Read jd002844 1..13 text version

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. D16, 8546, doi:10.1029/2002JD002844, 2003

Performance of a single-monochromator diode array spectroradiometer for the determination of actinic flux and atmospheric photolysis frequencies

Gavin D. Edwards1 and Paul S. Monks

Department of Chemistry, University of Leicester, Leicester, UK Received 14 August 2002; revised 13 February 2003; accepted 19 March 2003; published 2 July 2003.

[1] The performance of a new spectroradiometer instrument based on a diode array

technique during its first field deployment at the International Modeling and Measurement Intercomparison (IPMMI) has been evaluated. The instrument is a single-monochromator diode array spectroradiometer that was used to measure both actinic flux and derived values of j(O1D) and j(NO2). Spectra may be collected over the wavelength region l = 280­450 nm at variable integration times (typically 1 s for normal operation), and data may be converted to actinic flux/photolysis frequencies using National Institute of Standards and Technology traceable calibration standards. The relative accuracy of these measurements is compared to photolysis frequencies derived using similar and contrasting experimental techniques. The instrument performed adequately in the comparison, but the problem of poor stray light rejection in the single monochromator has the potential to propagate significant errors in the derived photolysis frequencies. A full error analysis together with instrument characterization has been performed and is reported. The feasibility of using such instruments in the field for the measurement of atmospheric INDEX TERMS: 0317 Atmospheric Composition and photolysis frequencies is also discussed.

Structure: Chemical kinetic and photochemical properties; 0368 Atmospheric Composition and Structure: Troposphere--constituent transport and chemistry; 0394 Atmospheric Composition and Structure: Instruments and techniques; KEYWORDS: photolysis frequencies, actinic flux, spectroradiometer, diode array, atmospheric photolysis, ozone photolysis Citation: Edwards, G. D., and P. S. Monks, Performance of a single-monochromator diode array spectroradiometer for the determination of actinic flux and atmospheric photolysis frequencies, J. Geophys. Res., 108(D16), 8546, doi:10.1029/2002JD002844, 2003.

1. Introduction

[2] The photodissociation of trace gas species in the troposphere is the driving force for much of the radical chemistry that occurs in the Earth's atmosphere. For example, ozone photolysis in the presence of water vapor is the primary production channel for the formation of the hydroxyl radical, OH, which initiates much of the oxidative removal of trace gases in the troposphere. Photochemical ozone production in the troposphere is also facilitated by photochemistry, namely,

À Á NO2 þ hvðl < 420 nmÞ ! NO þ O 3 P ; À Á O 3 P þ O2 þ M ! O3 þ M; À Á O3 þ hvðl < 340 nmÞ ! O 1 D þ O2 ; À Á O 1 D þ H2 O ! OH þ OH:

1

ð1Þ ð2Þ ð3Þ

In order to model the total HOx and NOx budgets within the troposphere accurately, validation of atmospheric photolysis frequencies with field measurements is often necessary. It has been shown by several authors [e.g., Madronich, 1987; Shetter and Muller, 1999] that photochemical processes are ¨ directly proportional to actinic flux rather than to pure irradiance. The actinic flux is defined as the spherical flux density, incident over all angles, that is capable of producing photodissociation in photoactively labile molecules. Irradiance is defined as the density of flux energy on a plane surface. Actinic flux and irradiance differ by the cosine of the angle of incidence, q, and are only equal for a parallel beam of light perpendicular to the irradiated surface, i.e., where cosine q = 0. For a molecule, A, the photodissociation of A is characterized by the photolysis frequency, j, where

j¼ 1 d ½ A : ½ A dt ð5Þ

ð4Þ

Now at Atmospheric Chemistry Division, National Center for Atmospheric Research, Boulder, Colorado, USA. Copyright 2003 by the American Geophysical Union. 0148-0227/03/2002JD002844$09.00

The photolysis frequency, j, is dependent on local actinic flux, F, the absorption cross section, s, and quantum yield, f, for molecule A. These molecular parameters themselves are functions of both temperature and wavelength. Hence a photolysis frequency can be evaluated as the integral of the

IPM

5-1

IPM

5-2

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

product of solar actinic flux, absorption cross section, and quantum yield with respect to wavelength:

Zl2 ji ¼

l1

si ðl; T Þfi ðl; T ÞF ðlÞdl:

ð6Þ

[3] In equation (6) the solar actinic flux, F, is the net radiation available to photodissociate a molecule over a solid angle of 4p sr, T is temperature, and i denotes the ith species. In order to evaluate photolysis frequencies the actinic flux from all directions must be accurately quantified. In equation (6), l1 and l2 are the wavelength limits of photolysis. Below $l = 290 nm, the so-called surface atmospheric ``cutoff,'' no solar actinic flux is able to reach the ground owing to the total absorption of radiation of shorter wavelengths by stratospheric ozone. For groundbased instruments the limit l1 is typically set to 290 nm, with l2 being the upper bound of photolysis for species A. Typically, l2 = 420 nm for j(NO2), and l2 = 340 nm for j(O1D). [4] Measurements of actinic flux and subsequent photolysis frequencies have been recorded in the literature over the last two decades using a variety of techniques. For example, a commonly used absolute method is chemical actinometry. This technique has been used to measure j(NO2) and j(O1D) by several authors [e.g., Bahe et al., 1979; Dickerson et al., 1982; Blackburn et al., 1992; Shetter et al., 1996]. The technique relies on the direct exposure of O3 or NO2 to actinic flux by way of a 2p sr tube or dome arrangement. For a j(NO2) chemical actinometer, detection is usually conducted via a chemiluminescence reaction of NO following photochemistry inside the reaction vessel. For j(O1D), changes in the conductivity of methanol have been used to detect artefacts in the photodissociation of ozone as described by Shetter et al. [1996]. Chemical actinometers do not require the use of molecular parameters in the evaluation of photolysis frequencies but do need accurate calibrations and are limited to one molecular species per instrument. The relatively long time responses and shear bulk of such instruments often means that they are often not practical for deployment on all platforms where the measurement of photolysis frequencies is necessary. [5] The development of a fixed bandwidth photoelectric radiometer by Junkermann et al. [1989] has led to the deployment of many such instruments for the measurement of photolysis rate coefficients. The use of so-called fixed bandwidth radiometers as usable field instruments for both ground-based and aircraft campaigns has been described by several authors [e.g., Hofzumahaus et al., 1992; Brauers and Hofzumahaus, 1992; Muller et al., 1995; Volz-Thomas ¨ et al., 1996; Edwards, 2000]. These are compact devices based on the original concept by W. Junkermann and coworkers consisting of 2p sr quartz dome/diffusers to collect the incident radiation [Junkermann et al., 1989]. Photons are then passed through a band-pass-filter set to transmit photons of the desired wavelength between the limits of l1 and l2 (see equation (6)). A photomultiplier tube (PMT) or similar device then detects the incident photons. The variation of actinic flux with time is then recorded as a function of the measured PMT voltage, the

