEOS have extensive improvements in the representation of the seasonal

zonalmean meridional circulation However misrepresentations of the equatorial upwelling

and p olar downwelling remain large enough that it is dicult to represent interannual vari

ability in longlived tracers

Much of the pro jected improvement in stratospheric meteorological elds is exp ected

to come from advances in mo deling analysis and sp ecication of improved error statistics

New data sources are limited primarily to improved temp erature sounders such as MLS

and GPS that have increased vertical resolution and accuracy Indirect improvements

to meteorological elds may also b e derived from the assimilation of longlived tracers

Stratospheric wind measurements from the Upp er Atmosphere Research Satellite UARS

have proven dicult to utilize and the DAO will follow the progress on impact studies b eing

carried at at the UKMO

Constituents

Building the capability of assimilating constituent observations is one of the longterm goals

of the DAO In such an assimilation scheme the constituent data are allowed to inuence

b oth the estimate of the tracer itself and by the coupling of the elds through the linear

transp ort equation the estimate of the ow eld Daley Riishjgaard Con

stituent assimilation in the present context means assimilation of observations of the minor

constituents of the atmosphere such as water vap or N O CO CH etc However one of

2 4

the main motivations for recent development in metho dologies for constituent assimilation

is the availability of ozone derived from satellitebased instruments Assimilation of these

data would b e useful for the atmospheric chemistry community since it would allow us to

build a multiyear sequence of threedimensional ozone elds consistent with atmospheric

dynamics available for each analysis time

From a meteorological p oint of view the ozone data is of interest b ecause the ow pat

terns around the trop opause level generates a very strong signal in total ozone observations

Conversely this implies that the total ozone measurements carry information ab out the

winds at these level information that over large parts of the glob e is not readily available

from alternative sources There are a numb er of dierent strategies that one can follow in

order to retrieve this information dep ending on the mo del underlying the data assimilation

pro cedure The basic ideas involved in constituent assimilation are intuitively app ealing

and conceptually simple However the actual design and implementation of a multivari

ate scheme in a general circulation mo del is technically a ma jor eort and there remain

a numb er of scientic issues to b e resolved Thus it is far from clear at the present time

which assimilation metho d EKF DVAR Fixedinterval smo othing assimilation through

a transp ort mo del or through a PV mo del etc will b e b est suited for the purp ose Also

even though some of these metho ds theoretically should b e insensitive to the initial sp ec

ication of the forecast error covariances this may not b e true for a particular numerical

implementation Since the partitioning of the information present in the data will b e deter

mined by the error sp ecied work towards improving our understanding of the errors b oth

of the observations and of the background mo del estimate is clearly needed and will b e an

integral part of constituent assimilation at the DAO

Assimilation Metho dology

Several approaches have b een prop osed and used to incorp orate b oth conventional and

satellite observations in data assimilation systems These metho ds will b e reviewed in sub

sequent sections following a brief review of statistical analysis and description of the DAOs

Physicalspace Statistical Analysis System PSAS We will fo cus here on assimilation of

remotely sensed data from satellites in PSAS although the concepts also apply to conven

tional data as well as other assimilation metho ds such as the Kalman lter

Statistical Analysis

The ob jective of statistical interp olation is to pro duce an optimal estimate of the atmo

spheric state given a set of observations and a rst guess usually in the form of a short

term forecast In the variational framework ie DVAR this can b e accomplished by

minimizing the likeliho o d functional

T

T

f f 1 f o o 1 o

J w w w P w w w hw R w hw

n f n

where w IR is a vector representing the D state of the atmosphere w IR is the

o p

forecast w IR is the observation vector and hw is an observation op erator that maps

the D atmospheric state into observables The rst term on the RHS of is weighted by

f n n

the inverse of the forecast error covariance matrix P IR IR while the second term is

o p p

weighted by the inverse of the observation error covariance matrix R IR IR Provided

these covariances are sp ecied correctly the analysis state obtained by minimizing J w is

S

f o

the mo de of the conditional probability density function pw jw w and is derived from a

maximumlikeliho o d principle assuming that forecast and observation errors are unbiased

normally distributed and uncorrelated with each other

Because the observation op erator hw is in general nonlinear the minimum of J w

can b e obtained by a quasiNewton iteration of the form

h i

1

f T o o f f f T

H P H R w hw H w w w w P H

i i i i i+1

i i

where

hw

H

i

w

w =w

i

a

The analysis vector w that minimizes J w is given by

a

w lim w

i

i!1

In PSAS a p p system of equations is rst solved at observation lo cations

f f T o o

w w H P H R x w hw H

i i i i i

i

p

for the vector x IR using a conjugate gradient algorithm da Silva et al Guo and

i

da Silva The rst term on the LHS of is called the innovation covariance The

state at iteration i is up dated by an additional matrixvector multiply viz

f f T

w w P H x

i+1 i

i

2

The computation required for the solution of the linear system is approximately O N p

cg

where N is the numb er of iterations of the conjugate gradient algorithm N dep ends

cg cg

on the conditioning of the innovation covariance matrix The matrixvector multiply in

requires O np oating p oint op erations The total op eration count to solve is ap

