A differential absorption light detection and ranging instrument (Differential Absorption LIDAR or DIAL) was installed on-board the Canadian Coast Guard Ship Amundsen and operated during the winter and spring of 2008. During this period the vessel was stationed in the Amundsen Gulf (71°N, 121–124°W), approximately 10–40 km off the south coast of Banks Island. The LIDAR was operated to obtain a continuous record of the vertical profile of ozone concentration in the lower atmosphere over the sea ice during the polar sunrise. The observations included several ozone depletion events (ODE's) within the atmospheric boundary layer. The strongest ODEs consisted of air with ozone mixing ratio less than 10 ppbv up to heights varying from 200 m to 600 m, and the increase to the background mixing ratio of about 35–40 ppbv occurred within about 200 m in the overlying air. All of the observed ODEs were connected to the ice surface. Back trajectory calculations indicated that the ODEs only occurred in air that had spent an extended period of time below a height of 500 m above the sea ice. Also, all the ODEs occurred in air with temperature below −25°C. Air not depleted in ozone was found to be associated with warmer air originating from above the surface layer.
 The first reported observations of Arctic tropospheric ozone depletion were made during the 1980s [Oltmans, 1981; Oltmans and Komhyr, 1986; Bottenheim et al., 1986; Barrie et al., 1989] using measurements at costal ground based monitoring stations. It was found that surface ozone mixing ratio periodically dropped from normal (30–40 ppbv) levels to near zero for periods ranging from hours to weeks at a time, starting in the spring, and coinciding with the polar sunrise. The currently understood mechanism facilitating these depletion events is the photochemical conversion of inert halide salt ions (e.g., Br-) into reactive halogen species (such as Br atoms) that deplete ozone in the boundary layer [Ridley et al., 2003]. The autocatalytic nature of this reaction leads to a marked increase in BrO levels during a depletion event. An overview of the current understanding of possible mechanisms for tropospheric ozone depletions is given by Simpson et al. .
 The majority of polar sunrise surface ozone depletion studies have focused on time series measurements from costal weather stations or research outposts, where measured depletions are generally thought to be the result of atmospheric transport of depleted air rather than local chemistry. Studies focusing on high resolution measurements of the vertical extent of depletion events have made use of ozonesondes attached to weather balloons [Tarasick and Bottenheim, 2002; Roscoe et al., 2001; Wessel et al., 1998], which have limited temporal resolution. Here we report on measurements in which a differential absorption LIDAR was operated continuously over March 2008 to provide a temporal and vertical resolution of ozone depletion events near the sea ice surface.
 The DIAL system was installed in a laboratory container that housed experiments for the OASIS (Ocean Atmosphere Sea Ice Snowpack) project [Shepson et al., 2003]. It was mounted on the top deck of the Canadian research icebreaker CCGS Amundsen for the Circumpolar Flaw lead (CFL) campaign [Barber et al., 2010] during the International Polar Year 2007–2009 (IPY). The ship was stationed in the Amundsen Gulf, immediately south of Banks Island at an approximate location of 71°N and 121–124°W. The vessel remained mostly stationary with respect to the sea ice, with occasional changes in position to accommodate sea ice studies. The location of the vessel at this time of year provided a unique opportunity to retrieve vertical profiles of ozone depletion events over the sea ice in high temporal and spatial resolution during the period of Arctic polar sunrise.
2. Measurement Technique
 The York University differential absorption LIDAR measured profiles of ozone abundance to a range of 6 km. The basic technique is to emit pulses of light into the atmosphere and record the backscatter signal as a function of time, or equivalently range. For a UV wavelength absorbed significantly only by ozone, the single wavelength backscatter signal is described as
where P(R, λ) is the instantaneous received power from range R. The backscatter coefficient, β(R, λ), represents the fraction of light scattered backward per unit length and per unit solid angle. The extinction coefficient, α(R, λ), is the fractional decrease in laser pulse intensity per unit length due to scattering. The product of the ozone number density and absorption cross section, σ(λ)n(R), is the fractional decrease in laser pulse energy per unit length due to absorption by ozone molecules. C is a system constant that takes into account characteristics such as transmitted laser pulse energy, receiver aperture area, system optical throughput, and detector efficiency.
 The DIAL emits multiple closely spaced wavelengths in the UV that lie on the broad Hartley ozone absorption band. The differential absorption between wavelengths with different absorption cross sections is employed to derive the ozone density. Ozone retrieval from the recorded signal is performed by calculating the slope of the logarithmic ratio of any pair of these signals and is described as
where ‘on’ denotes the wavelength with the larger ozone absorption cross section where the term 2(αm(R, λon) − αm (R, λoff)) is a correction factor to account for differential extinction due to molecular scattering.
