Evaluation of five dry particle deposition parameterizations for incorporation into atmospheric transport models

Abstract. Despite considerable effort to develop mechanistic dry particle deposition parameterizations for atmospheric transport models, current knowledge has been inadequate to propose quantitative measures of the relative performance of available parameterizations. In this study, we evaluated the performance of five dry particle deposition parameterizations developed by Zhang et al. (2001) (Z01), Petroff and Zhang (2010) (PZ10), Kouznetsov and Sofiev (2012) (KS12), Zhang and He (2014) (ZH14), and Zhang and Sao (2014) (ZS14), respectively. The evaluation was performed in three dimensions: model ability to reproduce observed deposition velocities, Vd (accuracy), the influence of imprecision in input parameter values on the modeled Vd (uncertainty), and identification of the most influential parameter(s) (sensitivity). The accuracy of the modeled Vd was evaluated using observations obtained from five land use categories (LUCs): grass, coniferous and deciduous forests, natural water, and ice/snow. To ascertain the uncertainty in modeled Vd, and quantify the influence of imprecision in key model input parameters, a Monte Carlo uncertainty analysis was performed. The Sobol' sensitivity analysis was conducted with the objective to determine the parameter ranking, from the most to the least influential. Comparing the normalized mean bias factors (indicator of accuracy), we find that the ZH14 parameterization is the most accurate for all LUCs except for coniferous forest, for which it is second most accurate (BNMBF = −2.31). From Monte Carlo simulations, the estimated mean normalized uncertainties in the modeled Vd obtained for seven particle sizes (ranging from 0.005 to 2.5 μm) for the five LUCs are 17 %, 12 %, 13 %, 16 %, and 27 % for the Z01, PZ10, KS12, ZH14, and ZS14 parameterizations, respectively. From the Sobol' sensitivity results, we suggest that the parameter rankings vary by particle size and LUC for a given parameterization. Overall, for dp = 0.001 to 1.0 μm, friction velocity was one of the three most influential parameters in all parameterizations. For giant particles (dp = 10 μm), relative humidity was the most influential parameter. Because it is the least complex of the five parameterizations, and it has the greatest accuracy and least uncertainty, we propose that the ZH14 parameterization is currently superior for incorporation into atmospheric transport models.

Download Full-text

Evaluation of five dry particle deposition parameterizations for incorporation into atmospheric transport models

Geoscientific Model Development ◽

10.5194/gmd-10-3861-2017 ◽

2017 ◽

Vol 10 (10) ◽

pp. 3861-3888 ◽

Cited By ~ 14

Author(s):

Tanvir R. Khan ◽

Judith A. Perlinger

Keyword(s):

Monte Carlo ◽

Particle Deposition ◽

Atmospheric Transport ◽

Current Knowledge ◽

Input Parameter ◽

Coniferous Forest ◽

Three Dimensions ◽

Transport Models ◽

Deposition Velocities ◽

Parameter Values

Abstract. Despite considerable effort to develop mechanistic dry particle deposition parameterizations for atmospheric transport models, current knowledge has been inadequate to propose quantitative measures of the relative performance of available parameterizations. In this study, we evaluated the performance of five dry particle deposition parameterizations developed by Zhang et al. (2001) (Z01), Petroff and Zhang (2010) (PZ10), Kouznetsov and Sofiev (2012) (KS12), Zhang and He (2014) (ZH14), and Zhang and Shao (2014) (ZS14), respectively. The evaluation was performed in three dimensions: model ability to reproduce observed deposition velocities, Vd (accuracy); the influence of imprecision in input parameter values on the modeled Vd (uncertainty); and identification of the most influential parameter(s) (sensitivity). The accuracy of the modeled Vd was evaluated using observations obtained from five land use categories (LUCs): grass, coniferous and deciduous forests, natural water, and ice/snow. To ascertain the uncertainty in modeled Vd, and quantify the influence of imprecision in key model input parameters, a Monte Carlo uncertainty analysis was performed. The Sobol' sensitivity analysis was conducted with the objective to determine the parameter ranking from the most to the least influential. Comparing the normalized mean bias factors (indicators of accuracy), we find that the ZH14 parameterization is the most accurate for all LUCs except for coniferous forest, for which it is second most accurate. From Monte Carlo simulations, the estimated mean normalized uncertainties in the modeled Vd obtained for seven particle sizes (ranging from 0.005 to 2.5 µm) for the five LUCs are 17, 12, 13, 16, and 27 % for the Z01, PZ10, KS12, ZH14, and ZS14 parameterizations, respectively. From the Sobol' sensitivity results, we suggest that the parameter rankings vary by particle size and LUC for a given parameterization. Overall, for dp  =  0.001 to 1.0 µm, friction velocity was one of the three most influential parameters in all parameterizations. For giant particles (dp  =  10 µm), relative humidity was the most influential parameter. Because it is the least complex of the five parameterizations, and it has the greatest accuracy and least uncertainty, we propose that the ZH14 parameterization is currently superior for incorporation into atmospheric transport models.

