Intercomparison of lidar , aircraft , and surface ozone 1 measurements in the San Joaquin Valley during the California 2 Baseline Ozone Transport Study ( CABOTS ) 3 4

Andrew O. Langford1, Raul J. Alvarez II1, Guillaume Kirgis1,2, Christoph J. Senff1,2, Dani Caputi3, 5 Stephen A. Conley4, Ian C. Faloona3, Laura T. Iraci5, Josette E. Marrero5,*, Mimi E. 6 McNamara5,7,§ , Ju-Mee Ryoo5,‡, and Emma L. Yates5,6 7 8 1NOAA Earth System Research Laboratory/Chemical Sciences Division, Boulder, CO 80305, USA. 9 2Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, CO, 80309, USA. 10 3Department of Land, Air, and Water Resources, University of California, Davis, CA, 95616, USA. 11


1
The San Joaquin Valley (SJV) of California is one of only two "extreme" ozone (O3) non-attainment areas 2 remaining in the United States with a 2016 ozone Design Value, i.e. the metric used by the U.S. EPA to determine 3 air quality compliance that is calculated as the 3-yr average of the 4 th highest measured maximum daily 8-h average 4 mixing ratio (MDA8), that is more than 20 parts-per-billion by volume (ppbv) greater than the primary National 5 Ambient Air Quality Standard (NAAQS) of 70 ppbv (https://www3.epa.gov/airquality/greenbook/hdtc.html). Such 6 high O3 concentrations are harmful to human health (U.S. Environmental Protection Agency, 2014) and impair plant 7 growth and productivity (Avnery et al., 2011a, b), adversely affecting both the $15 billion agricultural industry in 8 the SJV and the iconic forests of the nearby Sequoia and Kings Canyon National Parks (Panek et al., 2013). 9 10 The need to better understand the causes for the high surface O3 in the San Joaquin Valley has motivated several 11 major air quality studies over the years including the San Joaquin Valley Air Quality Study (SJVAQS) in 1990 12 (Lagarias and Sylte, 1991), the Central California Ozone Study (CCOS) in 2000, (Reynolds et al., 2010) and the 13 California Research at the Nexus of Air Quality and Climate Change (CalNex) field campaign in 2010 (Ryerson et 14 al., 2013;Brune et al., 2016). More recently, this issue was addressed by the 2016 California Baseline Ozone 15 Transport Study (CABOTS) organized and supported by the California Air Resources Board (CARB) 16 (https://www.arb.ca.gov/research/cabots/cabots.htm). CABOTS was designed to investigate the contributions of 17 background O3 (Jaffe et al., 2018) and the influence of stratospheric intrusions (Lin et al., 2012a) and long-range 18 transport from Asia (Lin et al., 2012b) on surface O3 concentrations in the SJV during late spring and summer.

19
Characterization of the vertical distribution of O3 in the lower and middle free troposphere above the SJV and 20 upwind regions with an accuracy of at least 10%, the nominal accuracy of ECC ozonesondes in the troposphere 21 (Smit, et al., 2014), was a key objective of the campaign, and O3 profiles were measured using three different 22 techniques (lidar, aircraft, and ozonesondes) in various parts of California. Integration of these datasets requires that 23 these measurements be intercompared (Ancellet and Ravetta, 2005;Beekmann et al., 1995;Kempfer et al., 1994; 24 Schäfer et al., 2002) and any differences between the various techniques understood and characterized. For pollution 25 studies, it is important that this validation includes the lowest 100 m, which is inaccessible to most ozone lidars 26 (Wang et al. 2017). In this paper, we compare O3 measurements from the NOAA ESRL multi-angle Tunable Optical 27 Profiler for Aerosol and oZone (TOPAZ) lidar with in-situ measurements from nearby regulatory and research 28 surface monitors, and from instruments flown aboard the UC Davis/Scientific Aviation Mooney (Trousdell et al., 29 2016) and Alpha Jet research aircraft based at NASA's Ames Research Center (Hamill et al., 2016;Yates et al., 30 2015). These comparisons, together with those from the multi-lidar (including TOPAZ) and ozonesonde Southern 31 California Ozone Observation Project (SCOOP) intercomparison conducted by the NASA-sponsored Tropospheric 32 Ozone Lidar Network (TOLNet) immediately after CABOTS (Leblanc et al., 2018), provide this validation. 33 34 2 California Baseline Ozone Transport Study (CABOTS)

