【作者】 孙立武；

【导师】 尚士友；

【作者基本信息】 内蒙古农业大学 ， 农业生物环境与能源工程， 2009， 博士

【摘要】 近年来日益恶化的草原生态环境对我国北方生态安全构成了极大的威胁,草原退化已成为重大环境问题。本文以西乌珠穆沁旗草原为试验研究区域,利用CBERS卫星数据监测、解译草原退化等级,通过GIS、GPS技术及地统计学方法,采用普通克里格插值方法处理土壤厚度试验数据,制作研究区域土壤厚度的空间分布预测图,将草原退化遥感数据与土壤厚度空间分布状态进行叠加分析,建立了一种土壤厚度空间分布状态与草原退化关系的分析方法,确定研究区域内不同退化程度草原的面积及位置,这将为制定草原科学管理和退化草原自然恢复的保护策略提供基础数据,为国家制定和实施沙源治理工程提供技术支持。论文主要结论如下:1.对土壤厚度预试验277个数据进行处理及误差分析,结果表明克里格插值法明显优于反距离加权插值法,并确定土壤试验样点合理间距选取原则,即:地势平坦的区域,草原土壤试验样点间距可由理论的19.5m适当调整到200m,局部区域甚至可以放大到500m,对于丘陵区域,则需根据地形地貌复杂程度,草原土壤试验样点间距可在19.5m～60m适当选取,最大不要超过100m。2.处理500个土壤厚度试验数据,结果表明土壤厚度试验数据接近于正态分布,直方图呈单峰,均值与中值大致相等,且有对称性;正态QQ图再次证实了土壤厚度试验数据是服从正态分布的,数据无须进行转化;土壤厚度试验数据经过趋势分析显示,可以用一个西南——东北向的二阶多项式对其进行拟合最佳;半变异函数/协方差函数云图表明土壤厚度试验数据中存在空间自相关,且知道数据集中没有离群值或错误的采样点,故选用普通克里格插值创建比较精确的土壤表面。3.应用不同半方差拟合模型进行插值后误差分析,结果表明采用Rational Quadratic model为半方差拟合模型,其预测误差的均值为0.012、均方根为7.384、平均预测标准差为7.426、平均标准差为0.001,均为其各自最小值,均方根标准预测误差为0.987,因此选择Rational Quadratic model为土壤厚度的最佳半方差拟合模型。4.通过草原退化现状与相应位置的遥感影像对比分析,确定草原退化不同等级的感兴趣区域,利用ERDAS IMAGINE8.6执行草原退化等级监督分类,经过分类模板误差矩阵评价、分类精度评价及实地验证,获得2006年～2008年草原退化等级解译图,其分类精度分别为90.00%、89.74%和87.50%,kappa系数分别为0.88、0.88和0.86。5.根据2006～2008年三年平均统计数据对比分析,在草原土壤厚度小于10cm的研究区域内,处于中度退化和重度退化程度的草原面积分别占总面积的44.9%和48.2%;在草原土壤厚度为10～20cm的研究区域内,处于中度退化程度的草原面积占总面积的70.3%;在草原土壤厚度为20～30cm的研究区域内,处于轻度退化程度的草原占总面积的64.8%;在草原土壤厚度大于30cm的研究区域内,处于无明显退化和轻度退化程度的草原面积分别占总面积的32.8%、53.3%。更多还原

【Abstract】 In recent years, the deteriorating ecological environment of grassland poses a great ecological security threats in northern China, and grassland degradated is the major environmenttal problems at home and abroad. The paper selected the Xiwuzhumuqin county grassland as the study area, monitored and classified the extent of degradated grassland by using the CBERS satellite data. Through the GIS, GPS technology and geostatistics methods, the Ordinary Kriging interpolation methods was adopted for processing the soil thickness test data, and the prediction map was produced which is the soil thickness spatial distribution to the study area. The research evaluated the degenerative grassland level by spatial overlay analyzing between the Remote Sensing data of degenerative grassland and soil thickness spatial distribution, and founded the analytical methods of the relation between the degradative grassland and soil thickness, ascertained the location and size of the different degradation levels at the grassland study area.This will offer the basic data for constituting the ecological environment protection strategy on the scientific management of grassland and the natural restore of degradative vegetation, and provides the technical support for the nation to formulate and implement sand sources control project.Main results and conclusions from the research are as follows:1.Through processing 277 samples of the soil thickness pre-test data and error analyzing, the results show that the kriging interpolation method is obviously better than the inverse distance weighted interpolation. So the Kriging interpolation method is selected test data interpolation. At the same time, the rearch ascertains the selecting principle of the reasonable distance between sample points in the soil thickness test. namely:To the flat region,the reasonable distance of the grassland soil test samples can be properly adjusted from the theoretic19.5m to 200m, the local area can even zoom into 500m. But hilly region, according to the complexity of the topography, the reasonable distance of the grassland soil test samples can be appropriately selected 19.5m ~ 60m, the largest do not exceed 100m.2.Dealing with 500 samples data of the soil thickness elicits the results: The data is closer to normal distribution and symmetry. The histogram shows a single peak. The data average is roughly equal with its median in value. QQ figure reconfirms that the soil thickness test data is subject to the normal distribution. So there is no need for data transformation. After extraction of the trend, the soil thickness test data shows a certain trend which can use the second-order polynomial of southwest-northeast to the best fit for its trend. The semivariogram/ covariance cloud figure shows that the soil thickness test data have the spatial autocorrelation. Because there is no data or erroneous outliers sampling points, the Ordinary Kriging interpolation can be used to create more precise soil surface.3 . Analyzing the different semi-variogram model fitting interpolation error: adopting the rational qadratic model as a semi-variance model, obtains that the mean value of the prediction is 0.012, the root-mean-square value is 7.384, the average standard deviation value is 7.426, the mean standardized value is 0.001. The property are minimum their respective. The root-mean-square standardized value is 0.987. Hence, the rational quadratic model is the best semi-variance fitting model to the soil thickness data.4.The research, comparatively analyzes the different degenerative vegetation level to grassland and the remote sensing image of the corresponding position, determines the region of interest of the different degenerative vegetation levels implements the supervised classification of the different degenerative vegetation level by ERDAS IMAGINE 8.6. Through the classification error matrix evaluation template, the classification accuracy evaluation and the field verification, receives the classification accuracy which are90.00%, 89.74% and 87.50% , kappa coefficients are 0.88, 0.88 and 0.86 on the degenerative vegetation level to grassland in 2006 ~ 2008.5.According to comparative analysis of statistical data on the average three years from 2006 to 2008,in the grassland soil thickness less than 10 cm of the study area, the moderate and severe degradation area are respectively 44.9% and 48.2% of the total area. The study area where is soil thickness of 10 ~ 20 cm, the grassland area of moderate degradation accounts for 70.3% of the total area. In the grassland soil thickness of 20 ~ 30cm of the study area, the area of moderate degradation grassland is the total area of 64.8 %. In grassland soil is greater than the thickness of 30 cm within the study area, the area of undegradation and moderate degradation grassland are respectively 32.8% and 53.3% of the total area.更多还原