Revision 2.02 on November 24, 2006.
─────────────────
Send correspondence to Jun-ichi Furumoto.
Email: furumoto@rish.kyoto-u.ac.jp
Telephone: 81 774 38 3861, Fax: 81 774 31 8463

1
The turbulence echo intensity observed by a wind-profiling radar is closely related to the vertical gradient of refractive index squared ( M
2
), which largely depends on the vertical humidity gradient in a moist atmosphere. We have developed a radar remote-sensing technique for determination of humidity profiles by using the turbulence echo characteristics. In this study we have combined the results collected with two co-located radars; the MU (Middle and Upper atmosphere) and Lower Troposphere
Radar (LTR) operating at 46.5 MHz and 1.3 GHz frequencies, respectively, and obtained humidity profiles at 0.3-7.5 km. The echo power profiles (signal-to-noise ratio,
SNR) with the two radars are connected smoothly in a height range between 1.5 and
1.95 km, by considering reduction of the receiver sensitivity for the MU radar due to leakage of the transmission signal. The retrieved humidity profiles show detailed time-height variations, which are compared with simultaneous Raman lidar measurements and radiosondes.
Keywords: humidity retrieval, atmospheric humidity, the MU radar, LTR, radar remote-sensing, meso-scale meteorological phenomena

2
Humidity is one of the most important driving forces of intense atmospheric disturbances via the latent heat release. However, regardless of weather conditions, it is important to develop new and accurate technique to continuously monitor the humidity profile. Global, in situ, radiosonde balloon humidity profile observations have for decades been the source of all knowledge of the atmospheric moisture distribution. A humidity sensor attached to a balloon can measure relative humidity from the surface to a height of 12 km, with a vertical resolution of several tens of meters. Radiosonde observation data are widely used as a reference for simultaneous comparisons with other measurements. Despite the advantages of the above methods, it takes about 90 minutes for a balloon to reach a height of 30 km, and a radiosonde cannot fully reveal the detailed structure of humidity, for example, within an individual rain cell. In addition, as it ascends radiosonde balloon drifts with the wind and thus provides measurement miles away from the event under observation.
Various active and passive remote sensing techniques have been developed to measure atmospheric humidity, including Raman lidar, differential absorption lidar
(DIAL), global positioning system (GPS), and microwave radiometers (MWR). Raman lidar produces precise humidity profiles from the Raman scattering of H
2
O molecules
( Whiteman et al.
, 1992). DIAL constructs humidity profiles by using two lasers that each have different rates of absorption by water vapor ( Grossmann et al.
, 1987). The vertical (about 100 m) and temporal (several minutes) resolution of lidar is good, but daytime observations are difficult due to strong background noise. The observation height is also restricted to below the cloud base during cloudy conditions.
A passive dual-frequency microwave water-vapor radiometer can measure

3 precipitable water vapor (PWV) and liquid water content from the intensity of microwave emissions from water vapor. Humidity and temperature profiles are estimated using multi-frequency MWR ( Ware et al.
, 2003). Improved humidity profiles can be obtained using a dual-frequency MWR in combination with complementary instruments such as a ceilometer and Radio Acoustic Sounding System (RASS) with wind profiling radar ( Stankov et al.
, 1995).
GPS, the satellite positioning technology, is also used for humidity measurements.
PWV can be derived from a ground based GPS receiver from propagation delays of the electromagnetic waves from GPS satellites ( Bevis et al.
1994). The GPS-tomography technique was also developed. Dense GPS receiver networks measure the three-dimensional distribution of humidity by delays of many propagation paths ( Flores et al.
, 2000; Noguchi et al.
, 2004).
Research groups from Kyoto University and NOAA/Environmental Technology
Laboratory (ETL) have developed a humidity observation method that uses the turbulence echo characteristics of wind-profiling radar. The turbulence echo intensity observed by wind-profiling radar is closely related to the vertical gradient of refractive index squared ( M
2
). In addition, M within the lower atmosphere largely depends on the vertical humidity gradient (e.g., Furumoto and Tsuda , 2001). This method enables continuous monitoring of humidity profiles regardless of the time of day or weather conditions; such data contribute to the elucidation of intense meteorological disturbances; however, the sign of M is unknown from radar observations and it is therefore necessary to determine the sign by a different method. Tsuda et al. (2001) estimated humidity profiles using the MU (Middle and Upper atmosphere) radar with
RASS (MU radar-RASS) observation using the correlation between M and the

4
Brunt-Väisälä frequency squared ( N
2
), assuming that the condition is an adiabatic process and ignoring horizontal advection.
Furumoto et al. (2003) improved the technique for estimating humidity profiles by using simultaneous GPS measurements of PWV to achieve a more robust estimation under various meteorological conditions. The results of this method are in good agreement with C-band meteorological radar data and GMS satellite observations. This method has also been applied to wind-profiling L-band Lower Troposphere Radar
(LTR) and Equatorial Atmosphere Radar (EAR) ( Furumoto et al.
2005, 2006a).
NOAA’s Earth Systems Research Laboratory obtained humidity profiles using wind-profiling radar. Stankov et al.
(1996) first obtained humidity profile by using 449
MHz wind profiling radar measurements of turbulence echoes to constrain the MWR
PWV observations within the statistical retrieval algorithm and Gossard et al. (1999) obtained humidity profiles by constraining the sign of M from GPS-derived PWV.
Stankov et al. (2003) integrated the vertical gradient of the humidity profile obtained from 913 MHz and 449 MHz wind-profiling radar turbulence echo measurements to obtain boundary layer humidity profiles.
In this study, data from two radars at different operational frequencies are combined by developing a humidity monitoring system that is applied over a wide vertical range from the boundary layer to the middle troposphere. In previous studies, humidity profiles have been successfully determined, using the MU radar, at heights between 1.5 km and 7.5 km ( Tsuda et al.
2001; Furumoto et al.
2003). Humidity profiles have also been estimated from L-band wind-profiling radar at heights between
0.3 km and 2.2 km ( Furumoto et al.
2005). In this study we expand on the earlier work of the Kyoto University research group by using radar from two radars aiming at the

