Multi-cyclone analysis and machine learning model implications of cyclone effects on forests

Past studies of cyclones (hurricanes, typhoons, tropical cyclones) disturbance showed that meteorological, topographical, and biological factors affect the patterns forest intensity but left open extent to which these findings were representative across different global cyclone regions. Using remote sensing data machine learning models, we examined how change over spatial scales assessed their consistency four major cyclones: Katrina (August 2005), Rita (September Yasi (February 2011), María 2017). Our results revealed best explained pattern varied Wind speed precipitation dominant contributing variation in impacts Katrina; terrain features, especially elevation, most Rita; pre-disturbance vegetation condition was significant predictors effects Yasi; played equal roles explaining María. A 40 m/s (144 km/h) wind threshold proposed split low- high-level intensity. Other than speed, few generalizations can be made on features multiple We built several generalized hurricane impact worked well with test from used for model development (R2 = 0.89). However, models did not have good predictions other cyclones, such as Michael (October 2018) Laura 2020). This study each interacted landscape a unique way challenges remained building model.