next up gif contents
Next: References Up: Analysis of Volume Imaging Previous: Summary of Wind
Return to the Publications. Return to the Index.


Data from the Volume Imaging Lidar coupled with advanced signal analysis methods provide direct mesoscale observations of cloud coverage, cloud base altitude, cloud top altitude, convective boundary layer mean depth, and vertical profiles of horizontal mean wind. The measurements are area-averaged over 25--70 km with a 3-minute to 1-hour time resolution. The statistical stability of mean values is generally much better than can be provided by point measuring or line averaging instruments.

All VIL data recorded during 1987 and 1989 FIFE programs are analyzed, providing information about boundary layer parameters under various atmospheric conditions. The vertical resolution of the results is defined by the elevation angle step size and is typically better than 45 m, while in the worst case it is 130 m. In the FIFE program, the scanning pattern was selected such that wind measurements can be calculated up to 2 km altitude and all other boundary layer measurements can be made up to 3.5 km altitude.

The automatically determined cloud bases, cloud tops, and CBL mean depths are compared with results obtained by a manual inspection of the VIL RHI-scans. The comparisons indicate very good consistency between automatic and manual results. The differences between the automatic and carefully examined manual estimates are within 50-100 meters, which are similar to the vertical resolution of the VIL volume scans. This small differences can be explained by statistical fluctuations due to smaller sampling coverage of the manual methods. The automatic convective boundary layer mean depth estimate is valid if the coverage of boundary layer clouds is less than 10%. If more clouds are present, the maximum aerosol backscatter variance represents the altitude of maximum cloud echo variance, which is usually higher than the CBL mean depth. The cloud top estimates are usually too low in overcast conditions due to lidar signal extinction in optically thick clouds and masking by low clouds. When a 15-minute maximum filter is applied to the cloud top estimates, there often are enough lines of clear sight through the cloud layer for reliable cloud top determination in up to 70% cloud coverage.

The CBL mean depths are also compared with radiosonde-based potential temperature measurements. This comparison indicates that the conventional point-measurements of the CBL depth may vary more than 100%, while the variations of the VIL measurements are on the order of 10%.

The presence of boundary layer clouds disturbs both CBL mean depth and cloud top estimation. The clouds may partly block the VIL backscatter signal; thus, the CBL mean depth may be calculated from an incomplete data set. Although most cloud echoes are removed when calculating the profile of horizontal backscatter variance, remains of the clouds may still disturb the detection of the lowest variance maximum, leading to errors in the CBL mean depth estimate. The cloud tops may not be detected when clouds cover a large percentage of the sky, since the lidar profile does not usually penetrate more than 200 m into the cloud. Nevertheless, the volume scanning enables the VIL to look through holes in the cloud base and detect tops of much thicker clouds than a conventional, vertically pointing lidar.

The VIL cloud mapping capability is demonstrated by comparing the horizontal cloud maps produced from the VIL data with satellite imagery. Consistency among the cloud images is evident. The three-dimensional cloud maps of the VIL provide good spatial and temporal resolution measurements.

This study provides enhancements to the original wind profiling algorithm [45]. A method is developed to objectively determine the reliability of each mean wind measurement. The VIL wind profiles from the 1989 FIFE data indicate that, by the most restrictive reliability definition (i.e. =1), 76% of hourly averaged wind estimates in the convective boundary layer are reliable.

The wind profiles correlate well with traditional wind measurements made with an aircraft-based instrument, radiosondes, and surface weather stations. The comparisons do not indicate any systematic errors in the VIL wind profiles. The VIL wind profiles are more stable than the traditional measurements, since both time and area averaging are performed, while traditional measurements are either time-averaged or line-averaged.

The deviation between VIL wind profiles and traditional measurements is dominated by the natural wind fluctuations in the traditional measurements. This makes it difficult to determine the error in the VIL measurements by comparing them to measurements made by other instruments. Therefore, one approach to estimate the errors in the VIL profiles is to use the internal consistency of the results. This is performed by comparing the wind estimates with the averages of the vertically adjacent estimates. The local root-mean-square errors of the hourly averaged wind profiles which have passed the reliability analysis are 0.03--0.1 ms in speed and 0.4--1 in direction. Since local deviations are partly due to real wind fluctuations, they may provide conservative error estimates.

The error in finding the peak of the cross correlation function in the presence of noise provides another estimate of the RMS errors in the wind results. These errors may be underestimated for two reasons. First, the noise variance of the correlation function is difficult to estimate, since the histogram equalization induces non-linearity in the noise statistics. Second, although comparisons with other instruments do not indicate any systematic errors in the mean results, it is possibille that the maximum peak of a cross correlation function does not always correspond to the real mean wind, causing errors in individual wind estimates.

Both error analyses indicate that the RMS errors of mean wind in the CBL are on the order of 0.1 ms in speed and 1 in direction, which are very small with respect to typical horizontal mean winds. The errors are too large to use a similar technique for determining vertical mean winds, which typically are on the order of 0.01 ms in the VIL scanning area. However, the VIL data can be used for wind divergence measurements, from which the vertical mean winds can be determined indirectly [46].

The 1987 and 1989 FIFE programs produced about 25 gigabytes of raw data consisting of about 125 hours of boundary layer volume scans. The data processing for this study is performed by an IBM Risc System 6000 Model 550 workstation with 512 MB main memory, 7 GB hard disk storage, and 26 MFLOPS (LINPACK double precision performance test [68] ) computing power. About 250 CPU-hours were dedicated for the wind profiling calculations and about 60 CPU-hours for the other boundary layer measurements.

The VIL measurements of the CBL from the FIFE program provide an excellent data source for atmospheric boundary layer studies. These measurements combined with radiosonde, satellite, aircraft, and surface measurements from the experiment provide a link between area-averaged and local-scale measurements.

next up gif contents
Next: References Up: Analysis of Volume Imaging Previous: Summary of Wind
Return to the Publications. Return to the Index.

Antti Piironen
Tue Mar 26 20:53:05 CST 1996