An evaluation of uncertainty and sensitivity of a regional scale air pollution and pesticide modeling system

A regional scale 3-dimensional coupled atmospheric transport, soil-air exchange and water-air exchange model was developed (REGAPS) to investigate organochlorine pesticides impact over the Northern American region. The coupled model was used to simulate -HCH (lindane) concentration and to study its impact over the Great Lakes (Ma et al 2003 Environ.Sci.Technol.37; Ma et al 2004 Environ.Sci. Technol.38). Here we present results from the statistical analysis of model experiments on the sensitivity and uncertainty associated with a few key dependent parameters of the model. We selected three meteorological variables and five model parameters that are strongly related to the dynamics of pesticide transport and dispersion and also to the surface-air exchange processes for the pesticide. These are horizontal wind, air temperature, and precipitation rate, mixing layer height, friction velocity, roughness length, dry deposition velocity, Henry’s law constant and octanol-air partition coefficient. The coupled model was run with lindane emission field and objectively analyzed data from the regional version of Global Environmental Multi-scale Model (GEM) - the numerical weather prediction model of CMC (Canadian Meteorological Center). This base experiment was run for the period of May to August 1998. Numerical model experiments are repeated by varying each parameter with a specified amount and its sensitivity to the model output is investigated in light of the base run results by applying statistical measures. Significant conclusions are the following.

• The modeled air concentrations are most sensitive to the wind speed among the meteorological inputs, while the soil concentrations at the exchange and reservoir layers respond most significantly to the air temperature. Relatively air and soil concentrations are not very sensitive to the changes in precipitation.

• The largest errors of the modeled air concentrations and the exchange layer soil concentrations are due to variation of the mixing layer height hb, followed by the octanol-air partition coefficient, Koa. Changes in the roughness length and dry deposition velocity do not show strong impact to the changes in the air and soil concentrations.

• The increase in the Henry’s law constant leads to significant reduction of the soil concentration in the reservoir layer (so does the air temperature), corresponding to the stronger volatilization and diffusion of -HCH from the deep soil layer to the upper soil layers.

• The model performance was further evaluated by comparing modeled (May to November 1998) daily air concentrations of -HCH with monitored data from the Integrated Atmospheric Deposition Network. Good correlation between the modeled and measured air concentration occurs for three out of five stations. Similarly good measures are noticed for model to observed standard deviation, t-test probability and relative error.

• Overall, the REGAPS model predicts well of -HCH air concentrations in the summer months and underestimates air concentrations in autumn and winter months.