县域耕地土壤速效钾空间预测方法的比较

【目的】筛选出测土配方施肥项目中最适宜的土壤速效钾空间预测方法，为准确预测土壤养分的空间分布提供依据。【方法】以陕西汉中汉台区耕地土壤速效钾含量为研究对象，分别采用反距离加权插值法、径向基函数插值法和普通克里格插值法3种方法，在选取3种方法最优模型参数的基础上，对试验区域内2 022个土壤速效钾样点进行空间插值,并对结果进行验证及评价。【结果】通过对3种空间插值方法进行交叉验证，径向基函数插值法均方根误差为267.349，小于其他2种插值方法；其相关系数为0.687，大于其他2种插值方法。利用分区符合度法和插值结果图进一步验证，结果表明，相对于反距离加权插值法和普通克里格插值法，径向基函数插值法更能反映试验区域土壤速效钾的分布格局。【结论】径向基函数插值法较反距离加权插值法和普通克里格插值法更适宜于试验区耕地土壤速效钾的空间预测；在测土配方施肥项目实施过程中，利用分区符合度验证法对土壤养分空间预测方法进行选择是切实可行的。

Comparison of spatial interpolation methods of available potassium in Hantai County

【Objective】 The study was to accurately predict spatial distribution,and to select suitable spatial interpolation methods of available potassium in Hantai county.【Method】 Taking soil available potassium of 2 022 sample sites as an example,based on optimum parameters selection of 3 methods and interpolation results validation and analysis,three interpolation methods were compared including OK(Ordinary Kriging),RBF(Radial Basis Functions) and IDW(Inverse Distance Weighted).【Results】 Through the cross validation of three spatial interpolation methods,the RMS of Radial Basis Functions was 267.349,smaller than the other two interpolation methods,and the correlation coefficient of the Radial Basis Functions was 0.687,bigger than the other two interpolation methods.Validation with the zonal similarity validation and interpolation graph indicated that RBF can reflect soil available potassium distribution pattern much better than OK and IDW in experimental zone.【Conclusion】 RBF,compared with OK and IDW,is more suitable in spatial interpolation of available potassium in Hantai County,and it also indicates that the method of zonal similarity validation is feasible in predicting spatial distribution of soil nutrients during the implementation of the program of soil testing and fertilizer recommendation.