2

proximately O N N p np where N is the numb er of outer quasiNewton iterations

o cg o

p erformed It is evident any typ e of data compression that reduces the numb er of observ

ables will signicantly reduce computation in PSAS At the DAO development of improved

metho dology to assimilate new data typ es with very high spatial or sp ectralresolution has

b een driven largely by consideration of computational costs

Direct radiance assimilation

In this do cument radiance is a general term meaning a directly measured quantity as

opp osed to a retrieved quantity For example radiance could refer to refractivity mea

sured by Global Positioning System GPS receivers backscatter from a scatterometer or

thermalreected radiation measured by a passive infrared or microwave sounder The

term radiance also applies to prepro cessed raw radiance eg cloudcleared radiance As

similation of radiances involves utilizing radiance measurements from a remote sounding

instrument as the observable and sp ecifying the radiance error covariance as the observa

tion error covariance in In addition the observation op erator h in that maps the

state variables to the radiances must b e sp ecied

For remote measurements h is an approximate radiative transfer or empirical mo del re

lating the atmospheric state in grid space to the radiance at the observation lo cation using

a set of parameters such as sp ectral line data andor calibration parameters To assimilate

radiances correctly an appropriate radiance error covariance must b e sp ecied that incor

p orates b oth detector noise and observation op erator error Practical implementation at

the DAO will use statistical mo deling from innovation sequences or some form of online pa

rameter estimation such as discussed in Dee to estimate radiance errors Observation

op erator parameter errors are usually systematic and may b e estimated and corrected for

in part by utilizing indep endent observations andor forecasts eg Eyre Susskind

and Pfaendtner

Radiance assimilation is computationally feasible with current instruments and analysis

schemes However the cost of this approach may b e prohibitive for future highsp ectral reso

lution sounding instruments such as AIRS and IASI with large numb ers of channels More

ecient approaches are currently b eing examined as an alternative to radiance assimilation

for such instruments

Traditional retrieval assimilation

Remotely sensed data have b een traditionally assimilated in the form of physicalspace

retrievals In this approach radiances are pro cessed oline by data pro ducers and familiar

data typ es such as temp eraturemoisture proles or wind vectors are used in the assimilation

o

system Sp ecifying the retrieval z as the observable w in the observation op erator h

is a linear interp olation op erator so that the iterated form of can b e reduced to

1

a f f T f f T z

w w P I z I w I P I R

D E

p z z z T

where z IR denotes the retrieved data R is the retrieval error covariance

n p

and I IR IR is the interp olation op erator used ab ove A more general form of that

D E

z f T

includes the retrievalforecast error crosscovariance denoted X is given by

1

a f f T T f T z T T f

w w P I X I P I R I X X I z I w

The assimilation of retrievals requires the sp ecication of the retrieval error covariance

z

matrix R and the retrievalforecast error crosscovariance X Op erational implementations

of retrieval assimilation are often based on statistically mo deled retrieval error covariances

under the assumption of stationarity and horizontal homogeneity and isotropy In addition

the retrievalforecast error crosscovariance matrix X is often neglected as a result of the

diculty asso ciated with mo deling it and accounting for it in a DAS At the DAO TOVS

retrieval errors have b een estimated using a tuning algorithm that separates horizontally

correlated comp onents of the error from uncorrelated comp onents as describ ed in da Silva

et al a

Consistent Assimilation of Retrieved Data CARD

Menard prop osed a p otentially less exp ensive alternative to radiance assimilation

that has a more theoretically sound basis than traditional retrieval assimilation This

metho d combines Ro dgers characterization of retrieval errors with Kalman lter

theory leading to the consistent assimilation of retrieved data CARD The implementa

tion of CARD may b e considerably less exp ensive than radiance assimilation for advanced

instruments where the numb er of radiance measurements is much greater than the numb er

of retrieved pro ducts ie the retrieval is a form of data compression The retrieval of

geophysical parameters is a nonlinear estimation pro cess Often the problem is illp osed and

in this case requires the use of prior information For example nadirviewing infrared and

microwave proling instruments use forecasts as prior information for interactive physical

retrievals Prior information could also come from climatology or a representative ensemble

of proles used to create a regression pattern recognition or neural net retrieval algorithm

The general CARD approach requires the sp ecication of error characteristics for the radi

ances as well as the prior information Because the statistical characteristics of the prior

information are often not known andor dicult to mo del and account for in a DAS some

mo dication to the retrieval may b e necessary in order to eliminate the eects of prior in

formation In the following two subsections examples of dierent CARD implementations

are given

Physical Space

If little or no prior information is used in the retrieval the retrieval errors may b e computed

by propagation of instrument error including detector noise and transfer mo deling errors

as describ ed in Ro dgers In this way the state dep endence of the retrieval error is

accounted for If the state dep endence is to b e prop erly accounted for instrument error

must b e sp ecied as in the case of radiance assimilation Once the retrieval error is

estimated the retrievals can then b e assimilated in a consistent manner This approach

has b een used by Menard et al to assimilate constituent data from the Cryogenic

Limb Array Etalon Sp ectrometer CLAES using a Kalman lter algorithm where retrieval

errors were rep orted by the instrument team

Phase Space

If a signicant amount of prior information is incorp orated into the retrieval the imple

mentation b ecomes more dicult as a result of the crosscorrelation b etween the prior

information and forecast errors X in Joiner and da Silva describ e several ways

to mo dify retrievals in order to remove the eect of the prior information thereby eectively

removing X These approaches involve ltering the p ortion of the retrieval aected

by the use of prior information ie nullspace ltering or NSF or p erforming a par

tial eigendecomp osition PED retrieval in which prior information is not incorp orated into

the retrieval These approaches compress radiance observations in a single sounding to a

small numb er of orthogonal functions that will impact the data assimilation system For

example Joiner and da Silva showed that over radiances observations from the

AIRS instrument could b e compressed into approximately pieces of relevant information

ab out the temp erature prole The observations may then b e compressed horizontally ie

combined into a sup erobservation or sup erob provided the retrievals are dened in terms

of a consistent set of basis functions The approaches describ ed here have b een demon

strated with simulated data in D thus far Full implementation of b oth direct radiance

assimilation and the CARDPED approach using TOVS data in PSAS is underway at the

DAO so that these metho ds can b e compared b oth in terms of cost and analysis quality

Implementation

Data ow and Computational Issues

One imp ortant consideration for incorp orating new data typ es an appropriate data assimi

lation metho d is the bandwidth required to accommo date data ow b etween the site where

the raw data is prepro cessed and the site where the data is assimilated For future high

spatial and highsp ectral resolution instruments such as MODIS and AIRS the required

bandwidths assuming no data compression can b e quite large Using level pro ducts

when p ossible sup erobbing andor the CARD metho dologies describ ed in section that

allow for data compression a signicant reduction in data ow as well as computation in

