The precipitation phase (rain or snow) strongly influences land-surface water and energy budgets, yet the thresholds used to distinguish rain from snow vary widely across regions and are not determined by air temperature alone. Because atmospheric humidity affects phase transitions through the cooling of falling particles, wet-bulb temperature has been proposed as an alternative to air temperature. However, spatial differences and long-term changes in wet-bulb thresholds across Japan remain insufficiently characterized. Using hourly surface meteorological observations from 158 stations across Japan (1989&ndash;2024), we analyzed the surface air temperature (<em>T</em>_s,50) and wet-bulb temperature (<em>T</em>_w,50) at which the probability of snowfall was 50 %, along with their difference, &Delta;<em>T</em>₅₀ (= <em>T</em>_s,50 &minus; <em>T</em>_w,50). Rather than treating &Delta;<em>T</em>₅₀ as a direct measure of cooling by evaporation and sublimation, we use it as a diagnostic index of humidity-related controls on the precipitation phase. This decomposition treats the conventional air-temperature threshold as the sum of a wet-bulb component and a humidity-related component (<em>T</em>_s,50=<em>T</em>_w,50 + &Delta;<em>T</em>₅₀). Across 115 stations, the mean <em>T</em>_s,50 was 2.63 &deg;C and the mean <em>T</em>_w,50 was 1.31 &deg;C. Inter-station variability in <em>T</em>_w,50 was less than half that in <em>T</em>_s,50, suggesting that a spatially uniform wet-bulb threshold is more broadly applicable. All three indices decreased with latitude and elevation, although these factors explained only a small fraction of the variability. &Delta;<em>T</em>₅₀ was associated with relative humidity during precipitation, with larger values at drier sites. Stratifying snowfall frequency curves by relative humidity, station pressure, wind speed, and precipitation amount revealed systematic shifts in the rain&ndash;snow transition temperature. These shifts were generally larger for air temperature than for wet-bulb temperature, suggesting that the latter accounts for much of the variability associated with meteorological conditions. At 44 stations with continuous records, <em>T</em>_s,50, <em>T</em>_w,50, and &Delta;<em>T</em>₅₀ decreased at more sites than they increased between 1990&ndash;2004 and 2005&ndash;2020. At more than half of the stations, wet-bulb and humidity-related components changed in opposite directions, indicating that at these sites the long-term changes in <em>T</em>_s,50 cannot be explained by a single factor. Because wet-bulb temperatures are widely available in reanalyses, climate simulations, and land-surface models, the proposed decomposition provides a practical basis for interpreting regional differences and future changes in the precipitation phase and associated hydrological responses.