IPMMI Workshop Summary
Calvert discussed the comparison of the measured photolysis frequencies.
j(O1D): cautioned the use of Talukdar’s O3 quantum yield formula; using the data to distinguish between JPL97 and Talukdar may not be possible; suggests using large zenith angles; suggests ULI look at their FR data for “stair-stepping” phenomenon.
j(NO2): Harder cross sections yield better comparison between actinometers and spectroradiometers than JPL97.
Actinic flux spectra: FZ higher at shorter wavelengths; concerned about ULI scattered light with their spectrometer (single monochromator vs double monochromator used by NCAR, FZJ).
Bais discussed comparison of modeled spectral data and some comparison of model and measurements.
There are 17 results, but not all independent since several used the same codes (TUV and libRadtran). Comparisons were performed on direct irradiance (normal to beam), global irradiance, actinic flux (downwelling) and j-values. He notes that some did not account for the O3 Chappius bands and some had plane-parallel atmospheres leading to problems at high zenith angles. The j(NO2) values agreed to ±15% near solar noon with larger differences at other times. The differences for j(O1D) were larger.
The model-measurement comparisons of spectral data were performed by applying the algorithm of Slaper et al. (discussed later) to correct for the instrumental slit functions and possible wavelength shifts. He notes deviations at short wavelengths and structure in the spectral ratios in the 370 to 390 nm region.
Crawford discussed other aspects of model-measurement comparison focusing on j-values at noon.
He notes that on 19 June data, the mean of the models for j(NO2) was 7.66 (±0.35 = 5%) x 10-3 s-1 with a ratio of the maximum value to the minimum of 1.16. The results for j(O1D) were 3.58 (±0.34 = 10%) x 10-5 s-1 with a max/min ratio of 1.43. There was a clustering of 8 models near the mean values. We need to try and find specific reasons for the clustering and for the high and low values.
In comparing the measurements, he suggests that the FR be treated as one source since they are tied back to the same calibration.
He recommends we use care with the model-measurement comparison for the temperatures used in the j(O1D) calculations, since the quantum yields are so sensitive to temperature.
There are some differences between FR and SR j(NO2) estimates. Are these due to cross section errors and/or other issues?
He suggests calculating j(O1D) from the modeled actinic flux spectra using common molecular data in order to learn something about the errors induced by the radiative transfer calculations alone.
He also points out that the modeled data submitting for IPMMI may have been more “rigorous” than the routine “normal” calculations done for atmospheric global or other models. We would like to understand the differences between “rigorous” and “normal” calculations.
Hofzumahaus discussed the results of a European j-value intercomparison called JCOM97.
It took place in Jülich in June 1997. There were three instruments deployed: the FZJ spectroradiometer (same as IPMMI), a flowing j(NO2) CA, and a static j(NO2) CA. The estimated accuracy of the flowing CA is 7% and of the static CA is 5%.
The slope of the static CA versus the flowing CA was 0.96. The slope of the flowing CA versus j(NO2) derived from the SR using Merienne cross sections (about 4% lower than Harder et al.) was 1.06. Thus the following ratios are derived for SR: CA-static: CA flowing = 1.0: 1.02: 1.06. He says the data suggest that more recent values of NO2 cross-sections lead to better results than JPL97. Using Harder et al. cross-sections leads to the following ratios: CA-flowing:SR = 1.02, CA-static:SR = 0.98. For IPMMI, similar ratios of SR(FZJ):CA(NCAR) = 0.98, 0.95, 1.01, 1.005 for 15, 16, 18, and 19 June, respectively.
Lefer discussed the chemical actinometers used in IPMMI.
He pointed out that they rely on a known, preferably zero, background albedo, and suggests we consider the best way to determine it. He discussed the various factors involved in the j-value determination using chemical actinometry. We discussed the potential of side reactions, the effect of surface-photolyzed NO2 and the effect of surface and light catalyzed conversion of NO to NO2 that has been shown to be a function of water vapor concentration.
The average ratio of j(NO2) NCAR: j(NO2) UMD was 0.927.
He noted hysteresis in j-values and hysteresis in the ratio of j(NO2) ratios, between morning and afternoon.
Hofzumahaus discussed issues related to j(O1D) FR measurements.
He showed that FR signal versus j(O1D) from SR is not linear. Likewise, FR/SR j(O1D) ratios vary with zenith angle. This is related to the FR filter function and PMT response. He has shown that modeled actinic flux spectra and measured filter functions and PMT response can be used to derive a response correction function as functions of zenith angle and overhead ozone column (basically slant ozone column) to lead to a slope for FR versus SR of 0.93 and a ratio FR/SR that does not depend on ZA. He points out the value of FR measurements having a high time resolution that can be used to get j-values between the slower SR measurements.
Several of these activities lead to the question: Can we use IPMMI data to make recommendations regarding the j(O1D) to the atmospheric chemistry and radiation communities?