instrument time response being usually of the order of 1 s. Although filter radiometers are usually more practical and versatile than chemical actinometers, they still rely on absolute calibrations made against chemical actinometers or radiometrically against primary standards to derive photolysis frequencies. The main limitation of filter radiometers is that the band-pass filter of each radiometer limits it to a specific bandwidth of l1 to l2, which makes these devices species-specific. [6] The desire for accurate evaluation of actinic flux and photolysis frequencies by instruments with rapid time responses, together with the flexibility to measure several j values simultaneously, has led to the development of the spectroradiometer (SR). This instrument determines actinic flux by scanning the spectral region between limits l1 and l2 and uses the solution of the integral shown in equation (6) to estimate photolysis frequency via the product of actinic flux and the relevant molecular parameters. The advantage of these instruments is that several photolysis frequencies can be derived simultaneously on a timescale limited to the scanning speed of the instrument. Such double-monochromator instruments have been successfully used by Muller et al. [1995], McElroy et al. [1995], Kraus ¨ and Hofzumahaus [1998], Hofzumahaus et al. [1999], and Shetter and Muller [1999]. ¨ [7] This paper describes the deployment of a recently commercially available single-monochromator diode array spectroradiometer instrument. This differs from the instruments described above since it detects atmospheric actinic flux by way of single monochromator and a 512-pixel diode array. The instrument was run for the first time at the International Measurement and Modeling Intercomparison (IPMMI) held in Boulder, Colorado [Cantrell et al., 2003; Bais et al., 2003; Shetter et al., 2003; A. Hofzumahaus et al., Photolysis of O3 to O(1D): Measurements and modeling during the International Photolysis Frequency Measurement and Modeling Intercomparison (IPMMI) 1998, submitted to Journal of Geophysical Research, 2003, hereinafter referred to as Hofzumahaus et al., submitted manuscript, 2003]. During IPMMI the diode array spectroradiometer was compared to a variety of chemical actinometers, filter radiometers, and double-monochromator scanning spectroradiometers. Data from the diode array spectroradiometer were used both in the evaluation of actinic flux and in the estimation of photolysis frequencies. A full characterization of the instrument has been performed, and the benefits and drawbacks of the use of a singlemonochromator diode array as field spectroradiometers are considered.

2. Diode Array Spectroradiometer

[8] The diode array spectroradiometer deployed by the University of Leicester (ULI) during IPMMI is a fast, monolithic single-monochromator spectral radiometer with a temperature-stabilized diode array detector. The device is a commercially available instrument (Meterologie Consult (METCON) Glashutten, Germany). (A full and extensive ¨ description of the instrument including technical information and a schematic diagram is found on the METCON Web site (http://www.metcon.net/).) This information is also summarized in Table 1.

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

IPM

5-3

Table 1. Characteristics of the Diode Array Spectroradiometer Used at IPMMIa

Type of Instrument Instrument dimensions Weight Power Detection optics Quantity measured Entrance optics Grating flat field monochromator type Wavelength range Wavelength accuracy Diode wavelength separation Resolution Detector Calibration Actinic flux detection limit Minimal detected j(O1D) Signal-to-noise ratio Temperature stability Stray light rejection ratio Dark current response Scan time

a

METCON Fast Monolithic Single-Monochromator Spectroradiometer 250 Â 150 Â 110 mm3 4.1 kg 115/230 AC, 12 ­ 36 DC quartz actinic diffuser actinic flux round-to-slit converter (70 Â 2500 mm) 248 lines mmÀ1 diffraction grating coupled to diode array in self-contained unit 285 ­ 450 nm >0.1 nm 0.83 nm $2 nm at FWHM 512-pixel diode array (Carl Zeiss) NIST traceable 1000-W tungsten halogen lamp $1 Â 1010 photons cmÀ2 nmÀ1 (±5%) $3 Â 10À7 sÀ1 (±13%) $4:1 $5 Â 10À4 nm KÀ1 $5 Â 104 $200 counts sÀ1 (at 303 K) user defined; 1 s scanning over l = 285 ­ 450 nm region during IPMMI.

IPMMI, International Photolysis Frequency Measurement and Modeling Intercomparison; METCON, Meteorologie Consult; FWHM, full width at half maximum; NIST, National Institute of Standards and Technology.

2.1. Optical Collection System [9] The ULI instrument collects incident photons by using a 2p sr quartz dome similar to dome/diffuser arrangements described by Volz-Thomas et al. [1996]. The basic principle of operation involves the use of a solid diffuser made up of several separate domes of sand-blasted quartz. Photons incident on the outer dome pass through the subsequent inner domes. As they pass through, photons undergo multiple refractions and reflections at the inner and outer surfaces. Thus photons entering the optics are detected across the entire 2p sr range. The outer dome has a height of $3.5 cm and width of $3 cm. The viewing angle of the detector is limited to 2p sr by the use of a matt black shadow ring, which was 1 cm above the instrument housing and extends to a diameter of 20 cm. The external portion of the diode array spectrometer is sealed off from the internal components by use of a rubber ``O'' ring. The instrument is regularly pressure tested up to 0.5 bar to insure good isolation of the diode array optics from the external atmosphere and is temperature stabilized at a user-defined level, typically 303 K for normal operation. [10] For an instrument of ideal sensitivity, all incident photons would have the same probability of detection by the diode array over the 2p sr solid angle. In addition, the instrument would have perfect rejection characteristics and detect no photons at polar angles >90°. The dome used for the detection of photons by the instrument was optimized by the manufacturer to be as close to these ideal criteria as possible. However, perfect isotropic detection and perfect angular rejection are not currently achievable in a working instrument. Estimates of the potential anisotropy in the 2p sr optics, together with angular rejection of the instrument, were obtained. The spectroradiometer was mounted on a platform that was capable of rotation through 180° in one plane in a ``light-box.'' This is an aluminium construction consisting of a rectangular box of 2 m length, 1 m width, and 1 m height. The box is separated into two sections of equal size, and the sections are separated from each other by a 2 mm thick aluminium divider of equal height and width

dimensions to the surrounding box. In one chamber sits a mounting point for the spectrometer. The second half of the box contains a lamp mounting connected to a suitable power supply. National Institute of Standards and Technology (NIST) traceable quartz tungsten halogen (QTH) lamps (200 W and 1000 W, Oriel, Ltd.) were used as the light source for this analysis. Light from these QTH lamps was allowed to enter the chamber containing the spectroradiometer through an iris that limited the illumination of the diode array to the 2p sr dome only. Photon counts were recorded in the usual way (see section 2.3). Light scattering from surfaces was minimized by painting all internal surfaces matt black. External light was also minimized by performing all experiments in a blacked out laboratory. [11] Rotation of the quartz dome through 180° in a simple plane, while monitoring the change in signal with the angle

Figure 1. Typical measured angular response for 2p sr quartz input optics. The legend for Figure 1 details two separate runs in different planes.

