Plants have evolved mechanisms to track seasonal variation in environmental resources, enabling them to time key life-history events to appropriate seasons. While the proximate cues for flowering initiation are well documented in the temperate region, it is still unclear what the flowering cues are in the tropics, especially in the subtropics. Our study compared first flowering dates (FFDs) predicted by eight hypothesized proximate cues concerning photoperiod, mean and directional changes in solar irradiance and warm/cool temperature, and rainfall with flowering dates observed over 19 years of weekly monitoring for 16 species in a subtropical rainforest. We observed considerable inter-annual variation in the median FFDs for the study species, ranging from 21 to 101 days. The early-spring flowering species tended to have greater inter-annual variation in FFDs than the summer flowering species. For 13 study species, temperature cues best explained inter-annual variation in FFDs. Cool temperatures in the previous fall/winter and warm temperatures in the current spring (or previous summer) might trigger the onset of flowering in these 13 species. Cues associated with photoperiod and irradiance also predicted inter-annual variation in FFDs with small root mean square error (<1.5 census intervals) for 12 species but generally had higher prediction errors than temperature-related cues. Cues associated with seasonal variation in rainfall failed to predict flowering times in any species. Our results suggest that future changes in temperature may alter flowering times for most species in subtropical forests, leading to changes in ecosystem processes and biosphere feedback to the climate system.

This dataset contains 19-year data of the first flowering dates (FFD) for 16 plant species in the Fushan Forest Dynamics Plot (FDP), Taiwan. The FFD data span from September 2002 to November 2021. A network of 106 0.5-m² litterfall traps was established along four 400 m trails within the Fushan FDP. We recorded the presence of flowers in each trap and identified most to the species level. Each unique species-trap-census combination constitutes a flower record. To ensure that multiple individuals were sampled and minimize bias when estimating first flowering dates, we selected species with ≥500 flower records in total and encountered in ten or more traps in at least ten years. We estimated the FFDs for each species by calculating the median date of each trap’s first census that recorded flowers within each species-specific phenological year.