MicroPulse DIAL (MPD) – a Diode-Laser-Based Lidar Architecture for Quantitative Atmospheric Profiling

Continuous water vapor and temperature profiles are critically needed for improved understanding of the lower atmosphere and potential advances in weather forecasting skill. Ground-based, national-scale profiling networks are part of a suite of instruments to provide such observations; however, the technological method must be cost-effective and quantitative. We have been developing an active remote sensing technology based on a diode-laser-based lidar architecture to address this observational need. Narrowband, high spectral fidelity diode lasers enable accurate and calibration-free measurements requiring 5 a minimal set of assumptions based on direct absorption (Beer-Lambert law) and a ratio of two signals. These well-proven quantitative methods are known as differential absorption lidar (DIAL) and the high spectral resolution lidar (HSRL). This diode-laser-based architecture, characterized by less powerful laser transmitters than those historically used for atmospheric studies, can be made eye-safe and robust. Nevertheless, it also requires solar background suppression techniques such as narrow field-of-view receivers with an ultra-narrow bandpass to observe individual photons backscattered from the atmosphere. We 10 will discuss this diode-laser-based lidar architecture’s latest generation and analyze how it addresses a national-scale profiling network’s need to provide continuous thermodynamic observations.

with the ancillary observations to which they are calibrated and are poorly suited to identifying the biases in those ancillary observations. 85 Diode-laser-based broadband DIAL has recently developed as profiling technology to measure water vapor (Newsom et al., 2019;Mariani et al., 2020Mariani et al., , 2021. It is an attractive alternative to narrowband DIAL as it is a simpler technological approach with less stringent spectral requirements on the laser source. Technologically, the primary difference is the spectral width of the transmitted laser light. Broadband DIAL emits a band of laser frequencies several tens to hundreds of GHz wide (much broader than a single water vapor absorption line), whereas narrowband DIAL emits a band no wider than a few hundred 90 MHz. Relaxing the spectral requirements of the laser comes at a performance cost along with other design trade-offs. First, the spectrally broadband source restricts using one of the best solar background rejection tools -narrowband filters in the receiver.
For narrowband DIAL, filter widths can be reduced to just a few GHz without rejecting the desired signal, limited only by the width of the Rayleigh-Brillouin scattered spectrum. With broadband lasers, narrowing the receiver spectrum becomes a zerosum game, where reducing receiver bandwidth serves to attenuate background and signal equally. The only way to overcome 95 this issue of detectability is to move to higher pulse energies and sacrifice the eye-safe classification. As a result, broadband DIAL will have reduced performance (e.g., reported maximum ranges are typically limited to less than 1.5 km AGL). As currently demonstrated, broadband DIAL also requires calibration. The laser spectral width term is nudged, independently for near and far range channels, to adjust the retrieved WV values to match ancillary measurements. Narrowband DIAL is a unique active remote sensing technique [.. 7 ]that does not require ancillary calibration and relies only on fundamental theory for its 100 retrievals. A further trade-off is the inability of broadband DIAL to be optimized to the atmospheric conditions. In contrast, one can spectrally tune a narrowband DIAL system to make equally high-quality water vapor measurements over a wide range of atmospheric conditions. Sideline tuning is the most straightforward method. It adjusts the absorption cross-section by changing the laser frequency while using accurate knowledge of the absorption lineshape from spectral databases (e.g., one may use Voigt profiles to estimate the absorption line shape as a function of altitude). In this manner, one can optimize the 105 measurement for a wide range of moisture conditions. Note that lidar researchers often use a one-way column optical depth of ≈1.1 as a rule of thumb for maximizing resolution in photon counting DIAL (Remsberg and Gordley, 1978). However, optimal performance depends on several instrument and atmospheric factors. For this reason, we typically determine when to adjust the wavelength of the MPD by monitoring the Cramer-Rao lower bound of the water vapor signal as a function of wavelength.
In practice, to optimize a narrowband DIAL, only a small amount of wavelength tuning is required. For example, for the MPD 110 at 828 nm, approximately 5 GHz (or several picometers in the infrared) of sideline tuning adjusts the absorption cross-section by one order of magnitude.

[.. 8 ]
3 Instrument overview 7 removed: in that it is self-calibrating 8 removed: MPD development history of the history, design trades, and testing that led to this instrument version are included in later sections of the paper.
The schematic is shown in two main parts. The subcomponents shown in blue represent the parts needed to make water vapor measurements; whereas the subcomponents in green are needed to make the temperature measurements. The added temperature profiling components allow for the measurement of differential absorption of oxygen at 770 nm and the simultaneous backscatter ratio (via the HSRL technique) using a potassium vapor cell in the receiver as described in 120 Stillwell et al. (2020). The work presented here focuses on general architecture changes that pertain to both the WV and the temperature profiling capabilities of the MPD. However, the specific subcomponent testing and instrument validation presented are for the water vapor measurements only. A few of the key changes from the Gen 4 transmitter design are highlighted here. As is common in DIAL instruments, each transmitter starts by emitting two spectrally narrow seed laser frequencies -one tuned to absorb at the species 125 of interest (online wavelength), and the other tuned to have minimal absorption (offline wavelength). A two-stage fibercoupled seed laser, which is more robust and provides more power in the amplifier has been added. The transmitter uses single-mode (SM) fiber and the correct polarization for subsequent components is maintained using paddle-style controllers that use stress-induced birefringence produced by wrapping the fiber around a spool to create half-wave 9 removed: The plates. The fiber is secured to the bench to avoid movement and maintain the alignment. The seed light is launched 130 into free space and coupled into a tapered semiconductor optical amplifier (TSOA) which is overdriven in pulsed-current mode operation. Following the TSOA an angled window is placed in the beam path to reflect a small portion of light onto amplified photodiode used to monitor the relative transmitted power level (note, these components are not shown in the schematic). The transmit path following the TSOA is the same as described in Spuler et al. (2015) -except for the remotely controllable shutter used for receiver scans, described below -and is only briefly reviewed here. The TSOA 135 amplified beam is expanded two times, shaped with an axicon pair, passed through a hole in an elliptical mirror, and expanded an additional 20 times with the transceiver optics (f=60 mm lens, and the 400 mm diameter, f/3 Newtonian telescope). The annular beam is expanded and transmitted to the atmosphere using the inner half of the telescope's diameter.
The receiver has a few significant changes. A multimode (MM) fiber is used as the field stop for the primary telescope 140 and it is combined with a fiber-coupled wide field-of-view (WFOV) telescope, which had been added next to the primary telescope. A spool of 5 m fiber, which is not shown in the schematic, is used before the filtering stage to attenuate cladding modes due to overfilling as described later in the paper. The fiber-coupled combined receivers use a shared filtering stage and are fiber-coupled to single-photon counting modules (SPCM). The photon counts are summed in a newly designed multi-channel scalar (MCS). As discussed later in the paper, these receiver design changes were made to improve the 145 technology readiness (i.e., make it more robust) and improve the performance closer to the surface.
At the time of this writing, five MPD units have been constructed as a network testbed. One unit of the five units is in the complete configuration shown in Fig. 1. One instrument in the network has added components that use a 780 nm HSRL with a Rubidium vapor cell as described in Hayman and Spuler (2017). The remaining three units contain the components required for water vapor profiling only. The MPD architecture is modular by design and can easily accommodate 150 these different configurations. Although not highlighted in this paper, we implemented many software and mechanical improvements to increase the functionality, stability, and reliability of the MPD design.

