We investigated meteorological conditions for frost occurrence using routine observation data during 1991–2010 from manned meteorological stations in Japan. The majority of the frost occurrence was observed when the daily minimum air temperature was in the range of − 5 °C to 5 °C. Due to strong radiative cooling of a frosted surface, frost was observed even when the daily minimum air temperature was above the dew-point temperature at a normal meteorological observation height. We proposed an analytical scheme to evaluate the probability of frost occurrence under given meteorological conditions. This scheme was based on a probability distribution of observed frost occurrence in a three-dimensional space of meteorological variables. The obtained distribution centered on − 1.9 °C for the daily minimum air temperature and 73% for the relative humidity, with an exponential decay with wind speed. The probability was halved when the daily minimum air temperature and relative humidity change by 3.3 °C and 24 percent-points, respectively, from the distribution center. Since this scheme gives probability of frost occurrence, it is advantageous for risk assessment under given meteorological conditions. Recent popularization of gridded meteorological datasets will help to expand the potential usage of the scheme.

中文翻译：

---

日本气象观测的霜冻发生概率

我们使用日本载人气象站 1991-2010 年的常规观测数据调查了霜冻发生的气象条件。当日最低气温在 - 5 °C 到 5 °C 的范围内时，观察到大部分霜冻发生。由于霜面辐射冷却强烈，即使日最低气温高于正常气象观测高度的露点温度，也会出现霜冻。我们提出了一种分析方案来评估给定气象条件下霜冻发生的概率。该方案基于在气象变量的三维空间中观测到的霜冻发生的概率分布。获得的分布集中在 - 1.9 °C 的每日最低气温和 73% 的相对湿度，随风速呈指数衰减。当配送中心的每日最低气温和相对湿度分别变化 3.3°C 和 24 个百分点时，概率减半。由于该方案给出了霜冻发生的概率，有利于在给定气象条件下进行风险评估。最近网格化气象数据集的普及将有助于扩大该计划的潜在用途。