Open Access
Add to BibSonomy
Abstract
We report on an airborne demonstration of atmospheric oxygen (O2) optical depth measurements with an Integrated Path Differential Absorption (IPDA) lidar using a fiber-based laser transmitter and photon counting detectors. Accurate atmospheric temperature and pressure measurements are needed for NASA’s Active Sensing of CO2 Emissions over Nights, Days and Seasons (ASCENDS) mission. Since O2 is uniformly mixed in the atmosphere, its spectrum can be used to estimate the dry mixing ratio of CO2. In its airborne configuration, the IPDA lidar uses an Erbium Doped Fiber amplifier, a frequency doubler and single photon counting detectors to measure O2 absorption at multiple wavelengths near 765 nm. This instrument was deployed in 2013 and 2014 aboard NASA’s DC-8 airborne laboratory as part of two campaigns to measure CO2 mixing ratios over a wide range of topography and weather conditions from altitudes ranging between 3 km and 13 km. In this paper we will review a summary of the results from our flights, discuss the errors that limit the precision and accuracy of the measurement and identify possible areas of improvement.
© 2017 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Atmospheric carbon dioxide (CO2) is currently recognized as having the largest radiative forcing of all greenhouse gases [1]. The current volume mixing ratio of CO2 is approximately 400 parts per million (ppm), a significant increase from the pre-industrial age level of 280 ppm [1]. CO2 has both natural and anthropogenic sources and sinks some of which are not very well understood or accurately measured on regional and global scales. The last US National Research Council (NRC) Decadal Survey for Earth Science [2] has recommended that NASA implements a laser based space mission called ASCENDS (Active Sensing of CO2 Emissions over Nights, Days, and Seasons) to measure global CO2 fluxes. The ASCENDS mission is planned to have sufficient accuracy to infer regional CO2 terrestrial and oceanic sources and sinks on a global scale. The ASCENDS working group has published a Science Definition Study [3] defining the science objectives of the mission. These are: “1) to quantify global spatial distributions of atmospheric CO2 on scales of weather models in the 2010-2020 era; 2) quantify the current global spatial distribution of terrestrial and oceanic sources and sinks of CO2 on 1 degree grids at weekly resolution; and 3) provide a scientific basis for future projections of CO2 sources and sinks through data-driven enhancements of Earth system process modeling”. The required measurement uncertainty for the dry air density of CO2 (XCO2) is 0.25% or 1 ppm.
In order to achieve this kind of precision and accuracy for XCO2, knowledge of the surface pressure, temperature, and water vapor are also needed [3]. Since Oxygen (O2) is a stable and uniformly mixed molecule in the atmosphere at 20.95%, it can be used to infer the surface pressure. Initial analysis by the ASCENDS working group and Crowell [4] shows that in order to keep the XCO2 error below 1 ppm, an error of ~0.2% (or ~2 hPa) will be needed for O2 over a 10 sec averaging period. The precision and accuracy requirement are very stringent and a lidar capable of making such measurements has to be able to address both random and systematic error sources.
Surface pressure information can be obtained from meteorological sensors and numerical weather prediction models. The European Space Agency A-SCOPE Mission Assessment report [5] had determined that for most of the globe, meteorological models could provide the necessary surface pressure information and there was no need for an Oxygen lidar. However, questions remain regarding the spatial resolution of the meteorological models and the lack of information in parts of the globe where sensors are sparse or not available. The ASCENDS working group has retained co-located Oxygen lidar measurements as a requirement. In this paper, we do not attempt to address the question whether numerical weather prediction models will provide adequate coverage and accuracy. Instead, we provide a summary of our results with an O2 Integrated Path Differential Absorption (IPDA) lidar from two airborne campaigns and discuss the errors that limit the precision and accuracy of the measurement and identify areas of improvement needed to meet the requirement.
2. Spectroscopy and line selection
Satellite-borne measurements of O2 with a laser using the A-band were proposed by Singer [6], Barton [7], and Mitchell [8] and were demonstrated with airborne lidar systems by Korb in 1983 [9], Schwemmer in 1987 [10], and more recently by Dobler [11] and Riris [12]. The selection of a suitable O2 line must meet several requirements: The line should not have interferences from other atmospheric constituents; it must not form a continuum with adjacent lines and completely extinguish the laser; it should have low temperature sensitivity; suitable lasers, detectors and other component technologies must be available in the selected spectral region to make the measurement possible. Two spectral regions at ~761 nm and ~1270 nm are best suited for O2 remote sensing measurements from space. The lines in these regions do not have interferences from other atmospheric constituents and have been used to make measurements from an airborne platform [11, 12].
For our lidar, we selected the 761 nm spectral region primarily because suitable lasers and photon counting detectors are available in this region. Figure 1 shows the two-way transmittance of the O2 A-band from space and the change in transmittance when the temperature in the boundary layer (assumed to be at 2 km) is changed by 1 K.
Fig. 1 Two-way transmittance of O2 A-band (black trace) and the change in transmittance x1000 for 1 K temperature change (red trace) in the atmospheric boundary layer (lowest 2 km). The minimum temperature sensitivity occurs near 760 nm and near 765 nm. The HITRAN 2008 [13] was used for the transmittance calculations.
Most O2 lines at shorter wavelengths are too strong and completely extinguish the laser. The minimum temperature sensitivity occurs near 760 nm and near 765 nm. The lines near 760 nm are not suitable for remote sensing from space since they are too strong and almost form a continuum. The lines near 765 nm are better suited for remote sensing from space. We selected an O2 doublet at ~764.7 nm for our lidar (Fig. 2). These transitions (P13Q12 and P13P13 at 764.6296 and 764.7407 nm respectively) are separated from adjacent lines, have no significant interferences from other atmospheric constituents and their temperature sensitivity is small.
Fig. 2 Two-way transmittance of the P13Q12 and P13P13 transitions of the Oxygen A-band at 764.6296 and 764.7407 nm respectively (black trace) and the change in transmittance x1000 (red trace) for 1 K temperature change in the atmospheric boundary layer (lowest 2 km). The two lines have clear separation from adjacent lines, have no significant interferences from other atmospheric constituents and their temperature sensitivity is smaller than other lines. The weaker, narrower lines are O2 isotope lines. The HITRAN 2008 [13] was used for the transmittance calculations.
