Code Package and Data of Deep Learning Driven Simulations of Boundary Layer Cloud over the US Southern Great Plains

These are the package of the deep learning model for boundary layer clouds (DLC: deep-learning-driven cloud), including input data, model outputs, the scripts. This research presents a deep learning framework that is designed to simulate boundary layer clouds (BLC) with the integration of three specialized deep neural network (DNN) models that capture the initiation, positioning, and vertical extent of BLCs. It all starts with a model for predicting cloud formation. This is followed by a model for positioning clouds. Last, the cloud coverage fraction is determined vertically within the cloud. This allows for a rich BLC dynamics representation. The model uses morning meteorological data and surface fluxes to predict BLC formation in TensorFlow. Normalization, L2 regularization, batch normalization, and dropout are all employed during training for efficient training and strong model robustness. The architecture is carefully designed to accurately simulate the BLC. Early stopping and learning rate adjustment are used in the training to achieve the optimized performance of the model. This approach produces a rich puzzle for simulating BLCs in detail, which is important to the understanding and prediction of cloud layer behaviors.