Semidiurnal and diurnal temperature tides (30-55 km): Climatology and effect on UARS-LIDAR data comparisons

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 101, NO. D6, PAGES 10,299-10,310, APRIL 30, 1996

Semidiurnal and diurnal temperature tides (30-55 km): Climatology and effect on UARS-LIDAR data comparisons PhilippeKeckhut, • M.E. Gelman, 2J.D.Wild,3F. Tissot, • A.J.Miller,2A. Hauchecorne, • M.-L. Chanin, • E.F.Fishbein, 4J.Gille,sJ.M.Russell III, 6andF.W.Taylor? Abstract. Very good agreementis shownfor diurnal and semidiurnaltemperaturevariations calculated from lidar measurements in southern France and from data of the microwave

limb

sounderof the Upper AtmosphereResearchSatellite(UARS). Tides inducetemperaturedeviations observedin southernFranceto be as large as+3 K, with a maximumat the stratopause.The amplitudesandphasesof the semidiurnalvariationchangesignificantlywith seasonand location. Seasonalchangesup to 2 K havebeenclearlyidentifiedfor the diurnalcomponent.An analytic model of the diurnal component,basedon sinusoidalfunctions,fits the data well, but is less successful for the semidiurnalcomponent.Substantialagreementis alsoreportedfor the diurnal componentbetweenthe resultsof our analyticalmodel and the publishedresultsof a twodimensionalglobal-scalewave model.In contrast,the semidiurnalcomponentis in total disagreement with numericalsimulationsthatreportvery small amplitudes,as comparedwith the observations reportedhere.The confidencein detectingbiasin datacomparisons is improvedif datausedare limited to periodsfrom April to Septemberandif time-of-dayadjustments are applied.Comparisonbetweenlidar andnearlycoincidentUARS temperaturemeasurements have revealed,systematically, for the4 experiments aboardUARS, a significantresidualmean differenceof up to 3 K around35-43 km. A comparison usingsimultaneous measurements suggests that thebias is associated with the variability of migratingtidesand/orthe presenceof nonmigratingtidesratherthaninstrumentalcharacteristics.

1. Introduction

and harmonics associated with local solar time. Some artificial

Satellite measurements allow a unique opportunity to provide global temperature fields. However, satellite radiometers need systematic validations and periodic calibrationsby ground references. Accurate validations require a selection

of

simultaneous

and

collocated

data

to eliminate

the effects of the spatial and temporal variability of the atmosphere. However, to increase the statistical sample, comparisonsare usually performed for some quite large spatial area of several hundred kilometers. Time separationsusually correspondto a significant fraction of the day. Owing to this nonperfect spatiotemporal coincidence, large variability, mostly due to planetary and gravity wave propagation, is attachedto the data comparisons,limiting the detection of true instrumental

difference

characteristics.

Moreover,

atmospherictides, due primarily to 0 3 and H20 insolation absorption, induce some oscillations with periods of 24 hours

biases can be expected when two nonsimultaneousdata sets obtained systematically at different local solar times are compared.

However, until recently, observations

of tides were

in the stratosphere, few

available.

Tidal

studies in this

rangehavethusfar beenlimitedalmostexclusively to a series of rockets. Temperature measurements fromrocketshadsparse coverages of the full day andalsowere subjectto errorsinduced by radiation from exposedcomponentsof the rocketsonde

instrument [Fingerand Wolf,1967;Hoxit andHenry,1973]. Wind measurements havebeenintensivelyinvestigated with good success[Reed et al., 1969;Reed 1972]. The LIMS

instrument (limb infraredmonitorof the stratosphere) aboard theNimbus7 spacecraft providedgoodestimates of day-night changes but wasnot ableto give a completeseparation of the tidal phasesand amplitudes[Hitchmanand Leovy, 1985]. Tides in the stratosphere have been intensivelystudied through numericalsimulations.The comparisonbetween

theory and rocket wind measurementsshows that Lindzen's i Service d'A•ronomie duCentre National delaRecherche Scientifique, V•rrieres-Le-Buisson,France.

[1967] simulations have successfully predictedthe major

2 ClimateAnalysisCenter,NationalMeteorological Center,National features of the tidal behaviorin the 30 to 60-kmregionwith WeatherService,WashingtonD.C. still some discrepancies. 3Research andDataSystems Corporation, Greenbelt, Maryland. The Upper Atmospheric Research Satellite (UARS), 4JetPropulsion Laboratory, CaliforniaInstitute of Technology, Pasadena. launched on September 12, 1991, which has a non-sun5NationalCenterforAtmospheric Research, Boulder,Colorado. 6 AtmosphericSciencesDivision,NASA LangleyResearchCenter, synchronous orbit, covers, for a single location, a full 24Hampton,Virginia. hour cycle in about 1 month. Temperature measurements ?Atmospheric Oceanic andPlanetary Physics, OxfordUniversity, Oxford, obtained from the limb scanning instruments aboard UARS England. allow derivationof a mean diurnal change.Dudhia et al. [1993] have done a preliminary study based on the improved Copyright1996 by AmericanGeophysical Union. stratosphericand mesosphericsounder(ISAMS) data and have shown some consistencybetween satellite and lidar analyses Paper number96JD00344 0148- 0227/96/96JD- 00344505.00 [ Gille et al., 1991]. Another study [Wild et al., 1995] 10,299