3. Measurement approach and airborne instrument description
Our approach uses Integrated Differential Path Absorption (IPDA), which has been used by several groups to measure CO2, CH4, and O2 from airborne platforms [14–24]. Our technique uses a sequence of laser pulses at increasing wavelengths, which sample the selected pair of absorption lines in the Oxygen A-band at 764.7 nm. The laser pulses are generated by an Erbium Doped Fiber Amplifier (EDFA) and a frequency doubler and are detected by single photon counting modules (SPCMs). Errors due to scattering from aerosols and clouds are mostly eliminated since our lidar directly measures the range from the aircraft to the surface [25]. Using the instrument in a sounding (surface reflection) mode enables integrated O2 measurements with relatively modest laser power. The multi-wavelength approach attempts to minimize systematic errors, such as etalon fringes and baseline structure that may be difficult to account for in a two-wavelength lidar [26, 27]. We validated this approach in several airborne campaigns with a multi-wavelength IPDA CO2 lidar. Since the multi-wavelength IPDA fits the entire lineshape it can account for the spectral shift of the absorption line with changing atmospheric pressure [28] and can retrieve mixing ratios above the boundary layer [29] and backscatter profiles [30].
Figure 3 shows the selected O2 absorption lines and the wavelengths we used in 2013 and 2014. The exact wavelength positions can be adjusted programmatically and do not need to be evenly spaced (in our 2013 flights they were not evenly spaced). Without considering effects like scattering or stimulated emission from ambient O2 [31] the differential optical depth (DOD) for a two-wavelength lidar (on and off the absorption line) for an atmospheric column R can be written as the ratio of the transmittance through the column at the two wavelengths:
where , is the normalized transmittance through the atmospheric column, and are the on and off line wavelengths, is the molecular absorption cross-section at the on and off wavelengths for each atmospheric layer of the column and is a function of pressure, p, temperature, T and the linestrength, S, R2-R1 is the total length of the atmospheric column R, and is the number density of the molecular species of interest (in this case O2). If the DOD can be accurately measured by measuring the ratio of the normalized transmittance through the atmospheric column and the cross sections and atmospheric conditions are adequately known, then N can be determined by inverting Eq. (1).Fig. 3 Transmittance of the Oxygen A-band at 764.7 nm using a 10 km US standard atmosphere and the HITRAN 2008 database (black) along with the wavelengths we used in our 2013 (red) and 2014 (blue) flights. The 2014 wavelengths were evenly spaced but the 2013 were not.
A simple way to illustrate the measurement requirement is to plot the OD for slightly different elevations. Increasing the elevation reduces the air mass (i.e. pressure) in the atmospheric column. For a US standard atmosphere, an increase of ~17 m in elevation will result in a change of ~2 hPa which is the measurement requirement. Figure 4 shows the OD from a 400 km orbit for zero and 17 m elevation and the corresponding OD difference, Δ(OD). The ODs for the two cases are virtually identical (the plots overlap). The OD difference Δ(OD), is plotted on an expanded scale on the right hand axis for clarity. The other obvious observation from Fig. 4 is that there is no real “off” wavelength. The wings (along with several O2 isotope lines on either side) extend far enough in wavelength (or frequency) to interfere with adjacent lines and the OD is different on left and right side. Our choice of “on” and “off” wavelengths is arbitrary and for our case is shown in solid squares in Fig. 4. We define the differential optical depth (DOD) as the difference in OD at the on wavelength and the average OD at the off wavelengths:
Fig. 4 Optical depth (OD) using a US standard atmosphere from a 400 km orbit with two different observer elevations (0 m and 17 m). An elevation change of 17 m will result in a change in pressure of ~2 hPa which is the ASCENDS measurement requirement for O2. The two plots (black and red trace) are virtually identical and overlap (left axis). The difference in OD, Δ(OD), is shown on the right hand axis in blue. Our choices for “on” and “off” wavelengths for the differential optical depth (DOD) calculations using Eq. (2) are shown in solid black squares.
It is clear from Fig. 4 the IPDA lidar must have the accuracy and precision to measure very small changes in OD and account for both random and systematic errors.
Our 2013 instrument (Figs. 5 and 6) was part of a combined CO2-O2 lidar. The transmit CO2 and O2 beams are combined by a dichroic beam combiner with a small angular separation (500 µrad) and the receiver telescope has two fiber-coupled fields of view (FOV) corresponding to each transmit beam. The O2 lidar used a continuous wave (CW) distributed feedback (DFB) diode laser (FITEL FRL15DCWD-A61-19600-C) operating at 1529.4 nm whose current and temperature were controlled by a commercial laser driver (SRS LDC502). The diode laser wavelength was rapidly scanned (at 500 Hz) over the O2 absorptions by applying a ramp waveform to the drive current. The frequency, amplitude, and shape of the wavelength scan waveform can be adjusted using a computer-controlled waveform generator. A wavelength calibration procedure using a heterodyne technique, a wavemeter (Burleigh WA-1650) and an acetylene cell was used to calibrate our wavelength scan. The inherent diode laser linewidth is very narrow (~1 MHz per FITEL specifications). The output of the CW diode laser was externally modulated with a fiber-coupled acousto-optic modulator (AOM) to yield relatively short (~250 ns FWHM) laser pulses with approximately 25 dB extinction ratio. The pulse frequency and pulse shape and width were easily adjusted by modifying the analog waveform applied to the AOM (EM4 part number: EM417). A master trigger, from a GPS receiver 1 pulse per second (pps) signal, initiated a wavelength scan with 20 laser pulses separated by 100 µs that were used to sample the oxygen absorption lines. The 100 µs time separation between pulses (equivalent to a range of ~15 km) ensured that all wavelength pulses in the scan were sufficiently separated in time so that there was no ambiguity on each pulse wavelength when it was detected by the receiver.
Fig. 5 Simplified functional block diagram of the 2014 IPDA lidar. In 2014, we replaced the single SPCM with eight SPCMs and the multichannel scaler with a PXI based data acquisition system using an FPGA digital counter. The output of the diode laser was externally modulated with a fiber-coupled acousto-optic modulator (AOM) to yield ~200 ns pulses, which are amplified by the EDFA and then doubled to ~764.7 nm. A 500 Hz wavelength sweep was used to scan over the oxygen absorption lines with 20 laser pulses separated by 100 µs.
Fig. 6 (Left) The O2-CO2 lidar transceiver with associated racks in the DC-8 during the 2014 campaign. The transmit CO2 and O2 beams are combined by a beam combiner and are separated by a small angular offset (500 µrad). The receiver telescope has two fiber-coupled fields of view, separated by 500 µrad. (Right) The anti-reflection coated transmitter and receiver windows mounted on the nadir port of the DC-8.
The optical pulses from the AOM were amplified by an Erbium Doped Fiber Amplifier (EDFA) made by NP Photonics Inc. (custom item). The EDFA output is directly fiber-coupled into a periodically poled Lithium Niobate crystal (PPLN) assembly (AdVR Inc. custom item) which frequency doubles the 1529.4 nm laser light to 764.7 nm. The free-space output from the doubling crystal is directed to the transmit optics assembly which includes two turning mirrors, a beam combiner, a beam expander to reduce the beam divergence to ~110 µrad and an integrating sphere with a detector (Thorlabs PDA36A) to measure the outgoing laser energy.