IPM

5-4

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

of incidence, provides a quantification of the angular response error. During these experiments the radiometer was rotated so that the angle of incidence changed from 0° at the starting position through to a maximum of 120°, where no data other than dark current were recorded. The experiment was then repeated from a 120° angle of incidence back to the starting position in a different plane (zero angle of incidence). A typical plot of incident angle versus relative response is shown in Figure 1. Both data sets showed similar behavior, specifically a deviation from 1:1 in the angular response that was on the order of ±2% or less for angles 80°. Also, between 80° and 90° the signal was observed to change by $6% before approaching 0 at angles >110°. These experiments were repeated for all wavelengths and at several incident geometries. All experiments showed similar profiles, indicating negligible wavelength and geometric dependence in the angular response of the instrument. 2.2. Spectroradiometer Wavelength Calibration Drift and Resolving Power [12] Atmospheric photons are collected by the 2p sr quartz dome as described above. Photons are then passed through a round-to-slit converter (70 Â 2500 mm) giving a spectral band pass of the order of $2 nm at full width at half maximum (FWHM). Photons entering the entrance slit are dispersed using a flat field diffraction grating (248 lines mmÀ1) and detected using a 512-pixel diode array (Carl Zeiss), capable of detecting photons from the range l = 285 ­ 710 nm. The Zeiss-METCON monochromator and detection array is a complete device with the grating and a detector encompassing one distinct unit. Pixel separation of the diode array is $0.83 nm. The manufacturer-quoted temperature-dependent wavelength drift for this arrangement is 5 Â 10À4 nm KÀ1. [13] The well-defined output of atomic emission lines can be used to estimate wavelength calibration drift and resolving power of the instrument. The manufacturer originally checked the wavelength calibration before the IPMMI experiment and reported no significant drifts from literature values. Similar checks were made against the output of a Hg Pen-Ray2 lamp, He-Ne laser, and sodium discharge lamp at the University of Leicester following the conclusion of the IPMMI experiment. During these experiments the instrument was exposed to the output radiation from these sources and data recorded in the usual way (see section 2.3). Data relating the maximum intensity of radiation and wavelength were then transposed to a secondary computer program. Fitting parameters were employed to analyze the peak attributes. No significant wavelength drift from the literature line was observed for any of the lamps under test. For example, with respect to the Na ``D'' emission lines, curve analysis showed that the spectroradiometer recorded a broad peak which was determined to be $2 nm at FWHM. This value is in agreement with the manufacturer-quoted resolution of 2.1 nm. The peak fitting routines also indicated the instrument recorded the maximum intensity of wavelength calibration with no greater resolvable accuracy relative to the literature values for these emission lines. These resolutions and wavelength drifts are not likely to present large errors over the typical photolysis wave band for j(O1D) or j(NO2) considered here. Madronich and Weller [1990] investigated numerical integration errors associated with

actinic flux model data, using grids of 1 to 10 nm resolution, and the reported errors in the model j values were no more than 5% for a resolution of <2 nm. This grid resolution is comparable to the spectral resolution of the spectroradiometer instrument described here, which suggests the errors in the measured actinic flux of this spectroradiometer should also be <5% over the entire wavelength range for the photolysis frequencies at this given resolution. [14] The potential error in the calculated actinic flux and photolysis frequencies owing to any drift in the wavelength calibration is large. The instrument wavelength drift was therefore measured by comparison both with emission sources and with the partly resolved Fraunhofer structure recorded in the solar spectrum as described by Slaper et al. [1995]. In the case of the partially resolved solar structure the instrument appears to be consistently within $1 nm of the reported literature actinic flux spectrum [Slaper et al., 1995]. However, any drifts in the wavelength calibration during IPMMI present difficulties in the analysis of data. Theoretical wavelength calibration drifts of 1 nm were introduced into the data workup in order to ascertain the error in calculated actinic flux associated with these wavelength drifts. If the magnitude of these theoretical errors were large, they could provide information on whether similar wavelength drifts occurred in the course of the IPMMI experiment. The resultant theoretical actinic flux (not shown) was found to vary only slightly from the normal. Ratios of normal versus wavelength-shifted data varied from 0.96 to 1.02 over the wavelength range l = 290­ 450 nm. These theoretical wavelength shifts do, however, result in a large propagation in the error observed in j(O1D). For a wavelength calibration drift of 1 nm the total error in j(O1D) is 15%. For a 2-nm theoretical wavelength calibration drift the error in j(O1D) is 40%. Since the photolysis frequencies calculated using spectroradiometer data show agreement within the calculated errors, it was concluded that the wavelength calibration for the instrument was <2 nm during the course of IPMMI. 2.3. Photon Detection and Instrument Flux Calibration [15] Spectra can be collected over the wavelength region l = 280­ 420 nm at user-defined integration times using this instrument. Typically, the integration times are set to 1 s for normal operation. All data described in this work refer to integration times of 1 s unless otherwise noted. During the normal operation of the instrument, data are recorded as an ASCII file containing a series of pixel ``counts'' corresponding to the individual diode response at each diode wavelength. These counts represent the photoelectric charge that is accumulated on the diode array while the array is exposed to atmospheric flux and are directly proportional to the amount of incident radiation. During data workup the data recorded in counts were converted to actinic flux and, subsequently, photolysis frequencies, using calibrations derived using NIST traceable (irradiance) standards. [16] Two separate calibrations of the diode array have been performed over the course of this work. The first was performed at NCAR (Boulder, Colorado) using a custommade light box. The instrument was placed at a fixed distance of 50 cm from a 1000 W NIST-calibrated QTH lamp (Oriel, Ltd.). For these standards an inverse square law

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

IPM

5-5

Table 2. Estimated Uncertainties in Diode Array Spectroradiometer Measurements at IPMMI

Source of Error Reproducibility of lamp spectra Inverse square law approximation Equivalence plane receiver assumptions Accuracy of lamp calibration Dark current stability Wavelength calibration Angular response error Slit function Stray light rejection error Total uncertainty in actinic flux Molecular parameters uncertaintya Uncertainty in j value determination

a

Uncertainty in Actinic Flux data, at 1s $2.5% over 290- to 450-nm range 1% 1% [from Hofzumahaus et al., 1999] <3% (from lamp manufacturer) 3% for times <0600 LT, 1% all other times <4% in UVB,<2% in UVA 2% <5% over whole spectral range [from Madronich and Weller, 1990] Variable for j(O1D), almost negligible for j(NO2) <9% in UVB, <8% in UVA $11% for j(NO2) and $10% for j(O1D) 14% for j(NO2) and $13% for j(O1D), plus additional variable stray light error (see text)

See Cantrell et al. [2003] or Kraus and Hofzumahaus [1998] for details.