10,300

KECKHUTET AL.:TEMPERATURETIDESAND UARS-LIDARDATA COMPARISONS

compareslidar data at differentlocal solartimes.US National Meteorological Center (NMC) stratospheric temperature analyses [Finger et al., 1993] representativeof a quasiconstant solar local hour over a single geographic location can be considered a good estimate of daily changes. These daily values have been subtracted to the microwave limb sounder(MLS) temperaturedata to eliminate daily fluctuations induced by planetary waves. Mean differencesbetweenlidars were shown to be reduced for some stations, proving the

importanceof tidesin data comparisons. However,UARS is an experimental satellite with limited duration. For at least a decade, no extended program is planned to ensure the continuity of such measurements and the possibility of monitoring continuouslytides in the upper stratosphereon a global scale.

The goal of this study is to estimatehow the variability in data comparisonsattachedto some geophysicalprocesscan be reduced

and thus instrumental

bias better described.

local hours, separated from each other by approximately 12 hours. Both measurements are similarly influenced by a 12hour

wave.

In

contrast,

the

derivation

of

the diurnal

component is not compromised, as the two measurements12 hours apart cover the opposite phase of a 24-hour wave. Trends have to be removed from the data to prevent aliasing effects by longer-term changes. U.S. National Meteorological Center (NMC) stratospheric temperature analyses [Finger et al., 1993] representative of a quasi-constant solar local hour over a single geographic location, were used as a good estimate of daily changes. Only five pressure levels are available from 10 to 0.4 hPa (31-55 km). The estimate of the daily changesfor the nine pressurelevels of UARS data on the same pressurerange requires interpolation of the NMC data. The procedureof removing temperature at a given time of day from the MLS data series decreases, by a factor of 2, the residual noise of the tidal components and improves the estimateof the semidiurnalamplitudesby a factor of 2-5.

In this

paper,3 yearsof MLS measurements (version3, level 3AT) are analyzed to establish a climatology of temperature tidal magnitudesand phases.First, we discussthe tidal retrieval method and compare the tidal characteristicsobtained from

2.2

Lidar

Data

On lidars in France, a semiautomaticmode to trigger a shutdown (in case of sudden cloud cover, rain,

laser

MLS data with the analyses of the long continuous lidar measurements obtained over France (44øN) in winter. The

deficiencies or fire) hasbeensetup to make easierregularlong continuousmeasurementsduring the night. Accurate hourly

spatiotemporalvariability of the diurnal and semidiurnaltides is investigated to point out the relevant parameters and to estimate the confidence of the mean climatology. Then, a parametricmodel over France (where the longest temperature database obtained with Rayleigh lidar exists) is established from tidal characteristic profiles deduced from MLS data analyses. This model is compared with recent published numerical simulations. Finally, we compare lidar data with temperaturemeasurementsfrom the four UARS temperature sounders,including time-of-day adjustmentsbasedon our tidal model. The purpose of such comparisons is to evaluate the improvementsgiven when tidal effects are consideredand to estimatethe quality of the temperaturemeasurementsavailable through UARS experiments.

profiles are available to study tidal effects between 30 and 80 km [Gille et al., 1991; F. Tissot et al., Diurnal and semidiurnal

2. Tidal Analysis Methods 2.1.

MLS

Data

The UARS orbit precessesrelative to the Earth-Sun axis by approximately 5 ø longitude per day, which allows limb instrumentsto cover one full 24-hour cycle in a 36-day period. Tidal effects should ideally be studied using a period shorter than 10 days so as not to be contaminated by seasonal migration of tidal phase and magnitude [Forbes, 1985]. The 36-day period used here would probably smooth amplitudes and phases, but we are not able to prevent such effects with UARS

data. The method

to extract

tidal effects from the MLS

measurementsis basically the same as the one described by Dudhia et al. [1993]. It consists of modeling time series independently,for each pressuresurface, by a least squaresfit of periodic functions of 12 and 24 hours. Uncertainties on the fitted parameters are calculated from the covariance matrix. Phases and amplitudes have been reported here with _+2 standard deviation. As we have considered a single location over France here, the local tidal characteristics are computed by selectingdata within a 5 ø latitude-longitudebox. Each day, measurementsabove the France area correspondto two distinct

tides in the middle atmospheredetectedby lidar, 1, Analyses method, submittedto Journal of GeophysicalResearch, 1996 (hereinafter referred to as F. Tissot et al., submitted manuscript,1996)]. The data are still limited to nighttime periodsdue to the luminance of the sky background.Since lidar

data are limited

in summer

to less than 6 hours of

operation, this does not allow an accurate analysis of tidal effects in summer.However, during winter, more than half a diurnal cycle (14 hours) can be continuouslycovered. The method to extract tidal changes from lidar data is similar to methodsusedwith mesosphericradar data. It is not possible, from a single night of measurement,to identify clearly the tidal signature because of the presence of other sources of atmosphericvariability (e.g., gravity and planetary waves). Moreover, the presenceof gravity waves of long period (8-10 hours)can be a sourceof pollution,mostly for the semidiurnal component.Changesinducedby planetarywaves of 10 K/day are frequent during winter [Hauchecorne et al., 1991], and short-termvariability of several Kelvin due to gravity waves occurscontinually[Wilson et al., 1991]. Long-term changes are estimatedby interpolationbetweenthe consecutivenights of measurement.The averagingof several nights of hourly measurementsreducesrandom atmosphericvariability. A least squaresfitting method, of a combination of 12- and 24-hour sine and cosine waves, and a constant value are fit to the

