Deep learning algorithms to improve the management of species at risk

点击翻译按钮获取中文摘要

NCASI, with Ontario Ministry of Natural Resources and Forests, and the University of Alberta, has developed large data sets that reflect at fine-scales the nutritional value to caribou of different types of vegetation. The next step is to analytically link this fine-scale information to habitat use, pregnancy rates, and body condition of wild caribou in Ontario. To address the inadequacy of current vegetation maps, NCASI has joined with Dr. Salimur Choudhury at Lakehead University to use current, high-resolution imagery to develop new maps of vegetation with higher accuracy and greater mapping resolution (e.g., 1/4 ha versus ~4 ha) and that are tailored to caribou nutrition than are currently available. Central to this effort are vegetation data sets collected at 500 half-hectare vegetation plots in northwestern and northeastern Ontario by NCASI scientists that describe the abundance of lichens and other caribou forage, and which provide detailed measures of overstory and other site characteristics. The goal of this project is to design computer-based mapping algorithms that will be used to construct maps of nutritional resources at larger scales (stand and landscape) for a variety of research and management applications.