relationship exists between the intensity of emission and distance from the source at distances greater than around 30 cm. Here the intensity of the lamp flux and detected spectroradiometer signal decrease with the square of the distance between the lamp and the spectroradiometer head. This enables the routine calibration of the spectroradiometer at distances of 50 cm and reduces the uncertainty in the radiation field at the position where the instrument is illuminated. Using the inverse square approximation, spectra of the QTH lamps were recorded as a function of the lamp output (in raw diode array counts) and converted into photon flux (photons cmÀ2 nmÀ1 sÀ1). The instrument was also recalibrated in the light box at the University of Leicester against a second primary QTH standard (200 W, NIST traceable lamp, Oriel, Ltd.) following the conclusion of the IPMMI experiment. [17] In this form of calibration the precision can be equated to the variation in the detected counts between each individual spectrum. The shot-to-shot standard deviation is combined with the errors associated with the inverse square law approximation and becomes the effective reproducibility of the measured lamp spectrum. The lamp itself was assumed to be unchanging in brightness, since power was supplied by an ultralow voltage ripple power supply (Oriel, Ltd.). The ``shot-to-shot'' variation for the diode array has a standard deviation of $13 counts in 500 (2.6% at 2s) over the signal. Using an uncertainty estimate for calibration precision of $2.5% in the actinic flux data workup gives a total error in the overall actinic flux determination of $2.8%. Using these assumed errors in the actinic flux determination, the error was used to test the sensitivity of the derived photolysis frequency to these errors. The error due to this changing precision was propagated through to the determination of photolysis frequencies, resulting in an error of 4% in j(O1D), while for j(NO2) the error was 2.5% when no additional errors were considered. [18] These uncertainties are coupled with other possible sources of error in the determination of photon flux, the magnitudes of which are shown in Table 2. These include errors arising from the inverse square approximation that was used in the calibration procedure ($1%) and the socalled ``equivalence plane receiver'' (EPR) error. The EPR error arises owing to the assumption that the hemispherical inlet optics behaves as a plane receiver that is irradiated at normal incidence. The consequences of this assumption are described in detail by Hofzumahaus et al. [1999]. Using the

assumptions described by Hofzumahaus et al., error propagation owing to the equivalence plane receiver assumption is $1% for this spectroradiometer. 2.4. Effect of Uncertainties in Molecular Parameters [19] Errors and uncertainties in the published molecular data are a major source of uncertainty in spectroradiometerderived photolysis frequencies [Kraus and Hofzumahaus, 1998; Hofzumahaus et al., 1999]. The effect of these errors with respect to IPMMI has recently been considered in detail by Hofzumahaus et al. (submitted manuscript, 2003) for j(O1D) and by Shetter et al [2003] with respect to j(NO2). The uncertainty in the ozone absorption cross section is usually quoted as 3% [Molina and Molina, 1986], while errors/uncertainties in the quantum yield for the formation of O(1D) can be up to 10% [Shetter et al., 1996; Hofzumahaus et al., 1999]. Using a maximum assumed error of 10% in the workup of quantum yield data [DeMore et al., 1997], together with a 3% error in the absorption cross-section data, the root mean square error propagated in j(O1D) is found to be 10%. For j(NO2) the molecular parameter errors s(NO2) from Harder et al. [1997] and f(NO2) from DeMore et al. [1997] were 4% for absorption cross section and 10% for quantum yield. Using these molecular data, the root mean square error in photolysis for j(NO2) is 11% where no additional instrumentation errors were considered. 2.5. Spectral Sensitivity and Detection Limit [20] The detection limit and overall sensitivity of the single-monochromator diode was investigated as part of the instrument characterization. The sensitivity of a spectroradiometer is defined as the ratio of the output signal in counts per second versus the input actinic flux, F, in photons per squared centimeter per second over a given bandwidth [Shetter and Muller, 1999]. For the diode array ¨ instrument described here the change in this sensitivity with respect to wavelength are shown in Figure 2. In Figure 2, sensitivity data show that for the first 60 nm the sensitivity is uniform at around 7.5 Â 10À11 photons detected per count detected. The sensitivity increases to 9 Â 10À11 photons per count between l = 345 and 380 nm before increasing rapidly at wavelengths >390 nm, owing to the rapid increase in ambient fluxes at these wavelengths. [21] The detection limit of the instrument is defined as the minimum measurable photon flux per unit count. Using data taken in the field, the signal-to-noise ratio for this spectror-

IPM

5-6

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

current precision error dropped to <1%. Working this uncertainty through to photolysis frequencies, the net j(NO2) is essentially unaffected by dark current uncertainties, while j(O1D) shows an influence owing to dark current uncertainty only for times that are between 0600 and 0800 LT. At these times the uncertainty propagation is 5% at 1s. 3.1. Stray Light Rejection, Estimation, and Incorporation [24] The importance of stray light (SL) rejection in monochromator systems and therefore photolysis frequencies has been described by several authors [e.g., Hofzumahaus et al., 1999]. The spectroradiometer used in this work has photodiodes sensitive to radiation with wavelength less than the atmospheric cutoff ($l = 290 nm). During a typical experiment, wavelength ranges between l = 285 and 450 nm were scanned. In order to make the necessary corrections the assumption that counts between l = 285 and 290 nm were entirely due to stray light and dark current was applied, with the spectral variation of SL across the whole spectrum assumed to be minimal (i.e., white). [25] From application of this spectrally white stray light (WSL) correction it is possible to correct each spectrum by subtraction of the average counts between l = 285 and 290 nm. The need for a method of accurate WSL subtraction is highlighted upon examination of Figure 3. In Figure 3, actinic fluxes measured by the diode array spectroradiometer with no stray light correction are compared to flux measurements made from a double-monochromator spectroradiometer from Forschungszentrum Julich (FZJ) on 19 August ¨ 1998 during IPMMI [Kraus and Hofzumahaus, 1998; Hofzumahaus et al., submitted manuscript, 2003]. The flux comparison shows that the actinic flux data measured by the single monochromator are significantly greater than data measured by the double-monochromator system in the region less than $330 nm. From comparison of measured flux ratios of one instrument to another it is clear that the University of Leicester diode array data show significant deviation from

Figure 2. Change in diode array spectroradiometer spectral sensitivity and detection limit with wavelength. adiometer was found to be close to 4:1 (for a 10-minaveraged actinic flux spectra taken at IPMMI). For this signal-to-noise ratio the relative change in this detection limit with respect to wavelength for the diode array instrument was calculated and is shown in Figure 2. These data show similar trends to the sensitivity plot, with a relatively poor detection limit of 1 Â 1010 photons cmÀ2 sÀ1 from 285 to 340 nm, a small change localized at 1.15 Â 1010 photons cmÀ2 sÀ1 and centered at 350 nm, before a rapid increase in the detection limit at longer wavelengths. [22] Using the experimentally determined signal-to-noise ratio, it is clear that the diode array has a sensitivity that is considerably lower than the equivalent double-monochromator spectroradiometers which have higher sensitivities and detection limits closer to 1 Â 108 photons cmÀ2 nmÀ1 sÀ1 at wavelengths $290 nm [Hofzumahaus et al., 1999]. These results suggest that even when other factors (e.g., stray light rejection; see later discussion) are considered, for wavelengths close to the atmospheric cutoff ($290 nm), a single-monochromator diode array spectroradiometer of this type is likely to overestimate the actinic flux when compared to an equivalent double-monochromator spectroradiometer. The ramifications of these factors are discussed later.

