Detection and Mapping of Invasive Alien Plant Water Hyacinth using Satellite Imagery and Machine Learning (Case Study: Rawa Pening Lake, Indonesia)

Abstract

Rawa Pening Lake, one of the 15 national priority lakes in Indonesia, faces a significant threat from invasive water hyacinth (Eichhornia crassipes). This plant once covered up to 70% of the lake's surface and continued to cause ecological and socio-economic impacts as of 2024, necessitating periodic monitoring to prevent future blooms. This study aimed to identify the optimal features to characterize water hyacinth, determine the most effective classification model, and map the plant’s distribution. Adopting the CRISP-DM framework, the study utilized Sentinel-1 (radar) and Sentinel-2 (optical) satellite imagery with multispectral band features, radar bands, and composite indexes. Feature selection was performed using Jenks Natural Breaks, and classification modeling was conducted using Random Forest and Convolutional Neural Network (CNN). The results demonstrated that the CNN achieved higher accuracy in distinguishing among land cover classes. The final mapping identified water hyacinth covering 34,775 pixels, 32,627 pixels, and 34,175 pixels in June, July, and August, respectively. This approach offers a reliable method for periodic monitoring of water hyacinths in Rawa Pening Lake.