This study reports airborne measurements of atmospheric CO2 column density using a 2-μm double-pulsed integrated path differential absorption (IPDA) lidar. This new 2-μm IPDA lidar offers an alternative approach to measure CO2 column density with unique features. The online frequencies of this lidar can be tuned to 1–6 GHz from the CO2 R30 absorption line peak. It provides high measurement sensitivity to the lower-tropospheric CO2 near the ground surface. This instrument was flown in the spring of 2014 in a NASA B200 aircraft. The results of these test flights clearly demonstrate the measurement capabilities of this lidar instrument. The CO2 column dry mixing ratio is compared to an in situ CO2 measurement by a collocated NOAA flight. The IPDA lidar measurement is determined to be in good agreement with a 0.36% difference, which corresponds to 1.48 ppm. It is the average difference between the IPDA lidar measurements and the NOAA air samples in the flight altitudes from 3 to 6.1 km.
Active sensing of near-surface CO2 concentrations will significantly increase our understanding of CO2 sources, sinks, and fluxes worldwide (NRC 2007; ESA 2008). The mid-IR wavelength regions at 1.57 and 2.05 μm are considered suitable for atmospheric CO2 measurements due to available technologies and the existence of distinct absorption features for the CO2 molecule at these particular wavelengths. Integrated path differential absorption (IPDA) lidars essentially measure the open-path laser absorption spectroscopy (Measures 1992). Two IPDA lidar instruments operating at 1.57 μm have been developed and deployed as airborne systems for atmospheric CO2 column density measurements in the United States. One instrument is based on an intensity-modulated continuous-wave (CW) CO2 laser absorption spectrometer approach (Dobbs et al. 2008; Dobler et al. 2013). The other instrument measures the CO2 absorption lineshape at several spectrally resolved points across the CO2 line using a pulsed multiple wavelength approach. This system utilizes high pulse repetition frequency (PRF) but relatively low pulse energy (Abshire et al. 2010, 2013). These airborne CO2 IPDA lidar systems operating at 1.57 μm utilize laser and detector technologies developed from the telecom industry. Using a 2.05-μm CW laser absorption spectrometer employing the coherent-detection method, airborne measurements of CO2 column abundance have been demonstrated (Spiers et al. 2011; Menzies et al. 2014). Since 2010, these three instruments have been flying together on NASA’s DC-8 aircraft for CO2 column density measurements.
At NASA Langley Research Center (LaRC), we have developed a new double-pulsed, high-energy, 2-μm direct-detection IPDA lidar instrument. Lidars operating in the 2-μm band offer high near-surface CO2 measurement sensitivity due to the intrinsically stronger absorption lines (Joly et al. 2008; Menzies and Tratt 2003; Joly et al. 2009; Caron and Durand 2009). The pulsed lidar approach inherently provides a mean for determining range to the scattering targets. With a pulsed laser and time-resolved receiver, the reflected signals can be resolved between aerosols, clouds, and topographical surfaces. It can directly eliminate contamination from aerosols and clouds to yield high-accuracy CO2 column measurements. The high-energy approach helps to improve the signal-to-noise ratio of the IPDA lidar measurement and therefore achieves high CO2 measurement precision.
This paper describes the airborne 2-μm double-pulsed IPDA lidar instrument techniques and the lidar system. Examples of airborne measurements of atmospheric CO2 column absorption from aircraft to ground or cloud tops are presented. A comparison between lidar measurements with model estimation and a collocated NOAA flight with in situ CO2 measurement is discussed.
2. IPDA lidar measurement technique
Differential absorption lidar (DIAL) is a powerful technique to measure atmospheric trace gases. In this technique, the laser transmitter emits at least two frequencies. One is called online frequency, which is at or near the absorption peak of the gas interested. The other is called offline frequency, which is at the vicinity of the absorption line but with relatively negligible absorption. To meet the measurement precision and accuracy requirements, the IPDA lidar technique is adopted (Menzies and Chahine 1974; Measures 1992; Ehret et al. 2008). The IPDA technique is essentially a special case of the DIAL technique, in that the reflected signal from a hard target is used to determine the integrated column density from the instrument to the hard target.
The optical depth (OD) (dimensionless) at on- and offline frequencies is the sum of all the optical depths due to molecular absorption (carbon dioxide and water vapor in this case) and aerosol extinction. Thus, the OD measured by airborne IPDA lidar with nadir view is
where is the optical depth for frequency i, which denotes for either on- or offline frequency; σcd and σwv (m2) are the absorption cross sections per molecule for carbon dioxide and water vapor, respectively, which is a function of frequency, temperature, and pressure; σa is aerosol extinction cross section per particle; , , and (m−3) are the molecular number density for carbon dioxide, water vapor, and particle density for aerosols, respectively. The integration is from airplane altitude Za to scattering surface elevation (SSE) Zs.
The key measurement parameter of an IPDA lidar instrument is the differential absorption optical depth (DAOD), DAOD = ODon − ODoff, which is defined as the optical depth difference between the on- and offline frequencies. The total measured round-trip DAODm contributed from carbon dioxide, water vapor, and aerosol through the atmosphere becomes
The contribution for the DAODm measurement due to aerosol scattering is cancelled in Eq. (2), since the optical depth at on- and offline frequencies is approximately the same for aerosol scattering.