12
The Bodega Bay and Half Moon Bay sites were located on the coast to sample the Pacific inflow, and the VMA was 13 chosen for the TOPAZ operations because of its central location in the SJV, the availability of the runway and 14 airspace for low approaches and aircraft profiles, and the presence of the co-located SJVAPCD wind profiler and 15 Radio Acoustic Sounding System (RASS) (Bao et al., 2008). The TOPAZ truck was parked on the west side of the The TOPAZ differential absorption lidar (DIAL) system was originally developed for the profiling of O3 and 26 particulate backscatter in the planetary boundary layer and lower free troposphere from NOAA Twin Otter aircraft 27 (Alvarez et al., 2011;Langford et al., 2011;Senff et al., 2010;Langford et al., 2012;Langford et al., 2010). The lidar 28 was reconfigured for mobile ground-based measurements in 2012 and deployed in this configuration to several field 29 campaigns including the 2013 Las Vegas Ozone Study (LVOS) (Langford et al., 2015) prior to CABOTS. The lidar 30 is installed in the back of a medium box truck (cf. Figure 2) equipped with a commercial UV absorption monitor for 31 in-situ O3 measurements (2B Technologies Model 205) that samples air 5 m above the surface and an Airmar 32 150WX weather station to measure temperature, pressure, relative humidity, and wind speed and direction. The 2B 33 Model 205 has been approved by the EPA as a Federal Equivalent Method (FEM) for surface O3 monitoring and has 34 a nominal (1s) precision and accuracy that is the greater of 1 ppbv or 2% for 10-s averages. Modified versions of 35 the same instrument were flown on both the Scientific Aviation Mooney and NASA Alpha Jet. Comparisons 36 between the NOAA 2B at the VMA and a mobile calibration source operated by CARB revealed a 3% low bias in 1 the recorded 2B measurements that has been corrected in the data used here.

3
The eye safe TOPAZ lidar is built around a low pulse energy (~100 µJ), high repetition rate (1 kHz) quadrupled 4 Nd:YLF pumped Ce:LiCAF laser that is re-tuned between each pulse to generate light at three different wavelengths 5 from 286 to 294 nm with an effective repetition rate of 333 Hz for each wavelength (Alvarez et al., 2011). The laser 6 pulses are transmitted and the lidar return signals collected by a coaxial transmitter/receiver equipped with a 7 commercial (Licel) photomultiplier-based dual analog/photon counting system. This hybrid data acquisition system 8 was installed in 2016 and replaced the original fast analog data acquisition system that was optimized for aircraft 9 operations (Alvarez et al., 2011;Wang et al., 2017). This modification increased the maximum useful range to ~6 km 10 during the day and to more than 8 km at night, depending on the laser power, atmospheric extinction, and solar 11 background light.

13
The truck-mounted version of TOPAZ incorporates a large scannable turning mirror above the vertically pointing 14 transmitter/receiver to allow profile measurements at different slant angles. These slant profiles can be combined to 15 create vertical profiles that start much closer to the ground (25-30 m) than conventional vertically staring lidar 16 systems (Proffitt and Langford, 1997). During CABOTS, the scanning mirror was moved sequentially between