5 expansion of the estimation height range. The observational height range of wind-profiling radar is mainly determined by radar operational frequencies; it is difficult to determine humidity profiles over the entire tropospheric range using a single wind-profiling radar. Atmospheric disturbances that occur in the boundary layer commonly spread to the free troposphere. For example, cumulus convection that appears in the boundary layer extends to the upper atmosphere due to upward wind motions caused by the release of latent heat. For a full understanding of the thermodynamics of this detailed meteorological phenomenon, it is very important to monitor humidity profiles over the entire height range from the boundary layer to the mid-troposphere. This paper describes a method for determining humidity profiles for heights between 0.3 km and 7.5 km by combining data and LTR data obtained at
Shigaraki MU Observatory.
The humidity retrieval algorithm using wind-profiling radars is widely described in detail in the literature ( Tsuda et al.
2001; Furumoto et al. 2003; 2005; 2006). The brief outline of the method is described here. The structure constant of refractive index ( C n
2 ) for the turbulence echo is related with the volume reflectivity (
 as

0.38

C n
2
, where
 is the radar wavelength.
 is derived from M
2
, the Brunt-Väisäla frequency squared
( N 2 ) and turbulence energy dissipation rate (
 as
     
( Gage et al.
1980).
Note that N
2
is calculated from the temperature profiles with the RASS measurements.
Since
 is calculated from a spectral width of turbulence echo and N , M
2
is calculated from the turbulence echo and RASS observations ( Tsuda et al.
2001). In addition, M in the moist atmosphere is primary determined with d q /d z (e.g., Furumoto et al.
2001).

6
Tsuda et al.
(2001) derived the q profiles from these relations as:
 
2



 z
0 z

1.65
T p
2
M

1 d T
7800 d z
 


 
2 d z
 q
0

0
2



, (1) where p , T ,
 q
0 and

0 are the pressure, temperature, potential temperature, dry adiabatic rapes rate, and boundary values of q and
 at a start height of the integration
( z
0
), respectively. The q profiles are estimated from the radar-derived M if temperature profiles are obtained from RASS observations. The present study used the method by
Furumoto et al.
(2003) for determining the sign of the radar-derived M .
In order to calculated

from the echo power intensity with the wind profiling radar, we adapted 8.0dB and 3.5dB as the noise figure of the MU radar and LTR. The loss factor is also determined as -4.0dB and -1.5dB for the MU radar and LTR, respectively.
Note that the value of | M | observed from wind-profiling radar with an active phased array system (| M
*
|) is a relative value as | M |= A | M
*
|, where A is a constant value.
A was estimated by comparing | M
*
| with | M | from the radiosonde results.
The sign of M was constrained from precipitable water vapor (PWV) obtained with
GPS observations (PWV
G
). Note that the PWV is the total precipitable vapor from the ground to the top of the atmosphere. The integrated water vapor in the radar estimation range (IWV) is calculated from the PWV
G
as IWV
G

PWV
G

R range
, where IWV
G and R range are the GPS-derived IWV and the ratio of IWV to the PWV calculated from radiosonde results, respectively. Note that R range is interpolated with time to the temporal resolution of GPS observations. The algorithm to determine the sign of | M | is described in detail in Appendix 1.

7
The experiment using combined MU radar-RASS and LTR-RASS was conducted over September 23-26 and November 20-21, 2002 at Shigaraki MU Observatory, Japan
(136.10E, 34.85N). During this period, a radiosonde was launched every three or six hours to obtain pressure, temperature, and humidity profiles. A GPS receiver (Ashtech
ZXII3) was also operated to monitor continuous PWV data at the radar site.
A Rayleigh/Mie/Raman lidar at Shigaraki MU observatory was simultaneously operated during the night time. This lidar transmits the laser beam of 532 nm with 30 W output power and receives scattered signals by a telescope with 82 cm diameter. The polychrometer detects 5-channels of two elastic (532 nm), two rotational Raman (531.1 and 528.5 nm) and a water vapor Raman (660 nm) signals, in order to measure temperature, humidity and particle properties ( Behrendt et al.
, 2004). Density of atmospheric molecules (N
2
and O
2
) is derived from two rotational Raman channels and water vapor mixing ratio can be estimated by comparison with the signal at the water vapor Raman channel. The data are stored with a time and height resolution of 1 minute and 72 m, respectively. It should be noted that the water vapor mixing ratio measured by a lidar without a bias once sensitivity difference of Raman channels is calibrated using a reference radiosonde profile, except for the region above thick clouds where wavelength dependency of extinction becomes significant.
During the experiment, combined MU radar-RASS and LTR-RASS measurements were undertaken. The detailed system descriptions of the MU radar and LTR are provided by Fukao et al.
(1985) and Hashiguchi et al . (2004), respectively. The MU radar is an MST radar that operates at 46.5 MHz frequency with maximum and mean transmitting powers of 1 MW and 50 kW, respectively. A total of 20 horn loudspeakers

8 developed for the MU radar-RASS and with a maximum sound pressure level of 140 dB are located at the radar site. An FM-chirped sound-wave swept around 100 Hz is employed to satisfy the Bragg condition over the entire height range. The LTR is a portable mono-static pulsed radar with the center operating frequency, maximum and average transmitting power of 1357.5 MHz, 2 kW, 428 W, and 100 m, respectively.
Details of the observational parameters and background conditions in the September campaign are provided here. The detailed observational parameters of the MU radar and
LTR are listed in Tables 1 and 2, respectively. For turbulence echo observations by the
MU radar and LTR, the vertical beam and four oblique beams steered to north, south, east, and west direction were employed. The turbulence echo parameters for the MU radar and LTR were estimated every 162 seconds and 86 seconds, respectively, with a height resolution of 150 m. The beam directions for MU radar-RASS observations were determined from real-time ray-tracing of acoustic wave-fronts ( Masuda 1988).
Wind-profiling radar operated on VHF radio frequencies can observe both isotropic turbulence echo and aspect-sensitive echo due to Fresnel reflection. An earlier study of the ratio of the echo intensity with the oblique beams to that of the vertical beam (aspect sensitivity) demonstrated that the intensity of a reflection echo rapidly decreases with increasing zenith angle from 0 degree to 10 degrees ( Tsuda et al. 1997). Considering these results, MU radar data were used at a zenith angle of 10 degree to remove the influence of specular reflection. LTR vertical beam direction data are also used in this study, as LTR operated at a 1.3 GHz center frequency does not observe anisotropic echoes.
Figures 1 (a) and (b) show time-height plots of the peak spectral density averaged for four oblique beams of the MU radar and the virtual temperature obtained with the MU

