1) Daytime random error characterization

The dryline case of May 22 shown in Fig. 5 has been used to characterize the random errors in the SRL water vapor mixing ratio data. For photon counting data, errors can be calculated assuming Poisson statistics using Eq. (6) from Part I, to be referred to as the water vapor error equation. However, as discussed in Part I, the water vapor mixing ratio measurements in general use a combination of photon-counting and analog measurements. In general, calculating statistics on a single profile of analog data requires that the square of the signal for each laser shot that goes into a summed profile be maintained (Whiteman et al. 1992). That information is not maintained in the current data acquisition electronics so another approach to determining errors in the analog data is required. The method used here is to first convert the analog signal to a virtual count rate scale using the glue coefficients determined through a regression analysis. The virtual count rate corresponding to the analog signal is used for the S terms in the water vapor error equation, and the background determined from the photon-counting data is used for the B terms. The implicit assumption is that the analog data converted to a virtual count rate scale behave according to Poisson statistics.

This method of determining the errors has been tested by comparing the results of the water vapor error equation with errors determined using spectral analysis techniques where the noise floor in a Fourier power spectrum (Senff et al. 1994; Linné et al. 2000) is determined as a function of height. If a portion of data is used when the atmosphere is stable—that is, where the real atmospheric variation is significantly less than variations introduced by counting statistics—then the noise determined by this Fourier technique can be used to quantify the instrument noise floor. The upper-right panel of Fig. 5 shows the comparison of the Poisson and spectrally determined random errors for the last 50 min of the measurement period (denoted by the horizontal brace on the time axis of the water vapor image). To improve the statistics of this comparison, the data have been used at their raw resolution (1-min temporal and 30-m spatial) instead of the smoothed resolution displayed in the image. In the final 50-min segment of the water vapor mixing ratio image, analog data are used for the water vapor signal between 0.3 and approximately 2.0 km and for the nitrogen signal from 0.3 to 5.5 km. Therefore, the water vapor mixing ratio is calculated using exclusively analog data below approximately 2.0 km for this 50-min segment. Above 2.0 km, the mixing ratio is determined using photon-counting data for the water vapor and analog data for the nitrogen. The plot in Fig. 5 shows that the two methods of determining the variance in the signal agree well except for two altitude ranges between 1.2–1.6 and 2.6–3.2 km, both indicated by horizontal arrows, where the spectrally determined variance exceeds that determined by Poisson statistics. The lower-altitude range corresponds to the top of the nocturnal boundary layer (confirmed by potential temperature analysis from radiosonde) while the upper-altitude range indicates the location of the residual layer. The increased atmospheric variability at these locations leads to larger variance in the spectral quantification of errors since the Fourier technique is quantifying both real atmospheric variation as well as variation due to the counting statistics. Therefore, this example illustrates that the Fourier and Poisson techniques for calculating errors agree well except in regions of increased atmospheric variability, thus supporting the assumption that the technique of calculating errors from the converted analog data assuming Poisson statistics is justified. It also demonstrates that this error comparison technique can be used to discern transition regions in the atmosphere.

The analysis of Fig. 5 validates the use of Poisson statistics to determine the random component of the errors in the water vapor mixing ratio calculation. Figure 6 now presents a comparison of how the errors determined using Poisson statistics varied during the 22–23 May 2002 dataset. Using the smoothed resolution presented in Fig. 5, the random error was calculated at three times in the dataset: 21.1, 23.5, 26.4 UTC. (The latter time indicates 2.4 UTC on 23 May 2002). The first two of these measurements were in bright daytime conditions while the last was in full darkness. These random error quantifications along with the boundary layer heights observed in Fig. 5 indicate that under all conditions, the random error in the mixing ratio measurement remains below 10% throughout the boundary layer at 2-min temporal and 60–200-m spatial resolution. During the daytime, the random errors increase steeply above the boundary layer where the water vapor content drops rapidly. However, under nighttime conditions, the random error does not exceed 10% below ∼6 km. This analysis also indicates that during the nighttime, SRL operations with 30-m spatial resolution and 15-s temporal resolution would possess less than 10% random error up to an altitude of 3 km. These high-resolution water vapor measurements permit boundary layer convective processes to be studied throughout the diurnal cycle as further described in Demoz et al. (2006).

