Future performance of ground-based and airbornewater-vapor differential absorption lidar.II. Simulations of the precision of a near-infrared,high-power system
In part I of this series1 an advanced water-vapordifferential absorption lidar �DIAL� system to per-form water-vapor measurements under a variety ofatmospheric conditions was proposed. It was shownthat this system must be a direct-detection systemoperating in a region near 940 nm. The equationsfor the simulation of its precision for ground-based,
airborne, and scanning configurations were derivedin part I as well.1
In this paper we use these equations to discuss theperformance of this DIAL system and to demonstrateits flexibility and its importance for the atmosphericsciences. This system is currently under develop-ment at Hohenheim University in Germany and atthe U.S. National Center for Atmospheric Research.A design goal of achieving 10 W of average power hasbeen set. We define average power as that at theon-line frequency or at the off-line frequency, whichare assumed to be the same. This DIAL system willbe equipped with a 1-m telescope on the ground anda 0.4-m telescope on an airplane.
By means of simulations we demonstrate that thesystem under consideration yields excellent resultson ground-based as well as on airborne platforms upto the lower stratosphere. Furthermore, the resolu-tion will be high enough to resolve, for the first timeto our knowledge, three-dimensional water-vaporfields in the boundary layer.
When this research was performed, the authors were with theAtmospheric Technology Division, National Center for Atmo-spheric Research, Boulder, Colorado 80307. V. Wulfmeyer wasalso with the Environmental Technology Laboratory, NationalOceanic and Atmospheric Administration, Boulder, Colorado80303. He �[email protected]� is now with the Insti-tute of Physics, University of Hohenheim, 70599 Stuttgart, Ger-many.
Received 20 November 2000; revised manuscript received 15May 2001.
Equation �27� of part I of this series1 estimates thedependence of the precision of a water-vapor DIALmeasurement on system and atmospheric parame-ters and reads as
�n
nH2 O� [ 1
����2
12mk ( Fdet
�DNs,off
�exp�2�� � 1�
�Fdet
�DNs,off
2*Nb
Ns,off
�exp�4�� � 1�
�2*B�Idet
2 � Iamp2 �
�BeG�DNs,off�2
�exp�4�� � 1� �1l �2�FT
�L
�2*Nb
2
Ns,off2
�FT
�f�exp�4�� � 1��)]1�2
(1)
in the range interval �Rmin, Rmax�. Here we define�n as the uncertainty in absolute-humidity measure-ment that results from uncorrelated noise and nH2O
isthe absolute humidity. Consequently, �n�nH2O
isthe relative error, which in what follows we call theerror of the humidity measurement owing to uncor-related noise. This error is given in percent. A lowerror corresponds to high precision.
In the derivation of relation �1� it has been assumedthat the overlap between laser transmitter beam andtelescope field of view is complete in the range inter-val. Relation �1� consists of four terms, which aredue to Poisson photon statistics of the backscattersignal, background noise, detector and amplifiernoise, and speckle noise. In what follows, we callthese terms Poisson, background, amplification, andspeckle terms, respectively.
All the terms depend strongly on the various opti-cal thicknesses �� nH2O
��R in the range cell,where �R is the range resolution and � is the water-vapor absorption cross section at the on-line fre-quency of the laser transmitter. The effects ofspatial and temporal averaging are taken into ac-count by the number m of independent range bins inthe range cell and the number k of independent shotsduring the averaging time T.
The factor D is due to optical and electronic dy-namic range reduction of the backscatter signal.For completeness we include this factor in the errorequation. In the simulations presented here we donot take into account dynamic range reduction; there-fore, D 1. Any dynamic range reduction will in-crease the error only in the near range, which turnsout to be lower in all simulations and is matched suchthat it reaches the predicted precision in the farrange. We assume that the full dynamic range ofthe backscatter signal can be captured by an ad-vanced detector and data-acquisition system in therange interval �Rmin, Rmax�. We also assume thatwithin this range interval the digitizer resolution issufficiently high that digitization does not limit theprecision of the measurement �for a further discus-sion, see Ref. 1�.
The Poisson term depends on the excess-noise fac-tor of the detector Fdet, the detector quantum effi-ciency �, and the backscatter signal’s photon number.The last-named property can be estimated from thebackscatter signal’s photon number at the off-linefrequency Ns,off, which is calculated from the lidarequation, and with the optical thickness � of watervapor at the on-line frequency taken into account. Ifthe precision is limited by the Poisson term, the pre-cision depends only on the average power of the lasertransmitter and not on the pulse energy chosen�while the same average power is maintained�.
The background term depends on background pho-ton number Nb. We have to distinguish betweentwo situations in the determination of Nb: For aground-based system pointing with a certain eleva-tion and azimuth angle, Nb can be calculated accord-ing to
Nb �I
h�f A��FOV�2�rec�t. (2)
I is the atmospheric radiation entering the detectorchannel, which is usually given in W cm�2 sr�1 m�1
�see Fig. 9 of Ref. 1�, h is Planck’s constant, and isfrequency. It has to be calculated by use of a radi-ative transfer program. �f is the FWHM of thetransmission of an interference filter used in the de-tector system, A is the aperture of the receiving tele-scope, and FOV is the half-angle field of view of thereceiving telescope. Note that FOV2 is proportionalto the solid angle �in steradians�. �rec is the trans-mission of the receiver and �t is the integration time,which in what follows is set equal to the duration ofthe laser pulse.
For a downward-pointing airborne system a simi-lar equation is used, which estimates the Earth’ssurface-reflected background that is entering the de-tector system channel.2 This equation reads as
Nb �L
h
�
�exp��2�atm����f A��FOV�2�rec�t. (3)
Here L is the spectral density of the solar irradiance,� is the Earth’s reflectivity, and �atm�� is the single-path atmospheric optical thickness. A quick assess-ment of Nb near the region about 940 nm shows thatit is of the same order of magnitude for a ground-based upward-looking and an airborne downward-looking system �see also Table 1�.
Detector and amplifier noise is taken into accountby the amplification term that includes their noisecurrents Idet
2 and Iamp2 , respectively. The estimations
of the noise currents also include the dark currents.The electron charge is e, and G is the gain of thedetector. We consider a matched filter design of thereceiver such that B, which is the receiver bandwidth,corresponds to the pulse width of laser transmitter�t. Therefore B 1��t.
