The growing-degree-days (GDD) model for pineapple was developed to predict flowering and harvest dates; however, it has not been adapted to the climate in Japan’s growing regions, where air temperatures fluctuate over a wide range, and the prediction accuracy is low. The present study aimed to develop models for predicting flowering and harvest dates with high accuracy by analyzing a large phenological dataset from Japan’s main (Nago) and warmer (Ishigaki) production areas. The number of days between budding and flowering decreased at air temperatures of up to approximately 25°C and remained constant above 25°C. The number of days between flowering and harvest decreased until approximately 23°C. The effect of day length on both days to flowering and harvest was small. The relationship between air temperature and the developmental rate after budding to flowering and after flowering to harvest was modeled using the GDD and exponential function models, both with upper limits. The GDD model with an upper limit temperature was more accurate at predicting flowering and harvest dates compared to the conventional GDD model. In particular, the prediction accuracy of the harvest date was dramatically improved. Because the relationship between the developmental rate until flowering and the air temperature was exponential rather than linear, the exponential function model provided a more accurate prediction of the flowering date. The root-mean-square errors of the most accurate models were 3.7–6.1 days for predicting the flowering date and 6.1–10.2 days for the harvest date. We believe that these models will be useful for planning shipments of pineapple in regions with wide temperature ranges, such as Japan, and for cultivation management in response to climate change.

菠萝的生长期（GDD）模型是为了预测开花和收获日期而开发的；但它尚未适应日本种植地区的气候，气温波动范围大，预测精度较低。本研究旨在通过分析日本主要（名护）和温暖（石垣）产区的大型物候数据集，开发高精度预测开花和收获日期的模型。当气温高达约 25°C 时，发芽和开花之间的天数会减少，而在 25°C 以上则保持不变。开花和收获之间的天数减少，直到大约 23°C。这两天的日照长度对开花和收获的影响很小。使用 GDD 和指数函数模型对气温与出芽后至开花后以及开花后至收获后的发育速率之间的关系进行建模，两者均具有上限。与传统的 GDD 模型相比，具有上限温度的 GDD 模型在预测开花和收获日期方面更加准确。特别是收获日期的预测精度得到了显着提高。由于开花前的发育速率与气温之间的关系是指数关系而不是线性关系，因此指数函数模型可以更准确地预测开花日期。最准确的模型预测开花日期的均方根误差为 3.7-6.1 天，预测收获日期的均方根误差为 6.1-10.2 天。我们相信，这些模型将有助于规划日本等温度范围较宽的地区的菠萝运输，以及应对气候变化的种植管理。