摘要:土壤磷素含量是反映农业生态系统土壤肥力的重要指标。准确预测磷素空间异质性是评价土壤生产力和质量的关键。本研究采用反距离加权法（IDW）、径向基函数法（RBF）、普通克里金法（OK）、全局多项式法（GPI）、局部多项式法（LPI）、地理加权回归（GWR）和地理加权回归克里金法（GWRK）等插值方法，分别预测了三江平原白浆土典型灌区八五三、七里沁以及大兴灌区土壤磷素分布，并运用交叉验证法，通过平均误差（ME）、均方根误差（RMSE）和改进效果（RI）对各种方法精度进行比较，以期确定同一土壤类型不同采样密度土壤中磷素空间异质性最佳插值方法。对比7种插值方法，在空间插值平滑性方面，LPI、GPI、GWR、GWRK表现较好，在插值速度方面，IDW、RBF、LPI、GPI、OK较快，GWR和GWRK方法运算复杂、速度较慢。IDW、RBF等6种方法与OK相比，根据RI判定，GWRK方法提高了磷素空间分布模拟精度，IDW、GPI和LPI方法降低了磷素空间分布模拟精度，RBF方法在提高磷素空间分布模拟精度上表现不一致。采样密度会影响预测结果，对于本文而言，不论采样密度如何，GWRK方法均方根误差（RMSE）均最低，为最佳插值方法，而RBF方法是在采样密度较低时一种可选方法。GWRK法在本文是最佳的插值方法，但其结果会受到辅助变量多少和各变量之间是否存在共线性的影响。
关键词:白浆土/
磷素/
插值方法/
交叉验证/
空间自相关性/
适用性
Abstract:In the late 1990s, the "Dryland to Paddy" project was implemented in the Sanjiang Plain. After planting rice in the Albic soil, the barrier soil layer turns into a favorable soil layer, the low-yield soil becomes high-yield soil, and the Albic soil phosphorus pool increases. After flooding, the availability of phosphorus (closed storage phosphorus[O-P] and iron-bound phosphorus[Fe-P]) increases with the decrease in soil redox potential (Eh) and the increase in pH, which substantially affects soil phosphorus heterogeneity. Therefore, we urgently need an optimal interpolation method to improve the prediction accuracy of total phosphorus in the Albic soil of typical irrigation areas of the Sanjiang Plain. This will help evaluate the impact of climate change and land use on the soil phosphorus pools and provide a reference for estimating future soil phosphorus pools. This study used the inverse distance weighting (IDW) method, radial basis function (RBF), ordinary Kriging (OK), global polynomial method (GPI), local polynomial method (LPI), geographic weighted regression (GWR), and geographic weighting regression to Kriging (GWRK) to predict the distribution of soil phosphorus in the Bawusan, Qiliqin, and Daxing irrigation areas of the Sanjiang Plain. The cross-validation method was used to obtain the mean error (ME), root mean square error (RMSE), and relative improvement (RI) to compare the accuracies of the various methods to determine the best interpolation method for assessing the spatial heterogeneity of phosphorus in the same soil type with different sampling densities. Based on the assumptions of regression analysis, this study incorporated 24 environmental variables for exploratory regression analysis, including elevation, pH, organic matter, exchangeable sodium, total nitrogen, available phosphorus, available copper, cultivated layer bulk density, and rice yield. According to the regression results, the auxiliary variables that were significantly correlated with phosphorus were selected for least square analysis. Finally, exchangeable sodium, cation exchange capacity, and available phosphorus were selected as auxiliary variables for the Bawusan irrigation area; organic matter, available zinc, and available boron were selected as auxiliary variables for the Qiliqin irrigation area; and cation exchange capacity, available zinc and copper were selected as auxiliary variables for the Daxing irrigation area. Compared to OK, RI indicated that the GWRK method with environmental auxiliary variables significantly improved the simulation accuracy of the spatial distribution of phosphorus. The IDW, GPI, and LPI methods reduced the accuracy of phosphorus spatial distribution simulation, whereas the RBF method was inconsistent. When comparing the mapping effect and interpolation speed of the seven interpolation methods, LPI, GPI, GWR, and GWRK had better mapping effects, whereas IDW, RBF, LPI, GPI, and OK were faster. The GWRK method had a better mapping effect, but it should be combined with environmental auxiliary variables, and the operation was complicated and slow. Sampling evenness also affected the prediction results. Nonetheless, GWRK had the lowest ME and RMSE, indicating that it is the best interpolation method. RBF is an optional method when the sampling evenness is lower. GWRK is the best interpolation method, but the results are affected by the number of auxiliary variables and collinearity between the variables.
Key words:Albic soil/
Phosphorus/
Interpolation method/
Cross validation/
Spatial autocorrelation/
Applicability
PDF全文下载地址:
http://www.ecoagri.ac.cn/article/exportPdf?id=8363f9db-ee6e-47c5-b4af-8e002f4c9ab7