The OD is related to the laser beam round-trip transmission according to the Beer–Lambert law. The laser beam round-trip transmission τ along the optical path from airplane height to the SSE is the negative natural logarithm of the optical depth,
In an IPDA lidar measurement, the round-trip transmissions, τoff and τon, are obtained by the return signal power (P) normalized to the transmitted laser energy (E) (from the laser energy monitor). Thus, the DAODm becomes
where the ti is the effective pulse width of the return signal at λon or λoff. The quantities on the right side of the equation are the lidar-measured parameters.
The column-weighted CO2 dry mixing ratio XCO2 can be derived from the lidar DAODm measurement and the knowledge of CO2 absorption spectroscopic parameters that are pressure and temperature dependent. DAODwv is the contribution due to water vapor absorption. The water vapor profile is also needed for CO2 dry mixing ratio derivation. The mixing ratio XCO2 is a dimensionless parameter,
The quantity inside the integral at the denominator of Eq. (5) is defined as the weighting function (m−1), which is the product of the differential absorption cross section and the dry-air number density (z),
Equation (5) implies that the airborne IPDA lidar instrument measures the average column CO2 dry mixing ratio from the instrument to the ground but with weighting that can be adjusted by controlling the transmitted online frequency. The online frequency of the transmitter could be tuned to weight the column measurement to the surface for optimum CO2 interaction studies or up to the free troposphere for optimum transport studies. This is achieved, for example, by tuning the online frequency to different positions relative to the R30 CO2 absorption line peak as shown in Fig. 1, where only integer values of frequency offset are shown for clarity. Except for the frequency offset 1 GHz from the center of the R30 line, the other frequency offsets are all weighted toward the surface, where the CO2 source and sinks occur. In fact, it is one of the advantages that the 2-μm pulsed IPDA instrument can provide.
3. CO2 line selection and spectroscopy
Unprecedented high precision and accuracy of atmospheric CO2 column density measurements at 0.5–1-ppm level require optimal operating lidar frequency selection and accurate knowledge of the spectroscopic parameters of the CO2 absorption line (ESA 2008; Miller et al. 2007; Dufour and Breon 2003). An optimal line shall meet the following criteria: 1) The line characteristics shall have minimum temperature susceptibility. 2) The value of the differential absorption optical depth shall be close to unity to minimize the measurement statistical errors (Remsberg and Gordley 1978; Megic and Menzies 1980; Bruneau et al. 2006). 3) The absorption feature at the sounding frequency shall have minimum interference from other gases. 4) The weighting function shall be favorable to the lower troposphere, so it senses the CO2 close to the ground, where the CO2 source and sinking occurs. 5) A laser will be available at this wavelength with energy, pulse repetition rate, and spectral properties needed for the measurement. Optimal wavelength selection helps to improve the lidar measurement accuracy by reducing both the random and systematic errors.
Although there are many lines in the CO2 transition bands, lines that can simultaneously meet all the above-mentioned criteria are limited (Menzies and Tratt 2003; Toth et al. 2008; Regalia-Jarlot et al. 2006). Many of the transitions in these bands have either inadequate absorption strength, high temperature susceptibility, or excessive interference from absorption due to other atmospheric species, in addition to technological availabilities. The spectroscopic parameters of the transitions in the (20°1)III ←(00°0)I CO2 vibrational–rotational band have been carefully determined in terms of line strengths, air-broadening coefficients, and variation with temperature. The R30 line in this band at 2050.967 nm (4875.7487 cm−1) has been found to be one of the most suitable lines for measurements of CO2 in the 2-μm wavelength region with regard to strength of the absorption lines, low susceptibility to atmospheric temperature variability, and freedom from problematic interference with other absorption lines (Menzies and Tratt 2003; Joly et al. 2009; Ambrico et al. 2000). At this line, the molecular line intensity is at 1.741 × 10−22 cm−1 per molecule per cm2; the lower state energy level is at 363 cm−1, which provides lower temperature sensitivity than most other lines in this band. The half-width of this line is ~0.068 cm−1. The accuracy of this line parameter is sufficient to reach a 1-ppm accuracy on CO2 total column content (Joly et al. 2009). We chose to operate the lidar on the long wavelength wing of the R30 CO2 line in the side line operation mode. By operating the lidar at side line mode, it moves the weighting function peak toward the desirable measurement region near the ground.
Figure 2 depicts the CO2 absorption cross section around the R30 line at ground surface and at an altitude of 10 km. The absorption cross sections for water vapor, the primary interference gas, at this wavelength region are also plotted. The cross sections are calculated by using the 2012 High-Resolution Transmission Molecular Absorption Database (HITRAN 2012; Rothman et al. 2013). The conventional Voigt profile was assumed in the calculation. However, some sophisticated profiles may be needed to achieve more accurate spectroscopic data (Casa et al. 2009). The vertical lines in the figure represent the frequency shift from the absorption peak from 1 to 6 GHz. The online frequency can be chosen and locked anywhere in this frequency range. The amount of frequency shift for the online frequency determines the expected optical depth value and the relative contributions of the various atmospheric layers to the total optical depth. The higher the online frequency shift from the absorption peak, the bigger the contribution from the lower part of the atmosphere. The online frequency shift from the peak of the R30 line shall be greater than a half-width of the line to yield sufficient round-trip transmission for adequate measurement sensitivity (Menzies and Tratt 2003). To help set the online frequency shift to obtain a desired DAOD value close to unity, we have calculated the DAOD as the function of altitude for the six different frequency shifts. It is tabulated in Table 1 and serves as a lookup table for setting the online frequency in airborne lidar measurements. The optical depth in the table is calculated using the U.S. standard atmospheric model and HITRAN 2012 (Rothman et al. 2013; Anderson et al. 1986). The guidance for setting the online frequency is to select the frequency shift with the boldface value in the table. Thus, the expected DAOD value is between 0.40 and 1.20. Nominally, the online frequency shift is set between 3 and 4 GHz from the absorption peak.
