基于像元二分模型的博斯腾湖西岸湖滨绿洲植被覆盖度变化研究

发布时间：2018-04-30 20:02

本文选题：植被覆盖度 + 像元二分模型　；参考：《新疆师范大学》2017年硕士论文

【摘要】：植被覆盖度在一定程度上反映了陆地生态系统中植被类型、数量和质量上的变化,是反映植被状况的一个重要指标。以博斯腾湖西岸湖滨绿洲作为研究区,以1990、2000、2015年三个不同时期的遥感影像为基本数据源,结合实地调研数据利用像元二分模型对研究区1990、2000、2015年植被覆盖度进行估算,并结合研究区实际植被覆盖情况对植被覆盖度等级进行划分。运用矩阵分析明确1990~2015年所有植被覆盖类型面积变化特征。利用差值法求对3期的植被覆盖度进行计算,分析1990~2015年间植被覆盖度空间变化特征。对比BP神经网络模型和LMBP神经网络模型。选取精度较高的模型对植被覆盖度进行预测。从自然因素和人文因素两个方面选取影响植被覆盖度变化的驱动因子,通过相关性分析,探讨极低植被覆盖度、低植被覆盖度、中等植被覆盖度以及高植被覆盖度等四种覆盖度类型变化主要驱动因素。主要研究结论如下:1.1990~2015年,极低植被覆盖度、低植被覆盖度面积减少,中等植被覆盖度、高植被覆盖度面积增多。近25a,高植被覆盖区面积变化明显,由71.93km2上升至361.17km2,增长近5倍。极低植被覆盖度面积则消减323.94km2,整体下降81.28%。2.近25a间,低植被覆盖度朝中等植被覆盖度转移率最大,转移率为59.01%,转移面积156.41km2。其次是极低植被覆盖度朝中植被覆盖度转移,转移面积227.12km2,转移率为45.51%,再次是极低植被覆盖度朝中植被覆盖度转移,转移面积达到142.10km2,转移率达到28.50%。最后是高植被度盖度朝极低植被覆盖区转移,转移率只有0.048%。3.1990~2015年,研究区植被覆盖度改善面积占了总面积的65.95%,面积达到558.47km2。明显改善面积为243.94km2,占总面积28.81%。改善区域主要表现在研究区的中部及东部,退化区域表现在研究区南部和西部小部分区域。近25a植被覆盖度总体呈现改善趋势。4.构建BP神经网络模型和LMBP神经网络模型,并进行对比分析。结果表明LMBP神经网络模型精度高于BP神经网络模型,LMBP神经网络模型预测出真实值与预测值的相关系数达到0.9400,相关性较高,说明运用LMBP神经网络预测研究区植被覆盖度是可行的。对研究区2020年植被覆盖度预测,结果表明植被覆盖度呈现退化趋势。5.植被覆盖度变化受自然因素和人文因素的共同影响。极低植被覆盖度与年均降水量呈现显著负相关(通过p0.05检验),相关系数为-0.969。低植被覆盖度与中等植被覆盖度区域面积变化与人口总数相关,相关系数分别为-0.972和0.888。高植被覆盖度区域面积与农业生产总值相关,相关系数为0.988.
[Abstract]:Vegetation coverage to some extent reflects the change of vegetation type, quantity and quality in terrestrial ecosystem, and it is an important index to reflect vegetation status. Taking the oasis on the west coast of Bosten Lake as the research area, taking the remote sensing images of three different periods in 1990 and 2015 as the basic data sources, combining with the field survey data, using the pixel dichotomy model to estimate the vegetation coverage in the study area of 1990 ~ 2000 and 2015. Combined with the actual vegetation coverage in the study area, the classification of vegetation coverage was carried out. The area variation characteristics of all vegetation cover types from 1990 to 2015 were determined by matrix analysis. The difference method is used to calculate the vegetation coverage in three periods, and the spatial variation characteristics of vegetation coverage from 1990 to 2015 are analyzed. The BP neural network model and the LMBP neural network model are compared. High precision models are selected to predict vegetation coverage. In this paper, we select the driving factors which influence the change of vegetation coverage from two aspects of natural factors and human factors. Through the correlation analysis, we discuss the extremely low vegetation coverage, the low vegetation coverage, the low vegetation coverage, the low vegetation coverage and the low vegetation coverage. The main driving factors of the change of the four types of coverage are middle vegetation coverage and high vegetation coverage. The main conclusions are as follows: 1. From 1990 to 2015, very low vegetation cover degree, low vegetation coverage area decreased, middle vegetation coverage degree and high vegetation coverage area increased. In the past 25 years, the area of high vegetation cover area changed obviously, from 71.93km2 to 361.17km2, the increase was nearly 5 times. The area of extremely low vegetation cover was reduced by 323.94 km2, and the whole area decreased by 81.28. 2. In the past 25 years, the transfer rate of low vegetation coverage to middle vegetation coverage was the highest, with a transfer rate of 59.01 and a transfer area of 156.41 km2. Secondly, the very low vegetation cover was transferred to the middle vegetation cover, with a transfer area of 227.12 km2, and the transfer rate was 45.51.The second was the very low vegetation coverage to the middle vegetation coverage, the transfer area was 142.10 km2, and the transfer rate was 28.50. Finally, the high vegetation coverage was transferred to the very low vegetation cover area, the transfer rate was only 0.048. 3.1990-2015. The improved vegetation coverage area accounted for 65.95km2 of the total area in the study area, and the area reached 558.47km2. The obvious improved area was 243.94 km2, accounting for 28.81% of the total area. The improvement area is mainly in the middle and east of the research area, and the degraded area is in the south and west of the study area. The vegetation coverage showed a trend of improvement in the past 25 years. BP neural network model and LMBP neural network model are constructed and compared. The results show that the accuracy of LMBP neural network model is higher than that of BP neural network model. The correlation coefficient between the real value and the predicted value is 0.9400, which indicates that it is feasible to use LMBP neural network to predict vegetation coverage in the study area. The vegetation coverage in the study area in 2020 was predicted, and the results showed that the vegetation coverage showed a degradation trend of .5. The change of vegetation coverage is influenced by both natural and human factors. There was a significant negative correlation between very low vegetation coverage and average annual precipitation (the correlation coefficient was -0.969 through p0.05 test). The regional area changes of low and moderate vegetation cover were related to the total population, and the correlation coefficients were -0.972 and 0.888, respectively. The area of high vegetation coverage is related to the agricultural gross domestic product, the correlation coefficient is 0. 988.
【学位授予单位】：新疆师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：Q948

【参考文献】