3. Importance of Stray Light and Dark Current

[23] As with most photoelectric devices, there is a systematic ``dark current'' offset in the diode array spectroradiometer. Using data collected overnight when there were no atmospheric light sources, signal levels are generally around 200 (±2% at 1s) diode counts at a fixed diode array temperature of 303 K. This dark current precision is significantly smaller than the error owing to stray light, and therefore it might be assumed that dark current is not as significant a factor in the error propagation as stray light. This hypothesis was tested by constantly subtracting a spectrally white 200-count dark current offset from each spectrum without any stray light corrections. A dark current uncertainty of 2% was found to produce a maximum error in actinic flux of 3% up to 0600 LT. This was the only time of day when the error was significant owing to the intrinsically low light levels, such that any variation in dark current has a relatively large effect. In comparison, at 0700 LT the dark

Figure 3. Actinic flux ratios for Forschungszentrum Julich ¨ (FZJ)/University of Leicester (ULI) with no stray light correction, International Modeling and Measurement Intercomparison (IPMMI) data, 19 June 1998.

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

IPM

5-7

0700-0710 data 1200-1210 data 0700 1200

for this variation in the ratio of ULI to FZJ flux at these wavelengths. 3.2. Comparisons to Model Data [27] The IPMMI experiment was an international comparison of both measured and modeled photolysis frequencies. Data obtained from the models were similarly compared to the diode array spectroradiometer data to further quantify errors/uncertainties associated with the instrument. Around 20 groups participated in the modeling phase of the IPMMI experiment [see Bais et al., 2003]. Essentially, the model groups were asked to provide data on actinic flux at solar noon in addition to photolysis frequency j(NO2) and j(O1D) via individual radiative transfer codes. Full details of the modeling part of the intercomparison are given elsewhere [Bais et al., 2003; Cantrell et al., 2003]. [28] Three models were used for the initial comparison in this work, denoted by ACD, BAS, and the Royal Netherlands Meteorological Institute (KNM). The Atmospheric Chemistry Division at NCAR (ACD) submitted the ACD data. The model used by this group was the pseudospherical eightstream discrete ordinate Tropospheric Ultraviolet Model (TUV) model as described by S. Madronich et al. (available at http://acd.ucar.edu/models/UV/TUV/index. html, 1998). The British Antarctic Survey (BAS) submitted the BAS model data to IPMMI. The model used was developed by BAS and based on the discrete ordinates radiative transfer (DISORT) code of Stamnes et al. [1988] and the pseudospherical adaptations of Dahlback and Stamnes [1991]. The KNM model was a radiative transfer model developed by van de Hulst [1980]. The current version used as part of the IPMMI experiment is described by DeHann et al. [1987]. All of the above models show similar trends when compared to the diode array data, as was also found in data comparisons with FZJ and ULI (see Figure 5). Specifically, a large overestimation of the actinic flux in the 290 ­ 310 nm wave band is observed with much better agreement in the >310­ 420 nm region. Plots of the ratios of the modeled and measured fluxes are shown in Figure 6.

Figure 4. Percentage contribution of stray light with respect to wavelength in spectroradiometer (SR) diode array data, IPMMI data, 19 June 1998.

the data of the double-monochromator instrument at times that are very early in the morning and/or near the atmospheric cutoff between l = 290 and 310 nm. Figure 4 shows a comparison of the percentage contribution of stray light to the detected diode array counts at 0700 and 1200 LT for diode array data taken on 16 June 1998. It is clear that although the detected counts at 0700 LT are much smaller than those at 1200 LT, the percentage of these counts that may be attributed to stray light is considerably greater than at noon. This analysis confirms the importance of a full and accurate correction for stray light when dealing with actinic flux data collected by single-monochromator diode arrays. [26] Actinic flux data worked up from the raw counts with the in situ corrections described above were compared to the data obtained from the FZJ double-monochromator spectroradiometer. Figure 5 shows the effect of applying the in situ SL correction. It is clear from Figure 5 that the SLcorrected data show improvement in this SL-affected region (<300 nm), but there are still considerable differences between these corrected fluxes and the data from the FZJ spectroradiometer. In addition, there appear to be variations in the relative agreement between these instruments at longer wavelengths. At $340 nm through to 370 nm the agreement is around 0.8 for the comparison of FZJ and ULI. This change is not related to stray light but is a reflectance of the relative sensitivity of the two instruments at these wavelengths. As shown in Figure 2, the single-monochromator instrument has variable sensitivity which increases from $7.5 Â 10À11 photons detected per diode array count at 345 nm to $9 Â 1011 photons detected per diode array count at 355 nm, before decreasing to $9 Â 1011 photons detected per diode array count at 375 nm. These effects are also reflected in similar changes in the detection limit with respect to wavelength (also shown in Figure 2), specifically a drop in the instrument detection limit in this wavelength region. Similar data with respect to the wavelength sensitivity for the FZJ instrument show almost a uniform sensitivity and detection at these wavelengths. It is therefore this changing sensitivity with respect to wavelength, rather than additional stray light influence, that is the likely reason

Figure 5. Ratios of spectrally derived actinic flux from the FZJ and ULI instruments with white stray light (WSL) correction (see text) applied to the ULI diode array instrument for 19 June 1998.

IPM

5-8

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

Figure 6. Modeled and measured (ULI with WSL correction; see text) actinic flux ratios for 19 June 1998. (a) ACS model data. (b) BAS model data. (c) KNM model data. [29] Similarly, there appears to be a change in the ratio of the measured to modeled ratio between all these models in the wavelength region 340 ­ 370 nm as was observed in the comparison of measured data from FZJ and the ULI spectroradiometer comparison described in section 3.1. Again, this is likely due to the poor/variable sensitivity at these wavelengths for the diode array instrument (compared to the model data). The problems identified in both the model and measured comparison of spectra is an area of potential concern, as these fluxes were used to calculate the photolysis frequencies j(NO2) and j(O1D) from the spectroradiometer. The standard deviation in stray light from one spectrum to another has also been estimated using the spectral region l = 285­ 290 nm data as a measure of SL, additive with a dark current, at solar noon. The noon data were used, as it was assumed there should be minimal changes in absolute light intensity at this time, and hence one additional variable of a rapidly changing incident flux could be removed. Using this assumption, the SL standard deviation at 1s was of the order of 1­ 1.5%. This held true for several different days and overhead cloud conditions during IPMMI. 3.3. Implications of Stray Light Rejection for Measured Photolysis Frequencies [30] Considering the stray light effect on j(NO2), it might be assumed that the influence of stray light and subsequent

