Measurements and Modeling of Solar Ultraviolet Radiation and Photolysis Rates during SCOS97 Laurent Vuilleumier Environmental Energy Technologies Division Presented at the SCOS97-NARSTO Data Analysis Conference February 14, 2001

Collaborators Nancy J. Brown, Berkeley Lab Robert A. Harley, UC Berkeley Jeffrey T. Bamer, UC Berkeley Steven D. Reynolds, Envair James R. Slusser, CSU David S. Bigelow, CSU Donald Kolinski, UCAR

Motivations Numerous sensitivity analysis studies* indicate large ozone (smog) formation sensitivities to NO 2, and HCHO photolysis rates. Monte Carlo study of ozone modeling uncertainties, Hanna et al. (2000, EPRI) report: Uncertainties in ozone predictions are most strongly correlated with uncertainties in NO 2 photolysis rate. *Falls et al., 1979, Milford et al., 1992, Gao et al., 1995, 1996, Yang et al., 1995, 1996, Vuilleumier et al., 1997, Bergin et al. 1998, Hanna et al., 1998, 2000

Outline  Uncertainty in photolysis rate coefficients  Optical depth variability during SCOS97  Modeling photolysis rate coefficients  Comparison between observations and predictions of NO 2 photolysis rate coefficients

Photolysis Reaction Rates X concentration rate of change due to photolysis reaction i Species X undergoes photodissociation. Reaction i:X + h  products X absorption cross section Reaction i quantum yield Spectral actinic flux Wavelength Action spectrum Reaction rate coefficient

Uncertainties in Photolysis Reaction Rates  Action Spectrum Experimental uncertainties reduced by better determination of cross sections & quantum yields  Actinic Flux (solar light flux available for photolysis) Depends on atmospheric optical properties that exhibit spatial and temporal variation Natural variability & Measurement uncertainty

Important Atmospheric Optical Properties  Optical Depth Measures light extinction along vertical path. Ex: constant atmosphere  Single Scattering Albedo Represents fraction of extinguished light that is scattered (remaining is absorbed). low SSA = high absorption Effect on light intensity is maximum when optical depth is high (extinction) and SSA is low (absorption). z (altitude)  Beam intensity  Incoming light beam Constant atmosphere

Total optical depth  (t) obtained by using relationship between irradiance at ground I(t), extraterrestrial irradiance I 0, and air mass factor m R (t). Optical Depth Computation m1m1 mimi mnmn ln(I i ) slope  i m slope  n mnmn mimi m2m2 m1m1 ln(R 2 I 0 ) slope  1

Measurements  Direct irradiance from UV multifilter radiometers: >Measurement at = 300, 306, 312, 318, 326, 333 and 368 nm. >2 nm nominal full-width half-maximum filters with integrated out-of-band light contamination less than 0.5%.  Data acquired at Riverside and Mt Wilson, CA from 1 July to 1 November >Riverside (260 m a.s.l.) characterized by frequent occurrences of severe air pollution episodes. >Mt Wilson (1725 m a.s.l.) located above much of the urban haze layer.

Optical Depth Variability After data selection (reject cloudy periods or low signal to noise ratio), 8,232 optical depths obtained at Riverside and 11,261 at Mt Wilson:

Accounting for Optical Depth Variability PCA attributes 97% and 2% of variability to 1 st and 2 nd most important components at Riverside, and 89% and 10% at Mt Wilson. Components correspond to light extinction by aerosols and ozone. aerosols ozone

Significant variability in atmospheric optical depth due to aerosols. Is it possible to reproduce it in models? What are the most significant sources of uncertainty?