The output signal from the energy monitor is integrated by a gated integrator (Signal Recovery Model 4121B) and then digitized by an analog to digital converter (ADC) to measure accurately the laser pulse energy. The outgoing laser energy after the transmit optics was ~1.0-1.5 µJ and it exhibited large fluctuations due to the large mode area (LMA) fiber of the amplifier. The transmitted laser pulses travel through the aircraft nadir window to the ground. The nadir port windows are anti-reflection coated (AR) coated for 765 nm and have a wedge to minimize unwanted etalon fringes and back reflections into the receiver. The Nominal Ocular Hazard Distance (NOHD) for the lidar during flight operations for unaided viewing was 285 m, slightly below the 305 m (1000 ft.) minimum separation between aircraft mandated by air traffic control (ATC). As an added precaution, the laser is always turned off any time the aircraft is below 914 m (3000 ft.) above ground level (AGL) or in the unlikely event, that another aircraft appears capable of passing beneath the NASA aircraft at a range that would present an eye-safety risk. If the laser energy were increased, the beam characteristics (beam size, divergence) will need to be modified to meet ATC requirements. The spot diameter from a 10 km altitude is 1.1 m and the separation between successive pulses (wavelengths) is 2-2.5 cm, using a nominal aircraft speed of 200-250 m/s. This separation minimizes the changes in reflectivity between successive wavelengths.
The reflected ground echoes are collected by a commercial 20 cm diameter receiver telescope (Vixen VC200L) and are coupled into an AR coated 400-µm core multi-mode fiber (Fiberguide CB18166). The receiver field of view (FOV) is determined by the telescope effective focal length (2 m), the fiber core size, and its numerical aperture (NA) and our FOV was 200 µrad. The fiber output from the receiver is collimated and directed through a narrow (0.5 nm FWHM) bandpass filter made by Materion (formerly Barr Associates), an adjustable iris to adjust the amount of light onto the detector, and then focused onto a single photon counting module (Perkin Elmer SPCM-AQRH-12). The fiber collimator, filter, iris, and focusing lens reside in a single opto-mechanical assembly to minimize alignment sensitivity and optimize the transmission of the bandpass filter. The SPCM output is sent to a multi-channel scaler (Quantar technology P7889) which produces a histogram of the return pulses as a function of time (or range) over the entire atmospheric column. The bin width for the histogram was 32 ns and the averaging time was 1 s. The computer averages, digitizes, and stores the histograms over 1 second. The averaging period is adjustable but is limited by the data transfer rate. The duty cycle for the data acquisition was 96%. By digitizing returns from the entire atmospheric column, we can separate contributions from clouds and the ground, and determine the range, R, to the ground using the time of flight (TOF) of the first laser pulse [25].
In 2014, we implemented several improvements to the IPDA lidar. The biggest improvements were in the receiver and the data acquisition. The dynamic range of the receiver is an important consideration since it has to respond to large changes in signal and background. The on wavelengths (the wavelengths that are coincident with the peak of the O2 absorption lines) are completely extinguished and do not contain any laser signal (only solar background and noise) but the off wavelengths in the wings contain both laser and background signals. Thus, our receiver needs to have a very large dynamic range and respond quickly to changes in signal and background. Our dynamic range is limited by the SPCM whose counts need to be corrected as the signal or the solar background increases (this correction is typically called dead time correction). However, in practice it is very difficult to implement a reliable correction at the 0.2% level over several orders of magnitude of signal and background observed during our flights. In order to increase the dynamic range of the receiver we split the output of the multi-mode receiver fiber into eight separate SPCMs using a 1x8 fiber optic splitter. Although there is a significant insertion loss in the splitter (~15-20%), the increased dynamic range improved our IPDA lidar performance by operating the SPCMs in their linear range.
In 2014, we also replaced the old data acquisition system with a National Instruments PXI system based on analog digitizer and a field programmable gate array (FPGA) digital counter. The PXI analog digitizer digitized the outgoing laser energy pulse and the counter, digitized the ground echo (return) pulses by summing the outputs of the eight SPCMs to produce the return histogram from the atmospheric column. Another data acquisition improvement in 2014 was the reduction of the integration time from 1 s to 125 ms (1/8 sec). Long averaging times can mask large but short-lived changes in signal and/or background signals that cannot be corrected in post-processing once the data has been averaged. By reducing the averaging time, we could better identify these large changes in signal or background and either exclude the data from the analysis or apply a dead time correction. We had also planned to increase the laser energy by almost an order of magnitude to ~10-15 µJ using a new power fiber amplifier (NP Photonics custom item). Unfortunately, the power amplifier was not available on time for the 2014 flights. It was delivered and tested shortly after the flights. The last modification we did in 2014 was to increase the receiver bandwidth by changing the width of the bandpass filter. The FWHM of the filter was increased to 0.8 nm from 0.5 nm. Although this design change seems counter intuitive since a wider bandpass filter would allow more solar background onto the SPCMs we were willing to accept higher solar background but reduce systematic errors that arose from a narrow band pass filter. Our wavelength scan was approximately 0.4 nm, which is comparable to the width of O2 absorption lines. Although a narrow filter reduces solar background, the wings of the measured lineshape are severely distorted by the filter (Fig. 7). With a careful calibration, it is possible to deconvolve the filter response from the measured lineshape. However, the filter response is sensitive to angle and temperature and in practice, it is difficult to achieve and maintain a <0.2% calibration precision during flight. By going to a wider filter, we were hoping to reduce the lineshape distortion by the filter.
Fig. 7 Our wavelength scan is approximately 0.4 nm, which is roughly equal to the width of O2 absorption lines. Although a narrow filter reduces solar background, the wings of the measured lineshape are severely distorted by the filter (red curve). To reduce the distortion the FWHM of the filter in 2014 was increased to 0.8 nm from 0.5 nm (blue curve). Although this design change would allow more solar background illumination by a factor of 1.6 we were willing to accept higher solar background in order to reduce systematic errors that arose from the lineshape distortion due to the narrow band pass filter.
The parameters of the airborne system in 2013 and 2014 are summarized in Table 1 below.
Table 1. Parameters of the airborne IPDA Lidar in 2013 and 2014
4. Results from 2013 to 2014 airborne campaigns
In 2013 and 2014, we participated in a multi-instrument airborne campaign sponsored by the NASA ASCENDS program to measure CO2 fluxes in the United States. The O2 lidar was part of a Goddard Space Flight Center (GSFC) instrument suite, which included a CO2 IPDA lidar and an in situ cavity ring down spectrometer (Picarro G1301-m). The instruments were installed on NASA’s DC-8 airborne laboratory based at Armstrong Flight Research Center Science Aircraft Integration Facility (SAIF) in Palmdale, CA.
