Abstract:

To avoid freezing damage to crops after frost, we used grassland temperature to predict frost. Based on hourly meteorological data from Lianyungang Meteorological Observatory during 2014-2016, including temperature, 0 cm ground temperature, dew point temperature, water vapor pressure, air pressure and 2 min average wind speed, which were the key factors affecting the grassland temperature in Lianyungang, we took these 6 elements as attribute features, and constructed a training sample set with grass temperature as a marker. The KNN data mining algorithm is combined to construct a grass temperature prediction model, and the frost occurrence is judged according to the grass temperature. The results showed that: the grass temperature prediction model based on the algorithm could achieve better prediction effect, the forecast average error was 1.2℃; according to the grass temperature, the accuracy of frost forecast reached 90.2%, which had a good indication especially for the prediction of the initial and the last frost. Therefore, the introduction of grass temperature as a prediction index of frost is of significance for avoiding crop damage.

Key words: crop, frost, grassland temperature, key factor, K-Nearest Neighbor (KNN), prediction model