hourly mean series. The major limitation of such measurements consists of the separation of the tidal components on noisy data partially covering the diurnal cycle. Numericaltestswith an artificial noisy signal [Gille et a/., 1991] suggestthat separatefits for diurnaland semidiurnal components,can lead to better estimates,dependingon both the signal-to-noiseratio and the fraction of the diurnal cycle covered. The determination of the best fit using different combination of 12-hour, 24-hour, and constant term is decided

by a chi-squaretest [F. Tissotet al., submittedmanuscript, 1996].

KECKHL• ET AL.:TEMPERATURE TIDESAND UARS-LIDARDATA COMPARISONS

Lidat analyses did not include every day, as opposed to the MLS analyses,which had daily coverage.Reflecting effects by

3. Comparisonsof the Tidal Observations A first comparisonbetween tides observedby the ISAMS experiment(UARS) andlidat analysiswas reportedby Dudhia

clouds, estimated to be of the order of 20% neat 30 km and

et al. [1993], showing reasonable agreement between both analyses. However, data selected in that study were not obtainedfor the sameyear, althoughthey representedthe same season.So the observed discrepanciescould be explained by the interannual and horizontal variability of the tidal characteristics.

The period selected in this previously mentioned study corresponds to one 36-day precession period of UARS (December 4, 1991 to January 14, 1992). For this winter period, many long nights of lidar measurements were available, allowing a good opportunity to observe tidal changeson a large fraction of the diurnal cycle (14 hours) and to compare them with satellite analyses.Lidat analyses use a data set composedof the averageof the measurementsfrom the lidat at the Observatoirede Haute Provence ((OHP) 44øN, 6øE)

negligible at 45 km [Grove, 1982], may also cause problems, as lidar measurements are only obtained during noncloudy nights. Considering the different characteristics of both methods and possible spatial and daily inhomogeneities,this comparison shows good agreement and reveals the high potential of both methods to detect tidal changesin the upper stratosphere.Moreover this study suggeststhat the chi-square test

44øN, løW), separatedby 550 kin. Increasing the number of nights improves the statistic in reducing the noise. During the period, MLS measurementscovereda full diurnal cycle. Mean tidal changesover the lidat siteswere computedfrom the MLS data and compared with lidar analyses. The

comparisonof magnitudesand phasesof the 12- and 24-hour componentsshows a reasonableagreementunder the error bars, as illustrated by the example reported in Figure 1. However, some discrepanciesin the comparisonsexist at some levels for the semidiurnal component. The largest differenceis observedaround 54 km and may be due to aliasing effect on MLS data analyses,as the 0.4 hPa (55 km) NMC is not

differences

as reliable

as the

other

around 33 and 36 km,

levels.

for

However,

the diurnal

is

an

efficient

criterion

to

detect

the

and

semidiurnal components,respectively, remain unexplained.

Solutions of Laplace's tidal equations for the latitudinal structure of the tide consist of some propagating modes that transmit energy away from the region of excitation and trapped modes in which energy decays exponentially away from the region. The transition, from trapped modes behavior (dominating at high latitude in which 0 3 heating is the main

driver) to propagatingmodes (prominent at low latitudes and mostly generated by tropospheric water vapor heating), is evident in both theory and wind observations[Reed et aI., 1969]. The superposition of both modes can be noted, by rapid changeswith altitude, in the phase profiles given by numerical simulations [Forbes, 1982]. However, in the region between 40 and 60 km, strong ozone heating excites mostly

the trapped modes. The main systematicchange of vertical heating structuresis due to seasonalchangesof insolation. This effect is expected to induce seasonal variations in the tidal amplitude. Temperature, wind, ozone, and water vapor

6

•'5

1

0

30

35

45

50

55

30

35

b

Altitude(Km)

......'/ ; ............. '....... :

12 '•

40

40

! ,:.....

/

6

45

50

55

Altitude (Km)

'.....t ....

,.i!"iiii! ........ :.iiiiii ....

9

0

0

30

C

35

40

45

Altitude (Km)

50

55

tidal

4. Tidal Variability

7

6

a

true

characteristics with lidars. In the next sections, only levels where tidal components, extracted from lidar analyses, are unambiguously separated according to this test will be presented.

and from another lidat at the Centre d'Essais des Landes ((CEL)

level

10,301

30

d

35

40

45

50

55

Altitude (Kin)

Figure 1. Tidal characteristics for (a) diurnalamplitude,(b) semidiurnalamplitude,(c) diurnalphase, and (d) semidiurnalphaseover the southof Franceas a functionof the altitude, given by lidar data analyses(meanand standarddeviationare represented by solidand dashedlines,respectively) and MLS data (meanand standarddeviationare representedby the solid circle and vertical line,respectively).

