Climate change is producing shifts in the species ranges, particularly for kelp, the trend is a decline in most of the populations in the last 20-50 years. Giant kelp is one of the most widely distributed species in the planet and the foundation of one of the most emblematic ecosystems in the world: the kelp forest. This ecosystem is highly dynamic: it shows a strong seasonal pattern, and interannual dynamics, being El Niño (ENSO) a major driver of its distribution. To study the effects of climate change over giant kelp forests, long-term monitoring is key. Satellite imagery is an essential tool to study the patterns over different spatial and temporal scales of giant kelp, however, these methods can be costly in terms of time and human work required to process it. In this work, we used an artificial intelligence network, convolutional neural networks (CNN) to automatically detect giant kelp in satellite images from California, US, and Baja California, Mexico. We tested 4 different CNN models to choose the best with the one overall performance for prediction, the maximum accuracy of the final model was 93%. This method is an automatic, simple, efficient, and open-source method for giant kelp detection in satellite images.
Managing Thesis and Internship Exchanges