overestimation of actinic flux in the 290­ 310 nm region would be relatively small owing to the j(NO2) spectral ``window,'' i.e., l = 290­ 420 nm. Overestimation of flux in the 290 ­ 310 nm portion of the ``window,'' owing to poor stray light correction, should be a small error when compared to the overall actinic flux in the NO2 photolysis range. This assumption was checked by comparison of the j(NO2) data obtained by the single-monochromator diode array with the photolysis frequencies obtained from the ULI filter radiometer. These filter radiometer data were shown to have good agreement relative to other IPMMI participants [Shetter et al., 2003; Cantrell et al., 2003]. Residual plots of change between the predicted output and the true output of the ULI spectroradiometer compared to the ULI filter radiometer for j(NO2) are shown in Figure 7a. [31] In Figure 7a the linearity of the scatterplots is analyzed. Figure 7a shows the change in the percentage ratio of predicted j(NO2)/true j(NO2) as a function of time. From Figure 7a it is clear that the largest deviation between a perfect correlation between the filter radiometer and spectroradiometer j(NO2) data occurs in the early morning and late evening where SL rejection is poor for the spectroradiometer. However, all data lie within a ±10% error band. This error of ±10% is on the order quoted for a perfect measurement of j(NO2) (that is, where the instrument errors are due to calibration and other errors quoted in Table 2 along with errors in the molecular parameters used in j(NO2) calculation). This seems to suggest that the overestimation of the actinic flux at the lower wavelengths owing to poor SL rejection does not affect the performance of the spectroradiometer j(NO2) data to any significant degree during the IPMMI study. Hence, as summarized in Table 2, the total error analysis suggests that during the IPMMI campaign the diode array spectroradiometer measured j(NO2) with an accuracy of !18%. [32] It would be expected that the calculation of j(O1D) would be most sensitive to the levels of stray light. Since j(O1D) is evaluated across such a narrow wavelength range, l = 290­ 325 nm, any overestimation of the actinic flux in the 290­ 310 nm window is likely to have a major effect on the overall j(O1D) data obtained from the integral of these fluxes with respect to wavelength. However, a high degree of correlation was repeated for all days of the intensive, with R2 values (where R2 denotes correlation coefficient for data calculated from a regression analysis) ranging from 0.961 to 0.993 over the course of the campaign when j(O1D) values from the ULI spectroradiometer were compared to other measured j(O1D) values using a variety of techniques (J. Calvert, private communication, 1998). [33] Residual plots of the change between the predicted output and the true output of the ULI spectroradiometer compared to the ULI filter radiometer for both j(O1D) were also used to test the effect of SL on the performance of the spectroradiometer with respect to j(O1D). These data are shown in Figure 7b. Here the predicted output from the residual analysis was divided by the spectroradiometer j(O1D) data in a similar way to Figure 7a. Figure 7b shows that the largest deviation from linearity of the scatterplot between the filter radiometer and spectroradiometer j(O1D) data again occurs in the early morning and late evening where SL rejection is poor for the spectroradiometer.

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

IPM

5-9

Figure 7. (a) Percentage deviation from ideal agreement for spectroradiometers and filter radiometer j(NO2); residual plot, 19 June 1998. (b) Percentage deviation from ideal agreement for spectroradiometers and filter radiometer j(O1D); residual plot, 19 June 1998. [34] By plotting the diurnal cycles, residual deviations between the diode array spectroradiometer and other instruments may be observed. Figure 8 shows the j(O1D) diurnal cycle data as measured by the diode array spectroradiometer and shows j(O1D) data from the NCAR chemical actinometer instrument [Shetter et al., 1996; Cantrell et al., 2003]. Overall, Figure 8 shows that there is good agreement between the two instruments, especially at solar noon. However, there were some marked deviations, clearly observed at times, where the intensity of light was low, i.e., early morning and late afternoon. Since the stray light influence is largest at such times (see Figure 3), it follows that the overestimation of j(O1D) at these times is probably the result of the influence of poor stray light rejection in the single-monochromator diode array combined with a poor detection limit. [35] A different way of visualizing any SL contribution is to calculate the contribution to overall photolysis frequency by wavelength. Photon flux was therefore separated into several wavelength bands in order to identify the most significant parts of the spectrum. There may be the potential for the overestimation of flux at the lower end of the photolysis wave band for j(O1D) to be counterbalanced with an underestimation of the flux at longer wavelengths. From a breakdown of the individual contributions to j(O1D), it was found that wavelength regions not affected by stray light or sensitivity problems had good agreement to the data from other instruments. The uncorrected contributions to j(O1D) for each wavelength for the diode array across the entire j(O1D) action spectrum are shown in Figure 9, which shows the wave band contribution to j(O1D) for the ULI diode array spectroradiometer. The strongest influence in j(O1D) shifts from between 290 and 299 nm early in the day to between 300 and 309 nm around solar noon. The detection limit for j(O1D) for the diode array was $3.26 Â 10À7 sÀ1, assuming a theoretical signalto-noise ratio of 1:1. This sensitivity increases to 1.18 Â 10À7 sÀ1 if a more realistic signal-to-noise ratio of 4:1 was

Figure 8. Comparison of National Center for Atmospheric Research (NCAR) actinometer and WSL-corrected ULI diode array SR j(O1D) data measured during IPMMI, 15 June 1998.

IPM

5 - 10

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

Figure 9. Contribution to j(O1D) by wavelength for uncorrected ULI SR data during IPMMI on 16 June 1998.

Figure 11. In situ spectrally corrected stray light spectra for the single-monochromator diode array spectroradiometer instrument (with dark current subtraction).

used. These results contrast to the FZJ and NCAR data sets where the contribution of the l = 300 ­ 309 nm wave band dominated at all times of the day, with the l = 290­ 299 nm range contributing little, even in early morning. A plot similar to those data shown in Figure 9, using data from the double-monochromator spectroradiometer of FZJ, is shown in Figure 10. [36] The problem of poor instrument stray light rejection in the single-monochromator diode array spectroradiometer deployed by the University of Leicester at IPMMI has implications for the accuracy of the derived photolysis frequencies. The spectrally white stray light correction procedures when applied to the data produced actinic fluxes and photolysis frequencies in better agreement with those from the double-monochromator-based instruments. However, from the above discussion it is also evident that there

are still further correction procedures required before the total stray light errors are fully accounted for.

4. Improved Stray Light Corrections

[37] In order to characterize the instrument further, an in situ stray light spectrum with respect to wavelength was measured following the conclusion of the IPMMI experiment. (Figure 11). The method of SL rejection calculation was similar to that employed by Shetter and Muller [1999], ¨ where the array was mounted so that the atmosphere served as a light source. The instrument was set to scan the range l = 285­ 450 nm, and the counts recorded were integrated over the wavelength range. The 2p sr optics of the instrument were then covered by an interference filter (Filter

Figure 10. Contribution to j(O1D) by wavelength for FZJ double-monochromator data during IPMMI on 16 June 1998.

