ABF: A data-driven approach for algal bloom forecasting using machine intelligence and remotely sensed data series

Abstract

This paper presents a fully automated framework for algal bloom forecasting in inland water by combining remote sensing data series and unsupervised machine learning concepts. In contrast to other methods in the specialized literature that usually employ pre-labeled data, the proposed approach was designed to be fully autonomous concerning pre-requisites, assuming as input only a time series of remotely sensed products to forecast algal proliferation. In more technical terms, the designed machine-intelligent methodology comprises the steps of pre-processing, feature extraction and modeling, and it learns unsupervised from past events to predict future scenarios of algal blooms, outputting algal insurgence maps.