10,302

KECKHUT ET AL.: TEMPERATURE TIDES AND UARS-LIDAR DATA COMPARISONS

density present some seasonal cycles that contribute to seasonalchangesof the tidal effects [Vial and Forbes, 1989; Hagan et al., 1995]. Rapid changes with latitude of tidal characteristics are expected by both theory and observations. In the equatorial regions, amplitudes of diurnal tides are expected to be

•. 24

maximum and to increase with altitudes, from 1 to 5 K, between 30 to 60 km [Forbes, 1982]. Also, a maximum around

•• 6 o

:::::::::

._

•'

265

the stratopause of 3 K due to forced modes by ozone is expected for all latitudes except around the tropics, where a minimum is found. The recent analysisof ISAMS data [Dudhia

'_•

et al., 1993] confirmed these results. Semidiurnal modes are

• 245

expectedto be small in the stratosphere,increasingabove the stratopause. However, Dudhia et al. [1993] observed amplitudesas large as 5 K in the latitudeband from 40ø to 70ø. Longitudinal variability has been assumedto be negligible, and global satellite data are averaged over a latitude band to increase the accuracy. Grove [1982] investigated ozone longitudinal dependence, and little evidence was found.

0

m

260

'"" 255

•250 •

240

30

60

90

120

150

180

210

240

270

300

330

360

Longitude

Figure 2. Diurnal amplitude (top) and phases(middle) at 1 hPa (48 km) at 45øN as a function of longitude. Standard deviationsare indicatedwith dashedlines.Mean temperatureat the samelevel (bottom) is also represented.

However, at that time, available ozone measurements were not

adequate for firm conclusions. Further, he found that longitudinal contributions arising from the lower reflecting layer introduce negligible effects. Other phenomenathat have been suggestedas potential contributors to variability of tides are topography, land-sea heating differences, or spatial structuresinducedby the dynamic of temperature,wind, ozone, or water vapor. These effects, which can cause longitudinal variations in the diurnal heating constituents, may generate higher-order tidal components (referred to as nonmigrating tides) traveling with a phase speed different than that of Earth's rotation [Wallace and Hartranfi, 1969]. The efficacy of a tidal model results in its ability to reproduce the observed tidal variability responsible for the largest energy. In the following sections, we propose to illustrate with both MLS and lidar data analyses such tidal variability on both longitudinal and temporal scales.

(+3 hours).Thesecalculationswere more randomlydistributed with longitudeand were seeminglynot related to any largescalepatternof the mean atmosphere.Local tidal analysesare then preferred here rather than a zonal mean, to consider

possiblesystematictidal patterns such as nonmigratingtides or other possible local effects.

4.2.

Temporal

Tidal

Variability

Tidal characteristics are expectedto changeon a daily and interannual basis. Unfortunately, the techniques presented here are not able to give us accessto the daily variability. Systematic seasonalchanges of the tidal characteristicsover the southof Francecan be notedfrom the MLS data(Figure3). Amplitudesof the diurnal componentchangesmoothly,with a maximum

in summer

and a minimum

in winter.

Time

of

maximum of the diurnal tide is nearly constantat 0.7 and 0.5 4.1. Spatial Tidal Variability hPa (52-55 km), varying from 1 to 2.5 hoursat 1-2.2 hPa (42Tidal characteristicswith longitude at 1 hPa (48 km), are 50 km), and varying more sharplybelow 2.2 hPa (42 km). A reported in Figure 2 for the December 1991 to January 1992 smooth pattern of seasonalchanges is not apparent for the period discussedin the earlier section. MLS data over 5øx30ø amplitudeand phaseof the semidiurnaltide (Figure 4), due to latitude-longitude boxes, shifting by 30ø along the latitude large interannual variability. The only quasi-systematic band centeredon the 45øN latitude line, have been analyzed patternswe can note here are a minimum of amplitudeobserved using the method describedpreviously. in summer (July-August) and the large amplitudes present Considering the good agreement with lidar reported in during equinox periods. In December and February, mean section 3, the small standard deviation and the smooth shifts amplitudeand phaseof the semidiurnalcomponentare more of the tidal characteristics along the latitude band, the difficult to estimatedue to the large variability. variability shown in Figure 2 is probably real. It could be Thelargetidal variabilityobserved with MLS dataanalyses induced by large mean winds or horizontal temperature duringwinter, may be explainedby the large atmospheric gradients due to the presence of planetary waves that had fluctuations, observed frequentlyat midlatitudes andcausedby induceda large horizontalanomaly,as illustratedby the mean the propagationof planetarywaves [Hauchecorneand Chanin, 1 hPa temperature field (Figure 2). Phases of the diurnal 1982]. The structureand amplitudes of the migratingtides,as componentpresent similar vertical profiles at all longitudes, predicted by classical tidal theory, can be altered by with a systematic abrupt phase shift (from 6 to 18 hours) background zonalwindsandmeridionaltemperature gradients. around3 hPa (38 km), previouslymentioned.The amplitude Also, nonmigratingtides can be suspected. Thesetidesexhibit profile reveals a maximum around 1-hPa (48 km), with a upward phasepropagationor a complete absenceof vertical 1991], which are variability of +0.5 K. However, very different signaturesare phase propagation [Lieberman, observed between longitudes 240ø and 330ø, with a smooth superimposed on trappedmigratingmodes.However,if large maximum around1 hPa (48 km) and no phaseshifts around3 amplitudesof thesetides are observedin the troposphere, hPa (38 km). Semidiurnal characteristics as a function of numerical simulations[Lieberman and Leovy, 1995] have longitudepresentedlarger variability for both the amplitudes indicated that their amplitude are attenuatedin the middle (from _+0.5to +3 K, dependingon the altitude)and the phases atmosphere.