Hall discussed the scanning spectroradiometers.
Besides the general good agreement, the following details were noted: the ratios of the NCAR and FZJ SR j(NO2) values shows a morning-evening hysteresis; there is a slope bias noticeable in the FZJ data in early morning and late evening data perhaps due to an incorrect interpolation procedure; there is structure in the ratio of NCAR/FZJ actinic flux spectra that needs explaining (due to instrument sampling interval?).
Monks discussed the diode array spectrometer.
The main issues related to the determination of j-values with diode array spectrometers are stray light, dark current, sensitivity and slit functions.
He discussed a method to attempt to correct for stray light by using measurements below 285-290 nm and assuming all the signal is due to stray light.
The detection limit for the ULI DA spectroradiometer in 1 second at 300 nm for a signal to noise ratio of unity is 1 ´ 1010 photons cm-2 nm-1 s-1. This compounds the accurate determination of stray light contributions.
Stray light does not affect j(NO2) much.
j(O1D) from DA SR agreement with other methods is poor, especially in the early morning. He points out the need to look at the spectral contribution to j-values to see reasons for agreement or disagreement. The so-called waveband analysis represents the fraction of contribution for various wavelength bands.
Since the single monochromator has poor stray light rejection (about 10-4), this instrument overestimates j(O1D) especially at high solar zenith angles. In order to correct for this problem, we need a better slit function correction as function of solar zenith angle.
Campos also discussed aspects of diode array spectrometers based on her experience in deploying them in the INDOEX campaign.
She suggests that temperature control of the diode is critical since there is a temperature shift of about 50 counts/sec/°C.
Determinations of j(O1D) are difficult (if not impossible) due to stray light problems particularly affecting the UVB spectral region.
Shetter discussed issues that limit the accuracy of j-value measurements for various instruments.
For chemical actinometers the issues are: field of view of the photolysis cell, cell angle relative to zenith, the horizon, and issues that affect the exposure time determination (cell volume, flow rate, cell pressure and temperature) as well as reactant and product determinations (for example j(NO2) requires an NO determination that depends on calibrations and flow determinations, a NO2 measurement that is based on permeation tube weight loss or NO2 conversion efficiency determinations, and evaluation of the effects of side reactions in the light and dark portions of the instrument that also includes uncertainties in the kinetic parameters involved). For the conditions of the NCAR chemical actinometers the gas phase chemistry side reactions lead to errors of less than 1%.
The issues associated with use of filter radiometers are: optical collection (angular response, head angle, horizon) and matching of the filter function with the cross-section and quantum yield product for the gas of interest.
Scanning double monochromators require consideration of: optical collection, wavelength registration, wavelength resolution, stray light rejection, selection of molecular parameters (NO2 cross-sections and O3 photolysis quantum yields in this case), use of high quality spectral intensity calibration standards (requires determination of the effective head distance).
In using diode array spectrometers one must consider: stray light rejection of the monochromator, wavelength resolution of the monochromator/diode array combination (due to diode bleed).
Madronich discussed issues related to accuracy of modeling j-values.
The first aspect relates to input data and includes: spectral data (extraterrestrial flux, O3 cross-sections, Rayleigh scattering cross-sections), vertical information (O3 profile, pressure & temperature profiles) including the earth-sun distance and its variation, location specific spectral information (amounts of aerosols and other absorbers), and location specific geometric information (horizon, obstructions, inhomogeneous albedo) including clouds (optical properties, amounts, blockage of direct beam).
The next aspect relates to the radiative transfer calculation. One can choose to use numerical methods (of various sorts) to solve the “exact” radiative transfer equation or an “approximate” radiative transfer equation. He points out that there are “established” models (tested, compared, in use for some time) and “less established” models (newer, perhaps prone to coding or other errors).
Finally, one must choose molecular data (cross-sections and quantum yields) to derive j-values from the calculated actinic flux. Interpolations are usually necessary and can be performed before or after the grid selection for the radiative transfer.
Complications arise due to several issues. Two stream models have systematic underestimates for overhead sun of 5-15% for j(O1D) and j(NO2) through the troposphere. There are also errors introduced by plane-parallel, Chapman or pseudo-spherical assumptions. Polarization can also play a role for some conditions.
Differences between models can sometimes be explained by: coding errors, effects of approximate radiative transfer methods and geometric factors.
In comparison of actinic flux spectra, he notes that STX appears to be low for wavelengths less than 310 nm, and that AES and KFA have spikes at 382, 360 and 345 nm possibly associated with use of the WMO extraterrestrial flux.
Crawford discussed issues related to the impact of clouds on j-values.
He showed a procedure to attempt to look at the relationship between the cloud effects on j(NO2) and j(O1D). Looking at the minute to minute changes in the j-values shows that j(NO2) changes are greater than j(O1D), and that the ratios of the changes vary with zenith angle. Can we use the IPMMI data to recommend to the chemical modeling community the relative and absolute effects of clouds on j(NO2) and j(O1D)?