Download Full-text

New Methodology Shows Short Atmospheric Lifetimes of Oxidized Sulfur and Nitrogen due to Dry Deposition

10.5194/acp-2020-1315 ◽

2021 ◽

Author(s):

Katherine Hayden ◽

Shao-Meng Li ◽

Paul Makar ◽

John Liggio ◽

Samar G. Moussa ◽

...

Keyword(s):

Monte Carlo ◽

Dry Deposition ◽

Deposition Velocity ◽

Monte Carlo Analysis ◽

Climate Impacts ◽

Atmospheric Pollutants ◽

Transport Models ◽

Deposition Fluxes ◽

Deposition Velocities ◽

Model Algorithm

Abstract. The atmospheric lifetimes of pollutants determine their impacts on human health, ecosystems and climate and yet pollutant lifetimes due to dry deposition over large regions have not been determined from measurements. Here, a new methodology based on aircraft observations is used to determine the lifetimes of oxidized sulfur and nitrogen due to dry deposition over (3–6) × 103 km2 of boreal forest in Canada. Dry deposition fluxes decreased exponentially with distance, resulting in lifetimes of 2.2–26 hours. Fluxes were 2–14 and 1–18 times higher than model estimates for oxidized sulfur and nitrogen, respectively, indicating dry deposition velocities which were 1.2–5.4 times higher than those computed for models. A Monte-Carlo analysis with five commonly used inferential dry deposition algorithms indicates that such model underestimates of dry deposition velocity are typical. These findings indicate that deposition to vegetation surfaces are likely under-estimated in regional and global chemical transport models regardless of the model algorithm used. The model-observation gaps may be reduced if surface pH, and quasi-laminar and aerodynamic resistances in algorithms are optimized as shown in the Monte-Carlo analysis. Assessing the air quality and climate impacts of atmospheric pollutants on regional and global scales requires improved measurement-based understanding of atmospheric lifetimes of these pollutants.

Download Full-text

On the Influence of Uncertainties in Estimating Risk Aversion and Working Interest

Energy Exploration & Exploitation ◽

10.1177/014459879601400103 ◽

1996 ◽

Vol 14 (1) ◽

pp. 13-45 ◽

Cited By ~ 2

Author(s):

J.A. MacKay ◽

I. Lerche

Keyword(s):

Monte Carlo ◽

Monte Carlo Simulations ◽

Probability Distributions ◽

Input Parameter ◽

Success Probability ◽

Risk Tolerance ◽

Relative Importance ◽

Hyperbolic Tangent ◽

Corporate Risk ◽

Parameter Values

The influence of uncertainties in costs, value, success probability, risk tolerance and mandated working interest are evaluated for their impact on assessing probable ranges of uncertainty on risk adjusted value, RAV, using different models. The relative importance of different factors in contributing to the uncertainty in RAV in analyzed, as is the influence of different probability distributions for the intrinsic variables entering the RAV model formulae. Numerical illustrations indicate how the RAV probabilities depend not only on the model functions (Cozzolino, hyperbolic tangent) used to provide RAV estimates, but also on the intrinsic shapes of the probability distributions from which are drawn input parameter values for Monte Carlo simulations. In addition, a mandated range of working interest can be addressed as an extra variable contributing to the probabilistic range of RAV; while negative RAV values for a high-cost project can be used to assess the probable buy-out amount one should be prepared to pay depending on corporate risk philosophy. Also, the procedures illustrate how the relative contributions of scientific factors influence uncertainty of reserve assessments, allowing one to determine where to concentrate effort to improve the ranges of uncertainty.

Download Full-text