A deep-learning model to predict thunderstorms within 400 km2 South Texas domains

A deep-learning neural network (DLNN) model was developed to predict thunderstorm occurrence within 400 km2 South Texas domains for up to 15 hr (±2 hr accuracy) in advance. The input features were chosen primarily from numerical weather prediction model output parameters/variables; cloud-to-ground lightning served as the target. The deep-learning technique used was the stacked denoising autoencoder (SDAE) in order to create a higher order representation of the features. Logistic regression was then applied to the SDAE output to train the predictive model. An iterative technique was used to determine the optimal SDAE architecture. The performance of the optimized DLNN classifiers exceeded that of the corresponding shallow neural network models, a classifier via a combination of principal component analysis and logistic regression, and operational weather forecasters, based on the same data set.

A deep-learning neural network (DLNN) model was developed to predict thunderstorm occurrence within 400 km2 South Texas domains for up to 15 hr (±2 hr accuracy) in advance. The input features were chosen primarily from numerical weather prediction model output parameters/variables; cloud-to-ground lightning served as the target. The deep-learning technique used was the stacked denoising autoencoder (SDAE) in order to create a higher order representation of the features. Logistic regression was then applied to the SDAE output to train the predictive model. An iterative technique was used to determine the optimal SDAE architecture. The performance of the optimized DLNN classifiers exceeded that of the corresponding shallow neural network models, a classifier via a combination of principal component analysis and logistic regression, and operational weather forecasters, based on the same data set.