自然资源学报 ›› 2018, Vol. 33 ›› Issue (3): 489-503.doi: 10.11849/zrzyxb.20170067

• 资源评价 • 上一篇    下一篇

黄河三角洲典型地区耕地土壤养分空间预测

厉彦玲a, b, 赵庚星a, *   

  1. 山东农业大学 a. 信息科学与工程学院, b.资源与环境学院,山东 泰安 271018
  • 收稿日期:2017-01-20 修回日期:2017-04-11 出版日期:2018-03-20 发布日期:2018-03-20
  • 通讯作者: *赵庚星(1964- ),男,教授,博士,研究方向为土地(土壤)资源与信息技术。E-mail: zhaogx@sdau.edu.cn
  • 作者简介:厉彦玲(1980- ),女,山东莒南人,讲师,博士研究生,研究方向为土地(土壤)资源与信息技术。E-mail: liyanling@sdau.edu.cn
  • 基金资助:
    “十二五”国家科技支撑计划项目(2013BAD05B06,2015BAD23B0202),国家自然科学基金(41271235)

Spatial Prediction of Cultivated Land Soil Nutrients in Typical Region of Yellow River Delta

LI Yan-linga, b, ZHAO Geng-xingb   

  1. a. College of Information Science and Engineering, b. College of Resources and Environment, Shandong Agricultural University, Tai’an 271018, China
  • Received:2017-01-20 Revised:2017-04-11 Online:2018-03-20 Published:2018-03-20
  • Supported by:
    The Sci-tech Pillar Project of the “Twelfth Five Year” Plan of China, No. 2013BAD05B06 and 2015BAD23B0202; National Natural Science Foundation of China, No. 41271235.

摘要: 掌握土壤养分的分布特点是实现养分优化管理的重要基础。论文选择黄河三角洲典型地区山东省垦利县为研究区,通过田间采样与实验室化验分析获取了1 278个样本(0~20 cm)的土壤碱解氮、有效磷、速效钾数据。在经典统计分析的基础上,用地统计学方法分析了土壤养分的空间变异特征,并拟合了养分的变异函数模型。利用普通克里格法(OK)、反距离权重法(IDW)、泛克里格法(UK)、径向基函数法(RBF)和局部多项式法(LP)5种方法进行空间插值,并采用独立数据集验证对插值结果进行精度评价,进而分析了各养分空间分布规律。为深入探索各方法的适用性规律,基于AN数据设计了离散、随机、聚集3种空间分布模式的数据,利用各模型的自动优化进行试验,对比分析了不同插值方法在土壤养分空间预测中的自适应性。结果表明:1)研究区碱解氮、有效磷、速效钾均为中等强度的空间变异和中等程度的空间自相关,其变异函数模型分别为球状模型、指数模型和球状模型,决定系数依次为0.951、0.892和0.787;2)在空间分布上,土壤碱解氮、有效磷、速效钾含量与地形和土地利用类型等有关,西南部地势较高,以水浇地和旱田为主,东北部沿黄农田受黄河淡水影响,耕地质量较好,而中部地区地势低平,以水田为主,养分含量偏低;3)相对于块金系数/基台值,Moran’s I是更为稳健有效的衡量土壤养分空间自相关性的方法;4)论文认为,空间分布模式、样本量、空间自相关性和空间聚集程度(最近邻比)均影响插值精度。在离散模式下,各方法自适应性均较差;在随机模式下,IDW与RBF自适应性优于OK和LP;在聚集模式下,各方法自适应性与样本量和空间自相关性有关,直至样本足够多时,4种插值方法精度接近。论文探明了研究区主要土壤养分的最佳插值预测方法,分析了土壤养分的变异特征和空间分布规律,为黄河三角洲典型地区耕地土壤养分利用管理和农业可持续发展提供了理论依据。

关键词: Moran's I, 黄河三角洲, 空间分布模式, 空间预测, 土壤养分, 最近邻比

Abstract: A good understanding of the distribution characteristics of soil nutrients is important to achieve the best management of soil nutrients. Kenli County, a typical delta area in Shandong Province in North China, was chosen as the study area of this paper. There were 1 278 soil samples (0-20 cm) collected, processed and analyzed in laboratory. Geostatistical analysis was conducted to elucidate the spatial variations of soil nutrients and interpolate key soil nutrients in space. The semivariogram models of alkali-hydrolyzable nitrogen (AN), available phosphorus (AP) and available potassium (AK) were estimated. The best fitting models were selected based on the residual sum of squares (RSS) and coefficient of determination (R2). The R2 of AP’s spherical model was 0.951, followed by R2 of AN’s exponential model (0.892) and R2 of AK’s spherical model (0.787). The interpolations were employed with Ordinary Kriging (OK), Inverse Distance Weighted (IDW),Universal Kriging (UK),Radial Basis Function (RBF) and Local Polynomial (LP) in GIS software. The prediction accuracy was validated with test data. In order to explore the self-adaptive ability of OK, IDW, RBF and LP, eight datasets with different spatial distribution patterns were designed with AN data. The default optimized model of each method was used on the eight datasets and the results were compared. It comes to the following conclusions: 1) The spatial variations and autocorrelation of AN, AP and AK in the study area are moderate. The best fitted semi-variogram models are spherical model, exponential model and spherical model, whose coefficients of determination were 0.951, 0.892 and 0.787, respectively. 2) The spatial distribution patterns of AN, AP and AK are closely related to the terrain and land use types in the study area, i.e., the contents of key soil nutrients are high in the southwest part of region where there are irrigated land and dry land and northeast part of the region where the cropland is along the Yellow River and affected by the freshwater of the river, and the nutrient contents are lower in the central region where there is paddy field. 3) Compared with Nugget/sill, Moran’s I, the most widely used index to describe spatial autocorrelation, is more robust and effective for measuring spatial autocorrelation of soil nutrients. 4) The influence factors of spatial interpolation are spatial distribution pattern, sample number, spatial autocorrelation and spatial clustered degree (measured with Nearest Neighbor Rate); in dispersed pattern, all methods perform bad; in random pattern, IDW and RBF perform better than OK and LP; in clustered pattern, the adaptabilities of the methods are related with sample number and spatial autocorrelation, and all methods perform almost the same when there are enough samples. This paper ascertained the best interpolation prediction method for the main soil nutrients in the study area, analyzed the variation and the spatial distribution of soil nutrients, and provided a theoretical basis for soil nutrients management and agricultural sustainable development in typical area of Yellow River Delta.

Key words: Moran's I, Nearest Neighbor Rate, soil nutrients, spatial distribution pattern, spatial prediction, Yellow River Delta

中图分类号: 

  • S153.6