海南岛土壤全氮时空变异特征的研究

发布时间：2018-04-20 01:33

本文选题：全氮 + 地统计　；参考：《中国农业科学院》2015年硕士论文

【摘要】：土壤氮素是土壤肥力高低和土壤质量的重要标志,是全球氮循环的重要源和汇。土壤全氮的空间变异特征研究,能够正确地评估土壤供氮能力,指导区域土壤资源利用和管理,对区域农业生产和环境管理有着十分重要的意义。土壤氮素在土壤中的分布并非均质,它受到成土母质、地形、生物、时间、气候等自然因素和施肥、耕作等人为因素的共同影响,表现出空间异质性。在预测土壤属性空间变异特征的研究中,传统地统计方法和混合地统计方法是目前应用最广泛和最有效的方法。以环境因子为辅助变量指导相关属性空间分布的预测和制图成为研究热点。但在大区域尺度有限的采样点条件下用不同地统计方法进行土壤属性空间分布预测的对比研究,选取适合预测模型预测大区域尺度土壤属性空间分布,这类研究并不多见。本研究选取海南岛全岛为研究区,结合地形与遥感指数等环境因子,从159个土壤样点中,每县随机选1-2点,全岛选取29个样点作为验证数据集,剩余的130个点作为预测数据集,采用传统地统计的普通克里格法和结合辅助变量的多元线性回归以及混合地统计方法的回归克里格法、协同克里格法共4种模型进行了土壤全氮分布的空间预测;利用29个验证数据集样点数据进行精度验证,并对比80年代土壤全氮空间分布,探讨了时空变异特征及其成因,结论如下:(1)除表层(0-5cm)土壤全氮预测最优模型为协同克里格法(CK)外,其他三层土壤全氮含量的预测最优模型均为普通克里格法(OK)。用较少网格样点对较大区域进行土壤全氮的空间分布预测时,普通克里格法总体精度最高,协同克里格法仅次于普通克里格法。协同克里格法预测精度优于或等于回归克里格预测精度。而回归克里格法(RK)的预测精度始终低于普通克里格法(OK)。协同克里格法、回归克里格法与多元线性回归模型和普通克里格法相比,在极值处存在着一定的消除平滑效应的效果。(2)海南岛4个土层全氮含量均有从东向西方向递减,高值区向外递减的分布趋势,高值主要分布在东北部和南部地区。耕层土壤(0-5cm和0-20cm)土层全氮含量集总体处于三级和四级中等水平。20-40cm和40-60cm土层土壤全氮含量处于六级较低水平,且该等级面积比例分别达到了51.5%和90.9%。空间同一位置的全氮含量随土壤深度的增加而降低。(3)不同土层土壤全氮含量与不同的环境因子呈现出不同程度的相关关系。0-5cm土层全氮含量与坡度、NDVI等因子均未达显著相关,仅与大田这一土地利用方式呈极显著相关。20-40cm、40-60cm土壤全氮含量与归一化植被指数、坡度呈极显著或显著相关,0-20cm土壤全氮含量与归一化植被指数呈显著相关。(4)时隔30年,海南岛土壤全氮含量整体分布趋势虽然相似,高值仍然保持在南部山区和东北部地区,但全氮含量整体呈下降趋势。以琼中县为例进行成因分析,全氮变化的主要原因为在坡度较大的山区进行人为开垦而造成的氮素随水土流失。
[Abstract]:Soil nitrogen is an important symbol of soil fertility and soil quality. It is an important source and sink of the global nitrogen cycle. The study of spatial variation characteristics of soil total nitrogen can correctly assess the ability of soil nitrogen supply and guide the utilization and management of regional soil resources. It is of great significance for Regional Agricultural production and environmental management. The distribution of soil is not homogeneous. It is influenced by natural factors such as soil parent material, terrain, biology, time, climate and other natural factors, such as fertilization, and cultivation and other factors, showing spatial heterogeneity. In the study of predicting the spatial variation characteristics of soil properties, traditional statistical methods and mixed statistical methods are the most widely used and most widely used at present. It is an effective method to guide the prediction and mapping of spatial distribution of related attributes with environmental factors as auxiliary variables. But a comparative study on spatial distribution prediction of soil properties is carried out with different statistical methods under the condition of limited sampling points in large regional scale, and a suitable pretest model is selected to predict soil properties in large regional scale. In this study, we select the Hainan Island whole island as the research area, and combine the environment factors such as terrain and remote sensing index. From 159 soil samples, 1-2 points are selected randomly in each county. 29 samples are selected as validation data sets in the whole island. The remaining 130 points are used as the prediction data sets, and the traditional Kriging method is adopted and the traditional Kriging method is used. Combined with the multiple linear regression of auxiliary variables and the regression Craig method of mixed statistical method, the spatial prediction of total nitrogen distribution in soil was predicted by 4 models of CO Craig law, and the accuracy was verified by 29 verifying data collection points, and the spatial distribution of soil total nitrogen in 80s was compared and the characteristics of temporal and spatial variation and its formation were discussed. The conclusions are as follows: (1) the optimal model of soil total nitrogen prediction in surface layer (0-5cm) is cooperative Kriging (CK), and the other three layers of soil total nitrogen content prediction optimal model is common Kriging method (OK). The general Kriging method has the highest overall precision when using less grid sample points to predict the spatial distribution of soil total nitrogen in larger regions. The Kriging method is second only to the common Kriging method. The precision of the cooperative Craig Fa prediction is superior to or equal to the regression Kriging prediction accuracy. The predictive accuracy of the regression Kriging method (RK) is always lower than the common Kriging method (OK). The cooperative Kriging method, the regression Craig Fa and the multiple linear regression model and the ordinary Kriging method, exist at the extreme value. The effect of smoothing effect is eliminated. (2) the total nitrogen content of the 4 soil layers in Hainan Island is decreasing from east to west, and the high value area decreases outward. The high value is mainly distributed in the northeast and south areas. The total nitrogen content of the soil layer of the plough soil (0-5cm and 0-20cm) is in the three and four grade medium level.20-40cm and 40-60cm soil. The total nitrogen content in the soil layer is at the lower level of the six level, and the total nitrogen content of this grade area is 51.5% and 90.9%., respectively, and the total nitrogen content decreases with the increase of soil depth. (3) the total nitrogen content in different soil layers and different environmental factors have different degrees of correlation between the total nitrogen content and slope of.0-5cm soil layer, NDVI No significant correlation was found between the other factors, only the land use mode of the field was significantly correlated with.20-40cm, the total nitrogen content in 40-60cm soil and the normalized vegetation index were very significant or significant correlation. The total nitrogen content in 0-20cm soil was significantly correlated with the normalized vegetation index. (4) the total distribution of total nitrogen in Hainan Island soil was generally distributed after 30 years. Although the potential is similar, the high value remains in the southern and northeastern regions, but the total nitrogen content has a downward trend. The genetic analysis is taken as an example of Qiongzhong County. The main reason for the change of total nitrogen is the nitrogen and soil erosion caused by artificial reclamation in the mountainous areas with large slopes.

【学位授予单位】：中国农业科学院
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S153.6

【参考文献】