KECKHUTET AL.: TEMPERATURETIDES AND UARS-L3DARDATA COMPARISONS 4

I3.2 hPaJ

•2

4

•.24 • 18• ß-.-'

ß

30

35

I 2 3 4 5 6 7 8 9 101112

0 40

ø

45

50

,a 15

circles). Fitted model results, discussed in section 5 are also reported (solid line).

The diurnal modes obtained during three Decembers (Figure 5) show a great reproducibility on the phase profile (+_1hours) below 40 km, where propagating mode with a phase progressionof 1.2 h/km can be noted. Above the transition region, around 40 km (which look likes a steady transition), larger variability is observed, and the smaller phase progression (0-0.8 h/km) suggeststhe dominant presence of trapped modes. Those characteristicsare in good agreement with reports for numerical simulation studies [Forbes, 1982]. The diurnal amplitudes (Figure 5) obtained during the three Decembers show a variability of +0.5 K (10-30%). This variability is under the range of the estimated uncertainties of the tidal analyses in most of the altitude range. However, observations of the local changes of the mean atmosphere (ozone, temperature, and wind) have revealed a possible connection with mean geostrophic zonal wind that increases above 10 hPa (30 km) by 2-4 m/s in correlation with tide amplitude at 0.7 and 0.5 hPa (50-54 km). The semidiurnal amplitude presents some similarities on an interannual basis, revealing forced modes with nearly constant phases from 6.8 hPa (33 km) to 1. hPa (48 km) and phase shifts of nearly 6 hours above and below this region. From May to October, variability is smaller. Tidal amplitudes vary under 0.5 K for the diurnal mode. Phase

6

?4

3.2hPa[

b30 35

..•

o

9

....

30

C

40

45

Altitude (kM)

50

o •ø

40

45

50

55

Altitude (kM)

9o

•- 0

• ..........

3s

o

o /•

• • o

E : •,o •,•'•'o •

o

•- 0

o o ,, e

o ,,0 .....................

55

Oo

Time (month)E

MLS over Observatoire de Haute Provence (OHP) (solid

•

& O

•t..............

a

--

Time (month)

0

•,

o .......

.

1 2 3 4 5 6 7 8 9 10111213

A

<

13.2 he,a j ß' ._.,m 6 1.5 hPaJ 0L.......

.....

o

o

•24[ 18 •

ß

•i 0....

10

•2

o

10,303

ss

ø ø ...............

ao

d

3s

o o

40

4s

Altitude (kM)

50

5s

Figure 5. Tidal characteristics for (a) diurnal amplitude, (b) semidiurnal amplitude, (c) Diurnal phase, and (d) semidiurnal phase as a function of altitude using MLS data for December periods. Tidal characteristics over the south of France are reported using data in December 1991 (open circle), 1992 (open diamond), and 1993 (open triangle).

variability is 1 and 3 hoursfor diurnal and semidiurnalmodes, respectively.However, successiveJuly-Augustperiodspresent larger differences that cannot be explained by dynamical background changes. During such periods, vertical propagationof planetary waves is blocked by easterly winds [Hauchecorne et al., 1991]. The diurnal amplitude profiles (Figure 6) for the successiveyears present similar shape, but amplitude increasesfor 1992 to 1994 at 5 hPa (36 km) and decreasesat 1 hPa (48 km) by about 1 K. One can note that the transition region at 3.2 hPa is located at the same level where phase transition is observed on vertical phase profiles. The semidiurnalcomponentpresentsa factor of 2 larger amplitudes in 1992 on the whole altitude range. The mean ozone profiles (measuredby MLS) of July-Augustmonthsdecreaseup to 2%

[1.5hP•

ß Diurnal O Semidiurnal

•.12

12

ß

3.1 2hPa• 9

'• 6

ß

ß

o

ß

o.1

ß

.... I

2

3

4

5

6

7

8

9

..

10 11 12 13

Time (month)

1

lO

Pressure level (hPa)

3

0 ''"'"•'•' ' " .... 1

2

3

4

5

6

7

8

9

'• 10 11 12 13

Time (month)

Figure 4. Seasonal changes of the semidiurnal amplitude andphaseat 1.5 hPa (45 km) and 3.2 hPa (38 km) observedby MLS over OHP (solid circles). Fitted model results, discussed in section5, are also reported (solid line).

Figure 6. Tidal amplitudesas a function of altitude using MLS datafor July-August periodsoverOHP. Tidal components correspondingto 1992 (high solar activity), 1993 (moderate solar activity), and (low solar activity) are representedby solid, dashed, and dotted lines, respectively . Diurnal and semidiurnalcomponentsare representedby solid and open circles, respectively.