25
The O3 profiles shown here were retrieved using two wavelengths (~287 and 294 nm) with 30-m range gates and a 26 smoothing filter that increased from 270 m wide at the minimum range (815±15 m) to 1400 m wide at the maximum 27 range (8 km). The effective vertical resolution increased from ~10 m near the surface to ~150 m above 500 m agl 28 and 900 m at 6 km. Profiles of the backscatter from aerosols, smoke, and dust were retrieved with a constant 7.5 m 29 resolution at 294 nm. The ozone profiles were computed using the O3 absorption cross-sections from Malicet et al. 30 (1995) and an iterative technique to correct for differential aerosol backscatter and extinction that assumes a 31 backscatter-to-extinction ratio of 40 and fixed Ångstrom coefficients of 0 for backscatter and -0.5 for extinction 32 (Alvarez et al., 2011). These values offer a good compromise for a wide variety of particulate types (Völger et al., 33 1996). The actual aerosol composition in the SJV was not measured during CABOTS, but measurements during the 34 2010 Carbonaceous Aerosols and Radiative Effects Study (CARES) typically found a mix of organics, sulfate, 35 nitrate, ammonium, and soil dust in the northern part of the valley (Zaveri et al., 2012). Smoke from the Soberanes 36 Fire near Big Sur dominated the aerosol mix in the SJV during the second IOP. We varied the aerosol backscatter 37 Angstrom coefficient between -1 and 1 and the aerosol extinction Angstrom coefficient between 0 and -1 for a 1 "worst case scenario" of a thin smoke layer with very high aerosol backscatter embedded in an otherwise clean 2 atmosphere to estimate the error in the ozone retrieval introduced by using these fixed parameters. The sharp aerosol 3 gradients at the smoke layer edges tend to magnify errors in the ozone retrieval if the aerosol correction is not 4 properly implemented. Temperature and pressure profiles interpolated from the 3-h National Centers for 5 Environmental Prediction (NCEP) North American Regional Reanalysis (NARR) using the grid point closest to the 6 TOPAZ lidar location were used to account for the temperature dependence of the O3 cross-sections and to convert 7 O3 number densities to mixing ratios. The total uncertainties in the 8-min ozone retrievals in the absence of strong 8 aerosol gradients are estimated to increase from ±3 ppbv below 4 km to ±10 ppbv at the top of the profile. When 9 strong backscatter gradients are present, the O3 uncertainty can potentially increase by another ±3 ppbv. 2301f Cavity Ring-Down Spectrometer (CRDS) to measure CO2, CH4, and H2O (Trousdell et al., 2016). The 2B 20 model 205 was used with the minimum integration time of 2 s, which corresponds to a mean distance of 150 m at 21 the typical level flight speed (the data stream was sampled at 1-s intervals). As noted above, the 2B has a nominal 22 accuracy of 2% for concentrations above 5 ppbv, and a precision of 2% for concentrations above 5 ppbv if 10-s 23 averages are used. If the limiting noise is randomly distributed, this implies a precision of 5 ppbv for 2-s averages.

24
Calibrations of the Scientific Aviation 2B using an external ozone source (2B, Model 306) found the instrument to 25 have offsets and slopes less than 1.5 ppb and within 4% of unity, respectively.

28
The NASA Ames Alpha Jet Atmospheric eXperiment (AJAX) (Hamill et al., 2016) sampled O3 and other 29 tropospheric constituents above California during CABOTS using a two-person jet based at Moffett Field, CA (MF, 30 37.415° N, -122.050° E). The Alpha Jet carried an external wing pod with a modified commercial UV absorption 31 monitor (2B Technologies Inc., model 205) to measure O3 (Ryoo et al., 2017;Yates et al., 2015;Yates et al., 2013) 32 and a (Picarro model 2301-m) cavity ringdown analyzer to measure CO2, CH4, and H2O (Tanaka et al., 2016). A 33 second wing pod carried a non-resonant laser-induced fluorescence instrument to measure formaldehyde (CH2O) 34 (St. Clair et al., 2017). The pod mounting kept the residence times of the sample inlets to less than 2 s. The aircraft is 35 also equipped with GPS and inertial navigation systems to provide altitude and position information, and the NASA 36 1 and 3-D wind data. The 2B O3 data, recorded every 2 s, are averaged over 10 s to increase the signal-to-noise ratio.

2
Ozone calibrations were performed before/after each flight using an external ozone source (2B Technologies Inc., 3 model 306 referenced to the NIST scale, certified annually). Raw flight O3 data were corrected using the linearity 4 correction factor and zero offset from the calibration closest in time to the flight. Overall accuracy of the O3 5 instrument is determined to be 3 ppbv or better at 10-s resolution, with uncertainty improving at lower altitudes, as 6 determined from pressure chamber tests; see Yates et al., (2013) for a more detailed error analysis.

