论文部分内容阅读
本文应用全国、31个省、6个典型地区和16个典型县的数据对粮食估产的“通道-概率模型”进行了系统性的验证和讨论。研究结果如下:(1)国家级估产由于地域空间尺度足够大,不同地区气象条件对产量影响的互补性强,所以估产误差小,因此国家级可以不使用小趋势修正和气候年型修正;省级、地区级和县级的估产由于同处一个气候区,因此气象条件对产量影响的互补性不强,必须使用小趋势修正和气候年型修正,县级估产还必须增加根据作物适时长势和专家经验的修正。(2)小趋势修正有两个公式:当预测误差小于10%时,使用Y×(1-K)修正;当预测误差大于10%时,使用Y/(1+K)修正。(3)估产单元气候年型可以自动划分,一般分为5级,波动大的预测单元可以使用7级,其中超丰年和超欠年的修正参数必须根据实时气象条件和作物实时长势具体确定。(4)研究表明:“通道-概率”估产理论和方法是科学的、实用的和准确的;在小趋势修正和气候年型修正基础上,如能结合作物长势调查和当地专家经验,估产误差可以达到3%以下。
This paper systematically verifies and discusses the “channel-probabilistic model” of grain yield estimation using the data of the whole country, 31 provinces, 6 typical regions and 16 typical counties. The results are as follows: (1) The national estimation can not use the small trend correction and the annual climate correction because the geo-spatial scale is large enough and the meteorological conditions in different areas have strong complementarities with the output. Due to the co-existence of a climatic zone, the complementarity of the meteorological conditions on the yield is not strong, and the small trend correction and the annual climatic revision must be used. The estimation of the county level must also be increased according to the timely development of the crop and Revision of expert experience. (2) There are two formulas for small trend correction: Y × (1-K) correction when the prediction error is less than 10%; and Y / (1 + K) correction when the prediction error is greater than 10%. (3) The annual climate of the estimation unit can be automatically divided into five levels. The prediction unit with large fluctuation can be used with seven stages. The correction parameters of the ultra-rich year and the extra-year must be determined according to the real-time weather conditions and the real-time crop growth. (4) The research shows that the theory and method of “channel-probability” estimation are scientific, practical and accurate. On the basis of small trend correction and annual climatic amendment, if we can combine the survey of crop growth with the experience of local experts, Estimated error can reach 3% or less.