10,304

KECKHUTET AL.: TEMPERATURETID•

AND UARS-IfDAR DATA COMPARISONS

the differencesobservedat 5 hPa (36 km) and 1 hPa (48 km)

large,the fits arederivedwith someconstraints (k3) to keep some homogeneitywith altitude and between phase and

for thediurnalcomponent andoverthewholealtituderangefor the semidiurnal components.This interannual change in

computed over the annual period, confidence for tidal

from 1992 to 1994. Changesin ozoneprofiles may explain

amplitude profiles. However, even if the fit functions are

ozone concentration could be attributed to solar activity

amplitude,especially during winter seasons,is poorer,

[Huangand Brasseur,1993] thathasdecreased according to

considering the large observed variability. The mean

10.7 cm solar radio flux from 132 to 99 to 80 units for 1992,

amplitude andthephaseof the tideas a functionof thedayof year (d) are estimatedby the absolutevalue X(d)of the

1993, and 1994, respectively.Also, the large aerosolloading followingthe MountPinatuboeruption(June1991) can be a

following formula:

possible cause.

X(d)-kl-}k2cos L-•-}r k3

5. Tidal Climatology Seasonal changes are clearly identified for the diurnal component. However, it is less clear for the semidiurnal components due to the large noise superimposedmostly in winter. Diurnal phases seem to present a great reproducibility on an interannual basis. Amplitudes present larger variability mostly in winter, probably connected to the variability of the mean atmospheric background [Lindzen, 1972, Lindzen and Hong, 1974; Vial, 1986; Forbes and Hagan, 1988] and/or aliasing with planetary waves for the semidiurnal characteristics.However, no clear evidence about the driving parametersof the interannual variability has been drawn. So at this stage, our tidal climatology model will be limited to seasonalcomponentsparametrized according to the altitude. A better understanding of such connections between tidal characteristicsand mean atmosphericbackgroundcould lead to a more sophisticatedmodel that could be investigated in the future in coupling such observations with numerical

(1)

The different coefficients k 1, k 2, and k3 are given in Table 1 for amplitudes and phases correspondingto the diurnal and semidiurnal modes. More sophisticated models have been tested but have revealed no major improvements. So we have chosenthe fit function having the simplest form. The small variability and the clear seasonal changes for diurnal tide characteristicsare well adapted to develop such a model (see Figure 3). The diurnal tides presenta maximum of 1.2-2.8 K (Table 1), peaking in summer at 1.0 hPa (48 km). The phaseshowsshifts smaller than a few hours with seasons above 2.2 hPa (41 km). Phasesduring winter are a few hours smaller than summer above 2.2 hPa (41 km) but nearly out of the phasebelow this level. At 4.6 and 6.8 hPa, it is not clear if the phase is increasing or decreasing during the transition between winter and summer. We have assumed a continuity with the altitudeand have decidedto choosethe sameform (k3) for the level above (3.2 hPa) and below (10 hPa). Seasonalchangesfor semidiurnalmodes are more difficult

simulations.

In this section, a tidal model is established from the MLS

data analysesobtainedduring the three successiveyears 19911993. The model is to be used in data comparisonsto adjust temporal coincidence between two nonsimultaneousdata sets during period with no tidal observations.This methodmay be expected to reduce the variability and the possible biases generated by systematic tidal changes. An analytic model based on sinusoidal functions has been developed to fit the observedtidal magnitudesand phasesas a function of season. As only five periods over each full year are available above France and as the interannual variability is sometimesquite

to figure out both for amplitude and phase, due to the high seasonal and interannual variability (see Figure 4). Mean phasesvary slowly in the model over the year by 4-8 hours. Distinct maxima of 1-3K in equinox periodscan be observed in the evolution of the amplitude with seasons.A minimum amplitude is observed during July-August, possibly a consequence of the transitionbetweenthe two distinctout-ofphasemaxima observedduring equinoxes.However, the high variability and the small seasonal coverage decrease the confidenceof the model for this 12-hour componentand limit the

conclusions.

Table 1. Coefficients for the Amplitude and Phaseof the Diurnal and Semi-Diurnal SeasonalModel Semidiurnal

Tide

Amplitude

Pressure

Diurnal

Phase

Tide

Amplitude

Phase

k1,

k2,

k3,

kl,

k2,

k3,

kl,

k2,

k3,

kl,

k2,

k3,

K

K

radian

hours

hours

radian

K

K

radian

hours

hours

radian

-0.5 -0.5 0.4 -0.2

2.5 1.5 2.4 1.9

rd4 rd4 rd4 rd4

7.0 7.0 5.5 4.0

•4 •4 •4 •4

0.0 0.0 2.5 1.0

-0.7

2.3

rd4

3.0

19.5

1.5

-0.4 -0.4 -0.3

1.3 1.6 1.15

rd4 rd4 rd4

3.0 0.5 4.5

•4 •4 •4

0.7 0.6 0.4

rd4 n/2 n/2 rd4 n/2 n/2 n/2

15.0 16.5 16.5 18.0

3.16 4.64 6.81

1.7 2.0 2.3 1.8 1.8 1.0 0.9 0.9

0.0 0.4 0.5 0.7

2.15

3.0 2.0 2.5 3.0 2.0 2.5 2.5 4.0

20.5 22.0 23.0

4.5 8.0 8.0

•/4 •/4 •/4 0 0 0

10.00

0.6

1.5

rd4

6.0

0.0

-

1.0

0.2

n/2

25.5

9.5

0

Level, hPa

0.46 0.68 1.00 1.47

•4

0.7

Model is calculatedbetween10.0 and0.5 hPa (30-55 km) andis obtainedin fiting MLS Data.