17
The NOAA 2B ozone monitor operated continuously at the VMA throughout the TOPAZ deployment with the 18 system response checked during each IOP by an external mobile calibration source operated by CARB. These 19 calibration checks revealed a 3% low bias in the NOAA 2B instrument that has been corrected in the data shown 20 here. shows that the in-situ concentrations measured at VMA were often much smaller than the concentrations measured 36 815±15 m away by the lidar, and even titrated to zero under some conditions. The data converge (filled black 37 circles) when the comparison is restricted to conditions when the two measurements are expected to sample a 1 common airmass, i.e. during the day after the nocturnal inversion has dissipated (0900 to 1830 PDT) and the winds 2 were southeasterly (125 to 145°) and greater than 2.5 m s -1 . The results of Orthogonal Distance Regression (ODR) 3 fits of these data are shown both in the figure and in Table 1. We use ODR fits, which assume that both variables 4 can have uncertainties, for our analyses instead of simple linear regressions which assume that all of the 5 uncertainties lie in the dependent variable. Fits of the filtered data give a slope of 1.00±0.03 and an intercept of -6 2.6±1.5 ppbv where the errors represent the 95% confidence limits of the ODR fits. 7 8 Figure 5 compares the 27.5 m TOPAZ O3 measurements to the regulatory O3 surface measurements from the 9 monitors at Visalia (8.5 km) and Hanford (24 km) described above, and from the more distant SJVAPCD monitors 10 at Parlier (34 km) and Porterville (43 km). The TOPAZ mixing ratios were slightly higher than those at Visalia and 11 Hanford, but lower than those at Parlier and Porterville, which are closer to the Sierra foothills and measure some of 12 the highest O3 concentrations found in the SJVAB. The degree of correlation decreased with distance as expected, 13 yet remained quite good more than 40 km from the VMA at Porterville. This suggests that the O3 measurements 14 acquired at the VMA during CABOTS can be considered representative of the central San Joaquin Valley.

17
Comparisons between the ground-based lidar and aircraft measurements are subject to much larger uncertainties 18 arising from spatial and temporal sampling differences compared to the comparison with nearby surface monitors.

19
During CABOTS, the fixed wing aircraft conducted both low approaches above the VMA runway (cf. Figure 2) and 20 spiral profiles around the airport, but never directly sampled the vertical column probed by the lidar. The  (Leblanc et al. 2018). Despite these caveats, we show that the lidar and 29 aircraft measurements usually agreed to within ±10%, the nominal accuracy of ECC ozonesondes in the troposphere 30 (Smit, et al., 2014), which is the generally accepted reference standard for ozone profile measurements.

33
The RLO flights were executed as a series of 2 to 3-day deployments with as many as 4 flights per day lasting 2 to 3 34 hours each between Fresno and Bakersfield. Two of these deployments, RLO2 (2)(3)(4), and RLO4 (24-26 July),

23
The scatter plots in Figures 8e and 8f show that the aircraft also measured much higher concentrations than the in-24 situ surface monitor during the night and early morning, in agreement with the lidar measurements in Figure 4. The 25 differences were smaller on 27 July than on 3 June, and also less pronounced than those in Figure 4. Closer 26 agreement between the aircraft and surface measurements might be expected since some of the aircraft 27 measurements were made within 200 m of the lidar truck (cf. Figure 2). The dark blue points show that the low bias 28 in the surface measurements decreased during the day after the surface inversion had dissipated (there were too few 29 measurements to effectively filter them by windspeed or direction). The mean ODR fit parameters based on the 30 measurements from both RLO2 and RLO4 listed in Table 1 are very similar to those found for the lidar which 31 suggests that the filtered surface measurements still have low bias that could be either instrumental or sampling 32 related.