1) Ice water content and particle radius retrievals

A newly developed Raman lidar technique for quantifying cirrus cloud ice water content (IWC) and generalized particle diameter (Dge) (Wang et al. 2004) makes use of simultaneous measurement of cirrus cloud scattering ratio and Raman scattering from ice. The technique was developed using measurements acquired at the U.S. Department of Energy Southern Great Plains Atmospheric Research Facility in northern Oklahoma where radar measurements were available for validation. This technique was used to retrieve IWC and particle size from these cirrus cloud measurements. The results are shown in Figs. 7c and 7d. These retrievals show that the region of intense scattering between the altitudes of 12 and 13 km and over the time interval of 0300 to 0500 UTC on 20 June (indicated as 27 and 29) is populated in general by small particles but high IWC consistent with our hypothesis of the evolution of the observed cirrus anvil. By contrast, later in the measurement period between 0800 and 1000 UTC, the retrieval of generally large particles in the 11–12-km altitude range and smaller particles above this is consistent with this being a region of cirrus precipitation.

2) Upper-troposphere humidification

To study the influence of the sublimating cirrus particles on upper-tropospheric humidity, the relative humidity with respect to ice (RH_Ice) was quantified at 2-h intervals during the time of the evolving cirrus cloud where temperature profiles from radiosonde were used to calculate RH_Ice from the lidar mixing ratio. The times at which RH_Ice was quantified were 0400, 0600, 0800, 1000 UTC and are indicated by the color-coded arrows in the image shown in Fig. 7b. The vertical profiles of RH_Ice corresponding to the times indicated by the arrows are shown in Fig. 7a. Careful study of the figure shows that subcloud RH_Ice values approximately double over the period of the measurements likely due to sublimation of the precipitating ice crystals from the cloud. Also observed is approximately a factor of 4 increase in the mean RH_Ice at the altitude of the lower scattering layer between 7 and 8 km.

3) Upper-tropospheric water vapor random error characteristics

Figure 7 demonstrates the upper-tropospheric water vapor measurement capability of the SRL during IHOP. The relative humidity profiles presented in Fig. 7 were analyzed using a routine that performs variable smoothing in both the spatial and temporal domain. The resulting vertical resolution of the data presented in Fig. 7 ranges from 60 m at 7 km to 600 m beyond 12 km. The procedure works as follows: the water vapor mixing ratio profiles are first vertically smoothed to the desired resolution, then the routine sums the number of profiles required to maintain the random error below a fixed value, chosen here to be 10%. At higher altitudes, more profiles are required to maintain a random error of 10% or less. A maximum number of 59 profiles, one acquired each minute, was specified for summing. This method of analyzing the data permits higher-resolution temporal features to be preserved in the lower altitudes of the profiles. On the left of Fig. 8, therefore, is shown the resulting random error as a function of altitude that is achieved using this routine. On the right is shown the number of profiles that have been included in the processed data shown in Fig. 7. Note that above some altitude, it was not possible to maintain less than 10% random error for the vertical resolution chosen. This altitude varies from ∼11.5 km between 0400 and 0600 UTC to 9.5 km at 1000 UTC. The reduction in this altitude at the later times is due to the increased attenuation of the laser beam by the lower scattering layer seen in Fig. 7 between 7 and 8 km.

4) Investigation of wave structure in lower scattering layer

The oscillations in the lower scattering layer seen in Fig. 7b suggest the possibility that energy from the bore event shown at the time/altitude of ∼0630 UTC/∼0.5 km in the water vapor image of Fig. 2 in Part I has propagated upward to ∼9 km and induced the oscillations seen in Fig. 7b. This possibility was studied by calculating the Scorer parameter, l²_s (Scorer 1949), which considers the balance between the atmospheric stability and wind shear as a function of altitude and can be used to identify regions of trapping for vertically propagating waves (Ralph et al. 1997; Shutts 1997). It is defined as

View Expanded

where z is the vertical coordinate, N is the Brunt–Väisälä frequency defined by

View Expanded

and U(z) is the component of the horizontal wind in the x direction. Disturbances can propagate vertically for l²_s > 0 and are trapped for l²_s < 0. The Scorer parameter calculated from radiosondes launched at 0602 and 0801 UTC (approximately 30 and 32 UTC in Fig. 7) from the Homestead site shown in Fig. 9a. The negative values of l² at approximately 3 km coupled with the near zero values of l² between 4 and 6 km do not support the hypothesis that upward-propagating energy from the bore thrust observed at 0630 UTC and 0.5 km in Fig. 2 of Part I was the source of energy for the oscillations observed in the lower scattering layer in Fig. 7. Furthermore, a careful examination of the image suggests that the oscillations in the lower cloud layer may have begun prior to the major bore thrust at 0630 UTC.