MPD development history
We take a step back here to give a brief history of the MPD development to provide context for why certain design choices have been made. The seminal research for atmospheric water vapor profiling with semiconductor-based lasers was done by 155 Nehrir et al. (2009Nehrir et al. ( , 2011Nehrir et al. ( , 2012. These papers discuss the early MPD development and span technology generations 1, 2, and 3. These instruments used external cavity diode lasers (ECDL) to seed a Tapered Semiconductor Optical Amplifier (TSOA).
They were 'laboratory only' instruments -primarily due to the temperature sensitivities of ECDLs. A collaboration with MSU and NCAR began in 2011 to enable the technology to move out of the laboratory. The first step was to work with an industry partner to develop a Distributed Bragg Reflector (DBR) laser at 828 nm wavelength to avoid temperature mode hopping issues 160 with the ECDL lasers. The MSU and NCAR partnership co-developed the Gen 4 MPD, creating an instrument capable of field deployments with significant performance improvements (Spuler et al., 2015). A prototype portable field enclosure (≈ 1m x 2m x 2m) was developed after this publication and integrated with the water vapor MPD instrument for subsequent field studies.
In 2016, after an extensive validation study of the Gen 4 performance by Weckwerth et al. (2016), NCAR and MSU began 165 constructing five MPD profiling instruments -to make a network testbed available to the science community. The initial field demonstration of the five-unit network occurred in Apr-Jul 2019 at the Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) atmospheric observatory in Oklahoma, USA 10 . The primary goal was to evaluate the engineering performance of the MPDs in a deployment setting. It also offered the opportunity to validate them against more mature instruments further and investigate atmospheric moisture variability. During this field project, the 170 'MPD05' unit was at the site's central facility, collocated with the SGP Raman lidar and radiosondes that were launched by the ARM staff approximately every 3 hours. The other four units were spaced approximately 50 km apart in a box pattern surrounding the center site. The placement of MPD05 provided many independent ancillary water vapor profile measurements.
An example of validation against the Raman lidar and radiosondes is shown in Fig. 2. The two plots have time on the x-axis (12 days) versus the height above ground level (AGL) on the y-axis (from the surface to 6 km). The upper figure shows the absolute 175 humidity measured by the MPD during this time. The seventy-seven radiosondes (Vaisala RS41) launched during this time are overlaid on the figure as vertical color bars. The lower figure shows the absolute humidity measured by the SGP Raman Lidar (also with the radiosonde overlays). The most notable difference between the two continuous remote sensing measurements is the white bar on the bottom of the MPD absolute humidity. The lowest ranges were not included since the Gen 4 version of MPD05 did not have a near-range receiver installed during the test. A frequency histogram of this time series is shown in 180 Fig. 3 for each measurement technique. Both active remote sensing techniques agree well with the radiosondes, with leastsquare linear fit slopes and correlation coefficients close to 1. It is worth noting that the correlation between Raman lidar and radiosondes is expected to be near 1 as the lidar is calibrated from the same sondes while the MPD retrievals never make use of the radiosonde data at any point and are truly an independent measurement. The network field campaign confirmed a very high correlation between the more mature measurement methods when the MPD data was confined to ranges 500 above ground 185 level. Furthermore, the field test was a significant milestone as the first demonstration of continuous, high-vertical-resolution water vapor profiles from a DIAL network. However, the testing also revealed that several instrument modifications were needed before the MPD testbed was made available to the larger scientific community and led to the Gen 5 design described below.
There is scientific value in measuring atmospheric variables throughout the atmosphere. Depending on location (latitude 190 and longitude) and topography, many important processes are confined to or prominently occur at low altitudes, including land-atmospheric feedback, turbulent and latent heat fluxes, convection initiation, and chemical transport. Equally, the upper troposphere state is critical to understanding inflow and outflow from mesoscale convective systems, long-range transport of aerosols, and cloud radiative effects. There is, therefore, a desire to measure the whole atmosphere, but a practical limitation is present in the design of all lidar systems. In the case of MPD, expanding scientific utility equated to extending the operational 195 range to lower altitudes. For the last several years -intertwined with the network testbed construction -we have focused on 10 Information about the project is available at https://www.arm.gov/research/campaigns/sgp2019mpddemo. Figure 2. Twelve days of data from the MPD Network Demonstration field test. Absolute humidity from 0-6 km AGL as measured by the Gen 4 version (prior to upgrade to Gen 5) of the MPD05 unit without a near range channel (top) and the SGP Raman lidar (bottom). Both Raman and MPD data are at 10 min temporal resolution. The radiosonde absolute humidity data is overlaid on both plots. addressing the scientific needs for better accuracy closer to the surface, while also improving the instrument architecture's technology readiness and enabling quantitative measurements of aerosols and temperature. Table 1 shows some of the key parameters of the different generations. The general trend has been to replace free space optics with fiber-coupled components to improve stability and reduce sensitivity to thermal drifts. [.. 11 ] 200 11 removed: Although not highlighted in this paper, we implemented many software and mechanical improvements to increase the functionality, stability, and reliability of the MPD. This paper details some key advances for diode-laser-based narrowband DIAL. between Raman lidar and radiosondes is expected to be near 1 as the lidar is calibrated from the same sondes while the MPD retrievals are run independent of the radiosonde information.

