Abstract:

We constructed a back-propagation neural network (BPNN) model with a P(O₃) sensitivity indicator,  Φ=k_HC+OH［HC］/k_NOx+OH［NO_x］, and influencial factors as its input variables. We employed the model to predict the control factors of ozone formation and its sensitivity character at ocean atmospheric boundary layer of Japan Hailikao island. Prediction results basically conformed with chemical computation results. It indicates that this model has better prediction performance. Our research demonstrates that artificial neural network (ANN) can serve as an effective tool to determine the sensitivity of P(O₃) and to formulate ozone abatement strategy.

Key words: tropospheric ozone, ozone production sensitivity, artificial neural network