Unlike the online frequency position, the offline frequency position has a much smaller impact on CO2 column density measurement accuracy with regard to laser frequency stability, variations of temperature, pressure, or scattering surface elevation. However, the offline position has an equal impact as the online frequency position in terms of water vapor contamination sensitivity. The impact of water vapor on the CO2 mixing ratio measurement can be seen in Eqs. (2) and (5). It not only contributes to the error in DAOD measurement as expressed in Eq. (2) but also in the CO2 dry mixing ratio derivation in Eq. (5). The water vapor interference is relatively low in the R30 spectral region as shown in Fig. 2. In addition, by appropriate selection of on- and offline frequencies, the water vapor interference shown in both Eqs. (2) and (5) can be effectively cancelled (Caron and Durand 2009). We chose the offline wavelength position at 2051.250 nm, which is at the minimum absorption between the R30 and R28 absorption lines. It is located at the long wavelength side of the R30 line peak, which is the same side as the online wavelength position.
4. Description of the IPDA lidar Instrument
A schematic of the 2-μm double-pulsed IPDA instrument is shown in Fig. 3. The key components of the IPDA instrument are the double-pulsed high-energy 2-μm laser and the wavelength locking and switching unit that will control the output pulse wavelength precisely at preselected on–off wavelengths. Each component of the IPDA instrument is briefly described in this section. The parameters for the 2-μm double-pulsed IPDA lidar are tabulated in Table 2.
a. Laser transmitter
The laser transmitter utilizes holmium (Ho) and thulium (Tm) codoped solid-state 2-μm pulsed laser technology (Barnes et al. 1996; Yu et al. 1998). Over the last 20 years, NASA LaRC has developed high-energy 2-μm lasers based on diode laser pumped solid-state laser technology (Singh et al. 2015). Greater than 1 J per pulse 2-μm laser has been demonstrated (Yu et al. 2006). An injection seeded single-frequency 2-μm laser with 250 mJ per pulse at 10 Hz was also integrated into a coherent wind lidar to make ground-based and airborne 3D wind measurements (Kavaya et al. 2014).
One of the unique features of the Ho:Tm:YLF laser is that it can generate multiple pulses within one pump cycle. Figure 4 shows a typical double-pulsed 2-μm waveform. Since the laser material is codoped with Ho and Tm atoms, the population at the upper-laser-level Ho 5I7 and Tm 3F4 level are in thermal equilibrium, where the stored energy is shared by those two manifolds. After the first pulse is emitted, the population at the upper laser level of the Ho5I7 is depleted to below the laser threshold. However, the energy stored at the Tm 3F4 level is intact and quickly transferred into the Ho5I7 upper laser level to repopulate it without external pumping. It produces the second pulse when the population at the Ho5I7 upper laser level reached above the laser threshold. The time interval between the pulses is about 150–200 μs, which is required to transfer the energy between the Tm and Ho atoms. This process can be repeated for a third pulse or even a fourth pulse if there is enough energy stored in the laser gain medium. The energy distribution between the multiple pulses can be adjusted. The energy of the first pulse can be higher, equal, or lower than that of the second pulse by adjusting the Q-switch trigger time sequence for the two pulses. A 600-mJ double-pulsed 2-μm laser was demonstrated using the characteristic of this laser medium (Yu et al. 2003).
In this IPDA instrument, the output energies of the laser transmitter are typically 90 and 45 mJ for the first pulse and the second pulse, respectively. This pulse pair repeats 10 times in a second. The double-pulsed IPDA lidar operation provides several advantages. First, it simplifies the lidar structure. Since a single laser can generate the required two pulses for the IPDA lidar, there is no need for two lasers in this instrument. It does not need additional beam combining and beam steering components. Second, not only is the system simpler but it is also more efficient. The second pulse is generated from the energy transfer between the Tm and Ho atoms. There is no external pump source required for the second pulse generation. Third, since the first pulse is at an online wavelength with energy twice that of the second pulse at offline wavelength, it compensates the higher attenuation at the online return signal due to absorption along the path. This improves the lidar system performance in regard to the overall system signal-to-noise ratio. Fourth, since the time difference between the first and second pulses is typically only 150–200 μs, the footprint overlap on the ground between the two pulses is high for an airborne flight platform. It mitigates the effect of the surface reflection and topography difference between the on- and offline pulses on the precision of the CO2 column density measurements. It helps that both pulses see the same atmospheric refractive turbulence and speckle realization. The percentage of the footprint overlap on the ground between the two pulses is a function of flight altitude, aircraft speed, the time interval between the two pulses, and the laser beam divergence. It can be expressed in Eq. (7) as
where R is the laser pulse footprint radius (m), d is the horizontal distance between the centers of the on- and offline pulse footprint, and θ = cos−1(d/2R) is the half of the section angle. The section angle is defined as the angle subscribed from the center of the footprint to the overlap points on the peripheral.
Figure 5 depicts the footprint overlap percentage of the on- and offline pulse at different flight altitudes. The aircraft speed is assumed at 120 m s−1 at the nominal cruise speed of a B200 aircraft, and the time interval between the two pulses is at 200 μs.