Speckle noise is considered in the speckle term.The number of independent speckles in an indepen-dent range bin and a single laser shot is l. It can be
calculated from the equations presented in part 1.1Furthermore, the speckle term depends on the ratioof the Fourier-transform limit �FT of the laser pulseand the real laser bandwidth �L. As speckle noisein the daylight background can play a role in precisemeasurements, the FWHM bandwidth �f of the re-ceiver is also part of the speckle term.
To estimate the off-line backscatter signal’s photonnumber we use the lidar equation
Ns�R� �E0
h
c�t2
�las�recO�R���R�AR2
� exp��2 �0
R
��r�dr� , (4)
where R is range, E0 is the laser pulse energy, is thelaser frequency, and c is the speed of light. The min-imum range resolution �Rmin can be approximatedby the range interval c�t�2. �las is the overalltransmission of the laser steering optics. O�R� is thenormalized overlap function. At full overlap be-tween the FOV and the laser transmitter, O�R� isnormalized to unity. � is the atmospheric backscat-ter coefficient, which is the sum of particle and mo-lecular backscatter. In the derivation of Eq. �4� itwas assumed that � does not change considerablywithin �Rmin. � is the atmospheric extinction coef-ficient that is due to molecular and particle extinctionas well as to eventual absorption by a gas.
If not otherwise stated, we are using the particle
and molecular backscatter model as introduced inpart 1,1 which uses data extracted from Ref. 3. It isclear that aerosol backscatter coefficient �par can varyconsiderably, depending on atmospheric conditions.However, it is beyond the scope of this study to per-form further simulations under a variety of condi-tions. We can estimate the change in the error byconsidering relation �1� and by making some assump-tions about changes in overall backscatter coefficient�. Relation �1� shows that, to first order, the errordepends on the inverse square root of �. In thelower and middle troposphere, � corresponds mainlyto �par. Allowing a change of �par of 1 order of mag-nitude would cause a change in the error profile by afactor of between 1��10 and �10. In the middletroposphere up to the lower stratosphere, in the worstcase there are no aerosol particles present. Thischange of � would cause an increase of the errorprofile by approximately a factor of �10 at 940 nmand by a factor of 10 at 1400 nm. Note that in themiddle troposphere to the lower stratosphere, whenthere are elevated aerosol particle layers, a decreasein the estimated error could occur. In conclusion,good approximations for error bars in the estimatedsystem performance are an increase in the error by afactor of �10 � 3.3 at 940 nm and 10 at 1400 nm inthe case of low �par and a decrease by as much as afactor of 1��10 � 3.2 in the case of high �par.
We consider a system with an average power of 10W at the off-line and at the on-line frequencies. Thisvalue is technically feasible if one takes into accountrecent advances in laser technology.4–6 The ex-pected precision of other systems with different av-erage power can be simulated from relation �1� aswell. Table 1 summarizes the range of values formost of the system parameters that we used in thefollowing simulations. The assumption for detectorand amplifier noise is a safe overestimation, as re-cently detector systems became available that haveeven less overall noise, of the order of 0.4 � 10�12 AHz�0.5.7 The prescribed telescope diameters are 1 mfor ground use and 0.4 m for airborne use. We per-form simulations only at wavelengths near 940 nmand 1400 nm, because these were the absorptionbands on which we reported in part 1.1
The relative error in a water-vapor measurement isinversely proportional to differential optical thick-ness �� but increases with overall optical thickness �.One needs to make a compromise to maintain a lowrelative error while performing measurements in along range. Remsberg and Gordley8 and Wulf-meyer9 demonstrated that the best compromises toallow for measurements with good spatial resolution,good signal-to-noise ratio, and long range are 0.02 ��� � 0.1 as well as � � 2 up to the maximum range.These considerations were taken into account in thesimulations described below.
We start by analyzing the order of magnitude oferror terms. To do this we calculate the rms noisecurrents of all error terms and compare them withthe currents produced by the on-line and off-linebackscatter signals. We consider the simple case of
Table 1. Parameters for Performance Simulationsa
Parameter Value
ConstantsPlanck’s constant h �Js� 6.626 � 10�34
Speed of light c �ms�1� 3 � 108
Electron charge e �C� 1.602 � 10�19
Wavelength �nm� 945 1400Atmospheric parameters for a ground-based system
Daytime background I
�W m�2 sr�1 nm�1�0.04 0.01
Atmospheric parameters for a downward-pointing systemDaytime background L
�W m�2 nm�1�0.8 0.4
Atmospheric optical thickness �atm 0.25 0.1Detector system parametersDetector gain G 200.0 0.9Detector quantum efficiency � 0.8 0.9Detector noise current Idet
�A Hz�0.5�0.4 � 10�12 6.0 � 10�15
aFor both 945 and 1400 nm, the following laser parameters arethe same: off-line average power, 10 W; pulse duration �t, 10–100 ns; transmission of steering optics �las, 0.98, and FWHM ofinterference filter ��F �nm� 0.2–1.0. The following atmosphericparameters are the same: nighttime background L, 0.0 W m�2
sr�1 nm�1; surface reflectivity �, 0.5. The following detector sys-tem parameters are the same: telescope radius Rt, 0.5 m on theground and 0.2 m on an airplane; field of view, 0.2–0.5 mrad.Transmission of receiver optics �rec, 0.9; detector noise figure Fdet,3.0; filter bandwidth of detector system B, 10.0–50.0 MHz; ampli-fier noise current Iamp, 20 � 10–12 A Hz�0.5; detector quantumefficiency �, 0.8; amplifier gain, 1000–100,000 V A�1.
a horizontally pointing system in the atmosphericboundary layer �ABL�, as all conclusions also hold forother measurement configurations. Table 2 sum-marizes the additional parameters used in this sim-ulation. Figures 1 and 2 compare the results at 940and 1400 nm. Starting at a range of approximately500 m, the signals decrease by as much as 6 orders ofmagnitude until they reach the rms noise level. Thedynamic range may be even 1 or 2 orders of magni-tude higher for a vertically pointing system or underdifferent atmospheric conditions. This variation inmagnitude alone poses a difficult measurement prob-lem and again demonstrates that extending the mea-surements to a far range of more than 15 km requiresdynamic range reduction. The factor 2*, which ex-presses the ratio of spatial averaging used for signaland background measurements, is assumed to be1.25, which means that the range for backgrounddetermination is four times larger than the DIALrange resolution. This condition is maintained forall simulations presented here.
