Changes in flowering phenology may affect species interaction in ecosystems like pollination, and lead to ecosystem-wide consequences. Therefore, it is important to monitor phenology and identify vulnerable areas. It is known that the major factor affecting spring phenology is temperature. The purpose of this study is to explain how the timing of flowering phenology varies in a city and how temperature affects phenology by developing a high-resolution dataset. The followings are research questions of this study: 1) How temperature gradient may influence flowering dates? 2) How can we utilize citizen science data to analyze phenology? We targeted cherry trees, generally distributed throughout Korea. We collected and processed citizen observations of cherry blossoms and temperature data in 1km resolution. Data was collected through Cada, a citizen science application, by asking participants to take pictures of cherry blossoms from March to May 2021 and 2022. S-DoT, urban sensors in Seoul, were used as temperature data. We considered first observed method, percentile of the distribution, and Weibull-parameterized estimates to estimate the flowering date of each grid cell. The hierarchical model was adopted to make inferences. We conducted correlation analysis and regression analysis to explore the relationship between temperature and flowering phenology.
Results/Conclusions
We collected 8,780 geotagged photos of cherry blossoms and the observation dates were varied from DOY(day-of-year) 79 to 120, in 2021. The mode of the data was 91 and the median was 95. As a result of statistical analysis, we found that the mean temperature of spring and flowering date has a weak but negative correlation. Also, the average flowering date between cold areas and warmer areas has significant differences. These results provide empirical evidence that as temperature increases, the flowering date of cherry blossom in the region may be faster. In terms of utilizing citizen science data, we found some bias and errors like false observations that observed different flowers and unbalanced sample sizes among regions involved in collected data. Therefore, it is suggested that careful design of citizen science projects should be made considering the overall process of data collection, filtering, and analysis process. Also, a hierarchical model that accounts for the variation of data may be helpful for handling citizen science data in case of spatially or temporally unbalanced samples.