PSAS can b e achieved To achieve this reduction in data ow the prepro cessing required

to pro duce appropriate retrievals and to p erform sup erobbing must take place at the site

where the data resides Furthermore quality control and in some cases covariance tuning

will have to b e p erformed at the data site in conjunction with data compression Interaction

with the EOSDIS pro ject will b e required to ensure that adequate facilities are in place to

accommo date data ow and data compression at the data site

Instrument Team Interaction

Memb ers of instrument teams are familiar with the intricate and unique problems asso ciated

with a particular instrument eg calibration and other sources of systematic error in the

observations and op erators In many cases instrument teams have develop ed the to ols that

are necessary to eectively use the data in a DAS eg observation op erators In the past

interaction b etween instrument teams and data assimilation teams has b een weak The

DAO and New Data Typ es group would like to foster stronger interaction with instrument

teams For this interaction to succeed strong commitment by b oth the DAO and instrument

teams is required A prototyp e for this interaction is planned with the MLS team

The current plan for instrument team interaction is that at least one representative from

an instrument team will work closely with the DAO for an extended p erio d of time of the

order of a few months to a year The time required to integrate the new data typ e into

the DAS will dep end on several factors including the amount of previous exp erience with a

similar data typ e the readiness of to ols to assimilate the new data typ e ie observation

op erators the quality of the data typ e including the ability to remove bias and the

amount of prepro cessing needed prior to assimilation In most cases instrument teams will

b e exp ected to provide the DAO with the to ols needed to prop erly assimilate the data such

as observation op erators and their derivatives The instrument team memb er will interact

Table Seminar Sp eakers for New Data Typ es Group

Sp eaker aliation Topic

Steve Blo om NASA Oceansurface Wind Retrievals

Al Chang NASA Snow Water Equivalent Retrieval

John Derb er NMCNCEP Radiance Assimilation at NMC

Anne Douglass NASA Ozone D ChemistryTransp ort Mo del

Gregory Gurevich USRAUMd Satellite Tomography

Paul Houser U Arizona Remote Sensing of Soil Moisture

Randy Koster NASA Mosaic Land Surface Mo del

Chris Kummerow NASA Precipitation Retrievals SSMI and TRMM

Venkataraman Lakshmi NASA Soil Moisture from SSMI

Andrea Molo d NASA Coupling of Land Surface Mo del at DAO

Howard Motteler UMBC Neural Net Retrievals SSMT and AIRS

Bill Seegar Ab erdeen Birdbased meteorological measurements

Larrab ee Strow UMBC Fast and hyp erfast IR radiative transfer mo deling

David Tobin UMBC Infrared sp ectral lineshap es

with DAO sta sp ecializin g in covariance tuning quality control and the PSAS interface

The instrument team memb er is exp ected to provide baseline mo dels for systematic error

correction and covariance mo deling and will assist with quality control and monitoring

Another asp ect of instrument team interaction involves teams utilizing the data assimi

lation system to diagnose and correct problems with the instrument ie calibration error

and observation op erators eg tuning the forward mo del An example of this is the work

done with the ERS scatterometer by Stoelen and Anderson This interaction will

b e part of the monitoring activity for any new data typ e to b e used in the GEOS system

As part of the DAO commitment to bring together instrument teams and data assimila

tion teams the DAO has agreed to host and help organize a satellite assimilation workshop

headed by Ron Errico and George Ohring in March The New Data Typ es group has

also invited several instrument team memb ers and scientists working on related pro jects to

give seminars Table lists sp eakers to date that have given seminars at the DAO Table

lists other visitors collab orators and consultants in contact with the DAO New Data Typ es

group

Priorities

During several meetings of the New Data Typ es group instruments on past present and

future platforms were reviewed in order to prioritize their usage in the GEOSDAS Data

typ es were selected to b e used either for assimilation analysis noncycled or validation

When two or more instruments provided redundant information generally one was selected

for assimilation allowing for the other to b e used as indep endent validation or p erhaps bias

estimation An initial selection of a data typ e for validation may change to assimilation

after monitoring if the data app ear to b e of higher quality than data currently used

for assimilation a commitment is made by an instrument team no commitment by

an instrument team is made but resources within the DAO p ermit Da Silva et al b

describ e the planned DAO OnLine Monitoring System DOLMS

Each data typ e was initially assigned a score on a four p oint scale based on an estimate of

the costtob enet ratio where b enet is dened primarily in terms of meeting the scientic

Table Visitors Consultants and Collab orators

Contact Aliation Sub ject

PK Bhartia NASA TOMSSBUV retrievals

Moustafa Chahine JPL AIRS retrievals

Jean Dickey JPL GPS groundbased H O retrievals

2

Steve Engman NASA Soil moisture from passive microwave

Evan Fishb ein JPL MLS retrievals

Larry Gordley GATS CLAES retrievals limb sounding radiative transfer

Christian Kepp enne JPL GPS groundbased H O retrievals

2

Arlin Krueger NASA TOMS retrievals

Steven Marcus JPL GPS groundbased H O retrievals

2

Piers Sellers NASA Sib LandSurface Mo del

Max Suarez NASA Mosaic Land Surface Mo del etc

Jo el Susskind NASA TOVSAIRS retrievals

Jo e Waters JPL MLS retrievals

goals describ ed in section and cost is estimated primarily in terms of p erson lab or and

to a lesser extent computational cost oline and online computation data storage and

transfer Factored into the benet are the spatial and temp oral coverage eg timep erio d

and coverage for which data typ e is available The cost estimate includes factors such as

instrument team commitment previous exp erience with a similar data typ e an estimate

of how much work is needed to pro duce the observation op erator known diculties with

calibration and systematic error and an estimate of the amount of prepro cessing needed

prior to assimilation The initial scores were translated into either a ranking of either

high or low priority These rankings are listed in following subsections along with relevant

information ab out each data typ e considered The prioritization is to b e used as a guide

