Abstract
Floral phenology is a key indicator of climate change. However, constructing a local level spatiotemporal record of cherry blossom blooming events is challenging because of the limited coverage of traditional stationary observations. To address this issue, citizen-based observation programs and remote sensing applications have been implemented. However, these strategies have still been limited by insufficient spatial and temporal observation data. In this study, we introduced a novel semi-automatic observation system framework for cherry blossom blooming by combining geographically and temporally dispersed street-level imagery and deep-learning models. We developed a YOLOv4 model for cherry blossom detection from street-level photos obtained from Mapillary, one of the main social sensing data repositories. The detection model achieved an overall accuracy, recall, and precision of 86.7%, 70.3%, and 90.1%, respectively. By using observation coordinates and dates attached to Mapillary photos, we mapped the probability of cherry blossoms occurrence daily using a spatial grid of dimensions of 10 m × 10 m. With sufficient observations, the start, peak, and cessation of blooming were estimated using time series analysis. A case study conducted in 2022 at the main campus of Saitama University confirmed the potential for automatically mapping the presence of cherry blossoms and their blooming timing with local variations. Given that our approach solely relied on geotagged street-level photos that could be taken by anyone with no prior knowledge of cherry tree species identification, it is likely to be easier to build spatial and temporal blooming records than using conventional stationary or citizen-based observations. This approach can be applicable to other study areas, as well as observing other species phenology and understanding the response of local environments.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have