雄安新区土壤氮空间分布预测方法

Prediction methods of soil nitrogen spatial distribution in Xiong'an New Area

土壤全氮(soil total nitrogen,STN)和速效氮(available nitrogen,AN)的空间分布是影响植被生产力的重要因素。以雄安新区东部91个取样点的上下两层共182个土壤样品为研究对象,采用土地利用方式为辅助变量的克里格(kriging combined with land-use,KLU)、普通克里格(ordinary kriging)、反距离加权法(inverse distance weighting,IDW)、径向基函数法(radial basis function,RBF)4种方法对STN含量、AN含量、土壤全氮密度(soil total nitrogen density,STND)、速效氮密度(available nitrogen density,AND)的空间分布分布进行预测,并以均方根误差(root mean square error,RMSE)、平均绝对误差(mean absolute error,MAE)和相关系数(R)为指标比较分析4种插值方法的预测效果,进而对研究区STN、AN空间分布格局和区域储量进行评价。结果表明:在垂直分布上,STN、AN含量和密度均随土层深度的增加而减小;水平分布上,不同土地利用方式STN、AN含量和密度之间均具有显著差异(P<0.05);预测STN、AN含量和密度空间分布的最优插值方法是KLU;在表层土壤中,由KLU得到的R分别为0.574(STN)、0.426(AN)、0.555(STND)和0.407(AND),比其他3种插值方法平均提高0.350、0.390、0.304、0.310;RMSE分别为0.2255(STN)、21.5902(AN),0.0901(STND),8.5365(AND),比其他3种插值方法平均减少0.0481、3.7874、0.0166、1.4236;MAE最接近于0;在下层土壤中KLU也具有相同的优势。

The spatial distributions of soil total nitrogen(STN)and available nitrogen(AN)contents are important factors influencing vegetation productivity.A total of 182 soil samples were collected from the upper and lower layers of 91 sampling points in the eastern part of the Xiong'an New Area.The distributions of STN content,AN content,soil total nitrogen density(STND),and available nitrogen density(AND)were predicted with four methods,i.e.,kriging combined with land-use(KLU),ordinary kriging,inverse distance weighting(IDW),and radial basis function(RBF).The predictive accuracy of these methods was evaluated by the root mean square error(RMSE),mean absolute error(MAE),and correlation coefficient(R).Furthermore,the spatial distribution patterns of STN and AN contents and the regional reserves were assessed.The results showed that the vertical distribution of STN,AN content and density decreased with increasing soil depth.From a horizontal perspective,there were significant differences for STN and AN content and density among different land-use types(P<0.05).The KLU was the opti-mal interpolation method for predicting the spatial distribution of STN,AN content and density.In the upper layer of soil,the R values of KLU were 0.574(STN),0.426(AN),0.555(STND)and 0.407(AND),respectively,which were 0.350,0.390,0.304 and 0.310 higher than that of the other three interpolation methods on average.The RMSE of KLU was 0.2255(STN),21.5902(AN),0.0901(STND),and 8.5365(AND),respectively,with an average reduction of 0.0481,3.7874,0.0166,and 1.4236 compared with the other three interpolation methods;while the MAE of KLU was the best.KLU showed advantages also in the deeper soil layer.