for allo cating resources and do es not necessarily indicate a commitment by the DAO to

assimilate or use a particular data typ e The actual use of a data typ e in the GEOS system

will dep end on resources available including instrument team commitments

Priorities group ed by science topic

New data typ es are rst group ed roughly by science topic The groupings b elow are not

listed in order of priority For a given topic all data typ es considered are listed For each

data typ e the exp ected use in the GEOS system is also listed ie whether the data typ e

will b e used in the rst lo ok analysis nal platform reanalysis only p o cket analysis etc

see Stobie or da Silva et al b for a summary of op erations Although we found

some data typ es to b e strong candidates for assimilation they may have b een selected for

validation on the basis of p ersonnel considerations

If available a sp ecic pro duct name is listed Data volumes are also listed These were

estimated from B Bass private communication most recent estimate when available or

from values rep orted in NASA These are given in Gigabyteday for level pro ducts

unless otherwise noted Caution should b e exercised in extrap olating from the gures listed

here for data ow calculations In some cases sup erobbing will b e p erformed to reduce data

ow For example the data volume from sup erobb ed or level MODIS data typ es will b e

more than a factor of less than the data volume of the level pro ducts In some cases

data typ es are pro duced less than once p er assimilation p erio d eg landsurface pro ducts

from MODIS to b e used as b oundary conditions pro duced once every or days and

need not b e retransmitted every assimilation cycle Estimates for data volume for level

pro ducts are not yet available

The notations used in the tables are dened as follows

L First Lo ok Analysis p erformed hours after data time used by EOS instrument teams

for retrieval algorithms etc

FP Final Platform including reanalysis of nal platform runs several months after data time

includes data from rst lo ok as well as data from the EOS platform

RA Reanalysis multiyear repro cessing of preEOS era and EOSera using frozen system

V Validation

OA Oine Analysis

B Boundary Condition

PA Po cket Analysis similar to reanalysis but for a limited timep erio d

H High priority

L Low priority

S Candidate for sup erobbingprepro cessing at data site

E Estimate



past

zfuture

ypast present future

Temp erature

There is an abundance of temp erature data available for assimilation and validation For

b oth the rst lo ok system and nal platform TOVS was given the highest priority The

other measurements with the exception of MLS were selected for validation Currently

in addition to conventional data retrieved temp erature height proles from TOVS from

NESDIS and GLA are assimilated in the traditional manner This approach has b een

implemented with covariance tuning as describ ed in da Silva et ala in GEOS

The advanced metho dologies radiance assimilation and phasespace retrieval assimi

lation describ ed in section are currently b eing implemented and evaluated with TOVS

Pathnder data in anticipation of future sounders such as AIRS and IASI For the rst

lo ok system as well as nal platform and reanalysis with GEOS and b eyond the most

appropriate metho d will b e selected on the basis of cost and quality of the analysis The

selected metho d will b e implemented with NESDIS radiancesretrievals for the rst lo ok

system If a reliable alternate data source for TOVS cloudcleared radiances and retrievals

is available in near realtime for the rst lo ok system a selection will b e made on the basis

of pro duct quality after monitoring has b een p erformed

The strategy for implementing new metho dologies with TOVS is as follows

A D simulation of the PED approach using TOVS and AIRS radiances was com

pleted in early The results showed that the data compression for b oth AIRS and

TOVS should signicantly reduce computation in PSAS Based on the D results the

theoretical foundation for this approach app ears sound

A fast forward radiative transfer mo del was extracted from the TOVS Pathnder

co de Susskind et al and an analytic Jacobian can b e used as TLM and adjoint

was added This mo del called GLATOVS is describ ed in Sienkiewicz This mo del

is currently b eing compared in several resp ects with RTTOVS

Systematic error correction to account for errors in the forward mo del and instrument

calibration is an integral part of assimilating data from any microwaveIR sounder In

early a simulation of forward mo del error was completed The results showed that

as exp ected systematic error can b e as large or larger detector noise In the rst part

of several systematic error correction schemes have b een compared in a simulation

environment Currently a physicallybased correction scheme is b eing implemented with

TOVS Pathnder data After the algorithm has b een tested validated and frozen the

results will b e do cumented The do cumentation should b e completed in summer of

After the systematic error correction scheme has b een implemented the radiance

error covariance will b e estimated using the approach describ ed in da Silva et ala

This activity will take place in summer of

Beginning in fall when observation op erators have b een implemented in PSAS

a comparison of the PED and the direct radiance assimilation approaches will b e completed

using TOVS Pathnder data for one season The results will also b e compared with the

current metho d of tuned retrieval assimilation One approach will b e selected on the basis

of cost and analysis quality for use in future systems

The selected metho d will b e implemented with NESDIS radiances in winter

The quality of the analysis with radiances from NESDIS and Pathnder will b e

compared If there is a signicant dierence in data quality the Pathnder data will b e

used for reanalysis

Stratospheric temp erature measurements from MLS were also given a high priority for

assimilation as a result of a commitment from the MLS team The availability of MLS

temp oral coverage makes it a candidate for p o cket analysis The rst meeting with the

MLS team will take place in late June of The pro ject design will b egin at that time

The pro ject is exp ected to b egin in late summer and should last approximately one

year

Moisture Assimilation

Conventional data has b een the primary source of moisture data in the GEOSDAS to

date Exp eriments with assimilating total precipitable water TPW from SSMI at the

DAO Ledvina and Pfaendtner have shown p ositive impact esp ecially in the tropics

where the GCM is known to have a dry bias Both TOVS and SSMI and TMI provide

information ab out TPW In addition TOVS provides some additional information ab out

