Cloud base height retrieval method based on muti-angle observations and data examples

This repository contains the essential resources to reproduce the cloud base height (CBH) retrieval algorithm presented in the study entitled "Retrieval of the base heights and cloud geometric thicknesses of clouds based on the PARASOL measurement." The algorithm is a machine learning-based model (deep neural network) trained on collocated POLDER muti-angle observations and active sensor (CloudSat-Calipso) data. It is designed to estimate CBH globally from passive satellite measurements. The repository includes: Data_Examples.csv: A minimal sample dataset containing input features and target variables for model validation and testing. DNN_Train.py: Python script for training the deep neural network model from scratch. DNN_Pre.py: Python script for performing cloud base height inference using the pre-trained model. Parasol_CBH_Model.pkl: The pre-trained DNN model weights for immediate prediction. scaler.joblib: Scaler object for data standardization, ensuring consistent preprocessing during inference.