KECKHUT ET AL.: TEMPERATURE TIDES AND UARS-LIDAR DATA COMPARISONS

10,305

Table2. Comparison of Seasonal Amplitudes andPhases of theDiurnalTideAboveSouthern France (44øN),Between theStatistical Fit ModelandtheGSWMNumerical Model DiurnalTidal Amplitude,

DiurnalTidal Phase,

Kelvin

Altitude, km Approximate Month

OurFitModel GSWM*

Pressure, hPa 33.7

39.1

50.0

hours

OurFitModel GSWM*

Model

Model

6.2

Jan.

0.5

0.2

6.5

4.5

6.7 7.6

April July

1.0 1.3

0.2 0.2

21.8 15.7

22.6 3.2

6.7

Oct.

0.8

0.1

2.0

12.3

2.9

Jan.

0.8

0.8

6.1

20.2

3.2 3.2

April July

1.1 1.6

1.0 0.7

19.8 15.5

19.1 20.6

3.1

Oct.

0.9

0.9

21.5

20.5

1.4

Jan.

1.3

1.2

17.3

17.4

1.6 1.9

April July

2.0 2.0

1.7 1.6

17.2 19.3

17.0 18.0

1.5

Oct.

1.7

1.6

19.0

17.3

0.7

Jan.

1.7

1.0

15.4

15.0

0.8 0.9

April July

2.1 2.7

1.6 1.8

14.4 17.0

16.3 16.5

0.7

Oct.

2.0

1.4

17.1

15.2

Global-scalewavemodel.From Hagan et al. [1995]

A recent update of a two-dimensional global-scale wave early morning is observed for the different seasons(Figure 7), model (GSWM), including improvementsof the background but its time duration varies from 7 to more than 12 hours, atmosphere and revised parametrizations of both thermotidal forcing and tidal dissipation, has been used to examine

leading to a diurnal change close to a 24-hour sinusoidal function only in the summer season. Some discrepancies seasonalvariability of the diurnal tide from the ground up to between semidiurnal tide observations in the mesosphereand the lower thermosphere[Hagan et aI., 1995]. A comparison theory have been explained by the gravity wave/mean flow between the GSWM simulations, for the latitude of 44øN interactions that can generate spurious tides referred as (interpolatedfrom simulationsat 42ø and 45øN from M. Hagan "pseudo-fides".This mechanism (reviewed by Vial and Forbes (personal communication, 1995)) and the fit model [1989]), in modulating the momentum deposition of gravity extrapolated for the altitude levels available from the waves, can induce tidal changes and may contribute to numericalmodel output, is presentedin Table 2. Good overall considerablevariability in the 24-hour harmonics. The energy agreement can be observed for amplitudes, phases, and deposition by gravity waves in the upper stratosphere is seasonalbehavior. However, at 34 km, larger amplitudesare probably smaller than in the mesosphere, but it could be obtainedfor all the periodscoveredby MLS data analyses.The sufficient to induce distortion of the diurnal changes phasesare in quite good agreementfor January and April but responsible for generation of tidal harmonics. The physical are nearly out of phase for July and October. For the upper explanation of these semidiurnal observations in the upper layer, both data analysesand simulation of the phase, as a stratosphere needs further investigation. However, the functionof the altitude, are observedto vary in the range from addition of both diurnal and semidiurnal components is 20 to 15 hours.The seasonalclimatologyis similar, and both necessaryto reproduce the observed diurnal changes that may observationand simulationreveal a maximum during summer. improve time-of-day adjustments in temperature data However, the GSWM maximum of the amplitudes is comparisons.

underestimated by 20-30%. The summerdiscrepanciescan be due to changes in ozone concentration. The 1994 values

(obtainedduringlow solar activity conditionsand 3 years after the Pinatubo eruption) are in closer agreement with the simulation. The good agreementobtained during winter is surprising,consideringthe large interannualvariability. The semidiurnal component has given some results in total disagreement with numerical simulations [F. Vial, personal communication, 1994], which predicted amplitudes always smaller than 0.5 K in the altitude range studied here. The comparisonbetween two very different analysis techniques, discussed in section 3, gives some confidence about the existenceof observed semidiurnal changes.A cooling in the

6. UARS and Lidar Comparisons The purpose of this section is to test time-of-day adjustmentsin evaluating whether our tidal model can improve data comparisons. Also, quality of the temperature measurements from UARS instruments is estimated by comparisonswith lidar data over France. In each case, a mean and a standarddeviation of the difference profile, are obtained. Standard deviation (SD) correspondsto the root-mean-square of the daily differences, after subtracting the mean difference. Mean difference is usually associated with systematic differences

or bias, while

standard deviation

can be considered

10,306

KECKHUTET AL.: TEMPERATURETIDESAND UARS-LIDARDATA COMPARISONS

10

satellite

-5

-10

•' løf o

-5

B.Februaryj ß eJ

Jß

t! ß

• -10

ßß

,e

ß

10

C. July-August

ß

validation.

