October 1991 J. Wang and Y. Mitsuta 587 NOTES AND CORRESPONDENCE Turbulence Structure and Transfer Characteristics in the Surface Layer of the HEIFE Gobi Area By Jiemin Wang Lanzhou Institute of Plateau Atmospheric Physics, Chinese Academy of Sciences, Lanzhou, China and Yasushi Mitsuta Disaster Prevention Research Institute, Kyoto University, Uji, Kyoto 611, Japan (Manuscript received 14 June 1991, in revised form 31 July 1991) Abstract A peculiar downward water vapor flux was observed over the Gobi desert in the daytime on the fine days of test observations of HEIFE in 1988 and 1991. However, further analysis of the characteristics of the turbulence structure over the Gobi desert has revealed that they are in good agreement with surface layer similarity theory and also with other observations in the past. 1. Introduction In the 'Sino-Japanese Cooperational Program on Atmosphere-Land Surface Interaction Processes Experiment at Heihe River Basin; HEIFE)' there are five stations located in the oasis, Gobi, and desert area. The study of these characteristics and their interrelation is a special feature of HEIFE. A bird'seye view to the whole HEIFE experiment region is mainly a large area of Gobi and/or desert, with some oasis dispersed along the river and irrigation canals. The surface state of the oasis is rather complicated due to the networks of cropland, windbreak trees and scattered residential areas. Comparatively, the Gobi surface is simpler, and is quite flat and open. So the HEIFE Pilot Observation (POP) in 1988 and Pilot Intensive Observation (PIOP) in 1990 were all carried out at the Gobi station. In addition, to test the whole observation system and methodology, we have made a preliminary study of the turbulence transfer characteristics over the Gobi desert. A peculiar phenomenon of downward water vapor flux over Gobi desert in the daytime which was found in POP (Wang et al., 1990 and Wang and Mitsuta, 1990) was notices also in PIOP in 1990. Here we C1991, Meteorological Society of Japan present a further analysis of the characteristics of the turbulence structure over the Gobi desert for both POP and PIOP. 2. Experiment and Data As the site and instrumentation are described in detail in Wang and Mitsuta (1990), only a further brief introduction is given here. The HEIFE Gobi station (100*06'E, 39*09'N,1480 m MSL) is situated in the south part of the Heihe river basin, approximately 1.5 km to the south of Linze oasis and 20 km to the north of Qilian mountain. The surface is quite flat and open, particularly to the south, and consists of coarse sand grains and small pebbles with very sparse scrub vegetation. Its geomorphic feature is consistent with conditions all along the south part of the basin. The HEIFE POP was conducted mainly in September of 1988, and PIOP was carried out in August, 1990. An investigation of the turbulence structure and turbulence transfer characteristics during POP and PIOP is presented in the following. About 88 runs in POP and 135 runs in PIOP are selected in the present analysis, in which various air stabilities and wind directions are included. Some runs are discarded according to the following criteria: 1) The
Journal of the Meteorological Society of Japan Vol. 69, No. 5 Fig. 1. Variations of * /u* versus stability. The dashed curve is Panofsky & Dutton's simulation (Eq. (1)). Slope 1/3 line is the free convection asymptotic. The solid line is the best fit for this experiment (Eq. (2)). mean wind azimuth deviation referring to the sonic anemometer coordinate system is larger than 45 degrees; 2) Wind speed is weak and lower than 1 m/sec and normally with a very low value of u'w', even a positive u'w' (where u' is alongwind fluctuation). These occurred mostly in the stability transient hours of morning and late afternoon, normally with an apparent non-stationary condition. 3. Similarity Analysis of Variances According to the Monin-Obukhov hypothesis, the standard deviations of the velocity components, normalized by the friction velocity, are functions only of z/l in the surface layer. Figure 1 shows the variation of */u*=* versus stability obtained at the Gobi station. The data distribution for both POP and PIOP are quite similar, so only the latter is presented. Compared to some observations in the literature i. e. Panofsky and Dutton (1988), data points in Fig. l show less scatter, particularly on the in unstable side. It is apparent that z/l being under -0.35,*/u* is in good agreement with the predictions of free convection that the variance of vertical velocity is in proportion to (z/l)1/3 (Priestley, 1954 and Monji, 1973). An empirical formula recommended by Panofsky et al. (1979) for the unstable condition, is also plotted in Fig. 1 by dashed