土壤砷和氮含量的空间变异及其相互关系研究

发布时间：2018-02-03 23:05

本文关键词： 土壤砷土壤氮空间变异 RBF神经网络协同克里格　出处：《华南农业大学》2016年博士论文　论文类型：学位论文

【摘要】：土壤是农业生产最基本的生产资料,全面了解土壤质量信息是展开土壤生态保护和修复等工作的基础,而土壤养分的丰缺程度是土壤质量高低的重要指标,同时土壤污染物含量的高低也是影响土壤质量的重要指标。本研究以广州市增城区为研究区域,在耕地地力调查成果的基础上,选取117个土壤样点,用不同的插值方法对土壤砷和土壤全氮含量的空间变异特征及其相互关系进行了研究。主要研究内容和结论如下:(1)通过对117个土壤样品数据的数理统计分析得出增城区土壤砷和土壤全氮含量的分布范围都较广,土壤全氮变异系数为38.97%属于中等变异,从偏度上看尽管大于零但数值比较小,其分布趋于对称;土壤砷变异系数为84.41%属于强变异,从偏度上看,土壤砷含量有较大的正偏离,其分布较正态分布向右偏离。(2)分别利用普通克里格方法和RBF神经网络方法对土壤砷含量进行空间插值,对插值结果进行对比分析发现在研究区范围内RBF神经网络方法在拟合能力和插值精度等方面都优于普通克里格方法。(3)增城区土壤砷含量范围主要分布在0~15mg/kg,增城区的东北部和南部土壤砷含量较高,且自东北至西南方向呈现逐渐降低的趋势,西部含量较低。土壤砷含量较高(9~15mg/kg)地区主要分布在派潭镇、正果镇、小楼镇、增江街道、荔城街道的东北部、石滩镇的中南部和新塘镇的南部,土壤砷含量较低(小于9mg/kg)地区主要分布在中新镇、朱村街道、荔城街道的西南部、石滩镇的西北部和新塘镇的北部。(4)分别利用普通克里格方法和RBF神经网络方法对土壤全氮含量进行空间插值,对插值结果进行对比分析发现在研究区范围内RBF神经网络方法在拟合能力和插值精度等方面都占优。(5)增城区土壤全氮含量高级水平(1~2.5g/kg)地区主要分布在新塘镇、石滩镇的西部、朱村街道和派潭镇的中南部,且自中部向东西两方向呈现逐渐降低的趋势,土壤全氮含量中级水平(0.75~1g/kg)地区主要分布在派潭镇、荔城街道和石滩镇的东部,土壤全氮含量低级水平(小于0.75g/kg)地区主要分布在中新镇、小楼镇、正果镇和增江街道。(6)研究区灌木林、有林地、果园、水田、旱地、水浇地中土壤砷和土壤全氮的含量都是呈现逐步上升的趋势,且土壤全氮和土壤砷含量在各土地利用类型上的相关系数R2达到了0.987,这种相关性不仅说明了土壤砷和土壤氮含量受人为活动影响较大,也说明了在研究区范围内土壤砷和土壤氮含量之间存在显著正相关关系。(7)以土壤砷含量为协变量对土壤氮含量进行协同克里格插值,协同克里格方法在模型的拟合能力和插值精度等方面较普通克里格方法和RBF神经网络方法都有一定程度的提高,也证明了土壤砷和土壤氮含量之间存在显著相关关系,土壤砷的含量会对土壤氮含量的空间变异造成影响。
[Abstract]:Soil is the most basic means of production in agricultural production. Comprehensive understanding of soil quality information is the basis of soil ecological protection and remediation, and the abundance of soil nutrients is an important indicator of soil quality. At the same time, the content of soil pollution is also an important index to affect soil quality. In this study, 117 soil samples were selected on the basis of the results of cultivated land fertility investigation in Zengcheng District of Guangzhou City. The spatial variation characteristics of soil arsenic and soil total nitrogen contents and their relationships were studied by different interpolation methods. The main contents and conclusions are as follows: 1). Based on the statistical analysis of 117 soil samples, it was found that the distribution range of soil arsenic and soil total nitrogen content in Zengcheng area was wide. The coefficient of variation of soil total nitrogen (TNA) of 38.97% belongs to medium variation, although the deviation is greater than zero, the value is smaller, and its distribution tends to be symmetrical. The coefficient of variation of arsenic in soil is 84.41%, which is a strong variation, and the soil arsenic content has a large positive deviation from the degree of deviation. Its distribution deviates to the right than the normal distribution.) the spatial interpolation of arsenic content in soil is carried out by using the ordinary Kriging method and the RBF neural network method respectively. Comparing and analyzing the interpolation results, it is found that the RBF neural network method is superior to the ordinary Kriging method in terms of fitting ability and interpolation accuracy in the study area. The range of arsenic content in Zengcheng area was 15 mg / kg. The content of arsenic in the northeast and south of Zengcheng area is higher and the content of arsenic is decreasing gradually from northeast to southwest. The content of arsenic in the west is relatively low. The content of arsenic in soil is higher than that in 15 mg 路kg ~ (-1) of soil. It is mainly distributed in the northeast of Pitan Town, Zhengguo Town, Xialou Town, Zengjiang Street and Licheng Street. The low arsenic content (less than 9 mg / kg) in the south of Shitan town and the south part of Xintang town is mainly distributed in Zhongxin town, Zhucun street and southwest of Licheng street. The general Kriging method and RBF neural network method were used to interpolate soil total nitrogen content in the northwestern part of Shitan town and the northern part of Xitang town respectively. The comparison and analysis of the interpolation results show that the RBF neural network method is superior in fitting ability and interpolation accuracy in the study area.) the higher level of soil total nitrogen content in Zengcheng area (. (1) the area of 2.5 g / kg is mainly distributed in Xintang Town. The west of Shitan Town, the street of Zhucun and the central and southern part of Pitan Town show a decreasing trend from the middle to the east and west. The intermediate level of soil total nitrogen content was 0.75g / kg) in Pitan Town, Licheng Street and the eastern part of Shitan Town. The low level of soil total nitrogen content (< 0.75 g / kg) was mainly distributed in shrub forest, woodland, orchard, paddy field and dryland in Zhongxin, Xiaolou, Zhengguo and Zengjiang streets. The contents of soil arsenic and soil total nitrogen in irrigated land showed a trend of gradual increase, and the correlation coefficient R2 of soil total nitrogen and soil arsenic content in each land use type reached 0.987. This correlation not only indicates that soil arsenic and soil nitrogen content are greatly affected by anthropogenic activities. It also shows that there is a significant positive correlation between soil arsenic and soil nitrogen content in the study area. The cooperative Kriging method can improve the model fitting ability and interpolation accuracy to some extent compared with the common Kriging method and the RBF neural network method. It is also proved that there is a significant correlation between soil arsenic and soil nitrogen content, and the content of soil arsenic will affect the spatial variation of soil nitrogen content.
【学位授予单位】：华南农业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：S153.6

【参考文献】