Lidar Technology for Remote Sensing Measurments of Ozone Concentrations

LIDAR TECHNOLOGY FOR REMOTE SENSING MEASURMENTS OF OZONE CONCENTRATIONS

By

Florence Bocquet

April 2003

Table of contents

Introduction…………………………………………………………………….……………….. / 1
Principles of lidar…………………………………………………………….………………….. / 2
Principle of the lidar and the DIAL technique…………………………….………………… / 2
Description of ozone lidar systems……………………………………….…………………. / 3
Description of the measurements and sampling strategy………………….………………… / 5
Ozone retrieval from the lidar signals…………………………………….…………………. / 5
Accuracy and vertical resolution of the measurement…………………….………………… / 6
OPAL application: experiment and results……………………………………………………… / 8
Data taking…………………………………………………………………..……………….. / 8
Differences in methodology between the airborne in situ measurements and the lidar measurements……………………………………………………………………………….. / 9
Figures of intercomparison…………………………………………………….……………. / 11
Statistics……………………………………………………………………….……………. / 11
Conclusion………………………………………………………………………………………. / 12
References………………………………………………………………………………………. / 13

Introduction

Since the mid-1960s, scientists have used lidar to study atmospheric particles and clouds. It is considered to be among the most accurate data sets attainable using remote sensing processes. Different types of lidars, such as Rayleigh lidar, Raman lidar, aerosol lidar or again, cirrus cloud lidar, measure different atmospheric properties. To measure ozone –and aerosols-, NASA Langley scientists in particular, use a specialized lidar called the airborne Ultraviolet (UV) DIfferential Absorption Lidar (DIAL). Because ozone plays important roles in the tropospheric –and stratospheric- chemistry (in particular, ozone is the source of the hydroxyl radical, which controls the abundance of many atmospheric constituents, including greenhouse gases such as methane and HCFC’s) and is a health risk near the surface, it is being intensively researched. It is known that tropospheric ozone arises both from ‘in situ’ photochemical production and downward transported from the stratosphere, but the relative importance of these two sources to the global budget is poorly understood.

Climate models are three-dimensional, but conventional ground-based ozone monitoring networks are only two-dimensional. For this reason, using lidars will add the third dimension to the data and enable researchers to fully, or at least to better understand ground level ozone concentrations.

Principles of lidar are presented in the first section. The second part of the report will show some data from an application and intercomparison of OPAL (Ozone Profiling Atmospheric Lidar) ground-ozone lidar with an airborne UV ozone analyzer, and finally, a conclusive section.

Principles of lidar

Principle of the lidar and the DIAL technique

The lidar (LIght Detection And Ranging) is a remote sensing instrument that uses laser light, i.e., operating in the optical range, in much the same way that sonar uses sound, or radar uses radio waves. Depending on the desired measurement, lidar systems use various light-matter interactions such as Rayleigh, Mie and Raman scattering or fluorescence. Measurements of atmospheric ozone, as well as temperature and aerosol are based on the first 3 processes. Generally, a lidar measurement consists in sending into the atmosphere a laser beam; a small part of this laser radiation is scattered back to the ground, where it is collected by a telescope, detected by a photomultiplier tube and analyzed by an electronic acquisition system. Range resolved measurements can be obtained using pulsed lasers. In order to measure the ozone vertical distribution, the Differential Absorption Laser technique (DIAL) is used. This technique requires thesimultaneous emission of two laser beams characterized by a different ozone absorption cross-section (Figure 1).

Figure 1: The absorption cross-section of ozone with the emitted wavelengths superimposed (wavelengths ‘on’ marked in red and ‘off’ marked in green).

Fig. 2: Absorption cross-section of ozone, sulfur dioxide and nitrogen dioxide in the ultraviolet.