Table Priorities for assimilating temp erature data

Instrument Satellite Observable UsePriority Vol GBday

yConventional NA ANC NOAA RDSONDS OBS L H

yTOVS ATOVS POES NESDIS ANC NESDIS TOVS THKN RETS L H

yTOVS ATOVS POES GLA TropStrat RA H

yMLS UARS Strat RA H

yGPS GPSMET TropStrat V H E

ySSMT DMSP TropStrat V H

yGOES sounder GOES Trop V L S



Nimbus Strat PA H LIMS



CLAES UARS Strat V H

zMODIS EOS AM Trop MOD V H S

zIMG ADEOS TropStrat V L S

the prole and has coverage over b oth land and water The capability to assimilate moisture

information from b oth TOVS as describ ed ab ove and SSMI will b e in place for GEOS

and b eyond The nal decision on which data typ e or combination thereof will b e made on

the basis of computational cost and analysis quality

A collab orative eort is underway with scientists at JPL and DAO to use groundbased

GPS data in the GEOS system The actual assimilation of GPS data into the GEOS

system will b egin in approximately two years after initial studies have b een completed The

numb er of GPS groundbased measurement systems is exp ected to increase dramatically in

the future and initial studies show the TPW from these systems to b e of high quality

Several other instruments received high priority for validation Although several of

these data typ es are strong candidates for assimilation in the GEOS system eg MLS and

SSMT resources in addition to those currently in place would b e required for this as

these data typ es have little heritage

ConvectivePrecip Retrieval Assimilation

Developing the capability to assimilate convective data including precipitation is an ongo

ing research pro ject at DAO The goal is to have this capability ready in fall of and

to compare this approach with the physical initialization approach used by Krishnamurti

et al in winter of The basic approach is to use a physical mo del RAS as an

observation op erator to p erform a D retrieval of water vap or which is then assimilated in

PSAS in the context of the CARDPED approach describ ed ab ove Currently a simplied

version of this algorithm has b een co ded and is undergoing tests This approach do es not

assume that either the data or the mo del are p erfect The rst implementation will use data

inferred from SSMI TRMM data will b e used when it b ecomes available In the future

other data typ es including inferred cloud top temp erature and OLR will b e assessed for

assimilation in a similar manner A summary of convective data typ es is given in table

Land Surface

The DAO is currently developing the capability to assimilate landsurface data At the

present time the Mosaic Koster and Suarez landsurface mo del LSM has just

b een coupled with the GEOS GCM and is undergoing tests Currently oline tests with

the Mosaic LSM are also b eing p erformed The Mosaic LSM do es not have the ability

to accept satellite data in order to sp ecify b oundary conditions It is planned that part

of the Sib eg Sellers et al LSM that accepts satellite data will b e integrated

Table Priorities for assimilating water vap or data

Instr Satellite Observable UsePriority Vol GBday

yConven NA ANC NOAA RDSONDS OBS L H

yTOVS ATOVS POES NESDIS ANC NESDIS TOVS RETS L H

yTOVS POES GLA Trop RA H

ySSMI DMSP ANC MSFC SSMI PRCP WATER L H

yGPS Ground TPW RA H

ySAGE ERBS etc Strat V H

yMLS UARS Strat V H

ySSMT DMSP Trop V H

yGOES GOES Trop prof V L S



UARS Strat V H HALOE

zTMI TRMM TPW FP H

zMODIS EOS AM Trop MOD L V H S

zIMG ADEOS Trop V L S

Note Volumes in GBday

Table Priorities for assimilating convective retrievals

Instrument Satellite Observable UsePriority Vol GBday

ySSMI DMSP Prec rate surf FP H

ySSMI DMSP Cloud liq H O V H

2

yISCCP POES Cloud Top frac V H

yConven NA Cloud base V H

yTOVS POES GLA Cloud top V H

yConven NA Radar Precip V H

yMSU POES Path Precip o cean V L



ERBS TOA ux V H ERBE

zTMI TRMM M TMI PROF LA FP H S

zTMIPR TRMM TRMM COMB LBM V H S

zPR TRMM PR Prof M V H S

zMODIS EOS AM Cloud MOD V H S

zCERES TRMM TOA cloud CER V H S

zCERES EOS AM TOA cloud CER V H S

zAMSU POES Cloud liq H O V L TBD

2

Table Priorities for assimilating landsurface data

Instrum Satellite Observable UsePriority Vol GBday

yConven NA T q ANC NOAA SFC OBS FP RA H

a a

yTOVS POES GLA T FP RA H

s

ySSMI DMSP Snow WatEq SWE FP RA H

ySSMI DMSP Surface Wetness FP RA H

yAVHRR POES ANC NESDIS NDVI B RA H S



Nimbus SWE RA H SMMR

zMODIS EOS AM T MOD L V FP H

s

zMODIS EOS AM NDVI MOD L DY B FP H GBdayS

zMODIS EOS AM LAI FPAR MOD L DY B FP H GBdayS

zMODIS EOS AM Evaptrans MOD L DY B FP H GBdayS

zMODIS EOS AM MOD BRDF L DY B FP H GBdayS

zMODIS EOS AM Land typ e MOD L DY B FP H dayS

zMODIS EOS AM LS res MOD L B FP H day

zASTER EOS AM T on demand only V L S

s

with the Mosaic mo del in GEOS When this is accomplished satellite derived pro ducts

such as the Normalized Dierential Vegetation Index NDVI and others listed in Table

from AVHRR and MODIS will b e ingested In addition to surface station observations

temp erature T and humidity q we are investigating the feasibility of assimilating surface

a a

skin temp erature T derived from IR instruments as well as snow water equivalent

s

SWE and surface wetness inferred from passive microwave instruments It is exp ected

that some combination of these data typ es will b e used in the nal platform system

Ocean Surface

Ocean wind data from b oth passive microwave SSMI and SMMR and scatterometers

ERS and NSCAT were identied as high priority data typ es for the system b oth

rst lo ok and nal platform Table lists the details of o cean surface data typ es considered

For the system the capability of assimilating data from b oth passive microwave and

scatterometers will b e in place However b efore one or more of these data typ es are used

the data will have b een monitored and corrected if necessary for systematic errors including

observation op erator error The sp ecic plans for use of NSCAT data are outlined b elow

