冬小麦农田不同高度极端气温预测[ ]

利用2019年1月—2022年6月冬小麦农田小气候自动观测站观测的数据和齐河县国家基本气象观测站同期观测资料,采用多元线性回归和BP神经网络方法,建立冬小麦农田30 cm、60 cm、150 cm日最高和日最低气温预测模型。结果表明:两种模型对农田气温的预测效果均较好,阴天条件下150 cm最高气温预测效果最好；晴天条件下30 cm最高气温预测效果最差。两模型模拟结果分层次看,农田气温的模拟精度150 cm>60 cm>30 cm；分天气类型看,多元回归模型农田各层气温的模拟精度阴天>多云>晴天,BP神经网络模型农田30 cm、60 cm最高气温的模拟精度多云>阴天>晴天,农田30 cm最低气温的模拟精度多云与阴天相同,均大于晴天,农田60 cm最低气温的模拟精度晴天>多云>阴天,农田150 cm最高及最低气温的模拟精度晴天与多云相同,均大于阴天；分要素看,30 cm最低气温的模拟精度高于最高气温、60 cm和150 cm最高气温的模拟精度高于最低气温。通过比较,BP神经网络模型的预测精度比多元线性回归模型的预测精度高。两种模型均能满足冬小麦农田气温的预测需求。

Forecast of Extreme Temperature at Different Heights in Farmland Winter Wheat

Using microclimate data of winter wheat field and surface meteorological observation data in Qihe from January 2019 to June 2022, the maximum and minimum temperature forecast model at 30 cm、60 cm、150 cm heights in farmland winter wheat was established with BP neural network and multiple linear regression. The results indicated that: :The prediction effect of the two models on farmland temperature is relatively good, the prediction effect of maximum temperature at 150 cm is best on overcast days; the prediction effect of maximum temperature at 30 cm is worst on sunny days. The simulation results of the two models are viewed in layers,the simulation accuracy of farmland temperature is 150 cm>60 cm>30 cm; By weather type, the simulation accuracy of? each layer temperature of farmland predicted by multiple regression model is overcast > cloudy > sunny; the simulation accuracy of the maximum temperature of 30 cm and 60 cm predicted by BP neural network model is cloudy > overcast > sunny,the simulation accuracy of the minimum temperature of 30 cm in farmland is the same as that on cloudy and overcast days, both greater than that on sunny days; the simulation accuracy of the minimum temperature of 60 cm in farmland is that on sunny > cloudy > overcast; the simulation accuracy of the maximum and minimum temperature of 150 cm in farmland is the same as that on sunny and cloudy days, both greater than that on overcast days.View by elements,the simulation accuracy of 30 cm minimum temperature was higher than that of maximum temperature, 60 cm and 150 cm maximum temperature is higher than that of minimum temperature.The prediction accuracy of BP neural network model is higher than that of multiple linear regression model. Both models could meet the forecast demand of winter wheat field temperature.