Figures 1 and 2 demonstrate that the Poisson andspeckle noise currents are particularly important inthe near range. The Poisson noise also sets anultimate error limit in the far range. It is obviousthat, when one is using a diffraction-limited beamin the near range, the speckle term is significantlyhigher than the Poisson term. At 940 nm, the lim-itation is set by daylight background and amplifiernoise. The daylight background starts to play animportant role in error analysis at 940 nm if aconventional receiver design �FOV, �0.5 mrad;��F � 1.0 nm� is used. A more-sophisticated sys-tem with a FOV of �0.2 mrad and ��F � 0.2 nm,which is feasible without posing too harsh require-ments for alignment and stability, provides an effi-cient reduction of the daylight background.Consequently, reaching the same daylight noise
level as amplifier noise requires a sophisticated re-duction of daylight background. Depending on thesetup for the measurement, it may be necessary toreduce speckle noise by degradation of the laserspecifications �see below�.
At 1400 nm the daylight background does not setan important limitation on speckle. The level ofamplifier noise strongly limits the measurementrange. This lower level serves for an unacceptablereduction of the measurement range by a factor of 2compared with performance at 904 nm. Note thatthis performance difference will be even worse foroperations in the upper troposphere and in thelower stratosphere because of the additional de-crease in the backscatter coefficient at 1400 nm.Figures 1 and 2 confirm the predictions made inSubsection 4.A.6 of part I1. These results lead to
Fig. 1. Comparison of the currents produced by the off-line andthe on-line signals of a 10-mJ horizontally pointing system in theABL at 940 nm. These currents are compared with typical rmsnoise currents produced by Poisson noise, speckle noise, daylightbackground noise, and detector–amplifier �det�amp� noise.
Fig. 2. Comparison of the currents produced by the off-line andthe on-line signals of a 10-mJ horizontally pointing system in theABL at 1400 nm. These currents are compared with typical rmsnoise currents produced by Poisson noise, speckle noise, daylightbackground noise, and detector–amplifier �det�amp� noise.
Table 2. Parameters for the Simulation of a Horizontally Pointing orScanning DIAL System in the ABLa
Variable
Altitude �nm�
940 1400
Atmospheric parameterLine center of absorption line
�cm�1�10 804.158 7702.545
Ground-state energy E �cm�1� 275.5 222.05Backscatter coefficient �
�m�1 sr�1�5 � 10�6 4.2 � 10�6
Atmospheric extinction coefficient ��m�1�
10�4 8.3 � 10�5
aFor both 940 and 1400 nm, the following laser parameters arethe same: pulse duration �t, 20–100 ns; laser spectral bandwidth�L, �FT; laser beam radius RL, 0.4 m; beam divergence excessfactor F�, 1; off-line pulse energy E0 �J� and repetition rate L �Hz�are given in Figs. 1–6. The following atmospheric parameters arethe same: absolute humidity �, 8.7 g m�3; water-vapor numberdensity nH2O, �2.9 � 1017 cm�3; line strength of absorption line S,10�24 cm. The following parameters for data analysis are also thesame: range resolution �R, 300 m; average time T, 10 s; differ-ential optical thickness ��, 0.03.
the recommendation that the DIAL system shouldoperate at 940 nm. In its final stage the detectorsystem should have an �0.2-mrad FOV, a ��F � 0.2nm, and the amplifier noise must be reduced tovalues of �2 pA Hz�0.5.
Another important issue is the trade-off betweenhigh pulse energy and lower repetition rate and be-tween low pulse energy and high repetition rate whilethe average power is maintained. Figures 1 and 2demonstrate that one can extend the range of opera-tion by increasing the pulse energy, as this increasedpulse energy will increase the signal current abovethe rms noise current in the far range. Particularlyfor the region near 940 nm, where even a 10-mJ pulseis often not eye safe �see Fig. 10 of Ref. 1�, this deci-sion will not have an effect on the design of the sys-tem, because in any event a safety radar has to beincorporated.
Another way to demonstrate the importance of thebackground, the amplification, and the speckle termsis shown in Figs. 3 and 4. Here the effects of theseterms on the relative error in a horizontal DIAL mea-surement in the ABL are presented. The parame-ters summarized in Table 2 also hold for thissimulation. At 940 nm, the amplification and thebackground term are of the same order if good sup-pression of daylight background is provided. Thisresult belies the statement that is often made in con-nection with DIAL system performance that a DIALsystem does not exhibit a significant difference indaytime and nighttime performance. Using a so-phisticated detector system, we could achieve a
14-km performance range while maintaining an errorof �10%.
At both wavelengths, the final limitation is set bythe amplification term. A clear performance differ-ence at 940 and at 1400 nm is evident and is due tothe low gain of the detector at 1400 nm. Thereforeat this point we proceed only with analyses at 940nm, as we consider the performance of a 1400-nmsystem not good enough for our applications.1 Thesame holds true for a system operating at 1900 nm, asthe detectors have the same specifications and thebackscatter signal is even less at this wavelength.
Figures 3 and 4 demonstrate that the speckle termdoes not play an important role in a 10-mJ systembecause of signal averaging. However, speckle isvisible at lower repetition rates as well as at higherrange and time resolution, so a separate discussion ofthis term when one is considering low-repetition-ratesystems is required. Another interesting effect is anincrease of the speckle term if the daylight back-ground photon number increases �see relation �1��.However, this happens beyond the range where theamplification and the background terms become im-portant.
A detailed investigation, with the parameters sum-marized in Table 2, of the speckle term at 940 nm ispresented in Fig. 5. Nighttime and daytime speckleterms for different settings of pulse energies areshown. In the first 10 km, the speckle error becomesan important error term if the repetition rate is de-creased. This is so because only a large number ofshots within the averaging time as well as an in-
Fig. 3. Error analysis of a 10-mJ horizontally pointing system inthe ABL at 940 nm.
Fig. 4. Error analysis of a 10-mJ horizontally pointing system inthe ABL at 1400 nm.
crease in the number of independent range bins canreduce the speckle noise. Otherwise, the speckleterm is independent of the signal photon number �seerelation �1��. Nevertheless, the speckle term is stilllow enough that operation at high pulse energy isbeneficial. It has to be taken into account that Fig.5 presents results for a Fourier-transform-limited la-ser spectrum and a diffraction-limited beam. An-other way to decrease the speckle term is to shortenthe laser pulse such that the number of independentrange bins in the range cell is increased.