 For the measurements in this study there was not a significant contribution to the signal by aerosol and cloud particles and the differential absorption and scattering due to aerosol had a negligible effect on the derived ozone concentration. Measurements where clouds were a significant factor were omitted when performing the analysis. The molecular scattering and extinction coefficients were calculated from atmospheric densities determined using data from meteorological balloons (radiosondes) launched at the Inuvik MET station (68.31°N, 133.53°W), located approximately 500 km from the Amundsen.
 A schematic diagram of the DIAL system is shown in Figure 1. The fourth harmonic of a Q-switched Nd:YAG laser (266 nm, 70 mJ per pulse, 20 Hz repetition rate) was focused into a gas cell filled with 140 PSI of CO2 to generate pulsed light at wavelengths 276 nm, 287 nm and 299 nm by stimulated Raman scattering [Nakazato et al., 2007]. The diameter of the multiwavelength output beam was expanded by a factor of three to reduce the divergence to about 0.2 mrad before it was directed into the atmosphere. Full overlap between the emitted laser pulse and the telescope field of view occurred at a range of approximately 300 m and the signal received from within that range was not used in the ozone analysis. The lower height limit for the measurements was reduced to 52 m above the LIDAR (∼72 m above sea level) using a removable mirror mounted to the exterior of the lab container, to direct both the emitted light pulses and the telescope field of view at an angle of 10° above the horizontal. The DIAL measurements alternated between zenith and slant viewing each day when conditions permitted.
 Backscattered light was collected with a 15 cm off axis parabolic mirror and a 1.5 mm diameter optical fiber positioned in the focal plane, 1000 mm from the mirror, to form a receiver field of view of 1.5 mrad. The three wavelengths were separated in the receiver by the transmittance and reflectance from interference filters tilted at angles of 5 degrees and having bandwidths of 1 nm. Photomultiplier detectors generated electrical signals that were recorded using photon counting for the weak signals from distances greater than 1.8 km and using analog to digital conversion for the strong signals in the near range. The raw data was recorded with a range gate of 7.5 m and an integration period of 1200 laser shots (1 min). During data processing the collected LIDAR profiles were averaged both spatially and temporally in order to reduce the measurement uncertainty. The amount of averaging applied is dependent on the desired measurement range. The 276/287 nm analog signals were used to a range of 1.8 km (height 332 m above sea level when slant viewing) and the 287/299 nm photon counting signals were used at greater ranges. Figure 2 shows an example of backscatter signals recorded using photon counting along with the corresponding derived ozone mixing ratio.
 Temperature profiles were acquired with a microwave Radiometer (TP/WVP-3000) capable of measuring with a vertical resolution of 50 m for heights below 1 km. [Güldner and Spänkuch, 2001; Liljegren et al., 2001]. Profiles were recorded once per hour and a continuous record exists for all but two days in March 2008.
 A contour plot of the LIDAR ozone measurements over March and early April of 2008 from 60 m to 1000 m above the sea ice is shown in Figure 3. These data are a composite of both zenith and slant measurements. The contour is composed of individual profiles that were averaged temporally over a window of 60 min as well as over vertical bins. The range resolution of the analog signals used below height 332 m was 22.5 m while the photon counting signals were averaged to a range resolution of 75 m. A number of ODE's were vertically resolved with the LIDAR and also measured in situ on the ship (Figure 3).
 The most complete ozone depletions, with mixing ratio less than 5 ppbv, occurred on March 12th, 13th–14th, 16th–18th, and 21st–23rd. These depletions, which lasted from 1 to 2 days, were quite shallow, with the most depleted pockets of air connected directly to the ground, and not extending past 250 m in height. Above this altitude the mixing ratio generally started increasing with altitude until background levels of 30–40 ppbv were observed. The greatest vertical extent was associated with the partial depletion in the period between March 14th and 19th, which reached a height of 1 km at times. One of the aims of the measurement campaign was to determine if there was evidence for local ozone depletion in air not in contact with the ice surface. No such evidence was observed with the LIDAR. All observed ODEs were connected to the surface. There were instances where an isolated patch of air partially depleted in ozone (e.g., 15 ppbv) was vertically separated from the main ODE that extended from the surface (e.g., March 21/22). This only occurred in the region of mixing between the ODE and the overlying air with background ozone density. There were no cases of an isolated patch of ozone depleted air that was not directly above an ODE at the surface. Also, there was no measurable diurnal correlation between local sunlight and ozone mixing ratio at the surface, or aloft. Diurnal variations of BrO concentrations over the sea ice were measured on board the Amundsen [Pöhler et al., 2010], indicating local generation of BrO, but the measured ozone concentrations did not exhibit similar diurnal variability.
