DOPPLER SODAR MEASUREMENTS OF VERTICAL WIND SPEED IN MOSCOW Mikhail A. Lokoshchenko1, Vitali G. Perepyolkin2, and Natalia V. Semenova1 'Department of Meteorology and Climatology, Geography Faculty, Moscow State University after name of Lomonosov, Lengory, Moscow, 119992, Russia. ' Phone: 007-095-939-1803; Fax: 007-095-939-4284; E-mail: [email protected]. 2Russian Academy of Science Institute of Atmospheric Physics after name of Obukhov.

Abstract The Doppler sodar measurements of wind speed vertical component W at Moscow University have been carrying out since 2000 continuously on base of a phase control of useful echo-signal, instead of an usual amplitude control. The common principle of the calculation scheme is described. The accuracy of sodar measurements of W is discussed. If a vertical direction of a sound beam is precisely controlled the Doppler W calculation may be accepted as reliable in the case of nearly calm conditions, a flat relief of a surrounding area, and accounting nonsymmetric intensity of thermal turbulence in upward and downward flows. The initial results for the lowest 200-300 m layer demonstrate an average tendency to upward diurnal movements, which accelerate with the height and, on the contrary, to downward nocturnal ones. Keywords: vertical component o f wind speed, phase feedback loop, comparison of direct and sodar data, accuracy estimation, buoyancy force, ground braking.

1.Introduction.

2.Methodical questions.

A wind speed vertical component W is an important parameter of the hydrodynamics equations. However, it is usually too small to be correctly measured by routine meteorological sensors. In synoptic scale processes it’s quite negligible, i.e. two or three orders less than a horizontal wind vector V. It allows to eliminate components containing W in equations, that leads to the static and, sometimes, even to the geostrophic approach in high-scale models. However, if we study meso-scale or local flows, the knowledge of W is considered to be necessary in any case. Indirect theoretical estimations of W can’t provide model calculations of a real success. Thus, we have to measure W, and the acoustic remote sensing seems to be one of the perspective tools in this field. At Moscow University the acoustic remote sensing of the lower 800-m air layer turbulent structure has been carrying out since 1988 by means of GDR vertical sodar “ECHO-1”. The operational frequency is 1666,6 Hz, the period of burst repetition consists of 10 sec, the power is 75 W, and the vertical resolution is 12,5 m. In 2000 the sodar was modernised and supplemented by a Doppler device for vertical wind measurements. Here we describe specific features of our scheme and present the initial data of W calculations within of two various months of 2000-2001.

The main possible errors of the sodar measuring and interpretation of the W values are considered to be the result of: a) The occasional deflections of a sound beam direction from zenith; b) The non-flat relief of a surrounding area (Mastrantonio et al. 1996; for the WTV data Coulter et al. 2000); c) The possible refraction of a sound beam - even if the antenna is directed strictly vertically; d) The different intensity of thermal turbulence in upward and downward flows (Mastrantonio et al., 1996; for the cr (W) data - Seibert, 1998, etc.). The factors (a) and (b) are presented for any method of the W measuring. Moreover, an error value depends on a wind speed horizontal component V. The more is V the more is its impact as a projection into a sound beam axis in case of its deflection from zenith (a). An inclined relief (b) represents a similar effect geometrically. As for the factor (a) we precisely test a horizontal position of a loudspeaker plane (that is a vertical beam direction) using the mechanical levelling. Its accuracy is nearly of 1°. As sin 1° = 0.017, a value of V can’t exceed more than ten times of a W value. According to this condition an error of the W measuring due to projection of V to a beam axis remains still one order less than the W value (< 0.2). Values of W are usually not more than some decades of cm/s. Thus, we should analyse separated

301

hourly-averaged profiles of W only in cases of a calm or very light winds (that is really in anticyclone centres, in ridge axis, etc.)- However, referring to monthly-averaged profiles o f W we can ignore this strong methodical principle as multiple occasional errors of the opposite sign tend to compensate each other. So, a great number of W measurements for long-term periods seems to be reliable statistically. As for the factor (b) we have nearly a flat relief around Moscow University like a plateau. In the radius of up to 400 m a slope here in any direction is less than 2° (1-1.5° in average). For more distance the relief is flat too except a sharp slope of the Moscow-river valley to the north, where a slope consists of 3.5° for 1 km distance, decreasing again for a more distance. Thus, errors due to inclined relief and beam deflections are evidently of the same order. A sound beam’s refraction (c), at least a thermal one, is quite negligible and can be ignored due to small beam’s deflections from zenith (£ 1°) and a short range of reliable W measurements. Even in case of strong inversions up to 20 °C of intensity in the 200 m air layer a beam inclined for 1° will increase its angle only by 0.05° due to the thermal refraction. The last factor (d) is specific for the acoustic remote sensing, basing on a sound scattering by small-scale turbulent inhomogenities. More intense thermal turbulence inside o f convective plumes in the comparison of downward zones can be a course of the overestimated values of W by sodar data. Fortunately, this effect seems to be not very strong due to a short range of W measuring (Mastrantonio et al. 1996), To test factor (d) we calculate precisely a number of single Doppler measurements of W for separated 12.5 m layers inside and outside of plumes for some periods of a strong daily convection. As a result, the average ratio o f single reliable W values in and out of plumes, detected by sodar echogram, was nearly of six to four up to the 200 m height. Evidently, the factor (d) is the most important among others, because it leads to systematic errors and, hence, to overestimated probability of upward flows - at least, at daily time. If a value of W is +X in plumes and -X (a downward flow) between them, an average error will be equal to nearly of +0.1 X. As it is known the average values of о (W) in the lower air layer usually has an order of 0.5 m/s (Mastrantonio et al. 1996, Coulter et al. 2000, etc.). Hence, we can suppose that an error due to the factor (d) usually has an order of +0.05 m/s, whereas the errors due to the factors (a), and (b) are ±0.05 m/s and ±0.08 m/s correspondingly (if W = 0.3 m/s and V = 3 m/s). So, the only factor (d) causes systematic errors, the rest ones seem to be smoothed in averaging. Hence, we can trust sodar estimations of W at least qualitatively, if values are not very small.

