摘要
空间插值方法与研究样区空间尺度会对土壤有机质和全氮含量空间变异特征预测精度产生重要影响。选取黄淮海北部未收割的夏玉米收获期50 m×50 m田块尺度(取80个土样)和1000 m×1000 m(取100个土样)农场尺度为研究样区,采用普通克里金插值法和径向基函数(RBF)人工神经网络插值法探究两种尺度研究样区土壤有机质和全氮含量空间变异特征及两种空间插值方法的预测精度。结果表明,两个研究样区土壤有机质和全氮含量分别为8.39~20.59和0.31~2.90 g·kg^(-1),块金系数为0.448~0.637,表明其呈现中等程度空间变异。土壤有机质和全氮含量在农场尺度研究样区大致呈现以西北部东南对称线含量较高向两边逐渐减少的空间分布趋势,在田块尺度上呈现东北部地区含量较高且整体向西南部递减的空间分布特征。基于RBF人工神经网络插值法的田块尺度研究样区均方根误差(RMSE)、平均绝对误差(MAE)和平均相对误差(MRE)均小于农场尺度,这表明土壤有机质和全氮含量空间分布预测精度受到采样尺度的影响,田块尺度预测精度优于农场尺度。基于RBF人工神经网络插值法在同一尺度研究样区土壤有机质和全氮含量空间分布预测的各项误差均有所减小,模型拟合决定系数R^(2)有所增加,且避免了普通克里金插值结果的“平滑效应”现象,其预测精度更高。研究结果表明,RBF人工神经网络插值法更适用于土壤有机质和全氮含量空间分布特征预测。
Precise predictions of spatial variability of soil organic matter and total nitrogen are largely depended on methods and spatial scales of the interpolation.In this study,two spatial scales including plot scale(50 m×50 m,80 soil samples)and farm scale(1000 m×1000 m,100 soil samples)were selected from nearly harvesting field of summer maize in the northern part of Huanghuaihai Plain.And two interpolating methods including ordinary Kriging and artificial neural network based on radial basis function(RBF)were compared to explore more accurate spatial variations of soil organic matter and total nitrogen content in the study area.Results of nugget to sill ratio(0.448-0.637)showed moderate spatial variations of both spatial scales with contents of soil organic matter and total nitrogen ranged from 8.39 to 20.59 g·kg^(-1) and from 0.31 to 2.90 g·kg^(-1),respectively.Spatial variations of soil organic matter and total nitrogen presented decreasing trends with higher contents distributed along the diagonal from northwest to southeast and lower contents distributed around the edges at farm scale,in contrast,higher and lower contents respectively located in northeast and southwest parts at plot scale.Concerning root mean square error(RMSE),average absolute error(MAE)and average relative error(MRE)of interpolation by artificial neural network based on RBF,smaller values indicated better predictions of spatial variations at plot scale rather than at farm scale.Without smoothing effect,increased R^(2) and decreased errors proved better performance of artificial neural network based on RBF rather than ordinary Kriging in modeling spatial variations of soil organic matter and total nitrogen at each spatial scale.It indicates that artificial neural network based on RBF is the better interpolating method for predicting spatial variations of soil organic matter and total nitrogen in this study.
作者
谢梦姣
王洋
康营
吴志涛
陈奇乐
刘琦
吴超玉
张俊梅
XIE Meng-jiao;WANG Yang;KANG Ying;WU Zhi-tao;CHEN Qi-le;LIU Qi;WU Chao-yu;ZHANG Jun-mei(College of Land and Resources,Hebei Agricultural University,Baoding 071000,China;State Key Laboratory of North China Crop Improvement and Regulation,Baoding 071000,China;College of Resources and Environmental Science/Hebei Province Key Laboratory for Farmland Eco-environment,Hebei Agricultural University,Baoding 071000,China;School of Information and Communication Engineering,Beijing University of Posts and Telecommunications,Beijing 100876,China)
出处
《生态与农村环境学报》
CAS
CSCD
北大核心
2021年第7期934-942,共9页
Journal of Ecology and Rural Environment
基金
国家重点研发计划(2017YFD0300905-6)
华北作物改良与调控国家重点实验室开放课题(NCCIR2020ZZ-20)
河北省自然科学青年基金(C2019204351)。
