The overexploitation of natural resources and pollution are urgent concerns affecting the Earth's global system. Earth Observation (EO) data can be used to analyze the environmental impact of human activities. However, extracting meaningful insights from EO time series data requires domain expertise. In this position paper, we propose a methodology to improve the accessibility and understanding of environmental trends for a wide audience. Using Machine Learning (ML) technologies, we detect and describe in the Semantic Web (SW) changes in EO time series.