The IPDA lidar technique requires accurate monitoring of the emitted pulse energy ratio between on- and offline pulse pairs. It is used as the normalization factor in the DAOD calculation as shown in Eq. (4). About 1% of the laser energy is split from the laser beam as an energy monitor reference before it leaves the laser enclosure. It is coupled to a 25-mm-diameter integrating sphere. An extended-range p-type–intrinsic–n-type (PIN) detector, the same type used in the lidar signal channel, is used to monitor the laser energy at the exit of the sphere. The use of the sphere helps to ensure beam uniformity and avoids detector saturation. The calibration and characterization of the energy monitor can be found in Refaat et al. (2015).
The temperature of the laser is controlled by two chillers: one controls the laser rod temperature and the other controls the temperature for the laser bench and pump laser diodes. All the optical mounts are custom designed and have spaceflight heritage. They are designed to be adjustable, lockable, and hardened to withstand vibrations that can occur during airborne operation. The laser transmitter is rigid and requires no further maintenance or tweaking during airborne lidar operation. The laser transmitter is 29 × 67.3 × 16.5 cm3 in size, and it weighs less than 27 kg.
b. Wavelength control unit
The double-pulsed on- and offline frequency accuracy and stability of the IPDA lidar are critical for making precise and accurate CO2 measurements. The first pulse is injection seeded with online frequency, while the second pulse is injection seeded with offline frequency. The ramp-and-fire seeding technique is applied to achieve frequency accuracy and stability requirements of the IPDA lidar system (Henderson et al. 1986). Three CW lasers are utilized to provide on/off frequencies. These CW lasers are diode laser end-pumped solid-state Ho:Tm:YLF lasers (Coherent Technology Inc., METEOR). As shown in Fig. 3, the centerline reference laser (CW “Center”) is locked on the center of the CO2 R30 absorption line in a CO2 gas cell by a frequency modulation spectroscopic technique (Koch et al. 2008; Yu et al. 2010). It established the frequency standard in the frequency control system. It then heterodynes with a second CW laser (CW “ON”) to generate online frequency. The online frequency is tunable from the R30 center. It is determined by the heterodyne beat signal between the two lasers. The online frequency tuning range can be set anywhere between a few hundred MHz and 7 GHz from the center of the R30 absorption line. An electronic control loop locks the online laser frequency at the user-decided tuning frequency from the line center. The capability of the online frequency tuning and locking allows the optimization of the optical depth for measuring atmospheric CO2 concentrations depending on flight altitude. Typically, the online frequency is locked at 3–4 GHz from the R30 line center. The offline CW laser (CW “OFF”) is locked against a high-precision wavemeter (Bristol 621). The wavelength locking accuracy is less than 2 MHz and 30 MHz for the online and offline frequencies, respectively. The locked online and offline laser outputs (CW ON and CW OFF, respectively) are coupled into an optical switch (CIVCOM, Free-X) that can be electronically addressed to select the on/off frequencies. The alternatively selected online and offline frequencies are then sent to injection seed the pulsed laser. The electronics unit controls the pulse laser injection seeding and Q-switch timing to generate the double-pulsed online and offline laser pulses at 10-Hz repetition rate. Three CW lasers, a CO2 cell, an electro-optic (EO) modulator, two detector units, an optical switch, fiber couplers, and connectors are all packaged in a custom designed three-unit (3U) 19-in. rack-mountable box.
The telescope is a custom designed Newtonian type with a 40-cm-diameter primary mirror size. The f# number of the primary mirror is 2.3. The primary mirror is made of aluminum with a diamond-turned machining technique. The secondary mirror for the telescope is made of fused silica. Both sides of the secondary mirror are dielectric coated for high reflectivity (HR) at 2-μm wavelength. One side of the secondary directs the optical signal collected by the primary mirror to the focal point of the telescope. The other side of the secondary directs the pulsed laser beam into the atmosphere. The HR dielectric coating is designed as a bandpass filter. It significantly blocks the light outside the 2-μm band to reduce the background noise from the solar spectrum. Therefore, a narrowband interference filter is not necessary for this lidar instrument. The telescope design is simple, easy to align, and able to focus the signal into the detector active area with a spot size of less than 300 μm in diameter.
A two-lens field image aft optics was designed to focus the lidar signal to photodetectors. The Hamamatsu InGaAs PIN photodiodes (model G5853–203) was fully characterized and used in this IPDA lidar instrument. The characteristics of this detector are listed in Table 2. The lidar signal from a hard target is divided into a high-gain channel and a low-gain channel: a high-gain channel with 90% of signal power and a low-gain channel with 10% of the signal power. Detector outputs are coupled to variable-gain, high-speed transimpedance amplifiers (TIA; FEMTO; DHPCA-100). The gain of the TIA is set at 103 V A−1 for the monitor channel, and between 103 and 106 V A−1 for the hard target return channels.
d. Data acquisition and processing unit
A 10-bit, 2 GS s−1 digitizer (Agilent; U1065A) was used for laser energy monitoring, and a 12-bit, 420 MS s−1 digitizer (Agilent; U1066A) was used for measuring hard target returns. The start time of data acquisition is synchronized with the Q-switch trigger signal for each pulse. For the energy monitor channel, 3000 data points sampled at a rate of 500 MHz are recorded for each pulse. The ground return signal recording length is usually set at 15 000 points for each pulse with a sample rate at 200 MS s−1. It covers a flight altitude up to 11.25 km. Each data file contains the return signal from 100 consecutive pulse pairs. The data can be either processed on a single-shot basis or averaged to reduce random errors. Time-gated receiver signal processing significantly reduces the detector noise and solar background included in the signal. Therefore, the signal-to-noise ratio of the lidar measurement is substantially improved.
