| 不同采样尺度下土壤碱解氮空间变异性研究——以榆树市农田土壤为例 |
| |
| 引用本文： | 刘吉平,刘佳鑫,于洋,田学智,徐艳艳.不同采样尺度下土壤碱解氮空间变异性研究——以榆树市农田土壤为例[J].水土保持研究,2012,19(2):106-110,115. |
| |
| 作者姓名： | 刘吉平 刘佳鑫 于洋 田学智 徐艳艳 |
| |
| 作者单位： | 吉林师范大学旅游与地理科学学院,吉林四平,136000 |
| |
| 基金项目： | 吉林省科技厅科技支撑计划项目(20080207);国家农业科技成果转化资金项目(2009GB2B100095);吉林师范大学研究生创新科研计划项目(201111) |
| |
| 摘 要： | 在精准农业的实施过程中,研究如何用较少的样本来反映田间信息的空间变异规律,再用科学的插值方法进行插值和预估是精准农业研究中的一个关键问题。以东北典型黑土区——吉林省榆树市为研究区域,在榆树市弓棚镇13号村内选择相对平整的地块进行土壤采样并测试其土壤养分。在对原始采样格网点按一定的样点间隔和布局进行抽取的基础上,利用克里格插值方法和BP神经网络方法分别进行空间插值,比较不同采样尺度(40m×40m,56m×56m,80m×80m,113m×113m,160m×160m五个尺度)对空间插值精度的影响。结果表明:(1)随着采样尺度的增大,碱解氮的空间结构系数C/(C0+C)有减小的趋势,表明采样间距以内的不可估计误差逐渐增大,其空间结构的表现能力在逐渐减弱;(2)Kriging插值精度总体优于BP神经网络,随着采样尺度的增加,两种模型的模拟精度都有所下降,BP神经网路的插值精度和Kriging模型的插值精度的差距逐渐减小;(3)两种模型在113m×113m尺度上插值精度都发生了突变,如考虑碱解氮的空间变异规律和经济因素,碱解氮的最佳采样尺度应在80～113m。
|
| 关 键 词： | 克里格插值 BP神经网络 土壤碱解氮 采样尺度 空间变异性 |
Study on Spatial Variability of Available Nitrogen in Different Sampling Scale——A Case Study on Cropland Soil in Yushu City |
| |
| LIU Ji-ping,LIU Jia-xin,YU Yang,TIAN Xue-zhi,XU Yan-yan.Study on Spatial Variability of Available Nitrogen in Different Sampling Scale——A Case Study on Cropland Soil in Yushu City[J].Research of Soil and Water Conservation,2012,19(2):106-110,115. |
| |
| Authors: | LIU Ji-ping LIU Jia-xin YU Yang TIAN Xue-zhi XU Yan-yan |
| |
| Institution: | (College of Tourist and Geoscience,Jilin Normal University,Siping,Jilin 136000,China) |
| |
| Abstract: | During the implementation of precision agriculture,it is a key issue to search for how to use the fewer samples to reflect the spatial variability of the field of information,and then use the scientific method of interpolation and the interpolation to estimate precision agriculture.In this paper,we choose Yushu City,a typical black soil area in northeaster China,which lies in Jilin Province as the study area to select relatively flat soil plots to sample and test soil nutrient in No.13 village of Gongpeng Town in Yushu City.The original sampling grid networks in accordance with certain intervals samples were taken and layed out based on the use of Kriging interpolation method and BP neural network for spatial interpolation,respectively,compared different scales(40 m×40 m,56 m×56 m,80 m×80 m,113 m×113 m,160 m×160 m) of spatial interpolation accuracy.The results show that:(1) with the increase of sample size,the spatial structure coefficient C/(C0+C) of the available nitrogen tends to decrease,indicating that the unpredictable error increases within sampling interval,the performance of its spatial structure the ability to gradually declines;(2) Kriging interpolation accuracy was better than BP neural network.With the increase of sample scale,the simulation accuracy of the two models have decreased,but BP neural network model and Kriging interpolation accuracy and precision of the difference are very small;(3) terms from the Kriging and BP neural network models,from 80 m×80 m to 113 m×113 m scale change process,the average relative error is suddenly decreased,indicating that in the 113 m×113 m precision interpolation scale mutations,considering the spatial variability of available nitrogen and economic factors,the optimal sampling size of available nitrogen should be between 80 m to 113 m. |
| |
| Keywords: | Kriging BP neural network available nitrogen of soil sampling scale spatial variability |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
