Biological invasions cause unprecedented disruptions to native ecosystems, and negatively impact health and economy. In the United States alone, the annual economic cost due to environmental damages and losses caused by such invasions is over $120B.
We are studying the spread of invasive plants in the Chitwan Annapurna Landscape (CHAL) of Nepal, which is part of a biodiversity hotspot. This problem of invasive species is an impediment to the achievement of multiple sustainable development goals drafted by the United Nations. CHAL has a rich diversity of flora and fauna, which is unfortunately threatened by the combined effects of climate change and increased human activities.
In this work, we explore the feasibility of applying modern machine learning techniques to the invasive species distribution problem. One option would be using convolutional neural networks (CNNs) to analyze visual imagery. Advances in machine learning and the availability of high-resolution imagery (satellites, drones, etc.), have made monitoring species, forests, and croplands using remotely sensed data a viable option. This method also opens up the possibility of retrodiction—the interpretation of past events—by using a time series of satellite imagery. Coupled with epidemiological models, this method can help provide forecasts and analyze the different pathways by which invasive plants can spread and establish in this landscape.