9 radar-RASS ( T vR
), respectively. The turbulence echo was continuously obtained above
1.5 km with the MU radar throughout the September campaign. MU radar-RASS observations were operated from 09:00 LT on September 23 until 19:00 LT on
September 26, while virtual temperature was monitored almost continuously by the MU radar-RASS between 1.5 km and 7.5 km height, although missing values occurred intermittently.
Figures 1 (c) and (d) show time-height plots of the LTR peak spectral density and the
Doppler velocity at the vertical beam direction, respectively. LTR data from 02:00 LT on September 23 to 16:00 LT on September 24 were not recorded due to strong interference from another L-band wind-profiling radar. The LTR was also not operated from 04:00 LT to 08:00 LT on September 25 and between 22:00 LT on September 25 and 10:00 LT on September 26. In Figures 1 (c) and (d), both strong turbulence echo and large downward velocity were simultaneously detected below 6 km between 15:00
LT and 22:00 LT on September 26. Precipitation was also observed at ground-level during this period, as shown in Figure 1 (e). This period was excluded from the analysis because the turbulence echo was contaminated by strong echo from raindrops. Data from both the MU radar and LTR were available for estimations for the periods from
16:00 LT on September 24 to 04:00 LT on September 25 and from 08:00 LT to 22:00
LT on September 25. During the former period, substantial missing LTR data resulted from contamination due to scatterings from insects or birds. It was eventually decided to choose the period 08:30-22:00 LT on September 25 for the analysis, as during this period the radiosonde was launched every three hours, as shown by the red squares in the bottom of Figure 1 (c). The radar-derived q can be compared with the radiosonde result at 18:00 LT on September 25.

10
Virtual temperature profiles derived from the MU radar-RASS data were employed above 1.5 km height for estimating humidity. The precision and accuracy of virtual temperature obtained using the MU radar-RASS is described in the next section.
Because RASS echo obtained with LTR-RASS is only estimated below 0.3 km, for heights below 1.5 km it was decided to use T v derived from the time-interpolation of radiosonde results instead of LTR-RASS data.
Here, the quality of T v data observed with the MU radar-RASS during September
2002 is considered. The left-hand panel of Figure 2 shows the averaged discrepancies between T vR and T v from the radiosonde results ( T vS
) in the September campaign; the standard deviation (s.d.) is shown by the error bars. The bias error averaged for 16 comparisons was less than 0.3 K throughout the height interval. The cross-correlation coefficient calculated at each height was greater than the rejection level at the 95% confidence level between 1.5 km and 7.5 km. The right-hand panel shows a scatter plot of T vS versus T vR
. The slope of the least-squared fitting result, obtained using the
Ordinary Least-Squares Bisector (e.g., Babu and Feigelson , 1992), is 1.006, with an s.d. of 5.6 10

6
. These results confirm the excellent precision of T vR
.
The turbulence echo power derived from the MU radar and LTR was examined prior to combining the data. The left- and right-hand panel of Figure 3 show the median SNR profiles at 08:30 LT and 21:50 LT on September 25 for the MU radar (SNR
MUR_obs
) and

11
LTR (SNR
LTR_obs
), respectively. Note that SNR
MUR_obs
and SNR
LTR_obs
were multiplied by appropriate factors to adjust each maximum value to 0 dB. Aspect sensitivity was detected above 2.5 km in the MU radar results; however, the SNR
LTR_obs
does not show the aspect sensitivity because wind-profiling radar operated at the L-band frequency is insensitive to partial Fresnel reflection. Only four oblique beams were used in this analysis to remove the effect of partial reflection on the MU radar. It is also possible to determine the azimuth anisotropy in both SNR
MUR_obs
and SNR
LTR_obs
. SNR
MUR_obs
in the east beam direction is smaller than that of other oblique beams between 1.5 km and
3.0 km height; the difference becomes as large as 3.0 dB at 3.0 km. In contrast,
SNR
LTR_obs from the east beam direction is larger than that of other beam directions above 2.4 km, where the peak spectral density for LTR is very weak, as shown in Figure
1 (c). This azimuthal anisotropy for LTR is probably caused by a decrease in the accuracy of echo power intensity with decreased SNR
LTR_obs
. Although the full reason for the azimuth anisotropy is not yet fully understood, LTR and MU radar data in the east beam direction were removed by applying the median SNR of each of the four oblique beams at each height, as described in the following section. The right-hand panel of Figure 3 shows that SNR
MUR_obs
is invariant with height below 1.95 km, although SNR
LTR_obs
decreases with increasing height. Monostatic radar, such as the
MU radar, has reduced sensitivity at these low heights to prevent leakage of the strong transmitted signal to the receiver. A previous study that estimated humidity profiles using EAR removed this effect by multiplying the correction factor calculated from many comparisons between radiosonde- and radar-derived | M | below 3.0 km ( Furumoto et al.
, 2006a). A total of 184 comparisons were averaged to calculate the meaningful recovery factor; however, in the present study the number of comparisons between

12 radiosonde and radar-derived | M | is insufficient to calculate a significant estimate of the recovery factor. We therefore developed a new method for estimating the recovery factors without simultaneous radiosonde results by assuming that transmitted radio waves from the MU radar and LTR are scattered by the same turbulence layer, with a refractive structure function. Note that, from Hocking (1985), the scale of intertial subrange is between 3x10
-2
m and 10
2
m below the height of 10 km. Because the Bragg scale of the MU radar and LTR is 3.2 m and 0.12 m, respectively, the both radar receive the scatter from the scales within the intertial subrange. The recovery factor for the MU radar was determined by comparing the turbulence echo power intensity of the MU radar and LTR above 1.5 km, where the recovery effect does not occur for LTR. As the turbulence echo was obtained every few minutes for the MU radar and LTR, a meaningful estimation of the recovery factor is expected from the numbers of comparisons of the echo power between the MU radar and LTR. It is important to state, however, that the SNRs obtained for the MU radar and LTR are relative values because it is difficult to obtain the absolute echo power of a wind-profiling radar using an active phased array antenna such as the MU radar or LTR. It is possible to estimate the theoretical ratio of SNR for the MU radar (SNR
MUR_theo
) to SNR for LTR (SNR
MUR_theo
) using the system parameters listed in Fukao et al.
(1985) and Hashiguchi et al. (2004) and the equivalent noise temperatures of the MU radar and LTR. The equivalent noise temperature of LTR was determined to be 525 K, after Hashiguchi et al. (2004). This analysis uses the diurnal variation in the equivalent noise temperature of the MU radar as estimated by Maeda et al. (1999). The theoretical ratio of the SNR for the MU radar to SNR for LTR was then calculated.
The solid line in the left-hand panel of Figure 4 shows the profiles of (SNR
MUR_obs
/