Figure 12. Ratios of spectrally derived actinic flux from the FZJ and ULI with spectrally correct stray light (SCSL) correction (see text) applied to the ULI diode array instrument for 19 June 1998.

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

IPM

5 - 11

Figure 13. Correlation between SCSL-corrected ULI SR and FZJ SR j(O1D) data measured during IPMMI. 59492, Oriel, Ltd.) that limited the transmission of photons to wavelengths of l > 500 nm. The manufacturer-quoted transmission of photons of this cutoff filter in the scanning range l = 285 ­450 nm was <1%. [38] The integral of the signals recorded over the spectral range 285 ­ 700 nm was divided by the difference of the counts recorded at 285 nm with and without the optical filter in place. Only stray light photons were recorded at 285 nm. From these experiments the SCSL rejection of the single monochromator was found to be 5 Â 104. This value is in good agreement with simulations for both single and double monochromators made by Hofzumahaus et al. [1999]. By recording the spectra where the filter was in place, effectively, an in situ stray light spectrum is recorded. That is, when the optical filter was in place, the recorded diode array data at wavelengths <500 nm represent the dark current and stray light counts at each diode array wavelength during this time. [39] These data represent a spectrally correct stray light (SCSL) spectrum and were used in order to improve the SL rejection for the instrument. The SCSL diode array counts could be subtracted from the ``normal'' flux counts when the array was collecting spectra without the filter in place and the wavelength dependence in the stray light then accounted for in this subtraction. It was assumed that the percentage reduction in the data due to the filter in this experiment was the same as if the experiment were repeated under IPMMI conditions. Hence the percentage contribution of stray light and dark current at each individual wavelength was evaluated and the necessary SCSL subtraction applied to the IPMMI data. [40] The application of this method to data recorded during IPMMI was found to improve the overall quality of diode array actinic flux and j(O1D) data when compared to that of other participants. Figure 12 shows the correlation between the ULI-resubmitted SCSL actinic data and data from the spectroradiometer of FZJ as a ratio plot similar to those data shown in Figures 5 and 6. The agreement at wavelengths longer than $305 nm appears to be closer to the 1:1 line than to the correction that assumed that stray light was spectrally white. Below $305 nm the agreement is still poor. It is likely that the main cause of the overestimation of the flux at these lower wavelengths by the diode array is due to its poor sensitivity and detection limit compared to the FZJ instrument. Although the influence of SL may still play a role at these wavelengths, the fact that the detection limit of the ULI spectroradiometer was $1 Â 1010 photons nmÀ1 sÀ1 compared with 1 Â 108 photons nmÀ1 sÀ1 for the FZJ instrument. This poor detection limit results in the large overestimation of SL corrected fluxes at low intrinsic light levels and large deviations from 1:1 agreement when the instruments are compared. [41] Despite these factors, the overall values of R2 in Figure 13 are in the order of 0.98 ­ 0.99, and the slope of the regression lines is unity within 5 and 15% for these 4 days. These values are within the quoted errors for the 4 days of the IPMMI campaign when using the SCSL data to subtract stray light signal from the raw spectra. Figure 13 also shows a comparison of the j(O1D) data measured by the doublemonochromator spectroradiometer of FZJ and the revised ULI spectroradiometer j(O1D) data. The plots suggest that

IPM

5 - 12

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

Figure 14. Percentage deviations of the data points from the regression line for ULI SCSL and FZJ SR j(O1D) data with solar zenith angle during IPMMI. for the 4 days of the IPMMI intensive there was good agreement between the SCSL j(O1D) and the photolysis frequencies measured by the double-monochromator instrument. Although there is a high degree of correlation (R2 > 0.9 in all cases), there is some residual scatter and deviations from the regression line observed during all 4 days. A plot of the residual deviation from these regression lines is given in Figure 13. In Figure 14, linearity of the scatterplots was analyzed by plotting the residual deviation against the given solar zenith angle. Figure 14 suggests that for the majority of IPMMI data the percentage deviation from ideal agreement between these two instruments is typically within ±20% for j(O1D) for the data where solar zenith angles are smaller than 70° and for measurements were made without the influence of clouds. The rapidly changing actinic flux associated with cloud modulation of direct and diffuse solar radiation results in the scatter from the regression line being more acute for such overhead conditions. From Figure 14, clouds can result in deviations of up to 40% for such cases. The full effect on actinic flux and photolysis frequencies of cloud modulations and other overhead conditions is described by Crawford et al. [2003]. have been investigated. Following an extensive evaluation of this instrument, used for the first time at the IPMMI experiment, it has been shown that single-monochromator instruments have both benefits and potential problems as working field spectroradiometers. Using the data taken during the intercomparison, it is clear that the instrument performed adequately with respect to the measurement of atmospheric photolysis frequencies and actinic flux. The spectroradiometer agreed reasonably well when compared to chemical actinometers, filter radiometers, and doublemonochromator spectroradiometers within quoted experimental error limits. The need for careful stray light and instrument characterization has been outlined. With the increasing drive for measurement devices capable of rapid time responses in atmospheric chemistry the use of singlemonochromator diode array spectroradiometers is likely to become much more widespread. For instance, for any long-term SR measurements it is desirable not to have any moving parts. The time resolution that diode array SR offers is not found in other scanning double-monochromator or chemical actinometers, thus making such instruments an obvious choice for the measurement of actinic flux and/or photolysis frequencies in the field. Therefore, in order for the data to be useful and accurate with respect to measurement of flux and photolysis frequencies, a full and extensive error correction regime must be employed to overcome the potential shortcomings of such instruments.

5. Conclusions

[42] In this paper, data collected by a diode array spectroradiometer during the International Photolysis Measurement and Modeling Intercomparison (IPMMI)

EDWARDS AND MONKS: PERFORMANCE OF A DIODE ARRAY SPECTRORADIOMETER

IPM

5 - 13

[43] Acknowledgments. The authors would like to thank all the participants of the IPMMI experiment for the use of data and for their input of useful comments and suggestions. The assistance of Gary K. Corlett with respect to IDL programming and of Gavin Salisbury and Andrew Rickard with respect to the additional suggestions is also gratefully appreciated. G.D.E. would like to thank the EPSRC for Ph.D. funding. The authors are also grateful to the anonymous referee, whose comments improved this paper. This work was funded by the UK-NERC under grant GST/02/1892 with additional funding from the EU.

