In order to join virtual sessions, you must be registered and logged-in(Were you registered for the in-person meeting in Denver? if yes, just log in.) 
Note: All session times are in Mountain Daylight Time.

Deep learning for environmental niche modeling of the cold-water coral Lophelia pertusa in the Gulf of Mexico

Authors: Jilin Hu*, University of West Florida, Zhiyong Hu, University of West Florida, Yaguang Zhou, Chongqing Institute of Surveying and Mapping, MNR
Topics: Geographic Information Science and Systems, Environmental Science, Coastal and Marine
Keywords: spatial point pattern, deep learning, environmental niche modeling, cold-water corals,
Session Type: Poster
Presentation File: No File Uploaded

Marine ecosystems are suffering from climate change, ocean acidification and overfishing. Cold-water coral ecosystems are vulnerable to these threats and slow to recover. In the NOAA coral database for the Gulf of Mexico (GoM), L. pertusa is the dominant species. However, the full geographic extent of L. pertusa is far from known.Hence, knowledge of their distribution is critical to assessing impacts from fisheries and environmental change, and is useful for conservation management plans. This study presents environmental niche modeling of L. pertusa in the GoM. First, spatial point pattern analyses were performed to explore the association of observed pattern of L. pertusa with covariates. It was found that the most important presence contributors are substrates superbly dominated by gravels, slope terrain, slope toward east, roughness, and slope degree. Next, an optimal deep neural network model was used to train the presence-absence-covariates data for the whole and the east half of the study area separately. Both models were used to predict probability of presence for the whole region. Threshold values which maximize the sum of sensitivity and specificity were used to classify the probability maps into presence and absence maps. Model performances were assessed using the reserved test data and Area Under Curve (AUCs) of receiver, operating characteristics (ROC) were used to assess the model performances. Finally, this study will also conduct three scenario analyses using the full model to predict the effect of environmental and climate change on the distribution of L. pertusa across the GoM.

To access contact information login