■

>

S

ШШ

9

cx/2

7 *

to+At,

x

Zn+c A t/2

О Fig.l. Change of scattering volume.

W = с (1 - fo/f) / (1 + f0 /f), where c is a sound velocity, f0 is an initial frequency; f is a frequency of scattered echo-signal. Our measuring scheme, created and realised at Moscow University, is based on a phase-control of an useful echo-signal. We can apply this way because a sodar makes a sliding averaging of an echo-signal from separated inhomogenities in limits o f a pulse scattering volume. During the initial signal period t « 1/fo an averaging zone will be changed only a bit (Fig.l), as a pulse duration т is two orders more than t. The horizontal transference of inhomogenities by a wind out of scattering volume also leads to negligible changes of an averaging zone due to extremely small Mach values in real atmosphere. Hence, generalised frequency of an echo-signal can’t be changed fast during t = 1/fo. On the contrary, in case of an occasional signal (that is noise) its frequency is occasional too and can change as fast as you like. So, the usage of a phase control is based on this conclusion. In Fig.2 the Doppler scheme is outlined by dotted curve as a part of common sodar electronics. So, a signal from quartz generator goes to multiplying of initial frequency (1) and, then, to the measuring counter (8). A part of received echo-signal from the switch goes to the Doppler scheme as well through the amplifier (2) and the band filter (3). Then, an amplified narrow-band signal goes to the phase feedback loop (4-6). Here the phase detector (4) compares the phases of an input signal and a signal of the voltage-controlled generator (6), which frequency is guided by the same detector through the special filter (5), which restricts the most possible rate of frequency change. Thus, the circle (4-6) is a feedback system with a resulted signal of the generator as an output. It imitates an input signal, smoothing quick jumps o f frequency due to the filter (5). As a result, an output signal is adapted to an input one in the phase. However, if frequency of an input signal changes too quickly due to the strong noises an output signal begins

3.Principal scheme of calculations. As it is known Doppler measurement of W can be represented in the following frequency ratio:

302

4.

Fig.2. Block-scheme of Doppler calculations. to be late since of the filter’s inertia. In this case the indicator o f a phase conformance (9) interrupts calculation as doubtful (a dashed arrow). A reliable echo-signal goes then to a frequency divisor (7). The counter (8) counts an amount of high-frequent pulses from (1) during a period of a low-frequent pulse from (7). This amount is evidently proportional to the fj/f value. The whole procedure is repeated separately for each layer o f 12.5 m height from 38 to 800 m. A traditional amplitude control of useful data was . realised too, but specific features of our scheme leads to its less efficiency here in comparison with a phase control. As a result of the comparative testing the phase principle o f noise elimination was taken as a basis of continuous Doppler measurements in Moscow.

300 m height. Firstly during one week from July the 26th to August the 3rd the sodar was operated in a routine vertical antenna position. As a result the W data, measured by a sonic anemometer, mounted on the mast on 73 m height, were compared with the ones, measured by the sodar in limits of the nearest the 12.5 m height interval. The agreement between both data was received as poor. The linear correlation co-efficient is positive, but statistically non-meaningful. This result was not unexpected for us, because similar comparisons often demonstrate non-meaningful connections as well (Seibert, 1-998). Of course, different spatial scales of the W measuring inside of sonic base and in limits of scattering volume above an antenna should be taken into account. Indeed, common average tendency to a movement of any sign, detected by the sodar, is accompanied by small-scale turbulent eddies and flows of an opposite direction, which can be measured by the sonic. Moreover, a distance between the sonic and the sodar is too great for an absolutely correct comparison (it could not be less really due to strong reflections from the mast). So, a poor connection can be a result of the both Doppler calculation errors or sonic errors, and complexity of turbulence nature. To test an ability of our Doppler scheme to measure real W data we needed direct inertial sensors, indicating average tendency of a wind, comparable with sodar data in terms of a spatial scale. Evidently, this tool is an usual mast vane. That’s why the next week from August the 3rd to August the 9,h the sodar antenna was inclined into the zenith angle of 30° (see the photography in Fig.3). Hourly-average values o f horizontal wind speed V, measured by routine vanes on two mast levels (73 and 121 m), were compared with sodar data, that is a radial projection of V to the beam axis in the nearest height intervals (having the middles correspondingly on 81 and 144 m along inclined axis). As an axis of the antenna was inclined to the south-east, southerly and easterly winds were detected by the sodar as downward flows, whereas northerly and westerly winds - vice versa, i.e. as upward flows. Radial projections on an axis o f the vane data of V were calculated with the same sign accordingly to a wind direction.

4.Reliability testing of sodar data. Sodar data of W, calculated by the described scheme, were compared with in-situ measurements on the meteorological mast in Obninsk, situated at 100 km to the South from Moscow. For this purpose a special experiment was carried out in summer of 2001. At this time our sodar was moved from Moscow University to Obninsk and installed at 150 m distance from the mast of

Fig.3. The inclined sodar in Obninsk.

303

3 -1 2

.я

С

! it

X

П *