The effect of a degradation of the laser spectrumabove the Fourier-transform limit and the beam di-vergence is shown in Fig. 6. It is simple to degradethe spectral performance of a laser as well as its beamdivergence, but it is surprising and—to ourknowledge—newly discovered that this degradationis beneficial and causes a significant reduction in thespeckle term of a high-pulse-energy system in the farrange.
Consequently, the main results of Figs. 1–6 arethat a 940-nm system is by far superior to 1400- and1900-nm systems with respect to range and resolu-tion. Operation with high pulse energy is beneficial,although the speckle term can become dominant inthe near range. The degradation of the laser band-width as well as the beam divergence is beneficial andmay be necessary, under certain measurement con-ditions, for reducing the speckle term. Also, a re-duction of the pulse duration is a way to reduce thespeckle term. If the detector system is designed
with a FOV of 0.2 mrad and ��F 0.2 nm, the rangeof a 940-nm system is limited by both amplifier andbackground noise. In what follows, these detectorsystem parameters are used for all the simulations.
3. Simulations
A. Ground-Based Long-Range Horizontal Measurementin the Atmospheric Boundary Layer
Here we analyze in more detail an important appli-cation of a future DIAL system. Horizontal pointingwill allow for the investigation of the line-of-sighthorizontal variability of water vapor. These simu-lations are the basis for the investigation of scanningapplications that are discussed in Subsection 3.B be-low. Figures 7–9 show the contributions of the mostimportant error terms and their variability withchanges in the pulse energy. The parameters sum-marized in Tables 1 and 2 were used for the simula-tions unless not otherwise stated.
During daytime, in all situations the backgroundterm limits the measurement range. The differencein daytime and nighttime performance is small, asthe error caused by the amplification term is close tothe effect of the background term. In the nearrange, the error is due mainly to Poisson statisticsand speckles. Increasing the pulse energy helps toextend the measurement range. An increase of thepulse energy from 10 mJ to 1 J causes an extension ofapproximately 4 km in the 10% error boundary. Atlower repetition rates, the speckle term sets the limitto the precision in the near range if the laser beam isdiffraction and Fourier-transform limited. Never-
Fig. 5. Error analysis of the speckle term in a horizontally point-ing system in the ABL. The laser pulse energy varies between 10mJ and 1 J. The detector system parameters are set to �f 0.5nm and a FOV of 0.5 mrad.
Fig. 6. Error analysis of the speckle term in a horizontally point-ing system in the ABL. The laser pulse energy is set to 1 J.
theless, even at lower repetition rates the speckleterm is low enough that the range is extended con-siderably at higher pulse energies. We may achievea further reduction of the speckle effect at high pulseenergy by increasing the laser spectral bandwidthand the beam divergence and reducing the pulse du-ration �see above�.
These statements are confirmed in Fig. 10, where itis shown that the speckle term in a 1-J system isreduced by a shorter pulse duration and a largerspectral bandwidth, as well as by greater beam di-vergence. We can reach a range of �19 km whilemaintaining an error of �10% and significantly re-ducing the error in the near range to �2%.
Figures 7–10 show that with a 10-W laser trans-mitter a range of more than 12 km is feasible withresolutions of 300 m and 10 s and a precision of lessthan 10%. During daytime this performance cannotbe achieved by any other remote-sensing technique.Inasmuch as these measurements are currently notavailable, we expect to see major contributions toatmospheric sciences from a DIAL system that em-ploys such a configuration and yields this kind ofperformance.
B. Ground-Based Scanning Measurement
Here we apply Eq. �35� from Ref. 1 to investigate theperformance of a ground-based scanning system inthe ABL. Again, we are using the parameters sum-marized in Tables 1 and 2 for the simulations. We
consider a scanner that makes a round trip in 60 sand plot the corresponding two-dimensional spatialarea, which must be averaged to reduce the error toless than 5% within one minute.
Figure 11 presents the results of these investiga-tions. To extend the measurement range, operationat high pulse energy is beneficial. This operationmode also reduces the difference in daytime andnighttime performance. A further significant im-provement, somewhat surprisingly, is made possibleby degradation of the laser spectral bandwidth andthe beam divergence and by shortening of the pulseduration. Figure 12 demonstrates a powerful andunique application of a ground-based DIAL system.Such a system will be able to measure, for the firsttime to our knowledge, the water-vapor distributionin a 360° polar-plane indicator or velocity-azimuthdisplay. It will complete a scan in 60 s to at least a15-km range with a spatial resolution of less than 1km. The resolution in the near range will be con-siderably higher. Consequently, in 10 min, ten dif-ferent planes can be scanned and the correspondingthree-dimensional water-vapor distribution can beinvestigated. This is a performance that is requiredfor many applications in atmospheric sciences.There is no other technique available that can comeclose to this performance. Possible applications arethe investigation of horizontal rolls, of the initiationof convection �from mobile platforms�, of fronts, and
Fig. 7. Error analysis of a horizontally pointing system in theABL. The off-line laser pulse energy is 10 mJ, and the repetitionrate is 1 kHz. The detector system parameters are optimized to aFOV of 0.2 mrad and ��F 0.2 nm. The other parameters usedare summarized in Table 2.
Fig. 8. Error analysis of a horizontally pointing system in theABL. The off-line laser pulse energy is 100 mJ, and the repetitionrate is 100 Hz. The detector system parameters are optimized toa FOV of 0.2 mrad and ��F 0.2 nm. The other parameters usedare summarized in Table 2.
of the inflow and outflow from thunderstorms, amongothers.
C. Ultrahigh-Resolution Horizontal Measurement in theAtmospheric Boundary Layer
In this subsection we discuss a configuration that isinteresting for turbulence research. It is achievedby tuning of the laser to a strong absorption line inthe ABL and by use of a short pulse of �10 ns. Table3 summarizes the parameters for this setup, and Fig.13 presents the result. The precision of the mea-surement is limited by Poisson noise, so increasingthe pulse energy while maintaining the same averagepower does not help to increase the precision. Nei-ther the amplification nor the background term playsa role in precision of measurement. A measurementwith resolutions of 1.5 m and 0.1 s can be made in a500-m range while an error of 20% is maintained.This performance is suitable for the resolution ofsmall-scale turbulent processes in the surface layer.It is similar to the precision and range achieved withthe Los Alamos Raman lidar.10
D. Ground-Based Vertical High-Resolution Measurement
Here we present a unique application of a futureDIAL system. The laser transmitter can be tuned toa strong water-vapor absorption line, so measure-ments in the ABL with resolutions of 30 m and 1 s tothe top of the ABL are possible. The additional pa-
rameters listed in Table 4 were used, and we assumea 2-km-deep boundary layer.