 Back trajectory calculations provided a basis for interpreting the origin and vertical structure of the observed ozone depletions. The calculations made use of the NOAA HYSPLIT model [Draxler and Hess, 1998], using GDAS (Global Data Assimilation System) meteorological data. The back trajectories were started at each hour for heights ranging from 100 m to 1000 m above sea level and were calculated for a 6-day period backward starting from the location of the Amundsen at the time of measurement. The calculated back trajectories near the ground agree well with the wind direction, wind speed and atmospheric temperatures that were measured independently on board the vessel.
Figure 4 shows the calculated back trajectories for six case studies that are indicated in Figure 3. In each case, whether or not an ODE was observed, the air had traveled from the north and over the sea ice before passing over the ship. An aspect of the back trajectories that had a more significant correspondence with the occurrence of ODEs was the amount of time that the air spent at heights below 500 m, where there would have been mixing with air that came in contact with the sea ice. For example, in case-A, when there was no ODE, the back trajectories that were started above the Amundsen at heights of 100 m, 300 m, 600 m, and 1000 m each spent the previous days well above height 500 m and descended only in the 12 h prior to passing over the Amundsen (Figure 5a). In case-B the ozone mixing ratio was less than 5 ppbv up to a height of 220 m and then gradually increased to the background level of 30 ppbv at height 500 m. In this case the back trajectories starting from 100 m and 300 m above the ice surface had spent significant time below 500 m in the previous days, while the back trajectories starting at heights 600 m and 1000 m did not spend any time below 500 m in the previous six days. In case-C the air that was partially depleted in ozone, lying above the main surface ODE, extended up to a height of 1000 m (Figure 3a). Figure 5c shows that in this case all of the back trajectories that started below 1000 m had spent at least three of the past six days at heights within 500 m of the sea ice. Case-D is another example where the ozone depleted air extended to a height of 1000 m (Figure 3a), and the back trajectory analysis indicates that the measured air had ascended from below height 500 m up to 1000 m in the three days prior to passing over the Amundsen. In case-E the ODE extended only to a height of about 300 m. Figure 5e shows that the back trajectories starting at 100 m and 300 m had each spent the previous 4.5 days below height 500 m, but the trajectory starting at 1000 m remained well above height 500 m during the previous six days.
 Despite that the uncertainty in back trajectory calculations is not exactly known, the case studies do show that there is a correspondence between the occurrence of ODEs and the altitude history of the air. By integrating the calculated altitude of the air masses over a 144-h period, it was found that lower average height was associated with lower ozone mixing ratios. An example of this for back trajectories calculated at 100 m is shown in Figure 6a. Depletion events were associated with air masses closely connected to the ground. During periods of strong depletions with ozone mixing ratios of 10 ppbv or less, the associated air mass was found to have a 6-day average altitude of 500 m and lower, close to the previously estimated height of 400 m for the boundary layer over Arctic sea ice [Bottenheim et al., 2002]. For the period of time between 9 March and 25 March 2008, a positive linear correlation of 0.65 was found (Figure 6b) between the LIDAR measured mixing ratios and the calculated 6-day historical mean altitude of the corresponding air masses.
 There are exceptions to the pattern in terms of the altitude history of the air. For example, the back trajectory that started at height 600 m in case-E did spend time at heights between 500 m and 300 m, but the ODE did not extend above 300 m. In case-F there was no ODE at any height and the back trajectories had all descended from greater heights over the past six days. The air passing over the Amundsen at heights 100 m and 300 m had been below 500 m for the previous two days and yet there was no ODE near the surface. This is another exception that implies there are factors other than height above the surface that affect ozone. It was found that temperature was another important factor.
 A contour plot of the potential temperature measurements from 80 to 1000 m is shown in Figure 7a. The contours were calculated from temperature profiles collected every 60 min by the zenith oriented microwave radiometer. Comparison between Figures 3 and 7 indicates that air with low ozone mixing ratio also has low potential temperature. (The correlation of ozone mixing ratio with absolute temperature was not as strong as with potential temperature.) This suggests it is the temperature that air aloft would have had while previously near the surface that is important. In cases B and E the potential temperature near the ground was less than −30°C and increased with altitude. In both cases, the strong depletion of ozone near the ground ended with a sharp increase in mixing ratio corresponding to a potential temperature of −27°C at altitudes of 200–250 m.