A simple real-time data processing algorithm is implemented to show the lidar return signal quality. It displays the laser energy monitor signal at on- and offline wavelengths, the raw data from the ground return signal, the range information from the lidar return, and the first-order estimate of the DAOD. These signals help to verify that the laser is operating correctly during the flight. A display also shows the ground images taken by a nadir-viewing video camera.
5. IPDA lidar airborne demonstration
a. Airborne flight overview
After the IPDA lidar was integrated into a class 10000 clean room environment, it was then installed inside a mobile trailer for ground testing with calibrated horizontal targets. The ground testing of this new IPDA lidar instrument was carried out at a lidar test facility in LaRC. Five targets with calibrated reflections at 2-μm wavelength were set up 858 m from the lidar instrument. The IPDA lidar was tested at various operating conditions, including five different target reflectivities ranging from 4% to 55%, online offsets from the CO2 R30 line center from 1–6 GHz, and different preamplifier gain and bandwidth settings. In addition, the IPDA-measured differential optical depth was converted to a CO2 dry mixing ratio using metrological data obtained from instruments that measure pressure, temperature, and relative humidity at the same test site. The measurement results were compared with an in situ CO2 and H2O gas analyzer (LI-COR; LI-840A) at the site. A general temporal profile trend agreement was observed between these two measurements. This verified that the IPDA lidar is sensitive to atmospheric CO2 concentrations (Refaat et al. 2015).
The objective of an airborne demonstration of the newly developed 2-μm pulsed IPDA lidar is to demonstrate the functionality and capability of the lidar instrument. The 2-μm CO2 IPDA lidar is designed for integration into a small research aircraft. The IPDA instrument size, weight, and power consumption were restricted to the NASA B200 payload requirements. This allows the system to be easily adopted in any larger airborne research platform, such as the NASA DC-8 aircraft, for future missions. In addition to the IPDA lidar, other housekeeping instruments were integrated into the B200 aircraft. These included a global positioning system/inertial navigation system (GPS/INS) for aircraft positioning, altitude, and airplane angle measurements, and a video recorder for target identification. In addition, aircraft built-in sensors provided altitude, pressure, temperature, and relative humidity sampling at the flight height position. Data collection was time-stamped using a GPS-disciplined timecode card. Figure 6 shows the integrated IPDA instrument inside the NASA B200 aircraft.
The 2-μm double-pulsed IPDA lidar has been flown 10 times in March and April of 2014. All flights were based out of NASA LaRC. Table 3 lists these flights and summarizes the flight objectives. Each flight has been designed with specific objective. Consistent improvements in data taking, data retrieving, and calibration have been realized in each successive flight.
b. Data examples and signal-to-noise ratio
Strong lidar return signals were obtained for both on- and offline wavelength at different altitudes. Figure 7 shows lidar signal examples at two different lidar operating conditions. There are two pairs of peaks in the return signal. The first pair of signal peaks is an on- and offline pulse signal reflected from the airplane window. The second pair of peaks is an on- and offline pulse signal from the hard target echo. The offline signals are intentionally offset from the online return signals in order to see the on- and offline return signals clearly. In fact, they are overlapped with each other in terms of range. Figure 7a corresponds to a lidar signal with flight at 1372 m with a preamplifier setting at 103. Since the airplane is at a low altitude, there is little absorption for the online pulse. Thus, the amplitudes of the online and offline returning signals are close to the amplitudes of the online and offline laser pulses recorded by an energy monitor. Figure 7b shows the lidar signal with an airplane flying at 6096 m with a preamplifier gain setting at 105. Because of the longer absorption pathlength relative to the conditions in Fig. 7a, the return signal amplitude from online pulse is reduced. Therefore, the lidar signals from on- and offline pulses are comparable as shown in Fig. 7b. The inserts in Fig. 7 are the enlarged ground return signals. It clearly shows that the ground return signal is strong with a high signal-to-noise ratio.
The signal-to-noise ratio and the measured and modeled DAOD values at different ground conditions are listed in Table 4. The SNR is defined as the ratio of the area integrated under the return lidar signal waveform divided by baseline RMS noise times the same time interval used to integrate the lidar signal. The modeled DAOD is calculated using Eq. (2), and the measured DAOD is derived from Eq. (4).
The major noise sources of a lidar instrument are shot noise, detector dark noise, solar background noise, TIA-introduced noise, and speckle noise. The mean current fluctuation for the various noise sources is expressed in Eq. (8) (Ehret et al. 2008):
where B is the electrical bandwidth; e is the electron charge; M is the detector internal gain; F is the detector excess noise factor; R is the detector responsivity; Ps and Pb are the signal power and background power, respectively; iD is the detector dark current; inA is the amplifier input current noise density; VnA is the amplifier input voltage noise density; KB is the Boltzmann constant; T is the temperature (K); RF is the feedback resistor; and Cdet is the equivalent capacitance of the detector. The first term inside the square root sign bracket is the shot noise due to signal and background. The second and third terms in the bracket are the noise current due to detector dark current and amplifier noise, respectively. The last term is the detector/amplifier capacitance noise current, which is significantly dependent on the electrical bandwidth. Table 5 lists the noise current calculated by Eq. (8) using the parameters and specifications of the PIN detector and TIA provided in Table 2. It can be seen that the dominating noise currents for the IPDA lidar instrument are from shot noise and the detector/amplifier capacitance noise.