Figure 14 demonstrates the potential of such a sys-tem. The pulse energy is set to 10 mJ and it is notvaried, as the error at the top of the ABL is limited byPoisson noise. Focusing on the error bars revealsthat a 10-W average power system will be capable ofmeasuring throughout the ABL with an error of �2%.Depending on the application, this high precisionmay permit a further increase in the range and timeresolution in the mixed layer. This performanceopens a large area of research applications. Profilesof turbulent variables can be investigated up to thefourth order with unprecedented accuracy and reso-lution. Also, entrainment processes can be resolvedwith high resolution. Scanning techniques will al-low for monitoring the two- or even the three-dimensional distribution of eddies in the convectiveABL. Gravity waves, Kelvin–Helmholtz instabili-ties, and exchange processes become visible even inthe stable ABL. Therefore we expect this system tohave a significant effect on ABL research.
E. Extension of Ground-Based Vertical Measurement tothe Upper Troposphere
Reducing the time and height resolution permits theextension of the measurement range to the uppertroposphere during daytime. Figure 15 demon-strates this important potential for climate research.
Fig. 9. Error analysis of a horizontally pointing system in theABL. The off-line laser pulse energy is 1 J, and the repetition rateis 10 Hz. The detector system parameters are optimized to a FOVof 0.2 mrad and ��F 0.2 nm. The other parameters used aresummarized in Table 2.
Fig. 10. Error analysis of an optimized horizontally pointing sys-tem in the ABL. The off-line laser pulse energy is 1 J, and therepetition rate is 10 Hz. The detector system parameters are aFOV of 0.2 mrad and ��F 0.2 nm. Furthermore, �t 20 ns,�L��FT 3, and F� 3. The other parameters used are sum-marized in Table 2.
The parameters summarized in Table 4 were used inthis simulation, except that the profiles shown inFigs. 1, 7, and 8 of Ref. 1 were inserted for the atmo-spheric parameters and that the pulse energy ischanged. The simulation shows that speckle effectsdo not play a role in the measurement error whenthese system parameters and these moderate timeand range resolutions are used. Figure 15 showsthat with a feasible reduction of the backgroundterm, even during daytime, a range of 9 km with anerror of less than 10% can be reached, based on arange resolution of 300 m and a temporal resolutionof 10 min. Consequently, during daytime, this sys-tem has approximately 2–3 times the range of theCloud and Radiation Testbed (Ref. 26 of Part I1) Ra-man lidar, and we believe that it will be the firstdaytime ground-based, upper-tropospheric humidityremote-sensing system.
The final limitation on the measurement range isset by the Poisson term. The error profile comescloser to this limit if the pulse energy is increased, asthe background term is reduced at higher pulse en-ergy. Further reduction of this term can be achievedonly by averaging. For applications in climate re-search, often reductions of the range and the timeresolution to 2 km and 30 min, respectively, are stillacceptable. In this case the range can be increasedto �13 km while an error of 10% is maintained.
The vertical range of a ground-based DIAL system
can be extended if the system is deployed on a moun-tain site, e.g., on Mauna Loa in Hawaii. At this sitethe laser does not need to penetrate the high humid-ity in the boundary layer, so a larger range as well asa higher resolution can be achieved in the upper tro-posphere. Equipping of ground-based reference sta-
Fig. 11. Error analysis of a scanning system in the ABL. Shownis the spatial resolution of the system with a scan speed of 360° in60 s. The resolution is chosen such that the precision is betterthan 5% in each range square. The results are shown for severalsettings of the pulse energy while the same average power ismaintained.
Fig. 12. Error analysis of an optimized scanning system in theABL. Shown is the spatial resolution of the system with a scanspeed of 360° in 60 s. The resolution is chosen such that theprecision is better than 5% in each range square. The bandwidthof the laser pulse is three times Fourier limited, the divergence isthree times diffraction limited, and the pulse duration is 20 ns.
Table 3. Parameters for the Simulation of an Ultrahigh-ResolutionHorizontally Pointing DIAL System in the ABL
tions with DIAL systems for satellite and in situsensor calibration is, for example, planned for theGlobal Water Vapor �GVaP� Project.11,12
Figure 16 demonstrates the performance of a DIALsystem deployed on a 3-km-high mountain site.Such a system has the potential to reach an altitudeof 11 km above the Earth’s surface for resolutions of300 m and 10 min. The measurement range is lim-
ited mainly by Poisson statistics, but a further im-provement of the pulse energy would still help thesystem to come closer to the Poisson limit. Depend-ing on the application, the range can be extended byreduction of the range and time resolution. For ex-ample, if the resolutions are decreased to 2 km and 30min, a range of ground to 16 km above ground can be
Fig. 13. Error analysis of an ultrahigh-resolution ground-basedhorizontal DIAL measurement with resolutions of 1.5 m and 0.1 s.The pulse energy is 10 mJ.
Table 4. Parameters for the Simulation of a Vertically Pointing, High-Resolution DIAL System in the ABL
Variable Value at 940 nm
Laser parameterOff-line pulse energy E0 �mJ� 10Repetition rate L �Hz� 000Pulse duration �t �ns� 20.0Laser spectral bandwidth �L 10.0 �FT
Atmospheric parameterABL height �km� 2Absolute humidity � �g m�3� 8.0 in the ABL; otherwise see Fig. 1 of Ref. 1Water-vapor number density nH2O �cm�3� �2.7 � 1017 in the ABL; otherwise see Fig. 1 of Ref. 1Line strength of absorption line S �cm� 1.05 � 10�23
Line center of absorption line 10,222.775Ground-state energy E �cm�1� 326.625Backscatter coefficient � �m�1 sr�1� 5 � 10�6 in the ABL; otherwise see Figs. 7 and 8 of Ref. 1Atmospheric extinction coefficient � �m�1� 10�4 in the ABL; otherwise, see Figs. 7 and 8 of Ref. 1
Parameters for data analysisRange resolution �R �m� 30Averaging time T �s� 1Differential optical thickness �� 0.024 in the ABL
Fig. 14. Error analysis of a vertically pointing high-resolutionDIAL system in the ABL with a pulse energy of 10 mJ as well asresolutions of 30 m and 1 s. All error profiles correspond to day-time measurements.
reached, so the range is extended to the lower strato-sphere.
