Machine Learning Approach for Predicting Persistent Contrails Formation Region

The proposal is motivated by the need for models for predicting the environmental impact of planned operations to enable aircraft operators to make tradeoffs between the benefit of reducing the negative environmental impact and the cost of deviating from fuel optimal routes and altitudes. These models will also provide metrics for policy makers to assess long term impact of the policies for reducing the environmental impact of aviation. nbsp;Modeling of two major sources of environmental impact of aviation: (1) persistent contrails and (2) emissions have been proposed. A machine-learning approach is proposed for forecasting regions of persistent contrails formation using features derived from atmospheric data and satellite images. Compared to earlier models that are point-based, our proposal is a region-based prediction methodology with reduced uncertainty in the prediction of regions of persistent contrails formation using clustering techniques. For improving the emissions estimates, we propose a computational procedure for estimating the takeoff weight considering in part information provided in the flight plan. Simulation of trajectories with the estimated takeoff weight along with regions of persistent contrail formation predicted by the machine-learning model provide environmental impact in terms of expected emissions and contrail formation.nbsp;nbsp; nbsp;