The ozone absorption cross-section spectrum (Fig. 1) shows three red vertical lines, which are the laser wavelengths of a conventional Raman-shifted DIAL system such as OPAL. The line at 266 nm is obtained by quadrupling the output of an Nd:YAG laser, and the lines near 289 and 299 nm are obtained by stimulating Raman scattering in high-pressure cells of deuterium and hydrogen. The ozone cross-section at 266 nm is so high that it seriously limits the lidar range. For an atmosphere at standard temperature and pressure containing 100 ppb ozone, the extinction coefficient at 266 nm is 2.4 km-1, which implies that the two-way transmittance to an altitude of 1 km is only 0.01. On the other hand, it is best for the wavelength pair to be on the steepest part of the cross-section curve in order to minimize errors resulting from aerosol backscatter. The Raman-shifted pair at 289/299 is on the tail of the curve, and in addition, it is susceptible to errors resulting from the presence of sulfur dioxide. Ideally, the operating wavelengths of the ozone DIAL system should be selectable and the GTRI (Georgia Tech Research Institute) lidar group is currently developing tunable laser technologies (Fig. 2 - see conclusion).

Description of ozone lidar systems

A lidar system includes basically one or several laser sources with optical devices to reduce the divergence of the beam, a telescope, which collects thelight scattered back by the atmosphere, an optical analyzing system with detectors such as photomultipliers to detect the optical signal, and an electronic acquisition system(Figure 3). The analyzing systems used to digitize the electronic signal provided by thephotomultipliers include photon counting and/or transient analyzers. In the case of the DIAL systems characterized by the emission of two laser wavelengths, the optical receiving system comprises spectral analyzing optics, such as interference filters or spectrometers.

Fig. 3: Schematic view of the principle of a lidar system.

To monitor atmospheric ozone with the DIAL technique, the choice of the laser wavelengths depends on the altitude range of the measurement (Mégie et al., 1985). The spectral range is chosen first in the ultraviolet where the ozone absorption is more efficient (266 nm as shown in figure 1), but the selected wavelengths differ according to whether the measurement is made in the troposphere or in the stratosphere - in the troposphere, the ozone number density is small so the laser wavelengths must correspond to a strong UV absorption, while for stratospheric measurements, the objective is to reach the stratosphere and to detect the high ozone concentrations there (Browell, 1989, Papayannis et al., 1990).

Description of the measurements and sampling strategy

The lidar signals cover a very high dynamic range corresponding to several orders of magnitude, which is not handled by the electronic acquisition systems. This requires the use of simultaneous photon counting and analogue acquisition for the low and the high signals or the separation of the optical signal in two parts (90% and 10%), corresponding to the high and the low altitudes respectively. For reasons of simplicity in terms of electronic acquisition, the latter solution is the most commonly used (Harris and Hudson, 1998). Depending on the power and the repetition rate of the laser, an ozone measurement lasts typically several hours, leading to a spatial resolution of the order of 100 km, depending on the atmospheric conditions. The vertical resolution ranges from several hundred meters in the lower range to several kilometers above 40 km. Finally, one main caveat of the lidar measurements is the requirement of clear sky meteorological conditions - laser radiation is rapidly absorbed by clouds - and only cirrus can be tolerated for accurate stratospheric measurements.

Ozone retrieval from the lidar signals

Assuming a monochromatic laser impulsion at wavelength λ, the received optical power corresponding to the light backscattering at altitude z, is given by (Measure, 1984):

P(z) = K(λ)β(λ,z)exp[-2τ(λ,z)]

(1)

where K(λ) is an instrument constant involving the telescope surface area, the emitted power, the optical efficiency of the receiving system and a geometrical factor depending on the alignment of the laser and the telescope axis, β(λ,z) is the atmospheric backscatter coefficient and τ(λ,z) is the atmospheric optical depth. τ(λ,z) depends on the following parameters:

(2)

where α(λ,z) is the atmospheric extinction coefficient, the ozone absorption cross-section, the ozone number density to be measured and the term corresponds to the extinction by other absorbers.

Applying this formula to the second wavelength and taking into account the background signal, one derives the ozone number density from the received lidar signals:

(3)

where Pbi is the background signal and

(4)

The laser wavelengths are chosen so that the term represents less than 10% of the measurement (Harris and Hudson, 1998). The derivation of the ozone number density from the laser signals shows thus that the DIAL technique is a self-calibrated technique, which doesn't need the evaluation of instrumental constants.