F. Airborne Horizontal Measurement in the AtmosphericBoundary Layer
It can be asked how an airborne system performsmeasurements in the ABL. We consider a horizon-tally pointing system, which will be, for example, ofhigh interest to scientists who are investigating theinitiation of convection. Table 5 summarizes the pa-rameters used for the following simulations.
Figures 17 and 18 compare the performance of anairborne system with pulse energies from 100 mJ to1 J. As we recognized above, the speckle term startsto play a role in the system performance. Thereforewe must consider a short pulse with �t 20 ns andwe must degrade the spectral bandwidth and thedivergence. In this case the range is limited mainlyby amplifier and Poisson noise. We learn from Figs.17 and 18 that measurements can be performed up toa range of �12 km with a range resolution of 300 mand a horizontal resolution of approximately 300 mwhile the error is maintained to less than 10% �as-suming that the horizontal speed of the airplane is�150 m s�1�. In the far range, at �10 km the av-eraging time as well as the range resolution needs tobe degraded. However, a range and a horizontalresolution of 600 m will still result in an error of lessthan 10% up to 15 km.
It is obvious that this performance is unique. An
airborne remote sensor as specified here will be ableto resolve structures in the mesoscale water-vaporfield that cannot be seen by other instruments. Forexample, horizontal rolls in the convective ABL aswell as gravity waves can be resolved. Also, data
Fig. 15. Error analysis of the daytime performance of a ground-based vertically pointing DIAL system in the troposphere with apulse energy of 100 mJ, a repetition rate of 100 Hz, and resolutionsof 300 m and 10 min as well as of 2000 m and 30 min.
Fig. 16. Error analysis of a vertically pointing DIAL system de-ployed on a 3-km-high mountain site with a pulse energy of 100mJ, a repetition rate of 100 Hz, and resolutions of 300 m and 10min as well as of 2000 m and 30 min. Again, the data show thedaytime performance.
Table 5. Parameters for the Simulation of an Airborne HorizontallyPointing DIAL System in the ABL
Variable Value at 940 nm
Laser parametersOff-line pulse energy E0 �J� See Figs. 17 and 18Repetition rate L �Hz� See Figs. 17 and 18Pulse duration �t �ns� 20.0Laser spectral bandwidth �L 10.0 �FT
can be provided that can be used for data assimilationand comparisons with mesoscale models.
G. Airborne Downward-Pointing Measurements
An airborne downward-pointing DIAL system is ofbroad interest to the atmospheric sciences commu-nity. It is a common setup for current DIAL sys-tems. This configuration is particularly suitable forDIAL measurements, as the backscatter coefficientas well as the humidity increase with range. There-fore the dynamic range of the backscatter signals isreduced and the degradation of the signal in the farrange is partially compensated for by an increase inthe backscatter signal as well as in the differentialabsorption. Furthermore, we can maintain highresolution over a long range by switching in flight tothe center and the wing of a water-vapor absorptionline. Although it does not reach the resolution of insitu sensors, this configuration has the advantage ofproviding range-resolved water-vapor profiles belowthe airplane so the two-dimensional turbulent andmesoscale structure of the water-vapor field can bestudied. This setup can be categorized in three con-figurations:
downward pointing from an altitude of 3–4 km forhigh-resolution ABL studies of such properties as en-trainment and turbulence,
downward pointing from the upper troposphere forthe investigation of the mesoscale troposphericwater-vapor field, and
downward pointing from the lower stratosphere forthe investigation of exchange processes between thestratosphere and the troposphere, including mea-surements of tropospheric water vapor down to thesurface.
The last-named configuration is of particular inter-est for the High-performance Instrumented AirbornePlatform for Environmental Research �HIAPER�,which will be acquired at the National Center forAtmospheric Research’s Atmospheric TechnologyDivision �see www.atd.ucar.edu�dir_off�hiaper�in-dex.html�. In all the following plots the dependenceof the error on the range from the aircraft is shownsuch that zero range is the position of the aircraft.
1. Downward Pointing from 4 km for ABL StudiesWe start with an analysis of high-resolution mea-surements in a 2-km-deep convective ABL. Theflight level of the airplane is typically chosen to be3–4 km. An important feature of this situation isthat turbulence measurements can be made with sig-nificantly lower sampling errors than from theground.
The parameters for these simulations are summa-rized in Table 6, and Fig. 19 presents the results. Asthe error is limited mainly by Poisson statistics, it is
Fig. 17. Error analysis of an airborne horizontally pointing sys-tem in the ABL. The range resolution is 300 m and the averagingtime is 2 s. Other system parameters are summarized in Table 5.All error profiles correspond to daytime measurements. The av-erage laser power is low.
Fig. 18. Error analysis of an airborne horizontally pointing sys-tem in the ABL. The range resolution is 600 m and the averagingtime is 4 s. Other system parameters are summarized in Table 5.All the error profiles correspond to daytime measurements. Theaverage laser power is low.
sufficient to show the performance when a 10 mJDIAL system only is used. Over the entire ABL theerror is less than 6% for resolutions of 60 m and 0.5 s.Figure 19 presents values for a future DIAL systemthat will enable us to resolve the two-dimensionalstructure of ABL processes with small sampling er-rors and unprecedented resolution.
2. Downward Pointing from 12 kmWe consider a system that can be switched in wave-length between an absorption line with a linestrength of the order of 10�24 cm in the far-range anda 20-times-stronger water-vapor absorption line inthe near range. Figure 20 demonstrates that such asystem can cover the entire range to the ground withresolutions of 300 m and 2 s while an error of lessthan 5% is maintained. Operation at higher pulseenergy reduces the background and amplificationterms slightly; the performance is limited mainly byPoisson statistics.
3. Downward Pointing from 18 kmDeploying a high-altitude aircraft such as HIAPERflying at a height of 18 km with a downward-pointingDIAL system will enable us to observe the two-dimensional water-vapor field from aircraft leveldown to the ground �Fig. 21�. Switching amongthree wavelengths with absorption line-strength ra-tios of 200:10:1 is required. To increase the preci-sion, operation with pulse energies higher than 10 mJis beneficial, particularly in the far range. For ex-ample, an increase of the pulse energy to 100 mJwould permit the error to get close to the limit set byPoisson noise.
At 10 mJ and using resolutions of 300 m and 3 s,
such a system will be able to measure water-vaporprofiles over the entire range with an error of �5%.It is clear that such a DIAL system will dramaticallyincrease our knowledge of the role of water vapor inmesoscale processes as well as in processes of ex-change between the stratosphere and the tropo-sphere.