Five science and one engineering flight in the continental US (CONUS) were carried out in 2013 and one engineering flight and four science flights in 2014. The O2 lidar collected data for all 2013 flights but only one 2014 flight. At the start of our 2014 flight campaign a laser current driver in our fiber amplifier failed. By the time we identified the problem, received and installed a replacement driver the ASCENDS airborne campaign was nearly complete and we collected data only during the last flight. In addition, data from the first leg of the 2014 flight (to Iowa) was compromised by telemetry issues with the adjustable iris. We managed to correct the problem during flight but we only collected data for the return segment of the flight (from Iowa to Palmdale). This unfortunate turn of events limited our ability to assess the improvements we implemented in 2014.
The flight paths and spiral locations for the campaign were selected to optimize science objectives for CO2 and were subject to air traffic control clearances. They typically included multiple segments at increasing altitudes from 3 to 13.5 km over varying topography, land cover, and atmospheric conditions. In addition, for most flights, a spiral descent from ~13.5 km to near the surface (30-70 m) was included in the flight plan in order to sample vertical profiles of meteorological parameters (pressure, temperature, humidity, etc.) using the aircraft’s on-board sensors and the CO2 mixing ratio profile using the in situ sensor. Figure 8 and Table 2 summarize all the 2013 and the 2014 flight that the O2 lidar collected data.
Fig. 8 Flight track summary of the 2013 and the one 2014 flight that the O2 lidar collected data. The flight paths and locations were selected to optimize science objectives for CO2 fluxes and they typically included multiple segments at increasing altitudes between 3 and 13.5 km over varying topography, land cover, and atmospheric conditions. In addition, for most flights, a spiral descent from ~13.5 km to near the surface (30-70 m) was included in the flight plan in order to sample vertical profiles of meteorological parameters (pressure, temperature, humidity).
Table 2. Flight Summary 2013-2014
Our retrieval algorithm (Fig. 9) first isolates the ground return and determines the range to the surface by correlating the first return pulse with the outgoing energy monitor pulse and measuring the time delay of the correlation peak [25]. Fixed time delays due to fibers and electronics are calibrated prior to flight during ground testing. The lidar range is also compared with the GPS altitude and the DC-8 on-board radar altimeter. This ensures that only ground returns are used in the O2 retrievals and cloud returns are flagged accordingly. Then the ground return pulses at each wavelength are integrated and a dead time correction is applied. The returns are normalized by the integrated transmitted pulse energy, the filter response, iris size, and other instrument calibrations. The data is further averaged in 10 s intervals and the algorithm then compares the experimentally derived transmittance with the theoretically calculated transmittance values and adjusts the fit parameters to minimize the root mean square error (RMSE). The theoretical predictions use the lidar range and DC-8 radar altimeter information to obtain the range to the ground, a Voigt lineshape, and the lineshape parameters from the HITRAN 2008 database [13], the vertical profile of the atmosphere, and line-by-line radiative transfer calculations [32]. The impact of Dicke narrowing, line mixing, collision-induced absorption and ambient airglow were not included in the calculations [33–35]. The meteorological data for the vertical profile of the atmosphere for the flights are obtained from the Goddard Modeling and Assimilation Office (GMAO) Modern Era Retrospective Analysis for Research and Applications [36]. The theoretical and experimental DOD values are then determined by the difference in optical depth (OD) at 764.684 nm (the on wavelength) and the average value of the OD at the off wavelengths (764.509 and 764.903 nm). The DOD is used as an estimate of the instrument’s performance.
Fig. 9 Our retrieval algorithm identifies and integrates the normalized ground returns and applies the necessary corrections and calibrations. The algorithm then compares the experimental with theoretically calculated transmittance values and adjusts the fit parameters to minimize the rms error.
Figures 10(a) and 10(b) show an example of the DOD time series and the corresponding scatterplot for the lidar DOD and model DOD prediction for our first flight on 22nd of February 2013, near Blythe, AZ and southern Sierra Nevada. The lidar DOD agreed well with the model DOD prediction and the linear fit of the scatterplot had a slope of 0.95 and an offset of 0.02. The R2 value was 0.96. The lidar DOD for our flight on 26th of February 2013, in Railroad Valley, NV also agreed well with the model DOD prediction and the linear fit of the DOD scatterplot had a slope of 1.08 and an offset of −0.03. The R2 value was 0.94 and a 10 sec averaging period was used.
Fig. 10 (a) Time series of the model DOD prediction and the lidar DOD for our first flight on 22nd of February 2013, near Blythe AZ. (b) Scatterplot of the same data. A linear fit of the scatterplot had a slope of 0.95 and an offset of 0.02. The R2 value was 0.96. A 10 sec averaging period was used.
Figures 11(a) and 11(b) show another example of the results from the February 28 2013 flight in Central Valley, CA. The lidar DOD agreed well with the model DOD prediction and the linear fit of the scatterplot had a slope of 1.04 and an offset of −0.02. The R2 value was 0.97. The lidar DOD for our flight on March 7 2013, to southern WI and eastern IA agreed well with the model DOD prediction and the linear fit of the DOD scatterplot had a slope of 1.02 and an offset of −0.03. The R2 value was 0.93 and a 10 sec averaging period was used. The other two flights in Northern California and southern UT and western CO produced similar results. Finally, Figs. 12(a) and 12(b) show the results from the September 3, 2014 flight from eastern IA to Palmdale. A linear fit of the scatterplot had a slope of 1.00 and an offset of 0.00. The R2 value was 0.97 and a 10 sec averaging period was used.
Fig. 11 (a) Time series of the model DOD prediction and the lidar DOD for our flight on 28th of February 2013, Central Valley, CA. (b) Scatterplot of the same data. A linear fit of the scatterplot had a slope of 1.04 and an offset of −0.02. The R2 value was 0.97. A 10 sec averaging period was used.
Fig. 12 (a) Time series of the model DOD prediction and the lidar DOD for our flight on 3 September 2014, from Iowa to Palmdale. (b) Scatterplot of the same data. For this flight, we analyzed data only during the return leg of the flight (from Iowa to Palmdale). A linear fit of the scatterplot had a slope of 1.00 and an offset of 0.00. The R2 value was 0.97. A 10 sec averaging period was used.
The agreement between the DOD prediction and the lidar DOD depends on the IPDA lidar performance but also on the accuracy of the theoretical prediction at that location. If our model atmosphere is not an accurate representation of what is actually being measured by the lidar then the agreement will not be very good. This may be particularly problematic in areas where the topography varies rapidly or where our weather models do not accurately capture the local state of the atmosphere. For all of our flights the model DOD prediction and the lidar DOD agreed fairly well. The average slope of the scatterplots of all the analyzed flights was 1.018 and the average offset was −0.012. These results indicate that our measurement accuracy is within a few percent (<2%) of the predicted DOD value with a modest amount of laser energy. The precision of our measurement can be estimated from the difference between the model DOD prediction and the lidar DOD. The Central Valley flight is probably suited for the precision estimation since it was mostly over flat terrain, except for the portion of the flight northwest of Palmdale where it crossed the southernmost end of the Sierra Nevada. The standard deviation of the difference between the model DOD prediction and the lidar DOD for that flight was 2.5% with a 10 sec averaging time.