Accuracy and vertical resolution of the measurement

The precision of a DIAL measurement is defined by the statistical error due to the random character of the detection process, which follows basically the Poisson statistics (Measures, 1984). The accuracy of the measurement depends on the approximations made in deriving the ozone number density from the received signals. It depends also on the linearity of the lidar signals.

According to the Poisson statistics, the statistical error on ozone is given by the following formula:

(5)

where ΔZ is the initial range resolution of the acquisition system, Pi,j corresponds to the lidar signal at wavelength i from altitude Zj, cj are the coefficients of the low pass derivative filter used to differentiate the signals, Ni is the number of laser shots at wavelength λi and Pbi is the background radiation at wavelength λi. The final statistical error ε on the measurement is the result of a compromise depending on the experimental system characteristics, the duration of the signal acquisition and the vertical resolution according to the following relation:

(6)

where A is the telescope receiving area, ΔZ the final range resolution, P0 the emitted power and Ta the acquisition time. Most teams choose low pass filters with varying number of points as a function of the altitude. The DIAL stratospheric ozone lidar profiles are thus generally characterized by a vertical resolutionvarying from several hundred meters in the lower stratosphere, to several kilometers around 50 km (see Figure 4 for an example of ozone lidar vertical resolution profile).

Figure 4: Precision and vertical resolution profile of an ozone measurement in the case of the OHP (Observatoire de Haute Provence –in France) lidar instrument. Both the precision and the vertical resolution profile depend on the experimental configuration. The precision can vary from one measurement to the other.

The accuracy of the measurement depends on the term (see equation 4) which corresponds to less than 10% of the value derived directly from the slope of the signals but still has to be corrected using ancillary measurements. It depends also on the accuracy of the ozone absorption cross-sections and on the approximation concerning the monochromaticity of the laser radiations. These error sources are summarized in Table 1, which indicates the residual error on the measurement after correction (Godin, 1987).

Table 1: Error sources for lidar measurements.

Error source / Residual
Ozone absorption cross-section
- Absolute value (Bass&Paur) / 2 %
- Temperature sensitivity / < 0.5%
Laser line width / < 0.3%
Rayleigh extinction / <0.6%
Other absorbers
- SO2 - normal conditions
- after major volcanic eruption / negligible
1%
- NO2 / < 0.3%
Aerosol backscatter and extinction
- Volcanic conditions:
- correction using ancillary size distribution measurements
- use of Raman channel / 30%
< 5%
- background / < 5%

OPAL application: experiment and results

This experiment was sponsored by California Air Resource Board (CARB) and the principal investigators were Yanzeng Zhao and R. Michael Hardesty (NOAA, ETL). The lidar was completed in March 1993, and modifications to the seatainer housing the lidar were completed just few days before the shipment. The objectives of the experiment were:

1. To verify the capability of the ETL ozone lidar for remotely sensing ozone distributions in the lower troposphere, and to evaluate the accuracy of the ozone lidar measurements.

2. To test the integrity of the lidar system and the mobile laboratory following long distance transport and to test its performance in a field environment.

The intercomparisons were performed between OPAL (Ozone Profiling Atmospheric Lidar) and the UC-Davis DASIBI UV ozone analyzer on board a Cessna 172 airplane, during the period of July 9 to July 23, 1993, near Davis, California. During the intercomparison, OPAL was staring vertically, the airplane spiraled from the surface to 3 km around the lidar beam (but not centered at the lidar). The lidar and airplane data taken during the ascending and descending of the airplane were compared. Considering the differences in methodology between the lidar and the airplane, each ozone profile measured by the airplane in one spiral (about a half hour) is matched with one to four contemporaneous lidar-measured profiles to perform as accurate an intercomparison as possible. In each lidar-measured ozone profile, only the one altitude section that matches the time of the airplane measurement is used. Thus, figures show discontinuities in the lidar-measured profiles. Examples of ten intercomparisons from July 15, July 19, and July 22 are shown in the Figures 5 to 14.

Data taking

The intercomparison experiment started on July 14, 1993. About 200 profiles were collected over seven weekdays. During each of those days, one to three aircraft flights were made; each flight consisted approximately of a 30-minute ascent and 30-minute descent. The flights took place at various times of the day between early or mid-morning and mid-afternoon. The lidar operated continuously during the flights, and usually gathered additional data before and after the flights.

