2020 journal article

Using Random Forest Algorithm to Model Cold-Stunning Events in Sea Turtles in North Carolina

JOURNAL OF FISH AND WILDLIFE MANAGEMENT, 11(2), 531–541.

By: J. Niemuth n , C. Ransom, S. Finn*, M. Godfrey n, S. Nelson n  & M. Stoskopf n

co-author countries: United States of America πŸ‡ΊπŸ‡Έ
author keywords: climate change; cold-stunning; green sea turtle; Kemp's ridley sea turtle; loggerhead sea turtle; stranding prediction; weather
Source: Web Of Science
Added: July 6, 2021

Abstract Sea turtle strandings due to cold-stunning are seen when turtles are exposed to ocean temperatures that acutely and persistently drop below approximately 12Β°C. In North Carolina, this syndrome affects imperiled loggerhead Caretta caretta, green Chelonia mydas, and Kemp's ridley Lepidochelys kempii sea turtle species. Based on oceanic and meteorological patterns of cold-stunning in sea turtles, we hypothesized that we could predict the daily size of cold-stunning events in North Carolina using random forest models. We used cold-stunning data from the North Carolina Sea Turtle Stranding and Salvage Network from 2010 to 2015 and oceanic and meteorological data from the National Data Buoy Center from 2009 to 2015 to create a random forest model that explained 99% of the variance. We explored additional models using the 10 and 20 most important variables or only oceanic and meteorological variables. These models explained similar percentages of variance. The variables most frequently found to be important were related to air temperature, atmospheric pressure, wind direction, and wind speed. Surprisingly, variables associated with water temperature, which is critical from a biological perspective, were not among the most important variables identified. We also included variables for the mean change in these metrics daily from 4 d before the day of stranding. These variables were among the most important in several of our models, especially the change in mean air temperature from 4 d before stranding to the day of stranding. The importance of specific variables from our random forest models can be used to guide the selection of future model predictors to estimate daily size of cold-stunning events. We plan to apply the results of this study to a predictive model that can serve as a warning system and to a downscaled climate projection to determine the potential impact of climate change on cold-stunning event size in the future.