Transmitter
The MPD transmitter architecture is founded on single-frequency laser diodes. These devices are much less powerful than narrowband solid-state lasers (crystal/glass media doped with rare-earth or transition metal ions) historically used for quantitative atmospheric lidar methods of DIAL and HSRL (e.g., [.. 12 ] Piironen and Eloranta (1994); Wirth et al. (2009);Späth et al. (2016)). The ≈ 100× reduction in power requires longer averaging times to achieve adequate SNR levels, and there-205 fore lidar instruments based on laser diodes are (currently) only suitable to observe atmospheric phenomena at longer timescales. However, diode lasers have the significant advantage of being lower-cost and more rugged for field operations, especially when fiber-coupled. For this reason, we have incorporated more fiber-coupling into the master oscillator power amplifier (MOPA) transmitter architecture used in the MPD. In the previous generation, light emitted from a free-space DBR seed laser was collimated with an aspheric lens, circularized using an anamorphic prism pair, isolated from optical feedback with 210 a Faraday rotator, polarization controlled with a half-wave plate, and coupled into a single-mode fiber with an aspheric lens.
These sub-components were held in mechanical cage mounts as shown in the solid model on the left panel of Fig. 4. The free- Semiconductor Optical Amplifier (TSOA), Traveling Wave Amplifier (TWA), Transmitter (Tx), Water Vapor (WV). a The Gen 4 MPD ran at 9 kHz for the 2014 FRAPPÉ field campaign but then reduced to 7 kHz starting with the 2015 PECAN field campaign.
b The reduction in pulse energy due to design changes to make the transmitted beam eye-safe with increased optomechanical stability.
c Gen 4 MPD instruments without the free-space near-range receiver.
space packaging was very stable in lab conditions and had shown adequate stability in the first generation MPD field enclosure environment during field tests. However, when field testing all five MPD units in 2019, some of the free-space modules had inadequate stability for multiple-month unattended operations. This instability required highly trained staff to service some of 215 the instruments since aligning the module required a high degree of expertise.
Following the network field demonstration, a newly available butterfly-packaged fiber-coupled DBR seed laser (Photodigm PH828DBR 020 BF-ISO) was tested as a replacement to the free-space packaging (see Fig. 4). The package contains a similar set of optical sub-components to the free-space version -including an isolator needed for stable single-mode operation.
However, the robustly aligned sub-components inside the sealed package have improved optomechanical stability and reduces 220 the technical skill required to service the instrument. This butterfly-packaged seed laser can provide adequate seed power into the final amplifier stage yet requires a high drive current to do so (due to higher insertion loss of the internal Faraday isolator), which could adversely impact the device lifetimes. Therefore we chose to add a booster stage -a butterfly-packaged traveling wave amplifier (TWA) -to allow the seed laser to operate at a lower drive current.  [.. 17 ] We tested the new seed laser architecture to ensure it has (1) sufficient power to operate the final amplifier stage in gain 235 saturation and (2) spectral properties -frequency stability, linewidth, and spectral purity -needed for the DIAL technique.
As established in the literature, to keep water vapor DIAL systematic errors below 10%, the transmitted laser must have wavelength stability of < ±200 MHz, a linewidth of <400 MHz, and a spectral purity -or the fraction of total laser power contained within the 'linewidth' -of >99.5% (Ismail and Browell, 1994). For temperature measurements using the DIAL technique, the transmitter requirements are more stringent -wavelength stability of ±100 MHz, linewidth <250 MHz, and 240 >99.9% spectral purity (Wulfmeyer, 1998). Throughout this paper, we will adhere to the more stringent spectral requirements for temperature. 16 Note we have conducted preliminary tests of fiber-coupled TSOAs. The early results look promising but more development is needed before a fully fiber-coupled transmitter is ready for routine deployment. 17 removed: Schematic of the Gen 5 MPD transmitter. DBR butterfly packaged lasers have been available for some time but only recently with internal isolators, allowing for stable narrowband operation. The 10% fiber tap connects to a wavelength meter used to monitor and stabilize or lock the laser wavelength.

Power output
In the latest transmitter design, the DBR seed lasers are operated just above 5 mW (well below the nominal 25 mW limit imposed by the maximum drive current). Roughly half of the light reaches the booster stage due to insertion losses from the 245 optical switches used to interleave the online and offline seed lasers. The TWA booster stage amplifies the output up to 20 mW -below the manufacturer's recommended 30 mW maximum to avoid optical damage to the fiber end. The typical current required for this amplification level is 50% of the allowed maximum. Following the 10x gain of the TWA booster stage, the light exits the single-mode fiber. It is collimated with an aspheric lens (f=4.6 mm, NA=0.53) and passes through a Faraday isolator, and is polarization aligned with a half-wave plate. The isolator typically has a 1 dB loss, so approximately 16 mW of 250 optical power can seed the TSOA. A pair of mirrors and an aspheric lens (f=4.5 mm, NA=0.55) couple the light into the 3 µm x 3 µm square input aperture of the TSOA. The amplified light, which exits a 3 µm × 190 µm output aperture, is astigmatic with a general divergence of 14 • x 28 • . Actual astigmatism and divergence are slightly dependent on the operational parameters due to minor changes in the TSOA substrate index of refraction that depend on the electronic pumping rate. The transmitter is typically operating at a variety of pulse durations (0.5 to 1 µs) and pulse repetition frequencies ([.. 18 ]7-10 kHz). Therefore, a 255 lens pair -an aspheric (f=4.5 mm, NA=0.55) and a cylindrical (f=50 mm) -are aligned to produce a collimated beam in both axes at these nominal operational conditions. The cylindrical lens is tilted slightly to avoid back-reflections to the TSOA.
We measured the input 'seed' power versus output power for a range of conditions to map out the TSOA performance. The results are shown in Fig. 5. For the plot shown, the TSOA was driven with 0.625 µs at 8 kHz repetition rate current pulses.
However, the general trend was the same when tested at [.. 19 ]1 µs and 7 kHz. From these results, we conclude that the final 260 stage amplifier output is well into the saturated gain region at seed input powers exceeding [.. 20 ]15 mW for all drive currents.
For the range of drive currents tested, 6 A to 12 A, there is a nearly linear increase in the TSOA output power with increasing drive current, at roughly 1 W peak power per 1 A drive current when operated in the saturated gain region. In this region, the final stage gain is >400x when using a 10 A drive current.