13
SNR
LTR_obs
) normalized by the ratio (SNR
MUR_theo
/ SNR
LTR_theo
). To remove the effects of azimuth anisotropy on the MU radar and LTR, it was decided to exclude the minimum and maximum ratios and averaged the two middle ratios at each height. The normalized ratio increases with increasing height due to the reduction in the sensitivity of the MU radar receiver below 2.1 km. Although a 3 dB fluctuation in amplitude with increasing height was observed above 2.1 km, the overall structure of the ratio is invariant. The recovery effects for the MU radar appear to be negligible above 2.1 km height. The normalized ratio averaged over 2.1-2.25 km became 6 dB, as shown by the dashed line in the left-hand panel of Figure 4. This indicates that the echo power intensity for the MU radar is smaller than the theoretical value due to the aging deterioration of the MU radar. The right-hand panel in Figure 4 shows the correction factor for the recovery effect for the MU radar. The height variation of the correction factor for the recovery effect is calculated by subtracting the normalized ratio averaged for 2.1 km and 2.25 km from the normalized ratio below 1.95 km. This factor was multiplied by SNR
MUR_obs at each height to correct the recovery effects.
Here a parameter is selected to combine the MU radar and LTR data.
 and | M
*
| are apparently inappropriate as merging parameters because they contain the system parameters of each radar. q or | M | should therefore be chosen as the combining parameter. There are three possible combining methods:
1. Two q profiles are independently estimated with the MU radar and LTR, given the boundary values from other instrument results. Two individual q profiles are then merged to obtain the final q profile.

14
2. Humidity profiles are first estimated using only one set of radar data. The estimation result is employed as the boundary condition for the estimation using data from the other radar at the overlapping height of the two radars.
3. The | M | profiles of the MU radar and LTR are merged before estimating humidity profiles between 0.3 km and 7.5 km height.
In humidity estimates based on the MU radar or Equatorial Atmosphere Radar (EAR), humidity profiles were estimated by solving the integral equation downward from the upper boundary height of 7.5 km ( Furumoto et al.
, 2003; 2006a). Boundary conditions of q
0
and

0
were derived from the interpolation of six-hourly radiosonde results at the top boundary, assuming that rapid fluctuations of the boundary values during the interval of radiosonde launch were small at 7.5 km height. We cannot use the time-interpolation of radiosonde results as boundary values for the humidity estimation from LTR, as rapid humidity fluctuations are very large in the observation height range of LTR. Furumoto et al. (2005) used humidity data from a kytoon moored at the observation site to obtain the lower boundary values of the estimation using LTR data.
Unfortunately, the present campaign did not operate a kytoon or other instruments to obtain boundary values of q
0
and

0
for the LTR estimation with high temporal resolution. Therefore, it was decided to use the second or third methods described above to merge the two radar datasets.
The second method outlined above is similar to the third method. In the second method, however, the humidity profiles are estimated separately with the MU radar and
LTR by using the GPS-derived IWV (IWV
G
) calculated for the height ranges 0.3-1.5 km and 1.5-7.5 km, respectively, whereas the third method requires only one IWV
G
for heights of 0.3 km and 7.5 km, respectively. As shown in Appendix 1, IWV
G
is

15 determined by multiplying the time-interpolation of the ratio of IWV of the radar estimation range to the PWV calculated from radiosonde results ( R range
), assuming that rapid temporal variations in R range are negligible. We investigated the values of R range calculated for different height ranges from radiosonde data from about 1500 launches at the MU Observatory since 1986. The upper left, upper right, and lower left-hands panels of Figure 5 show the histograms of R range calculated for the height of 1.5-7.5 km
( R
1.5
7 .
5km
), 0.3-1.5 km ( R
0.3
1 .
5km
) and 0.3-7.5 km ( R
0.3
7 .
5km
), respectively. R
1.5
7 .
5km and
R
0.3
1 .
5km are widely distributed between 0% and 80%, with s.d. of 14% and 12%, respectively. Bilateral asymmetry is also evident in the histogram of R
1.5
7 .
5km and
R
0.3
1 .
5km
. The histogram for R
0.3
7 .
5km shows a symmetrical distribution, with a mean and s.d. of 91% and 3.3%, respectively. The rapid variation in R
0.3
7 .
5km during the interval of radiosonde launch appears to be smaller than that for R
0.3
1 .
5km and R
1.5
7 .
5km
. The third method was eventually selected, involving combining | M | from the MU radar and LTR and constraining the sign of the radar-derived | M | using IWV
G in the height range 0.3 km and 7.5 km, because the estimation error related to rapid temporal variations in R range appears to be smaller than the error associated with the second method.
Here, the method of combining | M | profiles from the MU radar (| M
MUR
|) and LTR
(| M
LTR
|) is determined. Although data from the two radars should ideally be combined at the height where the recovery effect in the MU radar data is absent, turbulence echo intensity with LTR was very small above 2.4 km, and the data quality was insufficient for estimating humidity. Therefore, | M | for both radars was merged between 1.5 km and
2.1 km height with linear weighting functions. In the upper left-hand panel of Figure 6, the weighting functions for | M
MUR
| and | M
LTR
| are indicated by solid and dot-dash lines,

16 respectively. Note that plotted values are median values for 30-minute intervals. The maximum value of the radiosonde-derived | M | (| M sonde
|) at 14:44 LT at 2.0 km decreased over time until 17:48 LT. A large variation with height is evident in | M sonde
| below 2.1 km, demonstrating that rapid variations in | M | are large in the atmospheric boundary layer. | M
LTR
| below 2.1 km is generally in agreement with | M sonde
|, however, the peak height of | M
LTR
| at 17:48 LT is observed at a height that is 150 m lower than that of
| M sonde
|. Because the radiosonde transmitter is advected by horizontal wind, radiosonde and radar observed different horizontal sites. This difference appears to be due to the large spatial variation in humidity at the upper boundary layer. In the lower panels, the thin- and thick-solid lines show | M sonde
| and merged | M | (| M radar
|), respectively. The height structure of | M radar
| is generally in agreement with that of | M sonde
|, although the peak of | M radar
| is smoother than that of | M sonde
|.
Humidity profiles were estimated using | M radar
|. The sign of | M radar
| was constrained with IWV
G calculated from the GPS-derived PWV and the time-interpolation of radiosonde-derived R
0.3
7 .
5km
. As described in Section 5.2 no other measurements of humidity at the boundary height were undertaken at high temporal resolution. Therefore, time-interpolation of the radiosonde data at 7.5 km were used as the upper boundary values; the successive humidity profiles were retrieved by determining the sign of | M |, as described in Appendix 1. In this algorithm, six-hourly radiosonde results were used to construct initial profiles in the iteration process. Radiosonde data recorded at 15:00
LT and 21:00 LT on September 25 and 03:00 LT on September 26 were used for reference. The data from the radiosonde launched at 18:00 LT on September 25 were not used in the retrieval algorithm.