Solar background noise is not an important factor in this system. Solar spectral irradiance at the 2-µm wavelength region is relatively weak at less than 0.1 W m−2 nm−1 at the sea surface. The dielectric coating of the telescope secondary mirror mentioned in section 4c also helps to block radiation out of the 2-µm band. Compared to the relatively high noise characteristic of the PIN detector and TIA, solar background noise is negligible when assessing total lidar background noise. In fact, a 1.6-nm bandwidth optical filter centered at the 2-µm wavelength was designed but ultimately not implemented due to the negligible impact of solar background noise. By not using the narrowband filter, not only does the overall receiver efficiency increase but it also eliminates lidar DAOD measurement errors due to the temperature- and angle-dependent fluctuations of the filter transmission at online and offline frequencies.
Speckle noise contributes to the lidar-detected signal fluctuation. Assuming the speckle shape is circular, the correlation radius of the speckle can be calculated by rc = λ/πθdiv, where θdiv is the half-angle of the laser beam divergence (MacKerrow and Schmitt 1997). Thus, the number of statistically independent spatial speckle cells captured by the telescope per laser firing can be calculated. Speckle noise can be further reduced by pulse averaging. The calculation of the relative error introduced by the speckle effect is less than 0.4% when 100 pulses are averaged. Speckle noise is a contributor in this system, but it can be mitigated by averaging a large number of statistically independent speckle cells.
Ranging is an inherent capability of a pulsed lidar. Accurate range information between the aircraft and the scattering surface is very important. The range information is mandatory in calculating DAOD as shown in Eqs. (2) and (5). Accurate range measurement reduces the CO2 dry mixing ratio error due to surface topographic fluctuations and discriminates the return lidar signal from cloud and aerosols.
The range information x obtained from the 2-μm IPDA lidar is calculated from the time interval Δt of the signal peak from the airplane window reflection and hard target shown in Fig. 7,
where c is the speed of light; fd is the digitizer frequency; and SH and SW are the index of the signal peaks for window and hard target return, respectively. The ranging accuracy is laser pulse shape dependent (Gardner 1982). Since a typical online pulse width is around 200 ns, much smaller than the offline pulse width of about 310 ns, it is more accurate to determine peak positions from an online return signal and to calculate the time interval. The ranging accuracy also depends on SNR. Since the online pulse energy is 2–3 times higher than that offline pulse energy, it compensates the online return signal SNR reduction due to CO2 absorption. Thus, the online return signal is favorable even with the highest flight altitude in this campaign (~8 km). When the aircraft flies above 6 km, the preamplifier gain is usually set at 105. The return signal from window reflection could be saturated as shown in Fig. 7b. In this case, the time interval can be found using the window return signal from the low-gain channel. The digitizer sampling frequency is set at 200 MS s−1. Thus, the sampling resolution is 5 ns, corresponding to a range resolution of 0.75 m.
The ranging capability of the lidar is shown in Fig. 8. Figure 8a shows a range measurement from a ground lidar test facility. A hard target is set at 858 m from the lidar. With the hard target at fixed distance, it helps to determine the lidar ranging precision. The range measurements shown in Fig. 8a are spanned only in four range bins based on a single-shot signal calculation. The standard deviation of the measurement is 0.36 m. This represents the IPDA lidar range measurement precision under ideal conditions.
The GPS boarded on the aircraft records the flight altitude, the aircraft pitch, and the yaw and roll angles, along with timing information. The IPDA lidar measures the line-of-sight (LOS) length between the IPDA lidar and the hard target. The GPS altitude information can be converted to LOS range, x, to match with the lidar measurement using Eq. (10):
where Altitude is the GPS-recorded altitude, and the θPitch and θRoll are the aircraft pitch and roll angles, respectively. The yaw angle is ignored, since it does not affect the line-of-sight measurement.
Figure 8b shows altitude measurement from the onboard GPS of the aircraft, the LOS distance measured by lidar, and the converted LOS distance by using Eq. (9) from the GPS altitude data. It can be seen that the lidar-measured LOS distance agrees with the converted LOS distance using the GPS data. The aircraft flew at an altitude of 590 m shown in the figure. The peak shows that the range can increase by ~100 m due to turning of the aircraft.
d. DAOD measurement statistics
The lidar measures the backscattered signals from hard targets normalized to their emitted energy samples recorded by an energy monitor. The key parameter, DAOD, is calculated using Eq. (4). The accuracy of the DAOD measurements depends on the lidar signal and noise characteristics, and lidar system bias errors. Since the objective of the flights is to demonstrate the functionality of the newly developed instrument, many lidar instrument settings were adjusted during the flight. Adjustments include the preamplifier gain, the online frequency shift from the R30 absorption peak, the receiver bandwidth, and laser output energy. The instrument-measured DAOD is compared with a model-simulated DAOD value. The model used here is the U.S. standard atmospheric model with an assumed atmospheric CO2 concentration of 395 ppm (Anderson et al. 1986).