Modeling Photolysis Rates  Selected and modified TUV* program from Madronich (NCAR**) for implementation in AQM’s (UAM-IV, UAM-FCM, SAQM).  TUV allows consideration of: >absorption and scattering by aerosols, >absorption and scattering by gases (O 3, O 2, NO 2, SO 2 ), >ground albedo, >atmospheric pressure and temperature vertical profiles. * Tropospheric Ultraviolet-Visible, ** National Center for Atmospheric Research

Modifications to TUV  Increased modularity to enhance incorporating new science  Improved user interface for facilitating changing input variables  TUV can be called during AQM simulation with selected inputs depending on time and location: >Aerosol characteristics, >Ozone total optical depths, >Ground albedo (depends on location only).

Effect of Optical Depth Variability on TUV Predictions  TUV used to predict NO 2 photolysis rate (J NO2 ) for aerosol optical depths observed at times of high and low turbidity. low turbidity (  aer = 0) and high turbidity (  aer = 0.8 at = 340 nm, 95th percent.)  Predictions show differences between 15% and 40%.

Comparison of observed and predicted J NO2  SCOS97 J NO2 measurements (UC Riverside) used to assess correctness of TUV predictions. >Ground level data measured at Riverside with chemical actinometer on selected days >Required matching measurements of J NO2, aerosol optical depth and ozone column >Obtained 121 simultaneous observations and predictions of J NO2 over 14 non-continuous days.

J NO2 Predicted to Observed Ratio  Ratio of predicted to observed J NO2 reveals an average bias of 15 to 30% depending on single scattering albedo.  Daily profile reproduced, including variations due to atmospheric condition changes, resulting in low ratio standard deviation around average (±10%).

J NO2 Daily Profile  Predictions using constant average input (aerosol optical depth and ozone column) only show influence of solar zenith angle.  Predictions using time- varying input correctly predicts variations due to changes in optical depth.

Possible Sources of Bias  Single Scattering Albedo: Uncertainty in SSA can result in: >Bias in predicted J NO2 (uncertainty in average SSA) >Random uncertainty in predicted J NO2 (temporal variability of SSA)  Corrections used for J NO2 measurements: Quantum yield factor used for observed J NO2. >Impurities in carrier gas (N 2 ) have significant influence on quantum yield factor and can lead to bias in observed J NO2 *. * Dickerson and Stedman (1980) Environ. Science & Technol. 14,

Conclusions  Natural atmospheric variability has significant influence on photolysis rates.  In cloud-free situations, aerosols are responsible for most of the variability.  Aerosol single scattering albedo remains a significant source of uncertainty.

Conclusions (2)  Radiative transfer models can reproduce variability providing good input data are available: >Challenge at the scale of Air Quality Modeling. >Synergy between ground-based, air- borne, and satellite-based observation of troposphere may be key to success.

Additional Material

Langley Plot Calibration  If V(t) corresponds to I(t), V 0 corresponding to R 2 I 0 is obtained with a Langley plot method (1) applicable at time of low atmospheric turbidity.   is computed with: (1) Slusser et al. (2000) J. Geophys. Res. 105,

Optical Depth Data Selection  Clouds: >High photochemical air pollution is linked to stagnant high-pressure systems. >Times when clouds are present are rejected based on broadband visible irradiance.  Low signal: >Events where total minus diffuse irradiance is low are rejected to reduce electronic noise influence.

Optical Depth Correlations  Correlation between measurements at the seven wavelengths is strong.  Correlation is stronger between measurement at neighboring wavelengths.  Correlation is stronger at Riverside than Mt Wilson.  At Mt Wilson, two groups show stronger correlation: short (300, 306) and long wavelengths (312–368).

Correlation Matrix

Correlations at Mt Wilson

Correlations at Riverside

Principal Component Analysis  PCA is used to transform a set of correlated variables into a set of uncorrelated variables called components.  The most important components are linked to the physical causes of the observed variability.  The components are found by diagonalizing the correlation matrix.     a b PC 1 PC 2 Contribution from 1 to PC 1

Wavelength Contributions to the Components The wavelength contributions to the components suggest that the first two components correspond to absorption and scattering by aerosols and ozone, respectively.