17
In the lower panel of Figure 7, the solid line shows R
0.3
7 .
5km derived from the time-interpolation of radiosonde results, while the squares indicate R
0.3
7 .
5km at the times of radiosonde observations. R
0.3
7 .
5km derived from radiosonde data varies between 0.69 and 0.83, although rapid fluctuations were not observed during the analysis period. The black solid and dot-dash lines in the upper panel show IWV
G and the lower and upper thresholds of IWV, respectively, for constraining the sign of | M radar
|. The squares in the upper panel represent IWV calculated from radiosonde results (IWV
S
). Radar-derived
IWV (thick-solid line) is almost within the threshold of IWV except for the period
08:30-09:20 LT. This demonstrates that the constraint of the sign of M radar using IWV
G worked well.
Humidity profiles were successfully estimated by merging the MU radar and LTR data. The middle panel in Figure 8 shows the q profile from the radiosonde ( q
S
) launched at 17:48 LT on September 25 (solid line) and the radar-derived q ( q
R
) (dotted line) averaged over 30-minute intervals around the launch time of the radiosonde observations. The dot-dash line shows the saturated q profile. The general structures of q
S and q
R at 17:48 LT are in good agreement, indicating that humidity is very small above 2.4 km and sharply increases with decreasing height around the top of the boundary layer. The difference between q
S and q
R is plotted in the left-hand panel; the maximum difference is 2.3 gkg
-1 at a height of 1.95 km. In the right-hand panel, the radiosonde- and radar-derived M are shown by solid and dotted lines, respectively. A negative discrepancy of q
R at 1.05-1.35 km is evident. This occurs because the sign of radar-derived | M | is incorrectly determined at 1.35 km. This discrepancy is probably due to spatial differences between the radar and radiosonde observations. Note that IWV
S at
17:48 LT is larger than the upper threshold of IWV
G
. It is possible that the large

18 discrepancy between IWV
S
and IWV
G
could have caused the negative discrepancy between q
S and q
R at 1.05-1.35 km.
Figure 9 shows time-height variations in q
S
(upper) and q
R
(lower) between the heights 0.3 km and 2.5 km. A 30-minute running mean was used for q
R
. The overall structure of q
R is in agreement with q
S
, indicating a progressive decrease in q between
15:00 LT and 19:00 LT below 2.0 km height. q
R was used to interpret detailed fluctuations that are unclear in q
S
. In the following section, the detailed time-height structure of q
R is compared with the simultaneous Raman lidar results obtained during the November campaign period.
Here, the time-height structure of q
R is compared with q using Raman lidar observations ( q
L
). The observational parameters of the MU radar and LTR were the same as those in the September 2002 campaign. The beam directions of the MU radar-RASS observation were changed according to the real-time ray-tracing of acoustic wave-fronts. During the observation period, radiosondes were launched every six hours, all of which were used as the reference to make initial profiles of the retrieval algorithm.
Figure 10 shows time-height variations in q
S,
q
R
, and q
L at 21:00 LT on November 20 and 07:25 LT on November 21, 2002. The missing data in the Raman lidar results for heights above 4 km during the period after 01:00 LT on November 20 are caused by cloud coverage over the radar site. Lidar observation data are also missing after 06:00
LT due to the strong solar light. The overall structures of q
S,
q
R
, and q
L show good agreement. The detailed fluctuations of q
R are in good agreement with the lidar result, which are not clear in the six-hourly radiosonde observation results. In particular, the time evolution of q
R
at each height seems to follow q
L
excellently. However, the time

19 variation of height structure of the humidity seems to be slightly different between q
R and q
L
. For example, ascending structure between 1.5 and 2.0 km at 21:00 LT to 00:00
LT and descending structure at around 2.5 – 3.5 km altitude between 04:00 and 06:00
LT observed by lidar are less clear in the radar observation. This discrepancy appears to be partly due to horizontal advection of the radiosondes, as radiosondes launched from the surface are steered by background winds. A second reason is that the substantial range resolution of the radar observation is greater than the nominal values because the neighboring sampling volumes overlap each other in the radar observation.
Nevertheless, radar observation has an advantage in daytime estimations and humidity monitoring within clouds; in such conditions, Raman lidar observations are difficult to perform. In the present study we have determined the sign of M using the empirically determined threshold. The threshold derived from larger data-set will be useful for more precise estimation. Humidity profiles were estimated between 0.3 km and 7.5 km height by merging | M | from the MU radar and LTR with linear weighting functions. In a separated study, the variational assimilation method was applied for the estimation of humidity profiles with the wind-profiling radar ( Furumoto et al.
2006b). As an advanced work of this field, the method has potential for finding more sophisticated weighting to merge two radar dataset by considering their observation errors.
Humidity profiles are estimated within the height range 0.3 km to 7.5 km by combining data from the MU (middle and upper atmosphere) radar and LTR (lower troposphere radar) to expand the height range of humidity estimates. Humidity profiles were estimated from co-locating simultaneous observations by the MU radar and LTR.
The observations carried out from the Shigaraki MU Observatory during September and

20
November 2002 were then outlined, followed by selection of the analysis period on the basis of background conditions. We compared the turbulence echo power observed by the MU radar and LTR; this comparison showed that the receiver sensitivity of the MU radar is reduced due to the leakage of transmitted power. The turbulence echo power from the MU radar below 2.1 km was corrected on the basis of the ratio of the SNR of the two radars. It was decided to merge the | M | profiles of both radars at 1.5 km to 1.95 km height using linear weighting functions with height; humidity profiles were estimated from the merged M using the retrieval method. The estimation result showed detailed temporal variations in humidity which cannot be revealed by radiosondes launched at six- or three-hour intervals. Successive humidity profiles were compared with equivalent results derived from Raman lidar. Several detailed structures of humidity that were observed with the MU radar and LTR were not apparent from radiosonde results but were observed from Raman lidar results. In addition, the radar observations showed an advantage during daytime estimations or humidity monitoring during cloudy conditions; observations with Raman lidar are limited under such conditions.
The authors deeply appreciated Andreas Behrendt of Hohenheim University and
Michitaka Onishi of Osaka Gas corporation for their help to the lidar observations at the
MU radar observatory and their kind guidance for the data analysis of the lidar data. The authors also wish to thank Hiroyuki Hashiguchi of Kyoto University for his helpful guidance and suggestions concerning the LTR system and LTR data. The authors thank
Yoshinori Shoji of Meteorological Research Institute (MRI) and Yuichi Aoyama of the