H. Airborne Horizontal Measurement in the MiddleTroposphere and the Lower Stratosphere
Another interesting configuration can be a horizon-tally pointing system in the free troposphere. Table7 and Fig. 22 and 23 show the parameters chosen forthe simulation as well as the results. Using a pulseenergy of 100 mJ, one can reach a range of �12 km ata height of 6 km for resolutions of 600 m and 4 s. Afurther extension of range will be achieved at higherpulse energy. The question as to whether and whenthis configuration is better suited for the investiga-tion of atmospheric processes than is a verticallypointing system requires further discussion amongatmospheric scientists.
The same discussion is necessary for a horizontallypointing system in the lower stratosphere. Here arange of �11 km can be reached with resolutions of600 m and 4 s.
4. Summary
In this study we have illustrated the expected per-formance of an advanced water-vapor DIAL system
Fig. 19. Error analysis of an airborne downward-looking DIALsystem at a height of 4 km with a pulse energy of 10 mJ, a repe-tition rate of 1 kHz, and range and time resolutions of 60 m and0.5 s, respectively, per on-line frequency.
Table 6. Parameters for the Simulation of an Airborne DownwardPointing DIAL System in the ABL
VariableValue at940 nm
Laser parametersOff-line pulse energy E0 �J� See Figs. 19–21Repetition rate L �Hz� See Figs. 19–21Pulse duration �t �ns� 20.0Laser spectral bandwidth �L 10.0 �FT
Atmospheric parametersAbsolute humidity � �g m�3� See Table 4Water-vapor number density nH2O
�cm�3�See Table 4
Line strength of absorption line S �cm� 10�24
Line center of absorption line 10,804.158Ground-state energy E �cm�1� 275.5Backscatter coefficient � �m�1 sr�1� See Table 4Atmospheric extinction coefficient �
�m�1�See Table 4
Parameters for data analysisRange resolution �R �m� 60Averaging time T �s� 0.5Differential optical thickness �� 0.03–0.06
by using direct detection and operation near 940 nm.We considered a 10-W average power DIAL systemoperating at �940 nm equipped with a 1-m telescopeon the ground and a 0.4-m telescope on an airplane.This system is currently under development at Ho-henheim University, Germany, and the National
Fig. 20. Error analysis of an airborne downward-looking DIALsystem at a height of 12 km with a pulse energy of 10 mJ, arepetition rate of 1 kHz, and range and time resolutions of 300 mand 2 s, respectively, per on-line frequency.
Fig. 21. Error analysis of an airborne downward-looking DIALsystem at a height of 18 km with a pulse energy of 10 mJ, arepetition rate of 1 kHz, and range and time resolutions of 300 mand 3 s, respectively, per on-line frequency.
Fig. 22. Error analysis of a 6-km-high horizontally pointing DIALsystem with a pulse energy of 100 mJ, a repetition rate of 100 Hz,and range and time resolutions of 600 m and 4 s, respectively.
Table 7. Parameters for the Simulation of an Airborne HorizontallyPointing DIAL System in the Middle Troposphere and Lower
Stratosphere
Variable Value at 940 nm
Laser parametersOffline pulse energy E0 �J� See Figs. 22 and 23Repetition rate L �Hz� See Figs. 22 and 23Pulse duration �t �ns� 20.0Laser spectral bandwidth �L 10.0 �FT
Atmospheric parametersAbsolute humidity � �g m�3� To be determinedWater vapor number density nH2O �cm�3� To be determinedLine strength of absorption line S �cm� To be determinedLine center of absorption line To be determinedGround-state energy E �cm�1� To be determinedBackscatter coefficient � �m�1 sr�1� 2.5 � 10�7
Center for Atmospheric Research in the UnitedStates.
We proved the extraordinary potential of the DIALtechnique by using what is to our knowledge the mostextensive set of simulations that have been per-formed. The dependence of the simulation precisionon time as well as on range resolution and rangeincludes all kinds of known error, such as errors thatare due to Poisson statistics, daylight background,amplifier noise, and speckles. In particular, wehave not seen consideration of speckle errors indirect-detection DIAL systems in the literature, andwe show that these errors can be important whencertain configurations are used. Also, digitizationnoise has been discussed, and it has been assumedthat one can neglect this error by applying an ad-vanced detector system with high dynamic range aswell as optical or electronic dynamic range reductionor both.
It has been shown that high-resolution measure-ments are possible under a variety of atmosphericconditions. Starting with the surface layer, horizon-tal water-vapor profiles can be determined withultrahigh resolutions, of the order of 0.1 s and 1.5 m,up to a range of 500 m. Over this entire range, therelative error in the water-vapor profile will have avalue of �20%, which can be maintained for totallydifferent amounts of water vapor. By reducing therange and time resolutions we can considerably ex-tend the range of a horizontal water-vapor measure-ment in the boundary layer. Using a rangeresolution of 300 m and a time resolution of merely
10 s, we can measure water-vapor profiles up to arange of approximately 18 km while maintaining anerror of 10%. For the first time to our knowledge,the performance of a scanning ground-based DIALsystem has been investigated. If an error of 5% isprescribed, a dependence of the width of a square onrange has been estimated in which enough photonsare collected to maintain this high precision. It wasshown that a ground-based polar plane indicatorscanning system will have a square width resolutionof �1 km up to a 16-km range after a 60-s round trip.This performance cannot be reached by any otherremote-sensing technique and will provide uniquedata sets for research in atmospheric sciences. Inparticular, this outstanding time and range resolu-tion will permit the investigation of three-dimensional water-vapor fields with sufficient timeresolution.
We have also discussed the performance of aground-based vertically pointing system. With astrong water-vapor absorption line, investigations ofboundary layer processes are possible with resolu-tions of 30 m and 1 s while an error of 2% is main-tained throughout a 2-km-deep boundary layer.These data will provide important insights into tur-bulent as well as entrainment processes. Reducingthe resolutions to 300 m and 10 min, even during thedaytime, will permit a ground-based system to per-form measurements up to a height of 9 km whilemaintaining an error of 5%. Reducing the verticaland the time resolutions further to 2 km and 30 min,respectively, values that are often acceptable for cli-mate research, will increase the range to approxi-mately 12 km with the same precision. One canfurther extend the range of a ground-based verticallypointing system by deploying the DIAL system on amountain site. In this case the DIAL signal does notneed to penetrate the high absolute humidity in theboundary layer. For a mountain site height of 3 km,the system will perform measurements with resolu-tions of 300 m and 10 min up to 11 km above groundand up to 15 km with a resolution of 2 km and 30 min.Again, in both cases an error of 5% can be main-tained.