Starting in Decemb er the Synoptic Evaluation Group will start receiving NSCAT

science pro ducts which includes backscatter measurements Level data retrieved

ranked wind ambiguities Level data as well as gridded wind elds Level data

The op erational release of the data is scheduled to start ab out March although the

data format will change The initial plan is to assimilate NSCAT retrieved wind velo city

and p ossibly employ an ambiguity removal algorithm

The pro cess of assimilating surface wind data includes the following steps

A pro cess known as MOVEZ is used to create a suitable rst guess wind eld by

a reduction pro cess of the lowest mo del level wind eld to m consistent with the GEOS

PBL The multivariate wind and sea level pressure analysis is p erformed Subsequently

the analyzed wind eld is extended up to the lowest vertical mo del level and the analysis

increments are computed The analysis increments of the upp er air wind eld at hPa

and the lowest mo del level are then interp olated linearly in log P to the vertical mo del

levels in b etween This work will b egin approximately weeks after the GEOS with

PSAS is frozen

Table Priorities for assimilating o ceansurface data

Instrument Satellite Observable UsePriority Vol GBday

yConventional NA u v T q L H

air s

ySSMI DMSP Wind sp eed L RA H

o

yERS ERS sp eed dir L FP H E



Nimbus sp eed V L SMMR

zNSCAT ADEOS ANC NESDIS NSCAT WNDPDT L H

zNSCAT ADEOS ANC JPL NSCAT WNDPDT FP H

In order to maximize the inuence of the surface wind eld in the GEOS DAS the

NSCAT wind vectors are also extended by a pro cess analogous to MOVEZ to the mo del

lowest vertical level The observation innovations are computed and used by the multivariate

mass wind analysis A check is made of the atmospheric stability and only the observations

under the unstable or neutral conditions are used The D global analysis that makes use of

the data and propagates the information in the vertical according to the background error

vertical correlation function This particular pro cess has exp erimentally b een applied to

ERS scatterometer data within the GEOS DAS environment and presented in Tokyo in

It resulted in mo dest improvement of the resultant forecast exp eriments Within the

framework of PSAS it is exp ected to result in a more signicant impact of scatterometer data

due to a truly global three dimentional design of PSAS This step will b egin approximately

weeks after step

A simple surface wind ambiguity removal algorithm is in place in GEOS DAS It

will b e tested with the new GEOS DAS The algorithm compares the directions of available

ambiguous wind vectors with the background eld and cho oses the one closest to it In

the future an interactive pro cedure or a D Variational ambiguity removal algorithm might

replace the current one If done this work would b egin in Jan or Feb of

Constituents

The plans for constituent assimilation will b e given at a later time The data typ es con

sidered for ozone assimilation are given in table The data typ es considered for CO

assimilation are given in table

Wind prole

Currently wind prole data used in GEOS are obtained from radiosondes and derived cloud

track winds Several of the data typ es listed in table app ear to b e strong candidates

for assimilation eg water vap or winds similar to cloud track winds These new data

typ es can enhance the coverage of current wind prole data However b ecause there is

little heritage for assimilating these data typ es they are presently selected for validation

As resources p ermit some of these data typ es may eventually b e assimilated

Aerosols

Although consideration has b een given to aerosol assimilation Menard p ersonal notes

this topic has b een given an overall low priority at present Therefore all data typ es listed

in table were selected for validation

Table Priorities for assimilating ozone data

Instrument Satellite Observable UsePriority Vol GBday

yTOMS NMeteorADEOS total OA H

ySBUV NPOES prof V OA H

yConven NA prof V H

ySAGE ERBS etc strat prof V H

yMLS UARS upp er strat prof V H

yGOME ERS total prof V L

yTOVS POES GLA lower strat V L



CLAES UARS upp er strat prof V L



UARS upp er strat prof V L HALOE

zILAS ADEOS upp er strat prof V L

zIMG ADEOS lower strat prof V L

Table Priorities for assimilating CO data

Instrument Satellite Observable UsePriority Vol GBday



MAPS Shuttle prof OA H

zMOPITT EOS AM prof OA H TBD

yConven NA prof V H

Table Priorities for assimilating wind proles

Instrument Satellite Observable UsePriority Vol GBday

yConventional NA ANC NOAA RDSONDS OBS L H

yConventional NA ACARS ANC NOAA ARCFT OBS L H

ycloud track GOES ANC NESDIS GOES WIND MOTION L H

yRADAR NA tropstratmes VH

water vap or GOESother trop V H



UARS stratmes V L HRDI



UARS mes V L WINDI I



MODE Shuttle V L

zSWIPE TBD

Table Priorities for use of aerosol data

Instrument Satellite Observable UsePriority Vol GBday

ySAGE ERBS etc Strat V L

yTOMS Nimbus StratTrop V L



CLAES UARS Strat V L



UARS Strat V L HALOE

zMISR EOSAM prop erties V L

zILAS ADEOS Strat V L

Table Highpriority data typ es from POES satellite

Instrument Observable Use

yTOVS Temp Profrad L

yTOVS H O Profrad L

2

yTOVS T FP RA

s

yAVHRR ANC NESDIS NDVI B RA

Table Highpriority data typ es from UARS satellite

Instrument Observable Use

yMLS Strat T RA

yMLS Upp er TropStrat H O V

2

yMLS upp er strat O V

3



Strat T V CLAES



HALOE Strat H O V

2



upp er strat prof O V HALOE

3

Priorities group ed by satellite

This section categorizes high priority items by satellite The priorities for the POES UARS