line, which is obviously higher than our observations. Particularly in the range of -0.1 <z/l<0 the calculated values of Eq. (1) are 5*10% larger. On the stable side, the observed data points of */u* show a little increasing with stability. The same tendensity is also seen in the recent result by Shao and Hacker (1990). The solid curve in Fig. 1 represents the best fit to the present data, which can be shown by For neutral stratification, the present data show that /u* =1.14. This value is smaller than the * preliminary results of POP shown in the earlier paper (Wang and Mitsuta, 1990) and the commonly recognized value of 1.25 (Panofsky and Dutton, 1984). In fact, the latter was based on a simple average of about 10 observations conducted over both sea and land. The individual values of those observations were ranged. from 1.10 to 1.40. Recently, on the basis of an experiment in Lovsta, Hogstrom (1990) indicates that the near-neutral value of */u* increases with height above ground irrespective of similarity theory. In the Lovsta data the mean values of */u* at 3, 6, and 14 meters are 1.14, 1.23 and 1.33 respectively. Hogstrom thought that this increment is due to the so-called 'inactive' turbulence. A similar analysis for the along-wind and crosswind components were also conducted. The nondimensional values of */u*=*u and */u*=*v are about 2.65 and 2.22, respectively at near neutral conditions, which are a little larger than Panofsky and Dutton's (1984) results which are 2.4 and 1.9. Along with the increasing instability (when z/l < -0.1), the data points show a very large scatter, which can be explained from the spectral analysis shown in the next section. The variances of a scalar quantitiy normalized by the scaling parameter is also a universal function of z/l. Figure 2 shows the variation of */ T* =*
October 1991 J. Wang and Y. Mitsuta 589 Fig. 2. Variations of */ T* versus stability. The solid curve is the best fitted line of this experiment (Eq. (3)). (Data points in the stable part are very scattered, the constant value shown by the dashed line is for reference.) Fig. 3. Variation of *q/ q* versus stability. with z/l. The best fit function form for */ T* of the present study is In the strongly unstable case (z/l< -0.30), * is in proportion to (z/l)-1/3 as predicted in free convection (Priestley 1954 and Monji 1973). As in many other experiments, the data points in near neutral and stable stratification show a larger scatter. This result is quite similar to the recent result by Shao and Hacker (1990), derived on a local similarity analysis. Figure 3 shows the variation of *q/ q* =*q with z/l. Scatter of data points are much larger than former examples. Because of very low humidity over the desert, both *q and q* were of the order of 10-4 (Kg/Kg). So the accuracy of *q/q* is not sufficient in this experiment with a fine wire psychrometer. According to Monin-Obukhov similarity, *q has a similar form to *. We can roughly see a 1/3 slope in the strongly unstable region and a constant value in the near neutral and stable regions. But no definite functional relation is obtained. 4. Basic Characteristics of Turbulence Spectra Figure 4 to Fig. 7 show the power spectra of wind components u, v, w, and temperature T of five runs with different stabilities in the PIOP. The abscissa is the commonly used reduced frequency n= f z/u,
590 Journal of the Meteorological Society of Japan Vol. 69, No. 5 Fig. 4. Spectra of along-wind component of 5 runs in different stabilities; normalized by the Kaimal Scheme. Fig. 6. Same as Fig. 4 but for the vertical wind component. Fig. 5. Same as Fig. 4 but for the crosswind component. and the ordinate is the normalized spectra using Kaimal's scheme (Kaimal et al., 1972), By this scheme the inertial subrange spectrum, expressed in the frameworks of surface layer similarity (Kaimal et al., 1972), becomes, where a is the Kolmogorov constant, k the von Karman constant and * the dissipation rate of mechanical turbulence. When plotted on a log-log scale, the inertial subrange spectra collapse to a single straight line with a -2/3 slope independent of z/l. Fig. 7. Temperature spectra for the same runs as in Fig. 4. Five u spectra in the inertial subrange for various stability conditions shown in Fig. 4 apparently collapse to a single straight line with a -2/3 slope and can be approximated by Comparing this equation with Eq. (4), we can assume the combination of two constants a and k can be approximate by 0.5 and 0.4, as usually accepted. The power densities of other components in the inertial subrange are estimated to be larger by 4/3 than the u component (Kaimal et al., 1972). The present data show that they are larger than u, but the ratio is not exactly 4/3.
