We compared solid and liquid precipitation mass output from three categories of common model precipitation phase determination schemes (PPDS) to the recorded precipitation phase in a set of 45 years of 3-hour manual meteorological observations from 19 Swedish meteorological stations. In the first category of rain/snow thresholds, it was found that rain/snow air temperature threshold (ATT) is a better precipitation phase indicator than a rain/snow dew point temperature threshold. When a rain/snow ATT of 0.0 °C (a default value used in some recent models) was replaced by 1.0 °C, misclassified precipitation was reduced by almost one half. A second category of PPDS use two ATTs, one snow and one rain, with a linear decrease in snow fraction between. This category identified precipitation phase better than a rain/snow ATT at 17 stations. Using all observations from all the meteorological stations, a final category using an air-temperature-dependent snow probability curve resulted in slightly lower misclassified precipitation mass at 13 of the 19 stations. However, schemes from the linear decrease in snow fraction category had the lowest misclassified precipitation mass at four meteorological stations.