Frequency Stability 265
The [.. 21 ]stability of the MPD transmitter is driven by the seed lasers assuming adequate power at each amplifier stage.
The seed [.. 22 ]laser wavelength stability was measured using the operational laser locking routine developed for the MPD instrument. Several fiber-coupled seed lasers have been operated for months to verify stability within the required ±100 MHz.
While there are some differences in the characteristics of the fiber-coupled diodes, they all exhibit standard deviations on the order of 20-30 MHz 23 . Fig. 6 illustrates a histogram of data collected over ten days of continuous measurement from six seed  24 removed: These results were selected as they contained no user-defined system changes and were representative of time scales tested from one day to several months. 25 removed: The self-heterodyne technique was used to measure the linewidth and spectral purity of the transmitter. As shown in Fig. A1 a 'pickoff' optic reflected a small portion of the pulsed output from the transmitter's last amplifier stage into a fiber port that focuses the light into a single-mode fiber. A fiber-coupler split the light into two paths. The 90% leg passed through a polarization controller and a fiber-coupled acousto-optic modulator (AOM, Brimrose model TEF-1000-100-828-2FP-SM) to shift the frequency 1 GHz. A 50/50 splitter combined the frequency-shifted light with the 10% leg containing the originally sampled beam. One leg of the 50/50 splitter connects to Si PIN Amplified Detector (ThorLabs FPD310-FC-VIS) with 1 MHz -1.5 GHz bandwidth.
The amplified detector has a 2 mW damage threshold, so the pulses' peak power was maintained below that threshold. An Electrical Spectrum Analyzer (ESA) -operated at 50 kHz bandwidth resolution, 100 MHz to 1.9 GHz span, ten shot averaging, and log-power detector -measured the detector output. Since an etalon in the MPD receiver limits the bandpass to approximately 1.3GHz FWHM, the measured frequency spectrum is sufficient to ensure the instrument's spectral purity. 26 removed: Test setup to measure the linewidth and spectral purity. AOM: Acousto-Optic Modulator, MS: Mating sleeve, SMF: single-mode fiber, TSOA: Tapered Semiconductor Optical Amplifier 27 removed: A critical element of the MPD architecture is using the same amplifier for online and offline seed lasers and the same detector to measure both atmospheric profiles for DIAL measurements. As a result, the bandpass at the etalon's adjacent free spectral range (FSR) is a possible spectral purity Spectral purity considers the multimode structure of the laser and is also critical for the DIAL technique. For this work, we 285 define spectral purity as the amount of power in the center ±100 MHz over the remaining 1.8 GHz spectrum. The DBR seed alone and when used with the booster stage were both measured to be >99.9%, which was the resolution limit of the test setup. Fig. 8 shows a plot of the pulsed transmitter output's spectral purity as a function of input seed power for a range of operating conditions. The curves (colored black, blue, red, and magenta) are defined by the current applied to the final amplifier stage.
As shown in the figure, increased seed power is required to maintain high spectral purity with increasing electrical pumping 290 of the amplifier. So the transmitter spectral purity is a function of the TSOA drive current and the seed power coupled into the contamination source, i.e., online leaking into offline or offline leaking into online. Such leakage could result in a bi-modal seed, which would not be noticeable.
So, as described in the receiver section below, the instrument was designed for automatic receiver scans to measure both the etalon bandpass and the isolation between the online and offline wavelengths, which could leak into the adjacent etalon FSR passband. To ensure the spectral purity specification of 99.9% is maintained, greater than 30dB isolation between online and offline is monitored and verified. amplifier. Although higher drive currents allow for more power amplification (refer to Fig. 5), more seed power is required to maintain a high level of spectral purity. The new two-stage seed laser makes it easier to reach (and exceed) the required power levels. With the extra capacity in the components and current drivers, it is straightforward to remotely increase the power to maintain spectral purity and power during a long deployment [.. 28 ](e.g., in the case of component power degradation over 295 its lifetime), enhancing the system's operational reliability and robustness.
As previously mentioned, to maintain an allowable systematic DIAL error for atmospheric measurements in the lower troposphere, one can use the rule-of-thumb spectral purity requirements (Ismail and Browell, 1994;Wulfmeyer, 1998). However, one can also calculate a direct description of DIAL error based on the measured spectra and the model spectral absorption feature used for the DIAL measurement. The fractional error in absorption is thus the assumed absorption (where the laser 300 spectrum is a delta function) divided by the actual absorption ( 1) where σ est is the estimated absorption cross section for an ideal laser, σ act is the actual absorption cross section, l(λ) is the normalized laser spectrum such that ∞ 0 l(λ)dλ = 1, σ(λ) is the frequency resolved absorption cross section and λ 0 is the laser line center wavelength. line, which is likely to be a worst-case scenario for Rayleigh-Doppler effects [.. 30 ] (Späth et al., 2020). The accuracy for water vapor and oxygen absorption is 99.99% and 99.95%; respectively, for seed powers greater than 15 mW. Therefore, it seems reasonable to conclude that seed powers >15 mW provide more than adequate spectral purity properties for the MPD transmitter 310 for both water vapor and temperature measurements.
28 removed: , 30 removed: Späth et al. (2020) A critical element of the MPD architecture is using the same amplifier for online and offline seed lasers and the same detector to measure both atmospheric profiles for DIAL measurements. Therefore another aspect of spectral purity, that is important to the MPD design, is measuring leakage in the optical switches used to alternate the online and offline seed lasers. Cross-talk in the switches could result in a bi-modal seed, which would not be noticeable in the test setup described 315 so far, and since the receiver bandpass is designed to pass both online and offline wavelengths it is a possible spectral purity contamination source. So, as described in the receiver section below, the instrument was designed for automatic receiver scans to measure both the receiver bandpass and the isolation between the online and offline wavelengths. To ensure the spectral purity specification of 99.9% is maintained, greater than 30dB isolation between online and offline is The MPD receiver uses two solar background reduction techniques -a narrow field-of-view and narrow bandpass filtering -that enable the observation of individual photons backscattered from the atmosphere during the day and in all weather 330 conditions. There are inherent trade-offs with each of these design features. [.. 36 ]For example, as will be discussed in more 31 removed: The 32 removed: 's requirements 33 removed: Although the receiver has yet to be described, at this point, we introduce the schematic of the complete Gen 5 instrument shown in Fig. 1.
The transmit path following the TSOA is the same as described in Spuler et al. (2015) -except for the remotely controllable shutter used for receiver scans, described below -and is only briefly reviewed here. The TSOA amplified beam is expanded two times, shaped with an axicon pair, passed through a hole in an elliptical mirror, and expanded an additional 20 times with the transceiver optics (60 mm fl lens, and the 400 mm diameter, f/3 Newtonian telescope). The annular beam is expanded and transmitted to the atmosphere using the inner half of the telescope's diameter. added temperature profiling components allow for the measurement of differential absorption of oxygen at 770 nm and the simultaneous backscatter ratio (via the HSRL technique) using a potassium vapor cell in the receiver as described in Stillwell et al. (2020). One instrument in the network has added components that use a 780 nm HSRL with a Rubidium vapor cell as described in Hayman and Spuler (2017). The remaining three units contain the components required for water vapor profiling only. The MPD architecture is modular by design and can easily accommodate these different configurations. 36 removed: Because of the use of fixed telescope optics, lidar systems typically must choose a telescope focus. It is often near infinity, meaning that the systems are out of focus and inefficiently capture light at low altitudes.
detail later, a narrow field-of-view receiver is inefficient at capturing light from ranges close to the surface. Some lidar systems subvert this problem with a second receiver or multiple channels to see both low and high or using beam scanning units (e.g., Radlach et al. (2008); Reichardt et al. (2012); Newsom et al. (2019)). [.. 37 ]There is a delicate trade-off between observational capability, cost, system complexity, and scientific utility. This trade-off is magnified when developing a lidar 335 system for an eventual large-scale network. As will be described below, improving the instrument collection efficiency at lower ranges while maintaining high solar background suppression is a way to enhance measurements closer to the ground surface. However, pushing the measurement capabilities closer to the ground while keeping instrument costs low requires straightforward methods (i.e., without introducing the complexity and cost of a beam scanner or duplicating filter/detector stages). H ≈ D F OV , a distance of 400 mm 115µm ≈ 3.5 km is expected. 43 Light rays from shorter ranges arrive at steeper angles, which are rejected by a narrow field-of-view design. For the MPD, this effect results in low collection efficiency, <1%, at ranges below 500 m. Although the overlap function cancels out in the absorbing species' measurements -the DIAL technique uses a ratio of two signals within a shared receiver field-of-view -the low receiver efficiency close to the surface increases uncertainty. To 350 improve performance closer to the surface, the instrument designer wants to improve the receiver light collection efficiency at lower ranges without increasing the overall solar background levels.