5. Discussion and areas of improvement
The high accuracy and precision needed for the measurement pose many challenges for the instrument design. An IPDA lidar like all spectroscopic instruments regardless of the approach will have systematic and random error sources that limit accuracy and precision. The total error is an aggregate of random and systematic errors.
The random errors for an IPDA lidar has several contributions that have been addressed in detail by several groups, and will not be reproduced here (see for example Ehret [16], Sun [37], Kiemle [38], and Refaat [39]). Typically they include detector excess and/or dark noise, signal shot noise, thermal noise, solar background, amplifier circuit noise, speckle noise, laser noise, etc. Most of these noise terms have normal or 1/fα distributions (typically 1<α<2) and the Signal to Noise ratio (SNR) can be improved by increasing the averaging time.
Our SNR is currently limited by the low laser energy. The SNR for a single wavelength of a pulsed IPDA lidar was derived by Sun [37]. Using Eq. (29) from Sun, and the IPDA lidar values in Table 1 for our off wavelength pulse (last pulse), with a 10 sec averaging period the best we could expect is a SNR of ~68 or a precision of ~1.6%. Thus, we can never expect our DOD precision to exceed the theoretical SNR limit for a single wavelength. After the flights, we received and tested the repaired power amplifier and scaled the 1529.3 nm energy up to 60 µJ with a corresponding energy increase at 764.7 nm up to 9-10 µJ, a factor of 6-7 increase. With the increased energy, we would expect to reduce the shot noise by a √6 and improve the theoretical precision from 1.6% to 0.67%. However, that is still more than factor of three above the requirement and we would need about 64 µJ to meet 0.2%. Recently demonstrated Raman fiber amplifiers have produced more than 400 µJ per pulse at 1572 nm [40]. If the Raman fiber amplifier design is modified for transmission at 1529.3 nm and it is coupled with an efficient frequency doubling crystal, then 764.7 nm energy could be scaled to 64 µJ or higher.
The solar background is another source of noise. It is proportional to the receiver optical efficiency, the width of the optical bandpass filter, the square of the FOV, the square of the receiver aperture and surface reflectance. We can reduce the solar background by reducing the FOV or the width of the bandpass filter. However, these tradeoffs have practical limitations that are rarely taken into account in theoretical analyses and often efforts to improve a random error may unintentionally exacerbate systematic errors. Reducing the FOV makes boresight alignment difficult unless the beam divergence can also be reduced and a dynamic boresight adjustment mechanism can be implemented. Reducing the width of the bandpass filter is possible but it comes at a cost. Our bandpass filter in 2013 was 0.5 nm wide (FWHM) which was roughly the span of our wavelength scan (Fig. 7). The filter’s narrow spectral width distorted the wings of the line shape. Although we calibrated the transmission of the filter in the laboratory, prior to flight, small changes in the incidence angle and temperature may introduce an additional bias. A change of 1 degree in angle can result in a ~55 pm shift in the center wavelength. Although we do not expect a 1 degree change during flight, even a small change of 1 pm can introduce a ± 0.002 absolute change in the transmittance lineshape (or 0.23%) that needs to be accounted for and is not easily calibrated. The retrieval algorithm in 2013 tried to account for a shift in the filter transmission peak but that introduced another variable into the fitting process. In 2014, we changed the bandpass filter to 0.8 nm to minimize the effect of this systematic error. Although the solar background increased by a factor of 1.6 ( = 0.8nm/0.5nm) we believe the distortion in the wings of the baseline was significantly less and the retrieval algorithm did not need to adjust the filter transmission curve.
The energy monitor is another source of systematic error that is almost never taken into account in most theoretical analyses. Refaat successfully addressed several calibration issues for a 2-µm IPDA lidar and proposed a self-calibration feature to reduce measurement uncertainty [41]. Ideally, the monitor should be a perfect representation of the outgoing pulse energy and can always be used to normalize the received energy. In 2013 one of the biggest issues we encountered was the large energy fluctuations of the EDFA due to movement of the LMA fiber. Energy changes at the EDFA fiber output are then amplified in a non-linear fashion by the doubler. The energy monitor is supposed to capture the energy changes to a precision of 0.2% or better. It is reasonable to expect 0.2% precision in the normalization process when the changes in energy are only 1-2 percent. However, the energy fluctuations we observed were 30% or higher and normalizing to 0.2% was extremely challenging. In 2014, we attempted to address this problem. First, we used a fast digitizer instead of an integrator to record the monitor pulses. The digitizer records the entire monitor pulse waveform rather than returning a single integral value. Unlike the integrator, it does not require manual offset settings that can bias measurements and undermine repeatability. Second, we enclosed the LMA fiber from the EDFA to the doubler in rigid tubing to minimize fiber movement and reduce energy fluctuations. Our stabilization approach was not as effective, as we would have hoped and the energy still varied by more than 5-10%. In the future, fiber and energy stabilization should be addressed at the amplifier design stage.
Even if the energy monitor were perfect various other instrument “drifts” due to etalon fringes, detector responsivity and linearity, temperature changes, polarization and energy changes of the transmitter, and other instruments effects degrade the performance and limit the averaging time of the instrument. While these “drifts” are often difficult to separate and model analytically it is important to quantify the overall “stability” and optimum averaging time of the instrument. The Allan variance is often used as a metric to estimate the stability and the optimum averaging time of a laser spectrometer [42, 43]. Laboratory calibration experiments and ground-based open path tests showed that our instrument is stable for up to ~60 secs. Averaging longer than 60 seconds increased the Allan variance and did not improve the SNR of the IPDA lidar. Similar results were obtained in flight although the flight data are harder to assess since the ground returns vary significantly because of varying reflectivity and range.
In 2013, the SPCM nonlinearity was the most significant systematic error. This was particularly problematic in flights when we flew over bright, highly reflecting clouds. The SPCM dynamic range could not accommodate the rapid change in background signal resulting in lower counts and distortion in the wings of the line shape. In 2014, we significantly reduced this problem by adding seven more SPCMs. We split the output of the multi-mode receiver fiber into eight separate SPCMs using a 1x8 fiber optic splitter and the dynamic range increased significantly. Figure 13 shows a comparison of the normalized correction factor vs. the Photon count rate between one and eight SPCMs from 0 to 100 Mcounts/sec. The eight SPCM correction factor remains below 2 up to 100 Mcounts/sec, whereas the one SPCM correction factor is above 2 at ~17 Mcounts/sec. Another option for the future would be to replace the SPCMs with a low-noise, high quantum efficiency, HgCdTe avalanche photodiode (APD) [44] that has been used in our airborne CO2 and CH4 IPDA lidars.