Detailed studies have dealt with the performanceof a DIAL system on an airborne platform. For hor-izontal pointing in the boundary layer and using res-olutions of 300 m and 2 s, we expect a range of 12 kmwith an error of 10%. Reducing the resolutions to600 m and 4 s, respectively, can extend the range to15 km while maintaining this high precision. Thesedata will provide unique insight into atmosphericprocesses such as the initiation of convection.
The downward-looking mode from an airplane isparticularly well suited for DIAL measurements, asthe dynamic range of the backscatter signal de-creases and the backscatter coefficient as well as thedifferential absorption increases in the far range. Ina downward-looking mode from 4 km, high-resolutionstudies of the atmospheric boundary layer can beperformed. Using a range resolution of 60 m and atime resolution of 0.5 s, respectively, produces an
Fig. 23. Error analysis of a 15-km-high horizontally pointingDIAL system with a pulse energy of 100 mJ, a repetition rate of 100Hz, and range and time resolutions of 600 m and 4 s, respectively.
error of less than 4% over most of the range. Flyingat an altitude of 12 km and switching between twoabsorption lines with a line strength ratio of 20:1used alternately for the near range and for the farrange will permit an entire profile to be measured tothe ground. An error of less than 5% can be main-tained over the entire range by use of a range reso-lution of 300 m and a time resolution of 2 s perabsorption measurement. We also studied the per-formance of a downward-pointing system flying on ahigh-altitude aircraft at 18 km. Here, switchingamong three absorption lines with a line-strengthratio of approximately 200:10:1 still permits mea-surement of a water-vapor profile to the ground. Byuse of a range resolution of 300 m and a time reso-lution of 3 s per absorption measurement, an error of6% over the entire range can be maintained.
Also, horizontally pointing modes have been con-sidered for aircraft height levels of 6 and 15 km.Typically, a range of 12 km can be achieved in bothcases for a range resolution of 600 m and a timeresolution of 4 s while an error of 10% is maintained.This performance cannot be achieved by another re-mote sensing system. Therefore such a system canbe considered the key component of new high-altituderesearch aircraft such as the HIAPER.
We expect that the water-vapor sensor described inthis study will dramatically enhance our knowledgeof compelling issues in atmospheric sciences, such astrends in atmospheric water vapor, quantitative pre-cipitation forecasting, initiation of convection, andboundary layer research. Significant gaps in ourunderstanding of the role of water vapor in atmo-spheric processes no longer have to be accepted asinevitable.
This study was supported by the U.S. Weather Re-search Program and was performed at the Atmo-spheric Technology Division of the National Centerfor Atmospheric Research. B. Rye and R. M. Hard-esty of the National Oceanic and AtmosphericAdministration’s Environmental Technology Labora-tory provided valuable comments on the error anal-ysis and the simulations. D. N. Whiteman of theGoddard Space Flight Center was so kind as to pro-vide data on the daylight background.
References1. V. Wulfmeyer and C. Walther, “The future potential of ground-
based and airborne water vapor differential absorption lidar.I. Overview and theory,” Appl. Opt. 40, 5304–5320 �2001�.
2. S. Ismail and E. V. Browell, “Airborne and spaceborne lidarmeasurements of water vapor profiles: a sensitivity analy-sis,” Appl. Opt. 28, 3603–3614 �1989�.
3. L. Elterman, “UV, visible, and IR attenuation for altitudes to50 km,” Environmental Research Papers, AFCRL-68-0153 285�U.S. Air Force Cambridge Research Laboratory, Bedford,Mass., 1968�.
4. S. Erhard, A. Giesen, M. Karszewski, T. Rupp, and C. Stewen,“Novel pump design of Yb:YAG thin disk laser for operation atroom temperature with high efficiency,” in Advanced Solid-State Lasers, Vol. 26 of OSA Trends in Optics and PhotonicsSeries �Optical Society of America, Washington, D.C., 1999�, p.26.
5. G. Ehret, H. H. Klingenberg, U. Hefter, A. Assion, A. Fix, G.Proberaj, S. Berger, S. Geiger, and Q. Lu, “High peak andaverage power all solid-state laser systems for airborne LIDARapplications,” LaserOpto 32, 29–37 �2000�.
6. A. K. Sridharan, T. Rutherford, W. M. Tulloch, and R. L. Byer,“A proposed 1.55 m solid state laser system for remote windsensing,” in 10th Conference on Coherent Laser Radar �Uni-versity Space Research Association, Huntsville, Ala., 1999�,pp. 241–277.
7. T. F. Refaat, W. S. Luck, Jr., and R. J. DeYoung, “Design ofadvanced atmospheric water vapor differential absorption li-dar �DIAL� detection system,” NASA Tech. Publ. NASA�TP-1999-209348 �NASA Langley Research Center, Hampton, Va.,1999�.
8. E. E. Remsberg and L. L. Gordley, “Analysis of differentialabsorption lidar from the Space Shuttle,” Appl. Opt. 17, 624–630 �1978�.
9. V. Wulfmeyer, “DIAL—Messungen von vertikalenWasserdampfverteilungen—ein Lasersystem fur Wasser-dampf- und Temperaturmessungen in der Troposphare,” Ph.D.dissertation �Max-Planck-Institut fur Meteorologie, Hamburg,Germany, 1995�.
10. W. E. Eichinger, D. I. Cooper, F. L. Archuletta, D. Hof, D. B.Holtkamp, R. R. Karl, Jr., C. R. Quick, and J. Tiee, “Develop-ment of a scanning, solar-blind water Raman lidar,” Appl. Opt.33, 3923–3932 �1994�.
11. World Meteorological Organization, The WCRP�GEWEXGlobal Water Vapor Project �GVaP�: Science Plan, Publ. 27�International GEWEX Project Office, 1010 Wayne Ave., SilverSpring, Md. 20910, 1999�.
12. World Meteorological Organization, The WCRP�GEWEXGlobal Water Vapor Project �GVaP�: Implementation Plan,Publ. �International GEWEX Project Office, 1010 Wayne Ave.,Silver Spring, Md. 20910, 1999�.