21
National Institute of Polar Research (NIPR) for their careful help with analyzing the
GPS data. Thanks are also due to Satoshi Iwai for helping the data processing of the
MU radar and LTR data. The present study was partially supported by a Grant-in-Aid for scientific research on Priority Area-764 (A04) , a Grant-in-Aid for scientific research
(B) 18340140 and a Grant-in-Aid for young scientist (B) 17740295 of the Ministry of
Education, Culture, Sports, Science, and Technology (MEXT) of Japan.
Here, the outline of the method to determine the sign of the vertical gradient of the refractive index ( M ) by Furumoto et al.
(2003) is shown. Because only the absolute value of M (| M |) is derived from radar observation, it is necessary to determine the sign of M for estimating specific humidity ( q ) profiles. Furumoto et al. (2003) developed an algorithm to constrain the sign of M from precipitable water vapor (PWV) obtained from GPS observations and the continuity of temporal variations in humidity. The algorithm consists of the following steps:
1. First, initial q profiles are made by interpolating 6-hourly radiosonde observations.
The signs of M profiles are determined to ensure that the radar-derived q profile is consistent with the reference.
2. As the initial q profiles are inferred from interpolated radiosonde results, they may not reflect short-term variations in humidity profiles. GPS-observed PWV (PWV
G
), whose temporal resolution is comparable to radar data, is applied to modify the sign of
M . The water vapor integrated over the radar estimation range (IWV
R
) is calculated, and compare it with the GPS results (IWV
G
). Since IWV
G
cannot be derived directly from
GPS measurements, IWV
G
is calculated from PWV
G
as follows:

22
IWV

PWV

R range
, (2) where R range is the ratio of IWV of the radar estimation range to the PWV calculated from radiosonde results. R range is interpolated to the temporal resolution of GPS observations, assuming that rapid temporal variations in R range are minor in scale.
Furumoto et al. (2003; 2005) estimated humidity profiles using the MU radar and LTR at 1.5-7.5 km and 0.2-2.2 km height, respectively, over a period when short-term variations in R range appeared to be small; however, during strongly disturbed conditions, such short-term variations may not be negligible. When the estimation range is expanded, R range becomes larger, and the error involved in estimating IWV
G that is related to short-term variations in R range decreases. When |IWV
R
IWV
G
| exceeds the threshold, the sign of M is modified to minimize the difference. This threshold was determined empirically from many trials using various threshold values. In estimating humidity profiles with the MU radar, we used the threshold of 3

IWV
, where

IWV is the standard deviation of IWV
G for a 30-minute period ( Furumoto et al.
, 2003).
3. During Stage 2 above, q profiles are inferred without considering the continuity of q values over time. We calculate temporal variations in q at each height, and the sign of M is changed when it exceeds the threshold. The test for time continuity is started at the upper or lower boundary where integration begins, and proceed along the direction of integration. When the sign is changed, |IWV
R
IWV
G
| can increase; if it exceeds the threshold defined in (2), the constraint procedure is applied again. Note that the sign of
M is not changed at the height where time continuity has already been checked; this ensures that time continuity is preserved.

A
F g
IWV
IWV
G
IWV
R
IWV
S
M
M
*
| M
LTR
|
| M
MUR
|
| M radar
|
| M sonde
|
N
2 p q
L q
R q
S
R range
PWV
PWV
G q q
0
Proportionality constant between observed and theoretical echo power
Volume-filling factor within the radar sampling volume
Gravitational acceleration
Integrated water vapor
GPS-derived IWV
Radar-derived IWV
Radiosonde-derived IWV
Refractive index gradient
Relative value of M
Absolute value of M observed by LTR
Absolute value of M observed by MU radar
Absolute value of M merged from MU radar and LTR
Absolute value of radiosonde-derived M
Brunt-Väisälä frequency squared
Atmospheric pressure
Precipitable water vapor
GPS-observed PWV
Specific humidity q at the boundary height
Raman lidar-derived q
Radar-derived q
Radiosonde-derived q
The ratio of IWV to PWV
23

z z
0

 turb
T
T v
T vR
T vS
R
0 .
3 1 .
5kk
R range calculated over the height interval 0.3 km to 1.5 km
R
0.3
7 .
5km
R range calculated over the height interval 0.3 km to 7.5 km
R
1 .
5 7 .
5km
R range calculated over the height interval 1.5 km to 7.5 km
SNR Signal-to-noise ratio
SNR
MUR_obs
SNR observed by MU radar
SNR
LTR_obs
SNR observed by LTR
SNR
MU_theo
Theoretical SNR by MU radar
SNR
LTR_theo
Theoretical SNR by LTR
Absolute temperature
Virtual temperature
T v derived by the MU radar-RASS
Radiosonde-derived T v
Height (km)
Boundary height
Volume refractivity
Volume refractivity by turbulence
24

25
Babu, G. J. and E. D. Feigelson 1992: Analytical and Monte Carlo comparisons of six different liner squares fits, Communications in Statistics – Simulation and Computation ,
21 , 533-549.
Basley, B. B. and K. S. Gage 1980: The MST radar technique: Potential for middle atmosphere studies, Pure Appl. Geophys.
, 118 , 452-493.
Behrendt, A., T. Nakamura, T. Tsuda 2004: Combined Raman lidar for the measurement of atmospheric temperature, water vapor, particle extinction coefficient, and particle backscatter coefficient: System upgrades, Appl. Optics , 43 (14), 2930-2939.
Bevis, M., S. Businger, S. Chiswell, T. A. Herring, R. A. Anthes, C. Rocken, and R.
H. Ware 1994: GPS Meteorology - Mapping Zenith Wet Delays onto Precipitable Water,
J. Appl. Meteor.
, 33 , 379-386.
Flores, A., G. Ruffini, and A.Rius 2000: 4D tropospheric tomography using GPS slant wet delays, Annales Geophysicae , 18 , 223-234.
Fukao, S., M. D. Yamanaka, N. Ao, W. K. Hocking, T. Sato, M. Yamamoto, T.
Nakamura, T. Tsuda, and S. Kato 1994: Seasonal variability of vertical eddy diffusivity in the middle atmosphere, Part I: Three-year observations by the MU radar, J. Geophys.
Res.
, 99 , 18973-18987.
Fukao, S., T. Sato, T. Tsuda, S. Kato, K. Wakasugi, T. Makihira 1985: The MU radar with an active phased array system 1. Antenna and power amplifiers, Radio Sci.
, 20,
1155-1168.
Furumoto, J. and T. Tsuda 2001: Characteristics of energy dissipation rate and effect of humidity on turbulence echo power revealed by MU radar-RASS measurements, J.
Atmos. Solar-Terr. Phys.
, 63 , 285-294.