TRMM ADEOS and EOSAM satellites are listed in tables resp ectively

Priorities group ed by use in GEOS

This section categorizes high priority items according to use in GEOS Data typ es are

group ed as either used for rst lo ok nal platform data typ es in addition to those used for

the rst lo ok or reanalysisp o cket analysis may include data typ es used in the rst lo ok

and nal platform in tables resp ectively Some of the highrisk data typ es such

as landsurface data have not b een included here Data typ es used for validation are not

listed here

Table Highpriority data typ es from TRMM satellite

Instrument Observable Use

zTMI PR Prof M FP

zTMIPR PR Prof M V

zPR PR Prof M V

zCERES cloud CER V

Table Highpriority data typ es from the ADEOS satellite

Instrument Observable Use

o

zNSCAT wind sp eed dir FP

yTOMS total O OA

3

Table Highpriority data typ es from EOS AM satellite

Instrument Observable Use

zMODIS NDVI B FP

zMODIS LAI FPAR B FP

zMODIS Evaptrans B FP

zMODIS BidirRe B FP

zMODIS Land typ e B FP

zMODIS LS res B FP

zMODIS Cloud V

zMODIS T V

s

zMODIS Tq Prof MOD V

zMOPITT CO Prof OA

zCERES TOA cloud CER V

Table Highpriority data typ es for rst lo ok system

Instrument Platform Observable Volume

yConventional NA ANC NOAA RDSONDS OBS

yTOVS ATOVS POES NESDIS ANC NESDIS TOVS THKN RETS T q radiance

yConventional ship buoy u v T q

air s

ySSMI DMSP ANC MSFC SSMI PRCP WATER

ySSMI DMSP Wind sp eed

yERS ERS sp eed dir E

zNSCAT ADEOS ANC NESDIS NSCAT WNDPDT

yConventional NA ACARS ANC NOAA ARCFT OBS

ycloud track GOES ANC NESDIS GOES WIND MOTION

Table Highpriority data typ es for nal platform

Instrument Platform Observable Volume

ySSMI DMSP Prec rate surf

zTMI TRMM M TMI PROF LA S

zNSCAT ADEOS ANC JPL NSCAT WNDPDT

zMODIS EOS AM NDVI MOD L DY GBdayS

zMODIS EOS AM LAI FPAR MOD L DY GBdayS

zMODIS EOS AM Evaptrans MOD L DY GBdayS

zMODIS EOS AM MOD BRDF L DY GBdayS

zMODIS EOS AM Land typ e MOD L DY dayS

zMODIS EOS AM LS res MOD L day

Table Highpriority data typ es for reanalysis andor p o cket analysis

Instrument Platform Observable Volume

yMLS UARS Strat T

yTOVS POES GLA T q

References

Andersson E Pailleux J Thepaut J N Eyre J R McNally A P Kelly G A and

P Courtier Use of cloudcleared radiances in threefourdimensional variational data

assimilation Q J R Meteorol Soc

Daley R Estimating the wind eld from chemical constituent observations Exp eriments

with a onedimensional extended Kalman lter Mon Wea Rev

Dee D P Online estimation of error covariance parameters for atmospheric data assimila

tion Mon Wea Rev

Dee D P and A Trenholme DAO QC Strategy Do cument DAO Oce Note Data

Assimilation Oce Go ddard Space Flight Center Greenb elt MD

Eyre J R A bias correction scheme for simulated TOVS brightness temp eratures ECMWF

Tech Memo

Guo J and A M da Silva Computational Asp ects of Go ddards Physicalspace Statis

tical Analysis System PSAS Preprints Second UNAMCRAY Supercomputing Conf on

Numerical Simulations in the Environmental and Earth Sciences Mexico City Mexico pp

Joiner J and A M da Silva Ecient metho ds to Assimilate Satellite Retrievals Based on

Information Content submitted to Mon Weath Rev

Koster R D and M J Suarez Mo deling the land surface b oundary in climate mo dels as

a comp osite of indep endent vegetation stands J Geophys Res

Krishnamurti T N Xue J Bedi HS K Ingles and D Oosterhof Physical initialization

for numerical weather prediction over the tropics Tel lus AB

Ledvina D V and J Pfaendtner Inclusion of Sp ecial Sensor MicrowaveImager SSMI

total precipitable water estimates into the GEOS Data Assimilation System Mon Weath

Rev

Menard R Middle atmosphere assimilation of UARS constituent data using Kalman lter

ing Preliminary results International symposium on assimilation of observations in meteo

rology and oceanography Tokyo Japan

Menard R Consistent assimilation of retrieved data CARD private notes and manuscript

in preparation

Molo d A H M Helfand and L L Takacs The climatology of parameterized physical

pro cesses in the GEOS GCM and their impact on the GEOS Data Assimilation System

J Climate

NASAGo ddard Space Flight Center Science Data Plan for the EOS Data and Information

System covering EOSDIS Version and Beyond Do cument version Ed M Schwaller and

B Krupp

Riishjgaard L P On fourdimensional variational assimilation of ozone data in weather

prediction mo dels accepted for publication in the Q J R Meteorol Soc

Ro dgers C D Characterization and error analysis of proles retrieved from remote sounding

measurements J Geophys Res

Pfaendtner J S Blo om D Lamich M Seablom M Sienkiewicz J Stobie A da Silva

Do cumentation of the Go ddard Earth Observing System GEOS Data Assimilation System

Version NASA Tech Memo No Vol Go ddard Space Flight Center Greenb elt

MD

Schub ert SD R B Ro o d and J Pfaendtner An assimilated data set for earth science

applications Bul Amer Meteor So c

Schub ert S CK Park ChungYu Wu W Higgins Y Kondratyeva A Molo d L Tkacs M

Seablom and R Ro o d A Multiyear Assimilation with the GEOS System Overview

and Results NASA Tech Memo Vol Go ddard Space Flight Center Greenb elt

MD

Schub ert S D DAOs Quality Assessment Team Denition and Planing Do cument DAO

Oce Note Data Assimilation Oce Go ddard Space Flight Center Greenb elt MD

Schub ert S D and R Ro o d Pro ceedings of the Workshop on the GEOS FiveYear

Assimilation NASA Tech Memo Vol Go ddard Space Flight Center Greenb elt

MD

Sellers P J Y Mintz Y C Sud and A Dalcher A simple b oisphere mo del SiB for use

within general circulation mo dels J Atmos Sci

Sienkiewicz M The GLA TOVS rapid algorithm forward radiance mo dules and Jacobian

version DAO Oce Note Data Assimilation Oce Go ddard Space Flight Center

Greenb elt MD

da Silva A M Pfaendtner J Guo J Sienkiewicz M and S E Cohn Assessing the

eects of data selection with DAOs Physicalspace Statistical Analysis System International

symposium on assimilation of observations in meteorology and oceanography Tokyo Japan

da Silva A and J Guo Do cumentation of the PhysicalSpace Statistical Analysis System

