Operating dates: 28-October-2016 12:10 to 18 March 2017 8:35; 23-May-2016 13:30 to 13-September-2017 13:50.
Location Spectrometer: Roof of 2-storey building at 2 Percy St, Auburn NSW 2144, Australia, -33.85472, 151.0373, 20.6 m above sea level at rooftop, height rooftop above street level = 6.72 m, measurement path above roof-top 1.2 m
Location Reflectors: Roof 3-story building, Cumberland City Council Auburn Office, 1 Susan St Auburn 2144, Australia, -33.85311, 151.0335, 40.8 m rooftop above sea level, height roof top above street = 12.8 m, height mirror above rooftop = 2.7 m to centre mirror
Distance instrument to reflector = 395.8 m one-way
Total return path-length = 793.6 m (includes 2x1 m internal reflection)
Measurement Path Slope: 5.3 Degrees; difference in Altitude = 20.9 m
Measurement Path Bearing: 295°52'09''
Instrument description:
A DOAS 2000 Differential Optical Absorption Spectrometer (DOAS; Thermo Environmental Instruments Inc., Franklin, MA, 02038, USA, Manufactured 1999) consisting of a 150W Xenon arc-lamp mounted in a telescope to act as emitting and receiving optics was used to make measurements in the Ultraviolet (UV) and Visible (VIS) regions. Light emitted from the telescope is returned via a retro reflector array, 150mm diameter, positioned 25-1000m away. Optic fibre is used to couple the telescope to spectrometer for analysis of the returned light. Light entering the spectrometer passes through an aperture and is sorted by wavelength using a grating. A 40nm region is then scanned at 1 angstrom intervals each detected using a photo multiplier tube (PMT). The resulting spectrum is then analysed as described in a following section.
The telescope unit was mounted onto a Gibraltar Heavy duty tripod assembly (Quickset International Inc., Illinois, USA), to provide coarse alignment to the retro-reflector. Fine alignment was achieved by built in alignment aids on the telescope unit. The focus was manually adjusted by moving the source position on a slid rail.
The measurement system is sensitive to a wavelength range of 200-650nm allowing for the analysis of many UV and VIS active pollutant gases in the atmosphere. Data is reported for Ozone (O3), Sulphur dioxide (SO2), Nitrogen dioxide (NO2), Formaldehyde (HCHO) and Nitrous acid (HONO). The precision for each species reported is; O3 = 3.9-4.8ppb, SO2 = 0.3-0.6ppb, NO2 = 2.4-4.1ppb, HCHO = 1.4ppb, HONO = 0.3ppb. Data for benzene and toluene were recorded for the first measurement period however the data was not of sufficient quality to be included.
Data collection is controlled by the manufacturer's software package called DOAS 2000. The software uses user defined method files for operation and in this instance the collection/analysis procedure was NO2 (430nm) -> HONO (355nm) -> HCHO (330nm) -> SO2 (300nm) -> O3 (283nm) -> Benzene/Toluene (262nm). The measurement procedure continuously cycles around until interrupted by user or external factors. Instrument hardware calibrations are also controlled through
the DOAS 2000 software.
Calibration: Hardware calibrations were performed every month while wavelength correction checks were performed every 4 hrs and stored for use in spectra processing. Hardware calibrations use a mercury lamp located at the end of the optic fibre feeding the spectrometer. Gas species were calibrated using background mole fraction values (O3, SO2 & NO2), determined from a portable air quality monitoring station (Office of Environment and Heritage, NSW), measuring like species and co-located at the DOAS 2000 telescope. The air quality monitoring station instruments are referenced to a standard gas mixture daily to ensure performance and data quality. Wind speed needed to be > 1m/s for the determined background values to be valid. Data were then scaled according to the ratio of the calibrated background v's the DOAS 2000 background. Zero offsets were applied to HCHO and HONO as there wasn't a like measurement for these species.
While the absolute accuracy of the measurements from the DOAS 2000 system can't be assured, based on the calibration strategy employed. However the differences have more confidence and are supported by a close match in scale and pattern of hourly averages from the portable air quality monitoring station.
Data Collection Rate: Each spectral window is scanned for 2 mins and the resulting spectrum analysed before moving to the next region. A cycle of six spectral windows took 12-15mins to complete while a 5 spectral window cycle took 10-12mins.
Spectral analysis: Spectral analysis was performed on-line by the DOAS 2000 software. In any UV-VIS measurement technique the resulting spectrum is dominated by the lamp or light source. To accurately retrieve information from the spectrum the lamp spectrum for the wavelength region needs to be subtracted. Prior to deployment the lamp was changed and new lamp spectra recorded to be used in analysis.
The residual spectrum is then compared to a reference spectrum (treated identically) to provide a path averaged mole fraction measurement of the gas of interest. Reference spectra used in this work were those supplied with the instrument in 1999.
Data QA: Data were filtered based on spectral fitting parameters that varied for each species but determined the quality of the fit. Restarting the collection procedure caused a reset of the stored wavelength adjustment corrections, therefore the correction was not made until a wavelength adjustment cycle was run on its regular programmed 4 hr interval. Data collected without the required wavelength adjustment correction were removed. This procedure assisted in removing effects due to alignment changes.
Interruptions and issues: Maintaining a stable alignment of the optical system in an outdoor environment proved to be a challenge and significant data were lost. Outages due to alignment loss lasted upto 12 hrs. During summer the loss of alignment was rapid and only relatively short periods (hours) experienced a rapid change in signal to noise. During winter the alignment loss was gradual over many hours with more measurements made at a high PMT voltage, introducing more noise and thus lowering the signal to noise. This is noticeably by the increased scatter in the second half of the data set.
Reliability of the spectrometer also caused data loss until a component failure on the 23/11/2016. Spectrometer reliability improved after the component replacement.
The following extended (> 1day) outages were experienced:
29/10/2016 - 2/11/2016 poor alignment, site visit to correct.
11/11/2016 - 14/11/2016 alternate instrument deployed, signal to noise poor, data deleted.
23/11/2016 - 9/12/2016 component failure. Identical component from spare instrument used.
23/12/2016 - 10/1/2017 holiday period shutdown, site access unavailable.
10/6/2017 - 13/6/2017 instrument access issues
18/8/2017 - 21/8/2017 strong winds
Monthly spectrometer hardware calibrations interrupted data collection for ~5 hrs once a month.
