Predicting Bird Populations Using the Fractal Dimension of Forest Canopies

Forest ecosystems are vital components of the Earth’s biosphere, harboring immense biodiversity and providing invaluable ecological services. Assessing and forecasting the health of these ecosystems is imperative for informing conservation efforts and developing sustainable management practices. Remote sensing is traditionally used to evaluate forest loss and fragmentation, but this approach struggles to assess spatial and temporal trends simultaneously. To counter this deficiency, we utilized techniques from fractal analysis and geometry, which excel at characterizing the complexity of irregular data over multiple scales. First, we used the fractal dimension to examine spatiotemporal changes in forest canopies in various areas of the United States. Next we applied machine learning methods to these results in order to develop a novel forecasting technique that we used to predict bird populations in Virginia. We find that including the fractal dimension in our technique allows for more accurate predictions of bird populations when compared to standard forecasting methods. Our findings support the use of fractal analysis when examining the intricate dynamics of forest ecosystems.