26
Furumoto, J. and T. Tsuda 2001: Characteristics of energy dissipation rate and effect of humidity on turbulence echo power revealed by MU radar-RASS measurements, J.
Atmos. Solar-Terr. Phys.
, 63 , 285-294.
Furumoto, J., K. Kurimoto, and T. Tsuda 2003: Continuous Observations of
Humidity Profiles with the MU Radar-RASS Combined with GPS and Radiosonde
Measurements, J. Atmos. Oceanic Technol.
, 20 , 23-41.
Furumoto, J., S. Iwai, H. Fujii, T. Tsuda, W. Xin, T. Koike, and L. Bian 2005:
Estimation of Humidity Profiles with the L-Band Boundary Layer Radar-RASS
Measurements, J. Meteor. Soc. Japan , 83 , 895-908.
Furumoto, J., T. Tsuda, S. Iwai, and T. Kozu, 2006a: Continuous humidity monitoring in a tropical region with the equatorial atmosphere radar (EAR), J. Atmos.
Oceanic Technol., 23 , 538-551.
Furumoto, J., S. Imura, T. Tsuda, H. Seko, T. Tsuyuki, and K. Kato, 2006b: The
Variational Assimilation Method for the Retrieval of Humidity Profiles with the
Wind-profiling Radar, J. Atmos. Ocean. Tech.
, to be submitted.
Gage, K. S., J. L. Green, and T. E. VanZandt 1980: Use of Doppler radar for the measurement of atmospheric turbulence parameters from the intensity of clear-air echoes, Radio Sci.
, 15 , 407-416.
Gossard, E. E., S. Gutman, B. B. Stankov, and D. E. Wolfe 1999: Profiles of radio refractive index and humidity derived from radar wind profilers and the Global
Positioning System, Radio Sci.
, 34 , 371-383.
Grossmann, B. E., U. N.Singh, L. J. Cotnoir, T. D. Wilkerson, N. S. Higdon, and E.
V. Browell, 1987: Raman-shifted dye laser for water vapor DIAL measurements, Appl.
Opt.
, 26 , 1617-1621.

27
Hashiguchi, H., S. Fukao, Y. Moritani, T. Wakayama, and S. Watanabe, 2004: A
Lower Troposphere Radar: 1.3-GHz Active Phased-Array Type Wind Profiler with
RASS, J. Meteor. Soc. Japan , 82 , 915-931.
Hocking, W. K.: Measurement of turbulent energy dissipation rates in the middle atmosphere by radar techniques 1985: A review, Radio Sci.
, 20 , 1403-1422.
Hocking, W. K. and P. K. L. Mu 1997: Upper and middle tropospheric kinetic energy dissipation rates from measurements of C n
2
- review of theories, in-situ investigations, and experimental studies using the Buckland Park atmospheric radar in Australia, J.
Atmos. Terr. Phys.
, 59 , 1779-1803,.
Maeda, K., H. Alvarez, J. Aparici, J. May, and P. Reich 1999: A 45-MHz continuum survey of the northern hemisphere, Astron. Astrophys. Suppl. Ser.
, 140 , 145-154.
Masuda, Y. 1988: Influence of wind and temperature on the height limit of a radio acoustic sounding system, Radio Sci.
, 23 , 647-654.
Nastrom, G. D. 1997: Doppler radar spectral width broadening due to beamwidth and wind shear, Annales Geophysicae , 15 , 786-796.
Noguchi, W., T. Yoshihara, T. Tsuda, and K. Hirahara 2004: Time-height distribution of water vapor derived by moving cell tomography during Tsukuba GPS campaigns, J.
Meteor. Soc. Japan , 82 , 561-568.
Ottersten, H. 1969: Mean vertical gradient of potential refractive index in turbulent mixing and radar detection of CAT, Radio Sci.
, 4 , 1247-1249.
Skolnik, M. I. 1990: Introduction to Radar Systems , McGraw-Hill.
Stankov, B. B., B. E. Martner, and M. K. Politovich 1995: Moisture Profiling of the
Cloudy Winter Atmosphere Using Combined Remote Sensors, J. Atmos. Ocean.
Technol.
, 12 , 488-510.

28
Stankov, B. B, E. R. Westwater, and E. E. Gossard 1996: Use of wind profiler estimates of significant moisture gradients to improve humidity profile retrieval, J.
Atmos. Ocean. Technol.
, 13 , 1285-1290.
Stankov, B. B., E. E. Gossard, B. L. Weber, R. J. Lataitis, A. B. White, D.E. Wolfe,
D.C. Welsh, and Strauch. R.G. 2003: Humidity gradient profiles from wind profiling radars using the NOAA/ETL advanced signal processing system (SPS), J. Atmos.
Ocean. Technol.
, 20 , 3-22.
Tsuda, T., M. Miyamoto, J. Furumoto 2001: Estimation of a humidity profile using turbulence echo characteristics, J. Atmos. Oceanic Technol.
, 18 , 1214-1222.
Tsuda, T., T. E. VanZandt, and H. Saito 1997: Zenith-angle dependence of VHF secular reflection echoes in the lower atmosphere, J. Atmos. Terr. Phys.
, 59 , 761-775.
VanZandt, T. E., J. L. Green, K. S. Gage, and W. L. Clark 1978: Vertical profiles of refractivity turbulence structure constant: Comparison of observations by the Sunset
Radar with a new theoretical model, Radio Sci.
, 13 , 819-829.
Ware, R., R. Carpenter, J. Guldner, J. Liljegren, T. Nehrkorn, F. Solheim, and F. Vandenberghe 2003: A multi-channel radiometric profiler of temperature, humidity, and cloud liquid, Radio Sci.
, 38 , 1-13.
Warnock, J. M. and T. E. VanZandt 1985: A Statistical Model to Estimate the
Refractivity Turbulence Structure Constant C n
2 in the Free Atmosphere , NOAA Tech.
Rep., Boulder, CO., U.S.A.
Weinstock, J. 1987: Vertical turbulent diffusion in a stably stratified fluid, J.
Atmos.
Sci.
, 35 , 1022-1027.
Whiteman, D. N., S. H. Melfi, and R. A. Ferrare 1992: Raman lidar system for the measurement of water vapor and aerosols in the Earth's atmosphere, Appl. Opt.
, 31 ,

