Examination of a Machine Learning Model for Predicting Riverrine Vegetation Distribution Across Several Rivers With Different Channel Characteristics

Puntos clave

• This research aims to develop a reliable machine learning model for predicting riverine vegetation distribution across diverse river systems.
• Developed a multi-stage machine learning model for vegetation distribution prediction.
• Evaluated 16 machine learning models to select the most suitable one.
• Incorporated additional explanatory variables for local terrain changes.
• Utilized data from six efficient river segments for model training.
• XGBoost emerged as the optimal model based on accuracy and computational load.
• The model achieved F-scores ranging from 0.73 to 0.93 for predicting vegetation across 15 rivers.
• Confirmed the effectiveness of the machine learning model in predicting vegetation distribution in many rivers in Japan.