Fig. 13 Comparison of the normalized correction factor vs. photon count rate between one (red curve) and eight SPCMs (blue curve). The eight SPCMs remain relatively linear up to ~100 MCounts where the single SPCM response is not linear and needs a correction at much lower count rates.
Even with the increased dynamic range of the eight SPCMs significant issues with the retrievals remain, because of the wide scan needed and the strength of the O2 lines. The SPCMs must still be able to respond to a large dynamic range of signals: strong laser ground returns from highly reflective surfaces (for the “off” wavelengths) and complete laser light extinction (for the “on” wavelengths) in a single wavelength scan. In addition, our decision to increase the width of the bandpass filter to minimize the wing distortion had the undesirable effect of increasing the solar background. Finally, the wavelength scan must be wide enough (~0.4 nm) to trace the O2 lines. We scan the wavelength by scanning the diode laser current. The diode laser scan is not perfectly linear and the larger the scan the more difficult it is to calibrate the wavelength. In addition, the diode laser output power changes as function of current. The wider the wavelength span the bigger the power changes. Large diode laser power changes result in large changes in the output of the fiber amplifier and even larger changes in the output of the non-linear doubler. These competing requirements could be reconciled if we used a narrower and weaker O2 line. A weak line that does not completely absorb the laser light will impose less stringent dynamic requirements on the detector. If the line is also narrower, we can reduce the wavelength scan avoid some of the issues associated with the wide scan described above. Finally, with a narrow line we could reduce the width of the bandpass filter without distorting the lineshape and reduce the solar background noise contribution.
Of course, such a line must have suitable temperature dependence and be within the EDFA emission range to be viable. We identified a suitable, weak and narrow O2 isotope line (O16O18) at 764.93 nm that has low temperature dependence. Figure 14 shows the expected transmittance from 400 km for a US standard atmosphere and the transmission curves for two narrow etalon filters, 50 and 100 pm, and the resulting O16O18 transmittance lineshapes using the two etalon filters. The distortion with the 100 pm etalon filter is small, the laser is no longer extinguished at the peak, and the wavelength scan needed to trace the line is only ~20-30 pm vs. 400 pm. Using a 100 pm etalon filter vs. 0.8 nm will also reduce the solar background by a factor of eight. The weaker line should reduce the dynamic range requirements for the SPCMs and the narrower wavelength scan from should make the wavelength calibration easier and reduce the effects associated with the wide scan.
Fig. 14 Expected transmittance from 400 km for a US standard atmosphere (black trace) and the transmission curves for two narrow etalon filters, 50 and 100 pm (blue and red dash traces respectively), and the resulting O2 transmittance lineshapes using the two etalon filters (blue and red solid line traces respectively). The distortion with the 100 pm etalon filter is minimal, and the minimum transmittance is ~30% and the wavelength scan needed to trace the line is only ~20 pm.
We were able to scan over the isotope line with our existing laser but we did not have a suitable etalon filter. Figure 15 shows the theoretical (red) and experimental lidar transmittance (blue dots) obtained from a 3 km open path ground test at GSFC and the transmission curve of our existing 0.8 nm bandpass filter. Our filter, which is centered near 764.7 nm, has a ~30% transmittance at 764.93 nm. However, we were able to retrieve the normalized isotope line transmittance with our lidar, demonstrating the feasibility of the idea. To assess the performance of the lidar for the isotope line a suitable filter would have to be procured and integrated in the instrument. Fitting this line should be significantly easier as the fit will not be as dependent on dynamic range effects that have proven so challenging. Of course, the isotopic fractionation, which can lead to variations in the relative abundances of the O2 isotopologues, must be well-understood [45] if this line is to be used. The 18O/16O ratio of atmospheric oxygen is higher than that of seawater, as first reported by Dole [46]. The difference has been attributed to several processes such as respiration, evaporation, photosynthesis etc [47, 48]. and models have been developed to explain [49] these differences. However, these observations and models address primarily geological timescales and are mostly applicable to paleo climatic studies not a space instrument that is going to operate only for 3-5 years. Over short time scales, the 18O/16O ratio remains constant to 0.03 ‰ and the seasonal variability is less than 0.002‰ [45]. That may be sufficiently small for a space IPDA lidar with a measurement accuracy requirement of 0.2%, over 3-5 years. However, if the isotope line were to be selected, a much better understanding of the 18O/16O ratio and its impact on lidar measurements must be established.
Fig. 15 Theoretical (red) and experimental lidar (blue dots) transmittance from a 3 km open path ground test at GSFC and the associated transmission curve of our existing 0.8 nm filter. Our filter, which is centered near 764.7 nm, has a 30% transmittance at 764.93 nm where the O16O18 isotope line is centered.
6. Summary
We reported the results of an airborne demonstration in 2013-2014 of an Integrated Path Differential Absorption (IPDA) lidar using a fiber-based laser transmitter and photon counting detectors to measure atmospheric oxygen (O2) optical depth. Since O2 is uniformly mixed in the atmosphere, it can be used to estimate the dry mixing ratio of CO2 for ASCENDS. Our multi-wavelength IPDA lidar uses an Erbium Doped Fiber amplifier and a frequency doubler to measure O2 absorptions near 765 nm.
Our average results show that the DOD measurement accuracy is within 2% of the predicted DOD values with a modest amount of laser energy (1-2 µJ) and the precision is 2.5% in a 10 sec averaging time. These results show that significant improvements are needed to achieve the 0.2% accuracy and precision required for ASCENDS. The laser power must by scaled significantly to reduce the random noise contribution but also various systematic errors must be adequately addressed and quantified to achieve the needed accuracy and precision. In 2014, we implemented several improvements to address several systematic errors. Unfortunately, due to equipment failures we were not able to evaluate adequately the impact of these improvements. Post-flight we were able to demonstrate a factor of six increase in the laser energy with a power amplifier. We also demonstrated possible solutions to some of the systematic errors (dynamic range, wavelength calibration and large wavelength scan) by using an isotope line at 764.93 nm. Currently the O2 GSFC IPDA lidar is in hiatus until the new Earth Science Decadal Survey is released and clarifies the need for O2 measurements.
Funding
This work was supported by the NASA Earth Systematic Future Mission program for ASCENDS (304029), the NASA Instrument Incubator Program (IIP) of the Earth Science Technology Office (ESTO) (NNH07ZDA001N-IIP), and the NASA GSFC Internal Research and Development (IRAD) program.
Acknowledgements
The authors would also acknowledge and express their appreciation to the DC-8 flight operations team at the Science Aircraft Integration Facility in Palmdale, CA for flight planning and support.
References and links
1. Intergovernmental Panel on Climate Change, Climate Change 2014–Impacts, Adaptation and Vulnerability: Regional Aspects, (Cambridge Univ. Press, Cambridge, U. K. and New York, 2014).