Wide field-of-view receiver
A free-space wide field-of-view (WFOV) channel was initially part of the Gen 4 MPD design to improve collection efficiency at the lower ranges (Spuler et al., 2015). Although useful for alignment, this configuration was not used for data analysis in 355 practice (Weckwerth et al., 2016). There were several problems with this approach. First, it was not straightforward to combine the two signals. Each dataset had to be processed independently, and the [.. 44 ]Gen 4 WFOV channel used the same primary receiver telescope and had a significantly higher background rate, which introduced uncertainty and limited the useful daytime 37 removed: Therefore, there 38 removed: scattered beam images 39 removed: area without clipping 40 removed: -the overlap function is unity in this region 41 removed: , which is fairly typical for a narrow field-of-view lidar receiver with a telescope of moderate diameter 42 removed: overlap function. 43 In practice, as discussed in Repasky et al. (2013), the MPD receiver focus is adjusted to decrease that distance to about 1.5 km. 44 removed: WFOV channel range. Second, the additional channel doubled the file size and data transfer needs. Finally, the separate filtering stage module and detector added optomechanical complexity. Overall this approach reduced the instrument technology readiness level with 360 minimal improvements to low range performance. We needed to reduce the background level introduced by the WFOV channel and develop a more rugged and elegant design.
This balance was achieved in the Gen 5 receiver with the addition of a small-diameter refractor telescope and a multimode fiber coupler. This design allowed the primary receiver to be ruggedly combined with a 'near-range' receiver and share the filtering stage and the photon counting module. The combined primary and WFOV receivers were designed to retain similar light 365 gathering capabilities (i.e., background levels) as the stand-alone primary receiver to maintain the same daytime performance.
We provide details of these [.. 45 ]design considerations and plots of the theoretical overlap function [.. 46 ] and the receiver aperture normalized signals in Appendix B. The WFOV receiver uses an off-the-shelf 80 mm focal length refractor telescope with a clear aperture of 45 mm (Thorlabs C80FC-B) mounted next to the primary receiver telescope. A 2x1 multimode fiber coupler (Thorlabs TM105R1F1A), 105 µm diameter, 0.22 NA, receives the light from each of the telescope 370 receivers in separate legs before entering a common filtering stage. To work for a photon-counting application, we added black aluminum tape to the multimode splitter's fiber sections to make them light-tight 47 [.. 48 ]. 49 The primary telescope receiver is connected to the 90%-leg of the fiber splitter, while the 10%-leg connects to the refractor telescope receiver. The combined signal fiber connects to a common path filter stage followed by a fiber-coupled single-photon counting module. A key feature of the design is the receiver signals pass through the filters at essentially the same angles and travel the same path since the 375 light launches out of a shared multimode fiber. Furthermore, the design avoids the complexity and expense of adding a second receiver [.. 50 ]. It is also easier and more repeatable to align and more compact than a purely free-space module.

Narrowband filtering
The MPD receiver uses a filtering stage constructed from two parts to isolate signal photons from background sunlight. The first stage is a solid fused silica Fabry-Pérot etalon that transmits several wavelengths, including those centered on the absorption  49 We have since engaged the manufacturer of this part on the potential of purchasing it in the future with light-tight fibers. 50 removed: and is 51 removed: rotation 52 removed: gives 53 removed: of wavelengths of light are essentially the same -relying on constructive/destructive interference of light reflecting between surfaces -and they can be tuned by tilting or changing the temperature (both filter passbands will shift to a shorter wavelength at higher incidence angles with the etalon passband additionally shifting to shorter wavelengths with decreasing temperature). These high-performance filters are custom designed for each filtering stage, i.e., the thin-film filters and etalons are created uniquely for the WV receiver and the temperature receiver.

390
For the MPD receiver, the light enters the filtering stage receiver via a 0.22 NA, 105 µm diameter, multimode fiber. The receiver optics -a fiber port mounted aspheric lens (f=11 mm) and a 2.25x expander (comprised of a negative 20 mm and positive 45 mm doublet lens pair) -collimate the light exiting the multimode fiber. The multimode fiber diameter is used as the field stop and limits the maximum entrance angle to the filter stage. At the filter, the beam divergence [.. 54 ]is θ = d 2f M −1 , where d is the diameter of the fiber, f is the focal length of the collimating lens, and M is the magnification of the expander.

395
Therefore, assuming a 100% fill condition of the fiber core, the maximum range of angles at the filter is 2.1 mrad. In practice, however, the receiver fiber can have an 'overfilled launch condition' when imaging distances close to the instrument. Therefore, it is expected that light backscattered from shorter ranges would more readily travel through the fiber cladding and generate higher order cladding modes. [.. 55 ]Increased attenuation of these higher-order modes occurs with increased fiber length [.. 56 ]so the MPD uses a 5 m length of multimode fiber before the filtering stage to minimize these higher-order cladding modes.

400
An example of an overfilled launch condition, with an effective fiber diameter of 125 µm, increases the max angle range to 2.5 mrad. This applies to both telescopes as they use the same receiver optics. Wulfmeyer and Bösenberg (1998) provided details on how interference filters may be a potential source of accuracy errors in the water vapor DIAL retrieval due to the angular-dependent bandpass as mentioned above. [.. 57 ]A recent publication by Späth et al. (2020) explicitly calls out the prior MPD lidar architecture, with its use of narrowband filters in the receiver, 405 as having "range-dependent sensitivities due to angle effects in the near range[.. 58 ]" citing a derivation in Wulfmeyer and Bösenberg (1998). However, the derivation [.. 59 ]cited omits the use of a field stop on the receiver system[.. 60 ]. The standard method to alleviate the concern raised [.. 61 ]is to simply add the stop. The Gen 5 MPD has the field stop prior to the filtering stage, thereby limiting the angles at the filter to <3 mrad. To experimentally verify that MPD does not have [.. 62 ]"rangedependent sensitivities"[.. 63 ], two sets of tests have been [.. 64 ]performed. First, we have run tests with fiber lengths up to 10 410 m and find no evidence of higher-order cladding modes causing changes to the observed WV field. Second, through receiver passband scanning, described below, we find that observed transmission functions of the filters and etalons can be explained 54 removed: , 55 removed: However, increased 56 removed: . The 57 removed: Citing this reference, a 58 removed: . 59 removed: in Wulfmeyer and Bösenberg (1998) 60 removed: , which is the 61 removed: . To 62 removed: the 63 removed: described 64 removed: run with a narrow set of angles, less than approximately 3 mrad, which is consistent with the expected angle limit based on the diameter of light contained within the fiber core. As a result, we find no evidence that the new MPD design has any inherent "range-dependent sensitivities" when properly assembled and aligned.