The

method

and

the

characteristics

of

these Rayleigh lidars have been reviewed by many [Hauchecorne and Chanin, 1980; Hauchecorne et al., 1991; Keckhutet al., 1993]. In the stratosphere,two main sourcesof errors (misalignment effects and the presence of aerosols) decreasethe confidenceof the lowest part of the profile (30-35 km). Technical improvements [Keckhut et al., 1993] at both sites have considerably reduced the effects of misalignment errors. However, during the UARS validation period, the presence of a large amount of volcanic aerosols could have induced a spurioussignal in the computed temperature around 10 hPa (31 km), basedon pure Rayleigh scattering.The NDSC (Network for Detection of Stratospheric Change) intercomparisoncampaign, held in summer 1992 at OHP with the NASA/Goddard Space Flight Center lidar, experimentally equippedwith some Raman channels,has shown [Singh et al., 1996] that such effects occurred sometimes in the lower altitude range (30-35 km). So data comparisonswith the lidar at 10 hPa (31 km) have to be considered with some cautions,

j;

and

no

Standard

conclusions

have

data obtained

been

made

at the two lidar

here

on

this

sites consist

level.

of a mean

temperatureprofile per night of measurement, with a vertical ,

,

I

3

,

,

I

,

,

6

i

,

i

9

i

12

,

,

i

,

,

15

.........

18

,

21

24

27

,

,

,

30

.....

33

36

resolution of 3 km, integrated over noncloudy times. However,lidar dataintegratedovera 1-hourperiodcenteredon the time of the nearest ISAMS

LocalSolarTime (Hr) Figure 7. Temperatureanomaliesat 1 hPa (48 km) above southernFrance as a function of solar local time, for three

measurements, have been

specifically produced. UARS provides global stratospheric temperature data

periods:(a) December,(b) February,(c) July-August. MLS data ( 30-minmean)areshownby solidcircles.Fit functions(solid lines)includeboththe 12- and24-hourcomponents.

through several instruments:the improved stratosphericsnd mesosphericsounder(ISAMS), the microwave limb sounder (MLS), the cryogeniclimb array etalonspectrometer (CLAES), and the halogen occultation experiment (HALOE). Different

as mean random fluctuations, including instrumentalerrors and spatiotemporal geophysical variability. The standard error (SE), definedasthe SD dividedby the squareroot of the number of available daily comparisons, is used to estimate the significanceof the mean calculated differences.

methods of measurements have been used by the different experiments: pressure-modulation radiometry, limb scanning of the millimeter-wavelength thermal emission, measurement of the infrared spectral emission from the limb, and solar occultation. Periodically, the UARS spacecraft yaws its axis by 180ø, so only northward looking 36-day periods allow some measurementsover southern France (44øN). Level 3AT data, used here, consist of temperatureprofiles interpolated on a common grid (4ø latitude interval and 6 levels per pressure decade) from the level 2 data file (geophysical parameters)

6.1. Description Comparisons

of Data

Available

for

using a first-order linear interpolation.Temperatureprofiles

Two Rayleigh lidars in southernFrance, at CEL and OHP, operated routinely during clear nighttime over the entire period of UARS operations.The two lidar stationsare at the same latitude (44øN) but are 550 km apart, which could be consideredessentially as the same site for satellite validation. These lidars, in providing absolute temperature measurements free of any adjustmentsor calibrations,are ideal candidatesfor

from ISAMS, MLS, CLAES, and HALOE were collected in the

nearest location of the lidar sites, in a latitude-longitude box of +10 ø. Best coincidencesfor each day have been selectedas the minimum of the quadratic sum of the zonal and meridional distances

between

measurements

of

the two locations. the

UARS

The

instruments

differences and

UARS

Satelhte

Lidar

Reference

Data

Number

of

Comparisons

Version

ISAMS

10

OHP and CEL

13

Spatial Separation, degrees

Integration Time, hours

1

HALOE •

16 3

14 16

3-8

CEL

CLAES

7

CEL

11

3-8

OHP, Observatoirede Haute Provence. CEL, Centre d'Essaisdes Landes.

Average time separation, hours

3-8

OHP

were

computed at the standard UARS pressure grid points. The

Table 3. Characteristics of UARS-Lidar Comparisons

Experiment

between

lidar

3-8

KECKHUTET AL.:TEMPERATURETIDESAND UARS-LIDARDATA COMPARISONS geometricaltitudes of the lidar measurementswere converted into pressurescale by using NMC geopotentialheight at 10 hPa (31 km) and integrating the temperatureprofile provide by lidar. Daily differences are computed by interpolating lidar temperatureto the standardUARS pressurelevel. CEL and OHP data have been mixed for the comparison between lidars and ISAMS to increase the number of possible coincidences, limited becauseISAMS operatedfor less than 1 year. For the other UARS experiments and according to the number of coincidences,only one comparisonwith either CEL or OHP is presentedhere. However, comparisonswith the two lidar data sets give similar mean differences (