2. Board, Space Studies, and National Research Council: Earth Science and Applications from Space: National Imperatives for the Next Decade and Beyond, (National Academic Press, 2007).
3. K. W. Jucks, S. Neeck, J. B. Abshire, D. F. Baker, E. V. Browell, A. Chatterjee, D. Crisp, S. M. Crowe, S. Denning, D. Hammerling, F. Harrison, J. J. Hyon, S. R. Kawa, B. Lin, B. L. Meadows, R. T. Menzies, A. Michalak, B. Moore, K. E. Murray, L. E. Ott, P. Rayner, O. I. Rodriguez, A. Schuh, Y. Shiga, G. D. Spiers, J. S. Wang, and T. Scott Zaccheo, Active Sensing of CO2 Emissions over Nights, Days, and Seasons (ASCENDS) Mission, Science Mission Definition Study, (April 2015), http://cce.nasa.gov/ascends_2015/ASCENDS_FinalDraft_4_27_15.pdf.
4. S. Crowell, P. Rayner, S. Zaccheo, and B. Moore, “Impacts of atmospheric state uncertainty on O2 measurement requirements for the ASCENDS mission,” Atmos. Meas. Tech. 8(7), 2685–2697 (2015).
5. ESA A-SCOPE Mission Assessment report, http://esamultimedia.esa.int/docs/SP1313-1_ASCOPE.pdf.
6. S. F. Singer, “Measurement of atmospheric surface pressure with a satellite-borne laser,” Appl. Opt. 7(6), 1125–1127 (1968). [PubMed]
7. I. J. Barton and J. C. Scott, “Remote measurement of surface pressure using the oxygen A-band of absorption,” Appl. Opt. 25, 3502–3507 (1986). [PubMed]
8. R. M. Mitchell and D. M. O’Brien, “Error estimate for passive satellite measurements of surface pressure using absorption in the A band of oxygen,” J. Atmos. Sci. 44, 1981–1991 (1987).
9. C. L. Korb and C. Y. Weng, “Differential absorption lidar technique for measurement of the atmospheric pressure profile,” Appl. Opt. 22(23), 3759–3770 (1983). [PubMed]
10. G. K. Schwemmer, M. Dombrowski, C. L. Korb, J. Milrod, H. Walden, and R. H. Kagann, “A lidar system for measuring atmospheric pressure and temperature profiles,” Rev. Sci. Instrum. 58(12), 2226 (1987).
11. J. T. Dobler, J. A. Nagel, V. Temyanko, T. S. Zaccheo, and B. Karpowicz, “Lidar measurements of atmospheric oxygen using a 1.27-micron Raman amplifier,” in CLEO: Applications and Technology (Optical Society of America, 2011), paper JTuE1.
12. H. Riris, M. Rodriguez, G. R. Allan, W. Hasselbrack, J. Mao, M. Stephen, J. Abshire, R. Kawa, and J. B. Abshire, “Pulsed airborne lidar measurements of atmospheric optical depth using the Oxygen A-band at 765 nm,” Appl. Opt. 52(25), 6369–6382 (2013). [PubMed]
13. L. S. Rothman, I.E. Gordon, A. Barbe, D.C. Benner, P.F. Bernath, M. Birk, V. Boudon, L.R. Brown, A. Campargue, J.P. Champion, and K. Chance, “The HITRAN 2008 molecular spectroscopic database,” J. Quantitative Spectroscopy Radiative Trans. 110(9), 533-572 (2009).
14. J. T. Dobler, F. W. Harrison, E. V. Browell, B. Lin, D. McGregor, S. Kooi, Y. Choi, and S. Ismail, “Atmospheric CO2 column measurements with an airborne intensity-modulated continuous wave 1.57 μm fiber laser lidar,” Appl. Opt. 52(12), 2874–2892 (2013). [PubMed]
15. J. B. Abshire, A. Ramanathan, H. Riris, J. Mao, G. R. Allan, W. E. Hasselbrack, C. J. Weaver, and E. V. Browell, “Airborne Measurements of CO2 Column Concentration and Range Using a Pulsed Direct-Detection IPDA Lidar,” Remote Sens. 6(1), 443–469 (2013).
16. G. Ehret, C. Kiemle, M. Wirth, A. Amediek, A. Fix, and S. Houweling, “Space-borne remote sensing of CO2, CH4, and N2O by integrated path differential absorption lidar: a sensitivity analysis,” Appl. Phys. B 90, 593–608 (2008).
17. H. Riris, K. Numata, S. Wu, B. Gonzalez, M. Rodriguez, S. Scott, S. Kawa, and J. Mao, “Methane optical density measurements with an integrated path differential absorption lidar from an airborne platform,” J. Appl. Remote Sens. 11(3), 034001 (2017).
18. G. D. Spiers, R. T. Menzies, J. Jacob, L. E. Christensen, M. W. Phillips, Y. Choi, and E. V. Browell, “Atmospheric CO 2 measurements with a 2 μm airborne laser absorption spectrometer employing coherent detection,” Appl. Opt. 50(14), 2098–2111 (2011). [PubMed]
19. A. Fix, C. Büdenbender, M. Wirth, M. Quatrevalet, A. Amediek, C. Kiemle, and G. Ehret, “Optical parametric oscillators and amplifiers for airborne and spaceborne active remote sensing of CO2 and CH4,” Proc. SPIE 8182, 818206 (2011).
20. R. T. Menzies, G. D. Spiers, and J. Jacob, “Airborne laser absorption spectrometer measurements of atmospheric CO2 column mole fractions: Source and sink detection and environmental impacts on retrievals,” J. Atmos. Ocean. Technol. 31(2), 404–421 (2014).
21. A. Fix, A. Amediek, C. Büdenbender, G. Ehret, M. Quatrevalet, M. Wirth, J. Löhring, R. Kasemann, J. Klein, H. D. Hoffmann, and V. Klein, “Development and First Results of a new Near-IR Airborne Greenhouse Gas Lidar,” in Advanced Solid State Lasers (Optical Society of America 2015), paper ATh1A–2.
22. A. Amediek, G. Ehret, A. Fix, M. Wirth, C. Büdenbender, M. Quatrevalet, C. Kiemle, and C. Gerbig, “CHARM-F-a new airborne integrated-path differential-absorption lidar for carbon dioxide and methane observations: measurement performance and quantification of strong point source emissions,” Appl. Opt. 56(18), 5182–5197 (2017). [PubMed]
23. U. N. Singh, T. F. Refaat, J. Yu, M. Petros, and R. G. Remus, “Double-pulsed 2-μm lidar validation for atmospheric CO2 measurements,” Proc. SPIE 9645, 961204 (2015).