Receiver passband scanning
The ability to automatically scan the receiver in frequency, and obtain an accurate knowledge of the etalon bandpass, is a key to enabling higher-quality water vapor and temperature measurements. We developed a method inspired by the technique developed and employed by the University of Wisconsin for HSRL to inject the seed laser light into the receiver to achieve this capability (Ed Eloranta, private communication). [.. 65 ]The TSOA is turned off for this step (i.e., there is no concern 420 about damaging back reflections or different amplification characteristics) and the small amount of seed laser light that leaks through the TSOA is backscattered from a shutter into the receiver and scanned to map out the receiver passband transmission. The specific software control procedure is to (1) step down the laser power by turning off the TSOA, (2) switch [.. 66 ]data acquisition to accumulate the continuous-wave [.. 67 ]seed laser signals, (3) alter the laser locking behavior to scan rather than lock to a single wavelength, (4) control the position of a rotary solenoid mounted shutter used to scatter 425 light back into the receiver, and (5) collect data from a frequency scan sequence. This design allows the receiver transmission as a function of wavelength to be mapped in a repeatable fashion without physically modifying the receiver/transmitter optics or overall optical alignment. This ensures that the operational characteristics of the receiver are closely matched to the data acquired during the scan.
An example receiver scan is shown in Fig.10. During this scan, the offline laser was active, and the online laser was turned   We assume the recorded photon counts are Poisson distributed (this is a valid assumption because the photon arrival rate is low enough to be treated linearly related to the incident photon flux). In order to obtain the maximum likelihood fit, the negative log-likelihood of a Poisson distribution is minimized where ψ = {w(x), F, δ 0 , g 1 , g 2 , b} is a set of all the fit terms and y n (λ) are the observed photon counts on the nth detection channel as a function of wavelength. the signal-to-noise, SNR), but that, in turn, limits how close in range the absorption measurements can be made. A simple model of the shortest observable range can be developed by linking the duration of the laser pulse, τ , the distance needed to make an absorption measurement ∆R (which is a two-way, out-and-back distance), and the finite duration of the photon accumulation time, ∆t M CS , set in the Multi-Channel Scalar (MCS). This absolute limit is depicted in Fig. 11. If we define the range, R, from the laser pulse center, the first range where the detector is not blind from the outgoing laser pulse is where c is the speed of light in air. The next range used to make the DIAL measurement occurs at R 2 , where R 2 = R 1 + ∆R 2 . 77 removed: In the case shown in Fig. 10, the isolation is measured at 31.6 dB. As a balance of performance and file size, the Gen 5 MPD instrument [.. 88 ]typically uses a differential range, ∆R = 150 m[.. 89 ], and a 250 ns photon accumulation time, but these values are adjustable. To allow lower ranges to be more 480 accurately measured the transmitted laser pulse length was reduced from 1[.. 90 ] µs to 625 ns in the Gen 5 design (a theoretical lower range improvement of ≈ 56 m). The peak power was not increased (i.e., the TSOA drive current was not increased), so this pulse duration change results in a ≈ 37.5% reduction in pulse energy. The pulse repetition frequency was changed from 7 kHz to 8 kHz to increase the average power and minimize the signal-to-noise impact. Using a shorter differential range would allow measurements closer to the surface. But, as this would increase the absorption 485 measurement uncertainty, it would likely require sophisticated processing techniques such as Poisson Total Variation (Marais et al., 2016). Another possible solution to push further down in range and reduce the SNR impact is to alternate short and long transmit pulses. But as described above, this would have limitations. We plan to investigate both methods in the future as measurements close to the surface remain a high science priority. A further limitation for measuring close in range -beyond the pulse duration -is the detector afterpulse probability. The 490 scattering from optical surfaces during the outgoing transmission laser pulse inevitably affects the detector for a short time, creating a small signal-induced bias. The low photon counts in the near range region can exacerbate this bias. The MPD instruments use a single photon counting module or SPCM (Excelitas, SPCM-800-12-FC, or the SPCM-850-12-FC). These devices are fiber-coupled to the receiver via a 105 µm diameter 0.22 NA multimode fiber (with a light-tight armored jacket) following the filter stage. The SPCM modules are specified to have a nominal 1% afterpulsing probability with a maximum 495 of 3%; although our units range from 0.1% to 0.9%. Before the inclusion of the second wide field of view telescope, we investigated correcting for the afterpulse to extend the range closer to the surface while maintaining relatively long pules (1µs in our case). We placed a cover over the telescope and collected data to measure the bias baseline -a small number of increased photons immediately after the transmitted pulse ends -and the baselines (for both the on and offline) were subtracted from the average atmospheric data before standard data processing. The results indicated some minor ability to remove a slight wet 500 bias in the lowest bin. Still, the technique was perhaps overcompensating and added as much error as it removed. Currently, no afterpulse corrections are applied to the MPD data. The detector gating function has been found to reduce the effect of afterpulsing slightly. The gate disables the detector output and active quench function [.. 91 ] (Cova et al., 1996). Based on test results, as discussed below, the signal-induced bias using the gated SPCM (without afterpulse correction) adds ≈ 50 m beyond the minimum range imposed by the pulse duration.

Multi-Channel Scalar (MCS)
For the Gen 5 MPD instrument, a new multi-channel scalar (MCS) was developed. It was implemented in a flexible, reconfigurable logic that improves system integration and simultaneously increases the system's overall functionality, simplifies the system (i.e., reduces part count), and reduces data latency. The original MCS was a commercially available product. It 510 required a proprietary software package wrapped by another software package (in our case, Labview). The new MCS uses highly standardized Ethernet data communications, specifically User Datagram Protocol or UPD, that can be accessed by several programming languages, including Python, C++, and Labview. No proprietary software is required.
The new MCS also has advanced features while reducing overall system part counts. First, the new MCS uses softwaredefined switches local to the system's field-programmable gate array (FPGA) to sort data from the same detector into different 515 buffers (e.g., online and offline transmit lasers). This feature allowed the removal of MPD Gen 4 radio-frequency (RF) hard-91 removed: when a TTL low level is applied to the module 92 removed: The transmitted laser pulse length was reduced from 1 µs to 625 ns in the Gen 5 design to allow lower ranges to be more accurately measured (a theoretical lower range improvement of ≈ 56 m). The peak power was not increased (i.e., the TSOA drive current was not increased), so this pulse duration change results in a ≈ 37.5% reduction in pulse energy. The pulse repetition frequency was changed from 7 kHz to 8 kHz to increase the average power and minimize the signal-to-noise impact. Although not tested yet, one possible solution to push further down in range and reduce the SNR impact is to alternate short and long transmit pulses. But as described above, this would have limitations. Using a shorter differential range would also allow measurements closer to the surface. But, as this would increase the absorption measurement uncertainty, it would likely require sophisticated processing techniques such as Poisson Total Variation (Marais et al., 2016). We plan to investigate both methods in the future as measurements close to the surface remain a high science priority.
ware switches and associated wiring -required to route the signals from the photon counting module to the MCS. Second, the new MCS has transmitter pulse power monitoring channels. This advanced feature allows for monitoring individual lasers (both online and offline) independently via the same FPGA software-defined switches. This capability was not possible with the original hardware configuration, which required an extra power monitor and proprietary software only capable of moni-520 toring average online and offline lasers power. The MCS change allowed for a simplification by removing the original power monitoring components and the additional software used to communicate with the third-party device. The new MCS system achieves upgraded speed, and by extension, measurement duty-cycle by switching from USB 2.0 to UDP over 1 Gigabit Ethernet. Additional benefits of the USB to Ethernet upgrade include (1) improved ruggedness of the Ethernet physical interface over USB, (2) increased system flexibility and remote device accessibility, (3) improved data streaming regularity, (4) ease 525 of software development using sockets programming and in-house API, and (5) increased transparency of data structure and improved diagnostic capability using standard network observation tools. Finally, a noticeable increase in overall system reliability is observed using ethernet rather than USB communications as 3rd party installed software drivers can be unstable given differing operating systems and update packages. This protocol speed, simplicity, and reliability is a significant increase in technology readiness from the previous system.