3068-3082.
29

30
Figure 1: Figure (a) shows a time-height plot of peak spectral density averaged for four oblique beams of the MU radar. Figures (b), (c), and (d) show the virtual temperature obtained with the MU radar-RASS and the peak spectral density and Doppler velocity monitored at the vertical beam with LTR, respectively. The red squares below figure (c) indicate the timing of the radiosonde launch. The precipitation rate is plotted in figure
(e).
Figure 2: The left-hand panel shows the mean discrepancy profile between virtual temperature determined from MU radar-RASS ( T vR
) and the equivalent radiosonde result ( T vS
). The error bars show the s.d. of the difference between T vR and T vS
. A total of 16 comparisons were employed to calculate the mean discrepancy and s.d. profiles.
The solid line in the middle panel shows profiles of the cross-correlation coefficient at zero lag between T vR and T vS
. The gray shaded area shows the rejection region at the
95% confidence level. The right-hand panel shows a scatter plot of T vR versus T vS
. The solid line shows the least-squared fitting result to the linear function using the Ordinary
Least-Squares Bisector.
Figure 3: The profiles of SNR for MU radar and LTR are shown in the left- and right-hand panels, respectively. The thick-solid, thin-solid, dashed, dot-dash, and dot-dot-dash lines show the median profiles at 08:30 LT and 21:50 LT on September 25 for the vertical, north, east, south, and west beam directions, respectively. The error bars indicate the standard error of SNR.

31
Figure 4: The solid line in the left-hand panel shows the profile of the ratio of observed SNR
MUR to observed SNR
LTR
. Note that the median values of SNR
MUR
and
SNR
LTR
for four oblique beams were used in calculating the ratio. To normalize the observed ratio, it was divided by the theoretical ratio of the SNR for MU radar to that for LTR (SNR
MUR_theo
/ SNR
LTR_theo
). The dashed line shows the averaged ratio between
2.1 km and 2.25 km height. The right-hand panel shows a profile of the correction factor for the recovery effect for the MU radar.
Figure 5: Histograms of R range calculated from radiosonde data of approximately 1500 launches from the MU Observatory. The upper left, upper right, and lower left panels show histograms of R range calculated for height ranges of 1.5-7.5 km, 0.3-1.5 km, and
0.3-7.5 km, respectively.
Figure 6: The solid and dot-dash lines in the upper left-hand panel show the weighting functions used to combine the | M | profiles of the MU radar (| M
MUR
|) and LTR
(| M
LTR
|), respectively. Thick-solid, dot-dash, and thin-solid lines in another upper panel show profiles of | M
MUR
|, | M
LTR
|, and | M | profiles derived from the results of radiosondes
(| M sonde
|) launched at 14:44 LT, 17:48 LT, and 20:42 LT, respectively, on September 25.
The lower panel shows | M | profiles after combining | M
MUR
| and | M
LTR
| using the weighting functions shown in the upper left panel.
Figure 7: The upper panel shows temporal variations in IWV
R
(thick-solid line) and
IWV
G
(thin-solid line) for the height range 0.3-7.5 km. The squares represent IWV calculated from radiosonde results. Note that a 30-minute running mean for used for
IWV
R
. The dot-dash lines show the lower and upper thresholds for IWV
G in the retrieval algorithm described in Appendix 1. In the lower panel, the black solid line shows R range

32 at 0.3-7.5 km height, as calculated from the time-interpolation of radiosonde results.
The squares indicate R range at the times of radiosonde launches.
Figure 8: The middle and right-hand panels show profiles of radar-derived q and M , respectively, together with simultaneously recorded radiosonde results. The data from the radiosonde launched at 17:48 LT on September 25, 2002, are shown by solid lines.
Dotted lines represent radar results averaged over 30-minute intervals around the timings of radiosonde launches. The dot-dash line in the middle panel shows q at the saturation. The difference between the radar- and radiosonde-derived q is plotted in the left-hand panel.
Figure 9: Time-height variations in radiosonde-derived q and radar-derived q are shown in the upper and lower panels, respectively. The red squares indicate the launch times of the radiosondes used in the humidity estimation, while red triangles represent the launch times of radiosondes that were not used in the estimates.
Figure 10: The upper, middle, and lower panels show time-height variations in q
S
, q
R
, and q
L
, respectively, at 21:00 LT on November 20 and 07:25 LT on November 21, 2002.
The red squares indicate the timings of radiosonde launches.

33
Observation parameters of the MU radar-RASS during September and November 2002.
The suitable beam directions for RASS measurements were determined from real-time ray-tracing of acoustic wave-fronts.
Turbulence echo observation parameters
Number of beams
Beam direction
(azimuth o , zenith o )
5
(0 o
,0 o
), (0 o
,10 o
), (90 o
,10 o
),
(180 o ,10 o ), (270 o ,10 o )
Observation mode
Actual elevation (km)
IPP(
 s)
Sub-pulse length (
 s)
Coherent integration
FFT points
Incoherent integration
Pulse compression (bits)
Low mode
1.20-10.65
400
RASS observation parameter
1
60
128
2
2
Observation mode
Actual elevation (km)
IPP(
 s)
Sub-pulse length (
 s)
Number of beams
Coherent integration
FFT point mark
Incoherent integration
Pulse compression (bits)
MU radar low mode
1.20-10.65
410
1
4
16
256
5
2 high mode
4.80-23.85
400
1
60
128
2
16 high mode
4.80-23.85
434
1
4
16
256
5
16
Sound source
Sound source equipment
Waveform
Repetition cycle (s)
Chirped frequency (Hz) loudspeaker
CW
1.0
72.3-109.2

34
Observation parameters of the LTR-RASS system in September and November, 2002
Turbulence echo observation parameters
Number of beams
Beam direction
(azimuth o
, zenith o
)
5
(0 o ,0 o ), (0 o ,9.8
o ), (90 o ,9.8
o ),
(180 o
,9.8
o
), (270 o
,9.8
o
)
Actual elevation (km)
IPP(
 s)
Sub-pulse length (
 s)
Coherent integration
FFT points
Incoherent integration
RASS observation parameter
Actual elevation (km)
IPP(
 s)
Sub-pulse length (
 s)
Number of beams
Coherent integration
FFT point
Incoherent integration
Pulse compression (bits)
MU radar
0.29-9.74
100.50310
1
32
128
1
0.29-4.94
50.250155
1
1
48
256
5
1
Sound source
Sound source equipment
Waveform
Chirped frequency (Hz) loudspeaker
CW
2886-3159