5
Overall, the agreement between the TOPAZ and Mooney profiles in Figure 9 is within ±10%, but there are some 6 notable discrepancies. Most of these arise from the coarser vertical resolution of the lidar retrievals, which smooth 7 out abrupt concentration changes such as those seen at the top of the boundary layer (~0.8 km agl) in   Figure   24 10b, and Figure 10c shows the NO2 and H2O profiles measured by the aircraft. The backscatter measurements show 25 that the TOPAZ retrievals are unaffected by strong backscatter gradients, which can create second-derivative like 26 inflection points in the DIAL O3 profiles (Kovalev and McElroy, 1994). The absence of a corresponding structure in 27 the aircraft NO2 and H2O profiles confirms that the high O3 layer seen in the lidar and aircraft measurements was not 28 an artifact caused by interferences from these species, which weakly absorb between 280 and 300 nm (Proffitt and 29 Langford, 1997

29
The results of the different O3 comparisons are summarized in Table 1. As was noted above, comparisons between 30 the lidar and aircraft profiles are subject to uncertainties arising from sampling differences introduced by the 31 intrinsic vertical smoothing of the lidar retrievals and horizontal displacements between the aircraft and lidar. The 32 potential impact of horizontal displacements on the comparisons when the O3 spatial variability is large is illustrated 33 by Figure 14, and a good example of the differences created by the lidar smoothing is seen near the top of the 34 boundary layer around 0.8 km in Figure 9a. These uncertainties can be reduced by averaging the measurements to 35 be compared over larger volumes. Figure 15 compares the lidar and aircraft measurements from the profiles plotted 36 in Figures 9 and 13, and from several other RLO and EPA flights not shown, with each individual profile averaged 37 over 1 km segments (0 to 1 km, 1 to 2 km, etc.). This averaging decreases the influence of O3 spatial variability, and 1 also reduces the statistical uncertainties in both the lidar retrievals and aircraft measurements, with the effective 2 temporal averaging of the AJAX and SciAv measurements increasing to about 2 and 4 minutes, respectively. Each 3 point in the scatter plots of Figure 15a and 15b represents the mean mixing ratio from one of these 1 km segments, 4 with the error bars showing the standard deviation of the mean. The intercepts and slopes derived from orthogonal 5 distance regressions of both datasets overlap with zero and unity, respectively, within the 95% confidence limits of 6 the ODR fits. The lower panels (Figures 15c and 15d) plot the same data as differences which show that the TOPAZ 7 and SciAv measurements (Figure 15c) agree to within 1 ppbv on average, and the TOPAZ and AJAX 8 measurements (Figure 15d) to within 4.2 ppbv. Neither plot shows evidence of a systematic altitude dependence in 9 the differences.   Figure 13b where the lidar retrieval is clearly smoothing 16 out vertical gradient compared to the aircraft measurements. If this measurement point is excluded, the mean TOPAZ-17 AJAX difference decreases to 3.9±2.6. In either case, the differences between the TOPAZ lidar retrievals and the in-

30
The coordinated lidar and aircraft sampling of O3 above the central San Joaquin Valley during CABOTS also 31 illustrates the synergy between the two types of measurements. Lidar can provide long time series of the O3 (and 32 backscatter) vertical distributions above a fixed location while the aircraft can place the lidar measurements within a 33 larger spatial context and measure other important parameters. This synergy is illustrated by the two time-height 34 curtain plots displayed in Figure 16. Figure 16a shows the continuous TOPAZ measurements from a 14-hour time 35 span on 25-26 July with the data from SciAv FLT 35, 36, and 37 superimposed. The aircraft measurements made 36 within 5 km of VMA are highlighted by colored squares outlined in white. Figure 16b is similar, but shows 10-1 hours of continuous TOPAZ measurements from 15 June with the AJAX measurements (AJX191) superimposed.

3
The CABOTS ozonesondes were launched too far away (>300 km) from the VMA to allow quantitative 4 comparisons with the lidar. However, TOPAZ was relocated to the NASA Jet Propulsion Laboratory (JPL) Table   5 Mountain Facility (TMF) in the San Gabriel Mountains immediately after CABOTS for the Southern California