7 Data Processing
The data processing developed for the Gen 5 MPD has been applied to all of the data shown in this manuscript (including Gen 4 data for consistency). Some of the elements used in this data processing chain have been described in Hayman et al. (2019) and Hayman et al. (2020) and will not be repeated here for brevity. A summary of improvements to the data processing and how these elements are combined is included.

535
Photon counts are binned onboard the MCS at approximately 2-second resolution. Before any processing, a nonparalyzable deadtime correction is applied to correct some of the nonlinearity in the captured photon count signal (Müller, 1973). This standard correction assumes that photon counts between range bins are uncorrelated (deadtime is much less than range bin width -in this case approximately a factor of 10) and that the photon arrival rate is constant over the bin accumulation interval. It is for the second assumption, that the correction is applied at the base acquisition resolution, 540 though it is unlikely this is sufficient to accurately recover photon arrival rates from clouds (where water vapor signals are not retrieved). After binning the deadtime corrected photon counting data to a 1-minute grid resolution, we employ a technique to optimally determine the mean number of photons in the signal through smoothing in time and range with a Gaussian kernel . This technique quantitatively finds the balance between error caused by statistical noise (under smoothing) and smearing (over smoothing) and is applied such that each altitude level is optimized independently 545 in time and each minute is optimized independently in range.
After background subtracting the smoothed counts, the water vapor number density, n wv , is calculated using a slightly modified version of the DIAL equation ton counts, N of f (R) is the background-subtracted offline photon counts and the range derivative operations are approximated using a three point, first order Savitsky-Golay filter. The reason for moving the derivative operation to the background subtracted profiles is that this form better accommodates noisy observational data by avoiding negative arguments in the log of the more conventional DIAL equation. In Eq. (6), undefined values only occur when one of the background subtracted profiles are exactly equal to zero, which is rare. This is, in turn, more accommodating to the uncertainty analysis described below.

555
The water vapor absorption cross-sections are calculated for each pixel in a day using a principal component approximation for the water vapor absorption profile as a function of temperature and pressure (Hayman et al., 2019). The temperature and pressure profiles are obtained from NCAR/NCEP reanalysis (Kalnay et al., 1996) for post-processing or using an assumed lapse rate from surface temperature and pressure data for real-time processing 93 .
[.. 94 ]When we calculate the water vapor field[.. 95 ], most of the high-frequency structure caused by backscatter is absent 560 from the profile, so a second, larger smoothing operation, for further noise suppression, can be applied without causing significant bias. This second smoothing operation uses a Gaussian kernel where the standard deviation is set to the desired time and range resolution. In the data presented here, the time kernel standard deviation is set to 10 minutes, and the range kernel standard deviation is set to 170 m.
[.. 96 ]The MPD lacks the temporal resolution to calculate water vapor uncertainty using the Lenchow method that is popular 565 in high power DIAL systems (Lenschow et al., 2000). In photon-counting systems, it is common to use linear propagation of error to estimate the statistical error in water vapor retrievals. We have found, however, that this error estimate rarely encapsulates statistical noise at low SNR, largely because of the highly nonlinear nature of the DIAL equation. Also, the assumption that the observed number of photon counts is an accurate representation of the mean photon counts, and therefore the statistical variance, becomes invalid at low observed counts. We have employed bootstrapping [.. 97 ] (Hastie et al., 2001;570 Marais, 2017) for estimating statistical uncertainty numerically, which gives a much more robust error estimate.
To implement bootstrapping on the MPD data, we split the photon count profiles into two statistically independent sets with identical mean values using Poisson thinning ([.. 98 ]for details on the Poisson thinning method see Hayman et al. (2020)).
We calculate the water vapor number density[.. 99 ], as detailed above, for both of the independent photon count sets. We then estimate the statistical variance as half the difference squared of the two statistically independent water vapor profiles.

575
This thinning and processing is repeated so that the final variance is approximated as the average of all the squared-difference results given by During the baseline intercomparison test, we removed the near-range receiver's contribution on a single unit for one night to isolate its effect. The MPD05 WFOV telescope was blocked from 30 Sep (23 UTC) to 1 Oct (14 UTC). During this period, a wet bias appears in the data, extending to just above 500 m. When the light received from the WFOV telescope was reintroduced (i.e., the telescope was unblocked), the bias disappears. This effect can be seen in the Fig. 12 time series. Profiles of these two periods, during and after the block, are shown in Fig 13. This result -where only the WFOV telescope is added/subtracted -is a 600 compelling demonstration that the fiber-coupled near-range receiver is a significant component in improving the measurements close to the surface.
To further understand the performance difference between the 2019 Gen 4 and the 2020 Gen 5 network intercomparison tests, the correlation coefficient as a function of the range was computed and is shown in Fig. 14. Overall there is a clear improvement at low ranges between revisions concerning the instrument agreement with one another. Although we implemented multiple 605 changes between the design stages, two are likely most responsible for this improvement: the WFOV receiver's addition, which increases the overlap function near the surface approximately 100x, and the two-stage seed laser, which provides more power in the amplifier. These improvements may be coupled together. Their contributions are not easily separated since a difference 101 removed: (we use 102 removed: in this analysis) . This 103 removed: each step of the processing 104 removed: Gen 5 Field Testing in online and offline beam pointing can be a source of error in a DIAL. Wulfmeyer and Bösenberg (1998) discussed how even slight deviations in the overlap functions between the on-and offline signals could give rise to considerable errors. Like 610 the system described in that work, one would expect the MPD online and offline overlap functions to be equal since a single amplifier generated both beams. Still, Wulfmeyer and Bösenberg (1998) warn that systematic errors may occur in regions of incomplete overlap even with very slight deviations in the direction of the two beams. As shown above, the improved seed power in the Gen 5 MPD more fully saturates the final amplifier stage. This saturation should minimize potential gain guiding in the TSOA, and therefore beam pointing differences would be minimized in the new design. Additionally, since the beam efficiency at lower ranges. As discussed previously, the improved collection efficiency also minimizes any signal-induced bias in the detector that affects the lowest ranges. We do not attempt here to deconvolve these factors but theorize that a combination results in improved measurements closer to the surface.  2. As mentioned previously, seventy-seven sondes were released by the DOE SGP staff during those twelve days. During the Oklahoma springtime, the atmospheric conditions had humidity variability in the boundary layer ranging from 2-15 g m −3 .
As seen on the left side of Fig. 16, the absolute humidity RMSE is around 1 g m −3 from 500 m to 5.75 km. Unlike what was shown earlier in this manuscript, the low-range masking is now removed. This was done to show the substantial wet bias below 640 500 m with the Gen 4 dataset that was collected without a near-range channel. The Marshall dataset was collected in the dry Colorado winter. The sample size is considerably smaller; only 10 of the 12 sondes released during the 90 days overlapped with the instrument operation. However, with these caveats, the results show that the Gen 5 instrument provides measurements that agree with the radiosonde's absolute humidity with an RMSE < 1 g m −3 extending down to 225 m AGL. The individual profiles from the Marshall radiosonde comparisons are shown in Fig. 15. These initial results further validate that the Gen 645 5 MPD design provides measurements closer to the surface. They also provide a nice example of MPD's ability, employing the narrowband DIAL technique, to accurately measure water vapor over a wide range of atmospheric conditions.

