玉米潜在花粉源强计算模型的建立、改进与检验

The establishment,improvement and validation of the potential pollen source strength of maize

潜在花粉源强是玉米花粉扩散模型的重要参数。为了确定玉米潜在花粉源强,本研究测定了逐日散粉量和逐时开花率,以此来确定逐时散粉量,作为模型中潜在花粉源强的输入值。以2009年溧水试验资料为基础,应用正态分布函数模拟逐时散粉量的日变化,通过气温和相对湿度参数订正逐时散粉量;利用2010年公主岭和溧水试验资料对玉米潜在花粉源强的正态分布模型进行检验。结果表明:公主岭试验中的"紫糯18"、"吉单35"和溧水试验中的"紫糯18"、"苏608"的逐时散粉量的模拟值与实测值之间的均方根误差分别为2.24×105、1.52×105、1.24×105和8.92×104粒·株-1,决定系数分别为0.330、0.545、0.365和0.606。利用相同的数据检验气温和相对湿度订正后的模型,各品种逐时散粉量的模拟值与实测值之间的均方根误差分别减少至1.25×105、9.03×104、5.14×104和4.22×104粒·株-1,决定系数则分别提高至0.656、0.718、0.779和0.843;模型效果明显改善,从而为进一步精确模拟玉米花粉扩散奠定基础。

Potential pollen source is an important parameter in the maize pollen dispersal model. In order to determine the strength of maize potential pollen source, the daily pollen emission and hourly flowering rate were measured. The product was calculated to estimate the hourly pollen emission as the model input of the potential pollen source. On basis of the data obtained from Lishui experiment in 2009, the normal distribution function was used to simulate the diurnal pat- tern of hourly pollen emission and then air temperature and relative humidity were introduced to modify this model. Firstly, the data obtained from Gongzhuling and Lishui experiments in 2010 were used to validate the normal model. The measured hourly pollen emissions of ＂Zinuo-18＂ and ＂Jidan-35＂ in Gongzhuling experiment and ＂Zinuo-18＂ and ＂Su-608＂ in Lishui experiment were compared with the simulated values. Results showed that the root mean squared errors （RMSE） between the measured and simulated values of the four maize varieties were 2.24 ×10^5, 1.52×10^5, 1.24×10^5 and 8.92×10^4 grain · plant-1, and the determination coefficients （R2） were 0.330, 0.545, 0.365 and 0.606. Then the same data were used to validate the model modified by air temperature and relative humidity. Results showed that RMSE between the measured and sim- ulated values of the four maize varieties decreased to 1.25 ×10^5, 9.03 ×10^4, 5.14×10^4, and4.22×10^4 grain· plant-1, and R2 values increased to 0.656, 0.718, 0.779 and 0.843. Model was obviously improved, which could greatly increase the accuracy of the maize pollen dispersal model.