PSAS Part I The Conjugate Gradiant Solver version PSAS DAO Oce Note

Data Assimilation Oce Go ddard Space Flight Center Greenb elt MD

da Silva A M Redder C and D P Dee a Mo deling retrieval error covariances for global

th

data assimilation Proc Conf Satel l Meteor Atlanta GA in press

da Silva A Ekers K and A Conaty b Requirements for DAOs OnLine Monitoring

System DOLMS version DAO Oce Note Data Assimilation Oce Go ddard

Space Flight Center Greenb elt MD

Stobie J Data Assimilation Computing and Mass Storage Requirements for DAO

Oce Note Data Assimilation Oce Go ddard Space Flight Center Greenb elt MD

Stoelen A and D Anderson The ECMWF contribution to the characterisation inter

pretation calibration and validation of ERS scatterometer backscatter measurements and

winds and their use in numerical weather prediction mo dels European Space Agency Contract

Report

Susskind J Roseneld J and D Reuter An accurate radiative transfer mo del for use in

the direct physical inversion of HIRS and MSU temp erature sounding data J Geophys Res

Susskind J and J Pfaendtner Impact of iterative physical retrievals on NWP Rep ort on

the Joint ECMWFEUETSATWorkshop on the Use of Satellite Data in op erational weather

prediction Vol T Hollingsworth Ed

Takacs LL A Molo d and T Wang Do cumentation of the Go ddard Earth Observing

System GEOS General Circulation Mo delVersion NASA Tech Memo No

volume Go ddard Space Flight Center Greenb elt MD

App endix

App endix A Acronyms

CARD Consistent Assimilation of Retrieved Data

CTW Cloud track winds

DAO Data Assimilation Oce NASA Go ddard Space Flight Center Lab oratory for Atmospheres

DAS Data Assimilation System

EOSDIS Earth Observing System Data and Information System

GCM General Circulation Mo del

GEOS Go ddard Earth Observing System

EKF Extended Kalman Filter

LSM LandSurface Mo del

MTPE Mission to Planet Earth

NDVI Normalized Dierential Vegetation Index

NWP Numerical Weather Prediction

PSAS Physicalspace Statistical Analysis System

PV Potential Vorticity

SWE Snow Water Equivalent

TPW Total precipitable water

App endix B Instruments and Satellites

ADEOS Advanced Earth Observting Satellite Midlate

AIRS Atmospheric Infrared Sounder EOS PM

AMSU AB Advanced Microwave Sounding Unit POES EOS PM

AVHRR Advanced Very HighResolution Radiometer

ASTER Advanced Spaceb orne Thermal Emission and Reection Radiometer EOS AM

ATOVS Advanced TOVS HIRSAMSU POES

CERES Clouds and Earths Radiation Energy System TRMM EOS AM

CLAES Cryogenic Limb Array Etalon Sp ectrometer UARS

DMSP Defense Meteorological Satellite Program currently op erational

EOS AM Earth Observting Satellite AM June launch

EPS EUMETSAT Europ ean Meteorology Satellite Polar System

ERBE Earth Radiation Budget Exp eriment ERBS

ERBS Earth Radiation Budget Satellite

ERS Europ ean Remote Sensing Satellite Scatterometer channel IRVisible radiometer

GOES Geostationary Observational Environmental Satellite Imager and channel visible and

infrared sounder currently op erational

GOME Global Ozone Monitoring Exp eriment ERS

GPS Global Positioning System

HALOE Halogen Occultation Exp eriment UARS

HIRS HighResolution InfraRed Sounder POES

HRDI High Resolution Doppler Imager

IASI Infrared Atmospheric Sounding Interferometer EPS

ILAS Improved Limb Atmospheric Sp ectrometer ADEOS

IMG Interferometric Monitor for Greenhouse Gases ADEOS

ISCCP International Satellite Cloud Climatology Pro ject several IR and visible instruments

ab oard dierent satellite

LIMS Limb Infrared Monitor of the Stratosphere Nimbus

MAPS Measurement of Atmospheric Pollution from Satellites

MHS Microwave Humidity Sounder EOSPM

MLS Microwave Limb Sounder UARS

MODIS Mo derateResolution Imaging Sp ectrometer EOS AM

MSU Microwave Sounding Unit

MOPITT Measurement of Pollution in the Trop osphere EOS AM

NSCAT NASA Scatterometer ADEOS

POES Polar Orbiting Environmental Satellite Currently Op erational

PR Precipitation Radar TRMM

SBUV Satellite Backscatter Ultraviolet radiometer Nimbus POES

SAGE Stratospheric Aerosol and Gas Exp eriment ERBS

SMMR Scanning Multisp ectral Microwave Radiometer

SSMI Sp ecial Sensor MicrowaveImager DMSP

SSMT Sp ecial Sensor Microwave Temp erature sounder DMSP

SSMT Sp ecial Sensor Microwave Water vap or sounder DMSP

SSU Stratospheric Sounding Unit POES

TMI TRMM Microwave Imager TRMM

TOMS Total Ozone Mapping Sp ectrometer ADEOS Meteor Earth Prob e Nimbus

TOVS TIROS Op erational Vertical Sounder HIRSMSUSSU POES

TRMM Tropical Rainfall Measuring Mission summer launch

UARS Upp er Atmospheric Research Satellite some instruments in op eration

WINDI I Wind Imaging Interferometer UARS