Differences in methodology between the airborne in situ measurements and the lidar measurements

It is worth noting the methodological differences between the airborne in situ measurements and the lidar measurements before discussing the results of the intercomparison. The lidar was pointed vertically and operated at 2 Hz pulse rate. Each vertical ozone profile was an average over 7.5 min at a fixed horizontal location. Depending on wind speed, which can vary with altitude, this 7.5-min average could be equivalent to an averaged profile over wind trajectories of path lengths ranging from 0 to 10 km. However, the airplane spiraled up and down with a vertical axis not centered over the lidar and projecting a large horizontal square, each side of which was much greater than the altitude range. The airborne UC-Davis DASIBI data are 10-second averages and correspond to averages over 0.3 to 0.5 km of the flight path, which was horizontally 0 to 8 km away from the lidar. In addition, the airborne measurements have an altitude resolution of 20-30 meters, but hardware problems caused the lidar measurements in Davis to have a much lower resolution (200-300 m below 1000 m and above 2000 m). Temporal variations of ozone can be significant as shown in both the airborne measurements and lidar measurements (see Fig. 10). Horizontal variations may also be significant in certain circumstances, especially at higher altitudes. Consequently, temporal and spatial variations, range resolution, lidar errors and in situ inaccuracies are all included in the differences between the aircraft and lidar measurements.

Considering the above methodological differences, each ozone profile measured by the airplane in one spiral (about a half hour) is matched with one to four contemporaneous lidar-measured profiles to perform as accurate an intercomparison as possible. In each lidar-measured ozone profile, only the one altitude section that matches the time of the airplane measurement is used. Thus, figures show discontinuities in the lidar-measured profiles.

Click on any figure to view the full-sized version.
Fig. 5. 07:59-08:29, July 15, 1993 / Fig. 6. 08:49-09:00, July 15, 1993
Fig. 7. 14:45-14:57, July 15, 1993 / Fig. 8. 14:59-15:16, July 15, 1993
Fig. 9. 11:39-12:07, July 19, 1993 / Fig. 10. 12:12-12:35, July 19, 1993
Fig. 11. 14:45-15:17, July 19, 1993 / Fig. 12. 15:23-15:55, July 19, 1993
Fig. 13. 10:32-10:48, July 22, 1993 / Fig. 14. 10:53-11:20, July 22, 1993

Figures of intercomparison

Examples of ten intercomparisons from July 15, July 19, and July 22 are shown in Figures 5 to 14. When the atmospheric conditions were stable and the ozone profiles exhibited small variations, the agreement between the lidar measurements and the airborne measurements were good (e.g., the morning to noon observations like Figs. 5, 6 and 9). When atmospheric ozone concentration varied rapidly (e.g., the afternoon observations in Figs. 11, 12 and 14), greater differences between the aircraft-measured and lidar-measured ozone profiles were observed. In the afternoon, convective activities might have caused cells or columns that led to greater horizontal variations. It is interesting to note that when ozone concentrations fluctuated in higher altitudes as measured by the DASIBI, the lidar-measured profiles go straight through the middle of the fluctuations like vertically averaged profiles. A reasonable explanation is that the lidar has a lower range resolution in the altitude range from 2500 to 3500 m. When the aircraft-measured profiles were smoothed with a resolution similar to the lidar's, the results were closer. In fact, a running average of the airplane profiles (the averaging interval was ±20 points, equivalent to 500 to 600 m) at higher altitudes was carried out in Figs. 12 and 14. The smoothed airplane profiles have less fluctuations and are much closer to the lidar profiles.

Statistics

Statistical calculations for the intercomparisons were carried out for three altitude intervals (i.e., 0 to 1000 m, 1000 to 2000 m, and 2000 m to 3500 m). Twenty comparisons of the ozone profiles are included in the calculation for each layer as the differences in the ozone measurements (DO3 = O3lidar - O3air) are calculated with a 50-m interval (Zhao et al., 1997). Then the average value of DO3, aDo3, defined as