 For case-C and case-D the near surface temperature was colder than in the previous two cases, and the potential temperature inversion was less pronounced. In both cases partial depletions extended above 700 m in height. In each case an ozone mixing ratio greater than 25 ppbv corresponded to potential temperatures greater than −25°C and mixing ratios below that decreased linearly with potential temperature. This trend was also apparent in case-A in which an ODE was not observed and temperatures were relatively high (greater than −25°C) at the ground and aloft.
 The scatterplot (Figure 8) of ozone mixing ratio and potential temperature shows a strong statistical correlation with a positive linear correlation constant of 0.85. The strongest depletions at the ground occurred at temperatures below −25°C.
 As previously noted, the case on 29 March 2008 (case-F) did not follow the pattern established with respect to altitude history in the back trajectory. As in case A, there was no sign of ozone depletion, and this period of time is marked by a high ozone mixing ratio slightly above normal background levels. Unlike case-A, the calculated back trajectories indicated that the air measured at the surface had been below a height of 200 m for approximately 60 h, a condition associated with ODEs in the previous cases. This period of time was characterized by warm temperatures (greater than −20°C) over the sea ice, and is thus consistent with the previous cases in which the ODEs occurred only in air with potential temperature below −25°C.
 While the uncertainties in the back trajectory analysis are not quantified, comparison with the profiles of measured potential temperature gives confidence in the back trajectory analysis. The cases where air had risen from the surface (case-C and case-D) also have relatively low potential temperature up to greater heights. The cases where the back trajectory indicates that air had descended toward the surface (case-A and case-F) have relatively warmer potential temperature, again as expected. This also suggests that the correspondence of ODEs with low potential temperature is not separable from the correspondence with proximity to the sea ice surface. The air that was cooled substantially by prolonged turbulent heat exchange with the sea ice surface would also have low potential temperature. This then explains two remaining apparent exceptions to the overall pattern. In case-E the air arriving at height 600 m had spent time below 500 m in the calculated back trajectory, but the measured potential temperature was relatively warm at −22°C, which indicates that there was not substantial contact with the surface in the previous days. In early April there was a depletion despite that the average heights in the back trajectory analysis were well above 500 m. However, the measured potential temperature was relatively cold and well below −25°C, which suggests the air had spent time near the surface. It was found that the ozone depleted air did actually spend time near the surface when the back trajectory was extended beyond six days in duration.
 The Amundsen CFL/OASIS campaign during the 2007–2008 International Polar Year provided a unique opportunity to investigate Arctic boundary layer ozone depletion events directly over sea ice. The differential absorption LIDAR provided a continuous record of the vertical distribution of ozone above the ship. A number of full and partial ozone depletion events were observed during March 2008 and all were in contact with the sea ice. No depletions were observed in isolated layers separated from the surface. Backward trajectory analysis indicated that most of the ozone-depleted air had spent extended periods within 500 m of the sea ice. A strong positive correlation was also found between ozone mixing ratio and potential temperature. No ODEs were found to occur at any altitude when the potential temperature was higher than −25°C. Periods of normal, or higher than normal ozone concentrations were recorded when sampling warm (>−20°C) air masses originating from the free troposphere that had no recent connection to the ground.
 The observed ODEs correspond to two factors: 1.) prolonged periods within 500 m of the sea ice, and 2.) low potential temperature. We interpret that the source for the measured ozone depletions was long-range transport of cold air masses that had extended periods of contact with sea ice by turbulent mixing within the atmospheric boundary layer. The observed ozone depletion events were primarily the result of previously depleted air masses being transported over the ship, rather than local chemistry. These results are consistent with previous studies that have found a correlation between Arctic ozone depletion events and the length of time an air mass is connected to the sea ice [Bottenheim et al., 2009; Frieß et al., 2004], and low temperatures [Strong et al., 2002; Ramacher et al., 1999].
 Funding for this work was provided by the Canadian Natural Sciences and Engineering Council (NSERC) International Polar Year program, the Canadian Foundation for Climate and Atmospheric Science (CFCAS), and the Canadian Foundation for Innovation (CFI). Data from in situ temperature measurements aboard the ship were generously provided by Tim Papakyriakou and Brent Else of the University of Manitoba.