Nonlinear data assimilation for the regional modeling of maximum ozone values
2017
Božnar, Marija Zlata | Grašič, Boštjan | Mlakar, Primož | Gradišar, Dejan | Kocijan, J.
We present a new method of data assimilation with the aim of correcting the forecast of the maximum values of ozone in regional photo-chemical models for areas over complex terrain using multilayer perceptron artificial neural networks. Up until now, these types of models have been used as a single model for one location when forecasting concentrations of air pollutants. We propose a method for constructing a more ambitious model: a single model, which can be used at several locations because the model is spatially transferable and is valid for the whole 2D domain. To achieve this goal, we introduce three novel ideas. The new method improves correlation at measurement station locations by 10% on average and improves by approximately 5% elsewhere.
显示更多 [+] 显示较少 [-]