The characteristics of multi-source monitoring data and the causes of sudden temperature decline before and after the abrupt snowfall in southern Jiangsu Province on 28 March 2020, are investigated with the data of automatic meteorological station, dual polarization radar， microwave radiometer, NCEP reanalysis data and simulation results of mesoscale model WRF. The research shows that the time and area changes of temperature drop are consistent with the beginning time and moving direction of heavy precipitation in this sudden snowfall. The 0℃ layer height decline and the increase of liquid water content inversed by microwave radiometer can be used as the approach criterion to judge the surface temperature and precipitation phase transition. The time of ground temperature decrease lags behind the temperature drop in the 1-3 km height for about 2-3 hours. The radar reflectivity factor, the correlation coefficient and classified products of dual polarization radar products have obvious characteristics in the judgment of rain-snow conversion and snow expansion direction during this snowfall. The light zone and shape change of the 0℃ layer are consistent with the cooling zone and movement direction of the lower layer. In the correlation coefficients, the asymmetrical characteristic regions and morphological changes of different phase particles reflect the changes in height, thickness and movement direction of the melting layer. Asymmetric characteristics of mixed phase layer and wet snow monitoring signals of classified products appear 1-2 hours earlier than snowfall. The diagnostic analysis shows that the latent heat of melting during the subsidence of ice crystals and snowflakes and the latent heat of evaporation through the dry area have a positive contribution to the short and rapid decrease of temperature in the lower layer. The temperature change presents the spatio-temporal evolution characteristics from above to below and from west to east. The earliest temperature decline occurs at the height of 850-700 hPa. The near-surface temperature decline is delayed by about 2 hours. Therefore, this result suggests that the multi-source monitoring data can provide better decision support for the rain-snow conversion process, especially in the short-time and nowcasting forecasts.