To investigate other possible explanations for the presence of waves in the lower scattering layer, consider the wind speed and direction data from the 0602 and 0801 UTC radiosondes shown inFig. 9b. Large directional shear is observed in the wind field of Fig. 9 at the base of the cirrus outflow layer at ∼11 km consistent with the Geostationary Operational Environmental Satellite (GOES) loop, which indicated that the outflow from the thunderstorm that produced the upper layer of cirrus clouds observed in Fig. 7 was generally from the north and then veered toward the east as the measurement period proceeded.

Within the generally westerly flow in the midtroposphere, there still was significant variation in wind direction. The wind veered continuously from approximately 160° to 300° between the altitudes of 5 and 7 km. This implies the possibility of directional shear in the wind field within this altitude range. Considering that the Scorer analysis indicates that waves can vertically propagate in the altitude range of 6–8 km and the wind analysis indicates the possibility of directional wind shear between the altitudes of 5 and 7 km, a possible explanation for the waves observed in the lower scattering layer between 7 and 9 km in Fig. 7 is that waves induced by directional wind shear have propagated vertically to the altitude of the lower scattering layer.

To further investigate the properties of the scattering layers observed in Fig. 7, the optical depth and layer mean extinction-to-backscatter ratio were calculated and are presented in Fig. 10. The optical depth of the upper cloud layer varies from a maximum of approximately 3, the rough upper limit of the SRL's ability to quantify optical depth, to approximately 1.0 between approximately 2700 and 3400 UTC (0300 and 1000 UTC on 20 June). The lower cloud layer possessed mean optical depth of approximately 0.5. Both of these values are quite consistent with cirrus clouds. However, the extinction to backscatter ratio, S, of the two layers is markedly different. The mean S value in the upper layer of ∼20 sr is quite consistent with cirrus cloud values that have been measured using Raman lidar systems previously (Reichardt et al. 2002; Whiteman et al. 2004). The mean value in the lower layer of ∼70–80 is quite atypical of cirrus clouds and more indicative of smoke or absorbing aerosol. The Moderate Resolution Imaging Spectroradiometer (MODIS) fire product (http://modis-fire.umd.edu/products.asp) indicated that numerous fires were present in New Mexico, Arizona, and California near the path of the 3-day back trajectories obtained from the National Oceanic and Atmospheric Administration (NOAA) Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (http://www.arl.noaa.gov/ready/hysplit4.html) analysis at 7, 8, and 9 km, the altitude region of the lower scattering layer seen in Fig. 7b. Therefore, we take the lower cloud layer to likely consist of hygroscopic smoke particles that have been transported from fires to the west, which have served as seeds for ice particle growth.

There is also some indication in the lidar depolarization data that very light precipitation from the cirrus cloud may have helped to seed this lower scattering layer. Figure 11 provides both the volume and particle depolarization measurements of this cloud field. Although there is little indication of cirrus precipitation at 0300 and 0500 UTC below 10 km in the volume depolarization measurements, highly depolarizing precipitation reaching down to ∼9.5 km is observed in the particle depolarization measurements. Note also that at the base of the upper cloud layer, the volume depolarization slowly rises to peak values of ∼50% while the particle depolarization ratio indicates ∼50% depolarization at the very base of the cloud. The particle depolarization ratio permits the depolarization properties of the particles to be separated from the molecules providing both improved contrast and more accurate characterization of the particles' scattering properties.

5) Motivation of cirrus cloud modeling study

The cirrus case of 19–20 June presented here is motivating a modeling study at NASA GSFC with the goal of understanding the physical mechanisms that produce the layering observed in the upper cloud of Fig. 7 that has been interpreted as the cirrus mixed layer (Lin et al. 2005a). The simulation of this case through numerical modeling is a considerable challenge. One hypothesis that will be investigated is that vertical differential radiative heating is the dominant effect that determines the thickness of the well-mixed layer and stabilizes the lower layer of the anvil while the cloud microphysical properties evolve accordingly. Previous studies of this type include both 1D (Khvorostyanov et al. 2001; Sassen et al. 2002; Lin et al. 2005b) and 2D (Luo et al. 2003) simulations of cloud microphysical properties, which were compared with ground-based measurements. For the case of 19–20 June, we will compare the microphysical and optical properties derived from a 2D model with multi-ice-category bin microphysics to the lidar profile measurements of ice water content, particle size, and depolarization ratio. To aid this study, the technique for reducing the cross talk between the parallel and perpendicular channels will be optimized and a multiple scattering correction will be applied to the SRL depolarization data. We will also compare the retrievals of particle size using the Raman technique described here to one based on multiple scattering in the cirrus clouds (Whiteman et al. 2001a; Gambacorta et al. 2004).