October 1991 J. Wang and Y. Mitsuta 591 Fig. 8. Spectra of specific humidity (unstable cases). The spectral peak of the w spectra (shown in Fig. 6) shifts towards the low frequency side with decreasing stability in the range of z/l from 0.31 to -0.32. The nondimensional frequency of the w spectral peak for neutral stability is about 0.5. The spectral peaks of other components u and v are at lower frequencies than that of w component. The spectral peaks for neutral stability are 0.06 for u, 0.15 for v and 0.1 for T. Kaimal has reported the spectral peak frequencies of his Kansas experiment to be 0.05 for u, 0.2 for v, 0.5 for w, and 0.1 for T. In unstable air the u or v spectral density of different runs tend to cluster in a random fashion. Turbulent energy occupied by their low frequency region is clearly higher than that of w component spectra, reflecting the effects of higher altitude wind turbulence. In stable cases, the secondary peak often appeared in the low frequency region reflecting the existence of meandering of flow. The form of temperature spectra is normally in between the vertical and the horizontal wind components spectra. The power spectrum of specific humidity fluctuations is shown in Fig. 8, with a little different ordinate. Except for high frequency region, it resembles with temperature spectra with less deviations with stability. The q power density at the high frequency end, n>1 increase rapidly with slope of almost +1. This means the power density is constant with frequency, or this frequency range is already in the white noise region in which the signal level sinks into the noise level. Specific humidity measurement was made by a fine-wire thermocouple psychrometer in this case. The time constants of its dry and wet bulb were different, being 0.19 sec for the dry bulb and 0.7 sec for the wet bulb (Tsukamoto, 1986). These large and imbalanced time constants caused considerable noise at the high frequency end even though correction has been adopted in the original Fig. 9. Cospectra of water vapor flux for the same runs as in Fig. 8. data processing following the method proposed by Tsukamoto (1986). As an infrared hygrometer with quicker response will be used in the observational phase of the project, this deficiency will be greatly improved. However, as shown in Fig. 9, the cospectra of w'q' obviously drops down rapidly in this high frequency region. The noise of specific humidity shown in the bending up of the spectrum does not show correlation of the w component with turbulence, and does not have a notable effect in the eddy-correlation calculation of the water vapor flux. The cospectral peak of (w'q') is almost the same as that of the q spectral peak and is at about n=0.03 without respect to stability (-0.32 < z/l <0.12). Figures 10 and 11 show the cospectra of momentum (u'w') and temperature flux (w't') of the same runs. Compared with the power spectra, the falling off of both the high and low frequency ends is obviously faster. At the high frequency end, the slope may be approximated by -3/4 slope, as proposed by Wyngaard and Cote' (1972). While at the low frequency end, the secondary peaks in the power spectra are not seen in the cospectra; this shows that low frequency fluctuation is not an effective component in eddy flux. The cospectral peak frequencies are almost the same as those in power spectra. It can be seen from these figures that in case of the flux observation, the sampling period can be shorter, because fluctuations in the lower frequency part, which can be seen in the power spectra, do not contribute in eddy fluxes. Particularly in stable conditions, since more contribution is due to higher frequency eddies, a sampling time of 15 to 20 min could be enough for momentum, heat and water vapor flux measurements. As the cospectral density
592 Journal of the Meteorological Society of Japan Vol. 69, No. 5 sis of the turbulent structure of boundary layers in these observations following the surface layer similarity theory has shown that the turbulent statistics are in good agreement with surface similarity and also with other observations in the past. The humidity profile measurement in 1990 also supported the downward transport of water vapor in the daytime (Wang and Mitsuta, 1991). Therefore, we did not find an unusual phenomenon in the turbulence of the surface layer over the desert, except the cause of a humidity inversion in the daytime on a fine day. Acknowledgment The HEIFE program is supported by the National Natural Science Foundation of China, Chinese Academy of Sciences, and the Japanese Ministry of Education, Science and Culture. Fig. 10. Cospectra of momentum flux for the same runs as in Fig. 4. Fig. 11. Cospectra of sensible heat flux for the same runs as in Fig. 7. collapse in the inertial subrange very fast, the averaging time of measurements can be set in the middle of the inertial subrange such as n=2 or 3. This corresponds to 10 Hz sampling in a 5 m/s wind in this case. 5. Conclusion In September of 1988 a test observation (POP) of turbulent flux of HEIFE was made at the Gobi desert station and peculiar downward water vapor fluxes in the daytime on fine days were observed (Wang and Mitsuta, 1990). This phenomenon was also observed in the next text observation (IPOP) in August 1990 (Wang and Mitsuta, 1991). Analy- References Hogstrom, U., 1990: Analysis of turbulence structure in the surface layer with a modified similarity formulation for near neutral conditions. J. Atmos. Sci., 47, 1949-1972. Kaimal, J.C., J.C. Wyngaard, Y. Izumi and O.R. Cote, 1972: Spectral characteristics of surface layer turbulence. Quart. J. Roy. Meteor. Soc., 98, 563-589. Monji, N., 1973: Budgets of turbulent energy and temperature variance in transition zone from forced to free convection. J. Meteor. Soc. Japan, 51, 133-145. Panofsky, H.A. and J.A. Dutton, 1984: 'Atmospheric Turbulence', Jhon Wiley & sons, New York. Panofsky, H.A., H. Tennekes, D.H. Lenshow and J.C. Wyngaard, 1979: The characteristics of turbulent velocity components in the surface layer under convective conditions. Boundary Layer Meteor., 11, 355-361. Priestley, C.H.B., 1954: Convection from a large horizontal surface. Must. J. Phys., 7, 176-201. Shao, Y. and J.M. Hacker, 1990: Local similarity relationship in a horizontally inhomogeneous boundary layer. Boundary Layer Meteor., 52, 1-2, 41-68. Tsukamoto,O., 1986: Dynamic response of the fine wire psychometer for direct measurement of water vapor flux. J. Atrnos. Ocean. Technology, 3, 453-461. Wang, J., X. Liou and Y. Qi, 1990: A Preliminary study of turbulence transfer characteristics over Gobi Desert. Plateau Meteor., 9, 2, 120-129 (in chinese). Wang, J. and Y. Mitsuta, 1990: Peculiar dowanward water vapor flux over Gobi Desert in the daytime. J. Meteor. Soc. Japan, 68, 399-401. Wang, J. and Y. Mitsuta, 1991: Evaporation from the desert (to be published). Wyngaard, J.C. and O.R. Cote, 1972: Copectral similarity in the atmospheric surface layer. Quart. J. Roy. Meteor. Soc., 98, 590-603.