Conclusions
The atmospheric research and operational weather forecasting communities have continued to advocate for a ground-based network of instruments that profile water vapor and temperature at high temporal and vertical resolution in the lower atmosphere.

650
NCAR and MSU have developed an active remote sensor technology to meet this scientific and societal need. Our design approach is founded on the use of cost-effective, narrowband, high-spectral-fidelity diode lasers. These single frequency laser diodes are much less powerful than those historically used by the community for the quantitative atmospheric lidar methods of DIAL and HSRL. They require receiver designs with a narrow field-of-view and precise sunlight blocking filters. However, they have the significant advantage of potential lower-cost and are inherently more rugged for field operations, especially when 655 fiber-coupled. The mean error and RMSE [.. 108 ]were calculated assuming radiosonde data was the truth. Results from a pair of field studies are shown.
The plots in black are from the spring/summer field campaign (data from 19-Apr to 1-May 2019) at SGP with the Gen 4 system, without a near range channel. The plots in red are from the wintertime test at the Marshall field site with the Gen 5 system (data taken from 16-Oct-2020 to 14-Jan-2021).
We have constructed a five-unit MicroPulse DIAL network and performed an initial field demonstration test that allowed evaluation of the engineering performance in a deployment setting. The network, initially built on 4th generation MPD technology, was validated against more mature instruments. The MPD water vapor measurements were shown to compare very well with Raman lidar and radiosondes. Following the field test, we developed several advancements to the MPD technology 660 for the next generation. These improvements included a robust fiber-coupled seed laser. The transmitter operational parameters were mapped out and shown to meet all of the requirements for the DIAL technique. Further improvements such as a fibercoupled near-range receiver, the ability to perform quality control via automatic receiver scanning, advanced multi-channel scalar capabilities, and advanced processing techniques are discussed. Throughout we explain the balance of design trade-offs.
The new developments -which increase narrowband DIAL technology readiness -were shown to allow higher quality water 665 vapor measurements closer to the surface via intercomparisons within the MPD network itself and with radiosondes.
We developed the latest MPD generation in a modular fashion -to allow the straightforward addition of calibrated aerosol and temperature measuring capabilities to the network testbed units in the future. This work demonstrated the advances in water vapor profiling. As they share the same diode-laser-based architecture, the calibrated aerosol and temperature measurements also benefit from these developments. Work is proceeding to improve those advanced measurement techniques. We are also 670 actively investigating new methods and advanced signal processing techniques to measure closer to the surface and extend the range. The self-heterodyne technique was used to measure the linewidth and spectral purity of the transmitter. As shown in Fig.   A1 a 'pickoff' optic reflected a small portion of the pulsed output from the transmitter's last amplifier stage into a fiber port that focuses the light into a single-mode fiber. A fiber-coupler split the light into two paths. The 90% leg passed through a polarization controller and a fiber-coupled acousto-optic modulator (AOM, Brimrose model TEF-1000-100-828-2FP-680 SM) to shift the frequency 1 GHz. A 50/50 splitter combined the frequency-shifted light with the 10% leg containing the originally sampled beam. One leg of the 50/50 splitter connects to a Si PIN Amplified Detector (ThorLabs FPD310-FC-VIS) with 1 MHz -1.5 GHz bandwidth. The amplified detector has a 2 mW damage threshold, so the sampled pulse was attenuated so the peak power was below that threshold. An Electrical Spectrum Analyzer (ESA) -operated at 50 kHz bandwidth resolution, 100 MHz to 1.9 GHz span, ten shot averaging, and log-power detector -measured the detector 685 output. The optical filters in the MPD receiver limit the bandpass to approximately <1.3 GHz FWHM, so this measured frequency spectrum is sufficient to ensure the instrument's spectral purity.  geometrical optical relationship can be used to approximate the range to full overlap -the hyperfocal distance H ≈ D F OV . This simple relationship makes it easy to see that to efficiently collect light closer in range (and thereby closer to the surface), one could increase the receiver field-of-view[.. 115 ], F OV , and/or decrease the collection area diameter, D.
Increasing the field-of-view also collects more background light, so the wide FOV receiver diameter should be sized to compensate for this effect. To maintain a similar background collection efficiency to the [.. 116 ]original primary receiver, we 695 select a small-diameter telescope with a wide field-of-view. We further balance the background levels by selecting how the two receivers are combined. [.. 117 ] The background photon count rate, N B , at the receiver is given by where [.. 119 ]S B is the background radiance, Ω(F OV ) is the receiver solid angle as a function of the full-angle field of view 700 (FOV) and is equal to π F OV 2 2 , ∆λ is the spectral bandpass of the receiver, [.. 120 ]A is the receiver area, η is the receiver efficiency, and λ hc is the reciprocal of the energy per photon. The primary and [.. 121 ]WFOV telescopes are combined, so the background radiance, bandpass, and receiver efficiency for both receivers are the same. Therefore the difference in relative background collected between the two telescopes is proportional only to the receiver area and field-of-view where A 1 is the primary reflector aperture area, A 2 is the WFOV refractor aperture area, F OV 1 is the field of view of the primary reflector and F OV 2 is field-of-view the WFOV refractor, and η 1 is the [.  Figure B1. We now estimate the signal as a function of range. Since the background levels are matched, an improvement in signal is equivalent to an 720 improvement in SNR.
Appendix C: [.. 135 ] [.. 136 ]The relative signal count rate, N S as a function of range, r, is proportional to receiver area A and the overlap function, O(r), of each receiver, so the total signal, normalized to the original primary reflector aperture is where O 1 (r) is the overlap function of the primary reflector and O 2 (r) is the overlap function of the WFOV refractor The aperture normalized signal is shown in Figure B2 for the primary reflector telescope (η 1 = 1), the WFOV refractor telescope (η 1 = 0) and the combined configuration (η 1 = 0.9). From this figure, it is clear that the new combined telescope configuration provides an increased signal below 500 m as compared to the primary receiver alone.
124 removed: new combined receiver's relative background -the 90% fiber-coupled leg connected to the primary telescope and the 10% fiber leg attached

745
The NCAR authors wish to acknowledge and thank: Ben Crane and Kina Smith for technician support, Ed Eloranta and Amin Nehrir for helpful discussions about lidar design, and Tammy Weckwerth and Holger Vömel for useful discussions about the scientific application and internal review of the paper. We thank Dale Hurst, Patrick Cullis, Emrys Hall, and Allen Jordan of the NOAA Global Monitoring Laboratory for providing the sounding data from the Marshall field site.