Figure 9 depicts an example of the DAOD measurement results. The data shown were obtained in the morning on 27 March 2014 over land at a flight altitude of 6706 m. The land condition varies between rural areas with many trees or farm fields and residential areas in town. Figure 9a is a DAOD calculation based on a single-shot return signal. The online frequency is locked at 4 GHz from the absorption peak. The preamplifier gain is at 105. The DAOD mean value for the single-shot measurement is 1.0587 with a standard deviation of 0.0457. The model predicts the mean value of 1.0553. Random error can be reduced by shot averaging. Figure 9b shows the result with a 100-points moving average, which corresponds to a 10-s average. The standard deviation is improved to 0.0123 for the lidar data as shown in Fig. 9b. It appears to be a small drift in the measured DAOD value, or a gradient in the CO2 column value, due to land condition changes over the flight track.
e. Cloud slicing capability
It is well known that there is high cloud coverage, either optically thick cumulus clouds or optical thin cirrus clouds, on a global scale. It is important to be able to measure the CO2 column density in the presence of thin cirrus clouds, broken clouds, and cloud decks. In the presence of optically thick clouds, it affects a lidar’s capability to measure CO2 column density from instrument to ground, but the reflected signal from a cloud top may be used to retrieve the CO2 column density above the cloud. If a broken cloud is presented, the transmitted laser beam can reach the ground through the cloud holes. Thus, the total CO2 column density to the ground can be measured. The CO2 column density difference between the above cloud and the total column to the ground will provide the CO2 layer density information.
On 31 March 2014, the IPDA lidar flew along the coast of Virginia Beach, Virginia. The purpose of the flight was to fly over clouds to test the cloud slicing capability of the IPDA lidar. Figure 10 shows the IPDA-measured range data. It clearly shows the lidar return signals are from the thick cumulus cloud and from the sea surface through the broken clouds. The figure shows the leveled flight profile, the flat sea surface, and relatively constant cloud-top height. The average column length for ocean return and cloud return is 6805 and 5630 m, respectively. The column length of cloud returns range from 5560 to 5680 m. The DAOD retrieval to the tops of optically thick low clouds and to the sea surface through the broken clouds is calculated.
The DAOD model depends on the weighting function and the measured range. Given the lidar online and offline frequencies, the measured range to the backscattering surface—that is, cloud top and sea surface—and the temperature, the pressure, and relative humidity profiles taken by balloon data, the Line-by-Line Radiative Transfer Model (LBLRTM) forward model provided by Atmospheric and Environmental Research, Inc. (AER) was run to calculate the modeled DAOD value assuming a 395-ppm CO2 column average (Spiers et al. 2011).
The calculated and lidar-measured DAOD values for the return signal for cloud top and sea surface are listed in Table 6. Forty seconds of data for the ocean return and 40 s of data for cloud return are used to calculate the lidar-measured DAOD. The modeled and lidar-measured round-trip DAOD for sea surface are 1.094 and 1.0717 ± 0.0036, respectively. The modeled and lidar-measured round-trip DAOD for cloud tops are 0.782 and 0.7565 ± 0.0032, respectively. The measured DAOD is a result of 10-s averaging.
The ΔDAOD is likely to be more important for comparison, since the lidar system likely has an instrumental offset that must be carefully calibrated out before comparing measured DAOD values with the model-predicted DAOD values. The ΔDAOD measured from the lidar return is 0.315, and the ΔDAOD calculated from the model is 0.312. The difference between 0.315 and 0.312 is about 1%. The magnitude of the change is consistent with the calculated/predicted change. This is evidence that with this method, we can determine the boundary layer CO2 mixing ratio with high precision.
It is expected that cloud penetration depths at a 2-μm wavelength can vary over a wide range, depending on the cloud morphology. The cloud return signal width should be broader than that of the return from bare soil or from water. The effective lidar signal width can be expressed as (Ehret et al. 2008)
where the ΔtL is the pulse width of the laser, B is the electrical bandwidth of the detector/amplifier, Δh is the effective target depth within the footprint of the laser pulse, and c is the speed of light. The bandwidth is 10 MHz in the IPDA lidar design. At a relatively calm weather condition, the sea surface amplitude turbulence is small; thus, the lidar signal width is largely dependent on the laser pulse width. Fig. 11a depicts the laser pulse width and lidar signal width. Only the online laser pulse width and the online return signal width are displayed. The mean value for the laser pulse width measured by the laser monitor channel is 198 ns. The return signal pulse width is slightly larger with a mean value at 201 ns, which agreed well with the calculated Δteff using Eq. (11).
The width of the lidar return signal from a cloud is, however, significantly broadened. It is clear from Fig. 10 that there are “pockets” where the pulses penetrate more into the cloud, resulting in significant return pulse spreading. The laser pulse width should not change at cloud condition, as evidence in Fig. 11b. The return signal pulse width, though, is significantly increased from cloud reflection. It increased to a mean value of 246 ns with a maximum value at 630 ns as shown in Fig. 11b. It is the evidence that the lidar signal width is highly dependent on the cloud morphology.
f. CO2 mixing ratio measurement validation
It is required to have in situ CO2 mixing ratio data with a meteorological data profile to validate the IPDA lidar CO2 mixing ratio measurement capability. This was accomplished during flight 8, when the 2-μm IPDA lidar flew with a NOAA air-sampling flight at the same geophysical location. NOAA routinely conducts airborne air sampling at selected locations to collect atmospheric data. The programmable multiflask sampling system collects air samples. The samples were later analyzed in the laboratory to obtain air trace gas content, which can reach ±0.03-ppm precision for CO2 mixing ratio measurement per sample (Conway et al. 1994). It also measures temperature, pressure, relative humidity, and other trace gases. These data can be used to make direct comparison and validation for the 2-μm IPDA lidar CO2 mixing ratio measurements.