Lloyd discussed modeling related to use of the ER-2 CPFM instrument.
The method involves solution of the integral equation of radiative transfer including refraction. It also accounts for photolysis in the atmospheric window near 220 nm that accounts for a significant fraction of the photolysis of some molecules at 20 km. The spectrum is binned at 50 cm-1 intervals from 175 to 298 nm, at 0.5 nm from 298 to 320, and at 5 nm from 320 to 1000 nm. The CPFM instrument determines the overhead ozone column and the effective albedo beneath the aircraft to constrain the modeled j-values.
Swartz continued the discussion of CPFM derived j-values, and concluded that the modeled j(NO2) values were 6% higher than CPFM, and the j(O1D) values were 12% higher than CPFM during POLARIS.
Cantrell discussed issues related to statistical methods for comparing results including least squares when x and y variables have errors, weighting data based on their uncertainties, comparisons of frequency distributions, logarithmic transformations and other issues to serve as a basis for quantitative comparisons of the IPMMI results. The discussion that ensued led to the conclusion that several methods of comparison should be utilized including fits of one estimate versus another and statistics of ratios of the various estimates (much as Cavert has done for the measurements) as well as examination of the frequency distributions of the ratios in order to try to understand the differences.
Bais discussed the details of filtering measured data to compare with modeled data.
The process involves the use of the Slaper et al. algorithm. It is a four step process: (1) convolute a high resolution solar spectrum with the known instrument slit function; (2) calculate a correction spectrum; (3) determine a new deconvoluted spectrum; (4) repeat 10 times. It may also be necessary to shift the spectrum.
He noted that modeled actinic flux spectra deviate in the negative direction from the NCAR SR data for wavelengths less than 300 nm. Discussion led to the question: what is the effect of a dark current offset in the measurements on this filtering process? We concluded that he should try it after subtracting the dark current from the slit function spectrum.
He chose to correct the measurements to a rectangular slit function to compare with the model results. We suggested that he compare converting the model results to the actual instrument slit functions to this approach to see if there any differences. We also asked: how much effect does this filtering process have on the derived j-values?
Junkermann discussed broadband filter radiometers.
He suggested that some differences between his j(NO2) measurements and others could be due to zenith angle response imperfections and lack of temperature stabilization. He noted that the six-year-old detectors are losing sensitivity and suggests that they could be improved through temperature stabilization and calibrations that are more frequent.
The j(O1D) radiometer data first submitted had large errors because a zenith angle correction was applied in the wrong direction. The agreement with the other measurements is much better with the proper correction. Comparison of radiometers using filters even from the same batch can show large differences for zenith angles greater than 60 degrees due to small differences in the filters. To properly correct filter radiometer results, the response function of the entire filter/detector combination needs to be measured.
Discussion that followed led to the following questions: How sensitive is the filter radiometer correction to the actual ozone column amount? Could ozone column climatology be used with satisfactory results?
Other issues preceding the discussion of papers and special session.
ULI noted that their FR data on 19 June is bad since the detector temperature exceeded the control temperature.
Discussion of proposed scientific papers related to IPMMI.
See web page on IPMMI papers. We did not conclude as to which journal to which the papers should be submitted, but the following were suggested: Atmospheric Environment, Journal Geophysical Research-Atmospheres, and Journal of Atmospheric Chemistry. We invite discussion on this topic as the deadline for submittal approaches.
Protocols for changing data submitted to IPMMI referees.
If data submitted to IPMMI is changed and resubmitted, then the responsible PI must clearly explain the reasons for the change(s). The appropriate paper describing the results will use an appendix to show either the changed or the original data and explain these changes.
Proposal for a special session on IPMMI and other issues related to atmospheric photolysis processes.
We discussed requesting a special session at the spring AGU meeting in Washington DC (30 May – 3 June 2000). A draft proposal was discussed, refined, and submitted for consideration. The special session has been approved and details of the number of oral and poster papers will be forwarded to the community when it is available. The special session description is found on the AGU Special Session web page.
We also propose a one-half day workshop at the spring AGU meeting to update everyone on the status of papers.
It was proposed that one poster summarizing the IPMMI results be presented at the IRS meeting in St. Petersburg in July 2000.
It was suggested that a special session on IPMMI be proposed for the EGS meeting in Nice, France in April 2000.
Protocols and timing of release of data to the scientific community was discussed.
We propose a release date of 31 July 2000. Until then, data is to be available only to IPMMI investigators.
The data will be available on an FTP server at NCAR. The most recent, final data will be made released.
If anyone is to show data from IPMMI, they must get the approval of the PI responsible for the data before presentation or publication, and offer co-authorship if appropriate.
A future IPMMI (perhaps an airborne campaign) was put forward and discussed, but no firm conclusions were reached.