基于RS和GIS的成都平原LUCC模拟及预测研究

发布时间：2018-03-21 03:34

本文选题：成都平原　切入点：土地覆被/变化模拟及预测　出处：《四川师范大学》2017年硕士论文　论文类型：学位论文

【摘要】：土地覆被/变化模拟及预测,能够更好地了解土地转移机制和合理规划未来土地的利用。论文依托“教育部重点实验室”和“省国土资源协同创新中心”科研平台和国家自然科学基金(面上项目)等完成。研究利用RS和GIS技术,使用决策树和面向对象相结合的分类方法分别对2005a和2015a影像数据提取了成都平原土地利用数据,利用该数据分析了10a间研究区内各土地利用类型之间动态转移过程。利用相关性系数选出量化表达LUCC程度的最佳变量。分析了矢量化乡镇边界和13级渔网尺度下空间影响因子(土地利用面积、空间自相关性、离散度)与LUCC的相关性并利用主成分综合得分计算出最佳网格化尺度。利用主成分和线性回归构建了空间影响因子、地形因子(海拔、坡度、坡向)、距离因子(距道路距离、距水体距离、距林地距离、距耕地距离、距建设用地距离)、核密度因子(道路核密度、水体核密度、林地核密度、耕地核密度、建设用地核密度)与LUCC的回归方程,从而利用回归方程制作了各影响因子对LUCC影响的概率图。利用马尔可夫模型预测了2015a和2025a LUCC的数量,结合概率图预测2015a LUCC的空间格局,并检验结果精度,最后预测2025a LUCC的空间格局。研究表明:(1)所有表达研究区LUCC程度的变量中,剩余空间变化率最能表达LUCC的程度。乡镇边界和13级渔网尺度中270m×270m尺度的综合得分最高,因此定量化表达空间影响因子与LUCC关系的最佳网格尺度为270m×270m。在主成分和线性回归方程表达中,水体和耕地的主成分回归方程拟合度较高,分别为0.462和0.344,林地、草地和耕地的线性回归拟合度较高,分别为0.418、0.615和0.591。因此空间影响因子定量化表达LUCC的概率效果较好。(2)海拔、坡度与LUCC所构建的回归方程中6次项方程的拟合程度最高。海拔对水体回归方程拟合度为0.847,对林地0.703,对耕地0.448,对建设用地0.574;坡度对水体回归方程拟合度为0.947,对林地0.996,对耕地0.871,对建设用地0.866,因此海拔、坡度能够很好的定量化表达LUCC的概率。坡向拟合度较低。距道路距离中幂函数回归方程最高,均达到0.7以上,而距土地利用类型中拟合程度较高的回归方程类型不相同,主要为6次项和幂函数两种方程。对于整个研究区,构建核密度与LUCC的回归方程拟合度都较低,但分聚集对象单独构建方程的拟合度很高,因此土地利用类型在空间上聚集对象不同,其LUCC的程度也不同。(3)经预测,2025a研究区水体、未利用地和建设用地面积将持续增加,而林地、耕地和草地面积将持续减少。基于RS和GIS成都平原2015a LUCC的预测得到建设用地精度最高达75.11%,其次耕地为71.23%,林地69.97%,水体最低为43.5%,因此该土地利用模型对于在空间上聚集程度较高的土地利用类型预测精度较高,而离散型预测精度较低。
[Abstract]:Land cover / change modelling and forecasting, The paper relies on the key Laboratory of Ministry of Education and the Cooperative Innovation Center of Provincial Land and Resources, and the National Natural Science Foundation (NSF). Research and use of RS and GIS technology, The land use data of Chengdu Plain were extracted from 2005 and 2015a image data by using decision tree and object oriented classification method, respectively. The data were used to analyze the dynamic transfer process between different land use types in the study area for 10 years. The best variable for quantifying the degree of LUCC was selected by using the correlation coefficient. The vectorized boundary of villages and towns and the scale of 13 grade fishing nets were analyzed. Spatial impact factors (land use area, The correlation between spatial autocorrelation, dispersion and LUCC and the calculation of optimal gridding scale using principal component synthesis score. The spatial influence factors, topographic factors (elevation, slope, slope) were constructed by principal component and linear regression. Slope direction, distance factor (distance from road, distance from water body, distance from forest land, distance from cultivated land, distance from construction land), nuclear density factor (road nuclear density, water body nuclear density, forest core density, cultivated land nuclear density), The regression equation between the kernel density of construction land and LUCC is used to make the probability map of the influence factors on LUCC. The numbers of 2015a and 2025a LUCC are predicted by Markov model, and the spatial pattern of 2015a LUCC is predicted by combining the probability map. Finally, the spatial pattern of 2025a LUCC was predicted. The change rate of residual space can best express the degree of LUCC. The comprehensive score of 270m 脳 270m scale in the scale of township boundary and 13th grade fishing net is the highest. Therefore, the best mesh scale of quantitative expression spatial influence factor and LUCC is 270m 脳 270m.The principal component regression equation of water body and cultivated land has higher fitting degree in the expression of principal component and linear regression equation, which are 0.462 and 0.344, respectively. The fitting degree of linear regression of grassland and cultivated land was 0.418 ~ 0.615 and 0.591, respectively. Therefore, the probability of quantifying expression of LUCC by spatial influential factors was better. The fitting degree of the regression equation of slope and LUCC is the highest. The fitting degree of regression equation of elevation to water body is 0.847, to forest land is 0.703, to cultivated land is 0.448, to construction land is 0.574.The fitting degree of slope to regression equation of water body is 0.947, and to forest. Land 0.996, to cultivated land 0.871, to construction land 0.866, so altitude, The slope can quantify the probability of LUCC. The slope fitting degree is low. The power function regression equation is the highest in the distance from the road, and the regression equation is above 0.7, but the regression equation with higher fitting degree from the land use type is not the same. For the whole study area, the fitting degree of the regression equation of constructing kernel density and LUCC is low, but the fitting degree of constructing equation by dividing aggregation object alone is very high. Therefore, land use types are different in spatial aggregation, and their LUCC levels are also different. 3) it is predicted that the area of unused land and construction land will continue to increase, while the forest land will continue to increase, after predicting the water body in the study area of 2025a. The area of cultivated land and grassland will continue to decrease. Based on the prediction of LUCC in Chengdu plain of GIS and RS in 2015a, the accuracy of construction land is up to 75.1111, followed by cultivated land is 71.23cm, woodland is 69.97, and water is the lowest 43.5. therefore, the land use model can converge in space. The prediction accuracy of land use types with higher concentration is higher than that of other land use types. The accuracy of discrete prediction is low.
【学位授予单位】：四川师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237;P208;F301.2

【参考文献】