On 5 April 2014, NOAA conducted an air-sampling flight over the Atlantic Ocean off the coast of Cape May, New Jersey (CM; 38.83°N, 74.32°W). Flying the IPDA lidar over the ocean provides a target with near-consistent surface reflectivity, which tends to reduce measurement uncertainty compared to elevated continental grounds, which vary both reflectivity and scattering surface elevation. The NOAA flight collected data at seven different altitudes, starting from 6126 m and gradually descending to 912 m (6126, 5243, 3977, 3052, 2127, 1505, 912 m). It provided coarse vertical CO2 and meteorological data profiles. Because of airspace restriction, our flight flew over the same location half an hour after the NOAA flight. The IPDA lidar flew at the same altitudes as the NOAA flight. The online frequency was set at 4 GHz from the R30 line absorption peak for the flight altitude above 3052 m. The online frequency was changed to 3 GHz from the R30 line absorption peak below a flight altitude of 3052 m because of less absorption due to shorter range. At an altitude of 3052 m, the data with an online frequency shift at both 3 and 4 GHz was taken.
The profiles of CO2 mixing ratio, temperature, pressure, and water vapor from the ground to 8 km can be obtained by linear extrapolation of the NOAA data. According to Eq. (2), the DAODcd,c can be calculated using the profiles obtained from NOAA data and spectroscopy data from HITRAN 2012 (Rothman et al. 2013). Here the DAODcd,c is calculated from the NOAA-measured CO2 mixing ratio and meteorological data profile. Then the CO2 weighted-average column dry-air volume-mixing ratio, Xcd,c, can be derived from Eq. (5).
IPDA lidar measures the DAOD according to Eq. (4). The lidar-measured CO2 weighted-average column dry-air volume-mixing ratio, Xcd,m, is also derived using Eq. (5). The denominator in Eq. (5) to calculate the Xcd,m is the same as used to calculate the Xcd,c. Then, the lidar-measured Xcd,m can be directly compared with the calculated value Xcd,c (Refaat et al. 2016). The subscripts m and c represent the IPDA lidar measurement and the calculation result from the NOAA-measured CO2 mixing ratio and meteorological data profile, respectively.
Figure 12 shows the Xcd comparison measured by IPDA lidar instrument Xcd,m, and modeled from NOAA in situ instrument Xcd,c. Only the IPDA lidar operated at the online frequency of 4 GHz from the R30 absorption peak is included in the plot. The IPDA-measured Xcd,m value is the result of the 100 pulse average to reduce the error introduced from random noise. The difference between xcd and Xcd,c is between 3 and 4 ppm, resulting from xcd is for the local value at the flight altitude and the Xcd,c is derived from NOAA measurements to account for the weighted column–integrated value. The difference between xcd and Xcd,c is expected based on the weighting function definition. The Xcd,c value at a certain altitude is a weighted integration of xcd from that altitude to the surface. The direct comparison between Xcd,c, and Xcd,m, revealed that the column-integrated CO2 mixing ratio measured by the IPDA lidar instrument is higher than that derived from NOAA flask air sampling. The average difference is 1.4775 ppm, which corresponds to a 0.36% difference between the two instruments. This direct comparison between the two independent measurements validates the high-precision measurement capability of the 2-μm double-pulsed IPDA lidar instrument.
NASA LaRC has developed a new double-pulsed, 2-μm integrated path differential absorption (IPDA) lidar instrument for atmospheric CO2 measurement. One advantage of the pulsed IPDA remote sensing technique is that it provides a high signal-to-noise ratio measurement with accurate ranging. The 2-μm CO2 IPDA transmitter is capable of producing double pulses with 90-mJ energy for online frequency pulses and 45-mJ energy for offline frequency pulses at a 10-Hz repetition rate. High accuracy, stable and repeatable wavelength control, and switching units have been integrated within the transmitter. The IPDA lidar structure was compactly and ruggedly packaged to fit in the NASA B200 research aircraft. Ground and airborne testing of the 2-μm IPDA lidar was conducted at NASA LaRC through several validation procedures. It clearly shows the capability of the IPDA lidar to measure the CO2 density in the presence of cloud conditions. IPDA CO2 differential optical depth measurement results agree well with model prediction. With 10-s average, the standard deviation of the DAOD measurement is 0.0145. Compared to the CO2 mixing ratio measured by NOAA flask sampling data, the 2-μm IPDA lidar provided an accurate measurement with 0.36% difference.
We thank the NASA Earth Science Technology Office (ESTO) for funding this project (ATI-QRS-12-0002). The authors acknowledge Charlie M. Boyer, James Fay, Susan G. Johnston, and Luke S. Murchison for their excellent engineering contributions to building this instrument. The authors are particularly indebted to Dr. Robert Menzies at JPL for his invaluable input on the design of the instrument and for his help in data retrieval. We also thank the Engineering and Research Services Directorates at NASA Langley Research Center for their support. Thanks are also due to the dedicated efforts of the Research Systems Integration Branch, which made airborne flight testing possible. Acknowledgments are also due to the LaRC CAPABLE team and NOAA for providing public information that is significant for the science validation process. The authors also would like to thank Dr. Michael J. Kavaya, Dr. Syed Ismail, Dr. Yingxin Bai, and Tony Notari at LaRC for the very useful discussions, advice, and assistance.