24. T. F. Refaat, U. N. Singh, J. Yu, M. Petros, S. Ismail, M. J. Kavaya, and K. J. Davis, “Evaluation of an airborne triple-pulsed 2 μm IPDA lidar for simultaneous and independent atmospheric water vapor and carbon dioxide measurements,” Appl. Opt. 54(6), 1387–1398 (2015). [PubMed]
25. A. Amediek, X. Sun, and J. B. Abshire, “Analysis of Range Measurements from a Pulsed Airborne Integrated Path Differential Absorption Lidar,” IEEE Trans. Geosci. Remote Sens. 51(5), 2498–2504 (2013).
26. J. R. Chen, K. Numata, and S. T. Wu, “Error reduction methods for integrated-path differential-absorption lidar measurements,” Opt. Express 20(14), 15589–15609 (2012). [PubMed]
27. J. R. Chen, K. Numata, and S. T. Wu, “Error reduction in retrievals of atmospheric species from symmetrically measured lidar sounding absorption spectra,” Opt. Express 22(21), 26055–26075 (2014). [PubMed]
28. A. Ramanathan, J. Mao, G. R. Allan, H. Riris, C. J. Weaver, W. E. Hasselbrack, E. V. Browell, and J. B. Abshire, “Spectroscopic measurements of a CO2 absorption line in an open vertical path using an airborne lidar,” Appl. Phys. Lett. 103(21), 214102 (2013).
29. A. Ramanathan, J. Mao, J. B. Abshire, and G. R. Allan, “Remote sensing measurements of the CO2 mixing ratio in the planetary boundary layer using cloud slicing with airborne lidar,” Geophys. Res. Lett. 42(6), 2055–2062 (2015).
30. J. Mao, A. Ramanathan, J. B. Abshire, S. R. Kawa, H. Riris, G. R. Allan, M. Rodriguez, W. E. Hasselbrack, X. Sun, K. Numata, J. Chen, Y. Choi, and M. Y. M. Yang, “Measurement of Atmospheric CO2 Column Concentrations to Cloud Tops with a Pulsed Multi-wavelength Airborne Lidar,” Atmos. Meas. Tech. Discuss. (in review).
31. W. E. Sharp, T. S. Zaccheo, E. V. Browell, S. Ismail, J. T. Dobler, and E. J. Llewellyn, “Impact of ambient O2 (a1Δg) on satellite‐based laser remote sensing of O2 columns using absorption lines in the 1.27 µm region,” J. Geophys. Res. 119(12), 7757–7772 (2014).
32. S. A. Clough and M. J. Iacono, “Line-by-line calculations of atmospheric fluxes and cooling rates. Applications to carbon dioxide, ozone, methane, nitrous oxide, and the halocarbons,” J. Geophys. Res. 100, 16519 (1995).
33. D. A. Long, D. K. Havey, M. Okumura, C. E. Miller, and J. T. Hodges, “O2 A-band line parameters to support atmospheric remote sensing,” J. Quant. Spectrosc. Radiat. Transf. 111(14), 2021–2036 (2010).
34. D. A. Long and J. T. Hodges, “On spectroscopic models of the O2 A-band and their impact upon atmospheric retrievals,” J. Geophys. Res. 117, D12 (2012).
35. H. Tran, C. Boulet, and J. M. Hartmann, “Line mixing and collision-induced absorption by oxygen in the A band: Laboratory measurements, model, and tools for atmospheric spectra computations,” J. Geophys. Res. 111, D15 (2006).
36. M. M. Rienecker, M. J. Suarez, R. Gelaro, R. Todling, J. Bacmeister, E. Liu, M. G. Bosilovich, S. D. Schubert, L. Takacs, G.K. Kim, and S. Bloom, “MERRA: NASA’s modern-era retrospective analysis for research and applications,” J. Clim. 24, 3624–3648 (2011).
37. X. Sun and J. B. Abshire, “Comparison of IPDA lidar receiver sensitivity for coherent detection and for direct detection using sine-wave and pulsed modulation,” Opt. Express 20(19), 21291–21304 (2012). [PubMed]
38. C. Kiemle, M. Quatrevalet, G. Ehret, A. Amediek, A. Fix, and M. Wirth, “Sensitivity studies for a space-based methane lidar mission,” Atmos. Meas. Tech. 4(10), 2195–2211 (2011).
39. T. F. Refaat, S. Ismail, A. R. Nehrir, J. W. Hair, J. H. Crawford, I. Leifer, and T. Shuman, “Performance evaluation of a 1.6-µm methane DIAL system from ground, aircraft and UAV platforms,” Opt. Express 21(25), 30415–30432 (2013). [PubMed]
40. G. R. Allan, M. A. Stephen, A. Yu, J. B. Abshire, S. T. Wu, J. Chen, and K. Numata, “Optimizing Output Power through Temporal Pulse Shaping,” in CLEO: QELS_Fundamental Science (Optical Society of America, 2017), paper JTu5A–87.
41. T. F. Refaat, U. N. Singh, M. Petros, R. Remus, and J. Yu, “Self-calibration and laser energy monitor validations for a double-pulsed 2-μm CO2 integrated path differential absorption lidar application,” Appl. Opt. 54(24), 7240–7251 (2015). [PubMed]
42. P. O. Werle, R. Mücke, and F. Slemr, “The limits of signal averaging in atmospheric trace-gas monitoring by tunable diode-laser absorption spectroscopy (TDLAS),” Appl. Phys. B 57(2), 131–139 (1993).
43. P. O. Werle, “Accuracy and precision of laser spectrometers for trace gas sensing in the presence of optical fringes and atmospheric turbulence,” Appl. Phys. B 102(2), 313–329 (2011).
44. X. Sun, J. B. Abshire, J. D. Beck, P. Mitra, K. Reiff, and G. Yang, “HgCdTe avalanche photodiode detectors for airborne and spaceborne lidar at infrared wavelengths,” Opt. Express 25(14), 16589–16602 (2017). [PubMed]
45. R. F. Keeling, “The atmospheric oxygen cycle: The oxygen isotopes of atmospheric CO2 and O2 and the O2/N2 ratio,” Rev. Geophys. 33(S2), 1253–1262 (1995).
46. M. Dole, “The relative atomic weight of oxygen in water and air,” J. Am. Chem. Soc. 57, 2731 (1935).
47. M. Bender, T. Sowers, and L. Labeyrie, “The Dole effect and its variations during the last 130,000 years as measured in the Vostok ice core,” Global Biogeochem. Cycles 8(3), 363–376 (1994).
48. B. Luz, E. Boaz, E. Barkan, M. L. Bender, M. H. Thiemens, and K. A. Boering, “Triple-isotope composition of atmospheric oxygen as a tracer of biosphere productivity,” Nature 400(6744), 547–550 (1999).
49. G. Hoffmann, M. Cuntz, C. Weber, P. Ciais, P. Friedlingstein, M. Heimann, J. Jouzel, J. Kaduk, E. Maier-Reimer, U. Seibt, and K. Six, “A model of the Earth’s Dole effect,” Global Biogeochem. Cycles 18(1), 1–15 (2004).
