Cost-effective and accurate monitoring of flowering across multiple tropical tree species over two years with a time series of high-resolution drone imagery and deep learning

Abstract

<p>Detection of flowering and quantification of flowering <a href="https://www.sciencedirect.com/topics/earth-and-planetary-sciences/phenology" title="Learn more about phenology from ScienceDirect's AI-generated Topic Pages">phenology</a> are key to monitoring the reproduction of tropical trees and their response to global change. However, effective monitoring of flowering over various scales from individuals to forest ecosystem levels is lacking due to the relatively small sizes of flowers, diverse flowering strategies across species, and the short duration of flowering, making accurate flower detection difficult. Drone-based surveys require less time and human resources than traditional ground-based flower surveys and thus may be able to help address this in a cost-effective manner but remain underexplored in species-rich <a href="https://www.sciencedirect.com/topics/earth-and-planetary-sciences/tropical-forest-ecosystem" title="Learn more about tropical forest ecosystems from ScienceDirect's AI-generated Topic Pages">tropical forest ecosystems</a>. Here, we developed a method that integrated the <a href="https://www.sciencedirect.com/topics/computer-science/residual-network" title="Learn more about Residual Networks from ScienceDirect's AI-generated Topic Pages">Residual Networks</a> 50 (ResNet50) <a href="https://www.sciencedirect.com/topics/computer-science/deep-learning" title="Learn more about deep learning from ScienceDirect's AI-generated Topic Pages">deep learning</a> algorithm with high resolution imagery (<em>c.</em> 0.05 m) from monthly drone surveys done in a 50-ha tropical forest plot on Barro Colorado Island (BCI), Panamá, over 2018–2020 to detect a diversity of flowering species in this tree community and to track the timing of flowering throughout the year. We built a comprehensive training library of canopy components (flower, leaf, branch, and shade) from this forest plot throughout the study period, trained a single <a href="https://www.sciencedirect.com/topics/computer-science/deep-learning-model" title="Learn more about deep learning model from ScienceDirect's AI-generated Topic Pages">deep learning model</a> across all drone imagery, and validated it using five-fold cross validation at the pixel scale. We further generated image- and tree-crown-specific <a href="https://www.sciencedirect.com/topics/computer-science/supervised-classification" title="Learn more about supervised classifications from ScienceDirect's AI-generated Topic Pages">supervised classifications</a> to evaluate the deep learning model at the tree-crown scale. Our deep learning method accurately classified flowers (User’s accuracy = 95.3 %, Producer’s accuracy = 85.8 %) while maintaining high <a href="https://www.sciencedirect.com/topics/computer-science/predictive-power" title="Learn more about predictive power from ScienceDirect's AI-generated Topic Pages">predictive power</a> for the other three classes (Overall accuracy = 98.4 %). Our results also demonstrated high consistency against tree-crown-specific supervised classifications for flower (r² = 0.85), leaf (r² = 0.84), and branch (r² = 0.92) components, with lower agreement observed for the shade component (r² = 0.59). These results demonstrate the effectiveness of our method in advancing fine-scale flower monitoring in the tropics, with potential to be extended to other regions or other <a href="https://www.sciencedirect.com/topics/earth-and-planetary-sciences/telemetry" title="Learn more about remote sensing from ScienceDirect's AI-generated Topic Pages">remote sensing</a> platforms with frequent high-resolution monitoring. The method will allow us to better monitor flowering in tropical forests and improve our understanding of how phenology and reproductive success may be affected by climate change.</p><ul></ul>