References

Bahe, F. C., W. N. Marx, U. Schurath, and E. P. Roth, Determination of the absolute photolysis rate of ozone by sunlight at ground level, Atmos. Environ., 13, 1515 ­ 1522, 1979. Bais, A. F., et al., International Photolysis Frequency Measurement and Model Intercomparison: Spectral actinic solar flux measurements and modeling, J. Geophys. Res., 108(D16), 8543, doi:10.1029/2002JD002891, in press, 2003. Blackburn, T. E., S. T. Bairai, and D. H. Stedman, Solar photolysis of ozone to singlet D oxygen atoms, J. Geophys. Res., 97, 10,109 ­ 10,117, 1992. Brauers, T., and A. Hofzumahaus, Latitudinal variation of measured NO2 photolysis frequencies over the Atlantic ocean between 50°N and 30°S, J. Atmos. Chem., 15, 269 ­ 282, 1992. Cantrell, C., J. G. Calvert, A. Bais, R. E. Shetter, B. L. Lefer, and G. D. Edwards, Overview and conclusions of the International Photolysis Frequency Measurement and Modeling Intercomparison (IPMMI) study, J. Geophys. Res., 108(D16), 8542, doi:10.1029/2002JD002962, in press, 2003. Crawford, J., R. E. Shetter, B. Lefer, C. A. Cantrell, W. Junkermann, S. Madronich, and J. Calvert, Cloud impacts on UV spectral actinic flux observed during IPMMI, J. Geophys. Res., 108(D16), 8545, doi:10.1029/ 2002JD002731, in press, 2003. Dahlback, A., and K. Stamnes, A spherical model for the computation of the radiation available for heating and photolysis at twilight, Planet. Space Sci., 39, 671 ­ 683, 1991. DeHann, J. F., P. B. Bosma, and J. W. Hovenier, An adding method for multiple scattering calculations of polarized light, Astron. Astrophys., 183, 371 ­ 391, 1987. DeMore, W. B., S. P. Sander, C. J. Howard, A. R. Ravishankara, D. M. Golden, C. E. Kolb, R. F. Hampson, M. J. Kurylo, and M. J. Molina, Chemical kinetics and photochemical data for use in stratospheric modeling, evaluation 12, JPL Publ., 97-4, 1997. Dickerson, R. R., D. H. Steadman, and A. C. Delany, Direct measurements of ozone and nitrogen dioxide photolysis rates in the troposphere, J. Geophys. Res., 87, 4933 ­ 4946, 1982. Edwards, G. D., Aircraft studies of atmospheric chemistry over the North Atlantic, Ph.D. thesis, Univ. of Leicester, Leicester, UK, 2000. Harder, J. W., J. W. Brault, P. V. Johnston, and G. H. Mount, Temperaturedependent NO2 cross sections at high spectral resolution, J. Geophys. Res., 102, 3861 ­ 3879, 1997. Hofzumahaus, A., T. Brauers, U. Platt, and J. Callies, Latitudinal variation of measured j(NO2) rates over the Atlantic Ocean between 50°N and 30°S, J. Atmos. Chem., 15, 269 ­ 282, 1992. Hofzumahaus, A., A. Kraus, and M. Muller, Solar actinic flux spectrora¨ diometery: A technique for the measurement of photolysis frequencies in the atmosphere, Appl. Opt., 38, 4443 ­ 4461, 1999.

Junkermann, W., U. Platt, and A. Volz-Thomas, A photoelectric detector for the measurement of photolysis frequencies of ozone and other atmospheric molecules, J. Atmos. Chem., 8, 203 ­ 227, 1989. Kraus, A., and A. Hofzumahaus, Field measurements of atmospheric photolysis frequencies for O3, NO2, HCHO, CH3CHO, H2O2 and HONO by UV spectroradiometry, J. Atmos. Chem., 31, 161 ­ 180, 1998. Madronich, S., Photodissociation in the atmosphere: 1. Actinic flux and the effect of ground reflections and clouds, J. Geophys. Res., 92, 9740 ­ 9752, 1987. Madronich, S., and G. Weller, Numerical integration errors in calculated tropospheric photodissociation rate coefficients, J. Atmos Chem., 10, 289 ­ 300, 1990. McElroy, C. T., C. Midwinter, D. V. Barton, and R. B. Hall, A comparison of J values from the composition of photodissociative flux measurements with model calculations, Geophys. Res. Lett., 22, 1365 ­ 1368, 1995. Molina, L. T., and M. J. Molina, Absolute absorption cross sections of ozone in the 185- to 350-nm wavelength range, J. Geophys. Res., 91, 14,501 ­ 14,508, 1986. Muller, M., A. Kraus, and A. Hofzumahaus, O3 ! O(1D) photolysis fre¨ quencies determined from spectroradiometric measurements of solar actinic UV-radiation: Comparison with chemical actinometer measurements, Geophys. Res. Lett., 22, 679 ­ 682, 1995. Shetter, R. E., and M. Muller, Photolysis frequency measurements using ¨ actinic flux spectroradiometry during the PEM-Tropics mission: Instrumentation description and some results, J. Geophys. Res., 104, 5647 ­ 5661, 1999. Shetter, R. E., C. A. Cantrell, K. O. Lantz, S. J. Flocke, J. J. Orlando, G. S. Tyndall, T. M. Gilpin, C. A. Fischer, S. Madronich, and J. G. Calvert, Actinometric and radiometric measurement and modeling of the photolysis rate coefficient of ozone to O(1D) during Mauna Loa Observatory Photochemistry Experiment 2, J. Geophys. Res., 101, 14,631 ­ 14,641, 1996. Shetter, R. E., et al., Photolysis frequency of NO2: Measurement and modeling during the International Photolysis Frequency Measurement and Modeling Intercomparison (IPMMI), J. Geophys. Res., 108(D16), 8544, doi:10.1029/2002JD002932, 2003. Slaper, H., H. A. J. M. Reinen, M. Blumthaler, M. Huber, and F. Kuik, Comparing ground-level spectrally resolved solar UV measurements using various instruments: A technique resolving effects of wavelength shift and slit width, Geophys. Res. Lett., 22, 2721 ­ 2724, 1995. Stamnes, K., S.-C. Tsay, W. Wiscombe, and K. Jayaweera, Numerically stable algorithm for discrete-ordinate method radiative transfer in multiple scattering and emitting layered media, Appl. Opt., 27, 2502 ­ 2509, 1988. van de Hulst, H. C., Multiple Light Scattering: Tables, Formulas, and Applications, vols. 1 and 2, Academic, San Diego, Calif., 1980. Volz-Thomas, A., A. Lerner, H.-W. Patz, M. Schultz, D. S. McKenna, ¨ R. Schmitt, S. Madronich, and E. P. Roth, Airborne measurements of the ¨ photolysis frequency of NO2, J. Geophys. Res., 101, 18,613 ­ 18,627, 1996.

ÀÀÀÀÀÀÀÀÀÀÀ ÀÀÀÀÀÀÀÀÀÀÀ

G. D. Edwards, Atmospheric Chemistry Division, National Center for Atmospheric Research, Boulder, CO 80303, USA. ([email protected]) P. S. Monks, Department of Chemistry, University of Leicester, Leicester LE1 7RH, UK. ([email protected])

Information

jd002844 1..13

13 pages

Find more like this

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate

993429


You might also be interested in

BETA
jd002932 1..15
jd002844 1..13