基于优化光谱指数的草地生物量估算

发布时间：2018-01-24 01:43

本文关键词： 草地波段优化高光谱生物量　出处：《内蒙古农业大学》2015年硕士论文　论文类型：学位论文

【摘要】：遥感技术的发展为草地快速、无损的动态监测提供可能,本文通过对2013-2014年内蒙古天然草地和人工草地生物量及其冠层光谱数据分析,通过概念模型和从350nm-1150 nm所有可能两两波段组合的归一化算法和简单比值算法进行波段选择,从而提取对生物量变化敏感的波段组合,通过线性和非线性回归分析进行生物量估算模型的建立,探讨不同优化算法计算的光谱指数在估算草地生物量时的鲁棒性。结果表明：(1)相比现有常用估算草地生物量的植被指数,优化光谱指数显著提高了生物量的预测能力,不存在现有基于红光位置植被指数存在的饱和性问题,并且优化光谱指数的波段组合随着草地类型和牧草品种的不同而不同,其波段组合包括蓝光、红光、红边位置和近红外位置。现有植被指数与草地生物量的相关性随着草地类型和生长时期的变化而变化(R2=0.00-0.77),优化光谱指数克服了基于红光的光谱指数在高生物量条件下的饱和问题。(2)包含绿光和红边位置的优化光谱指数CI(Chlorophyll index).包含蓝光和红边位置的优化光谱指数NDSI(Normalized difference spectral index)和蓝光和红边位置简单比值植被指数RSI(Simple spectral index)目比现有植被指数都表现出与生物量较好的相关性(R2=0.36-0.80),其中波段组合为(R696-R652)/(R696+R652)的NDSI相比CI和RSI与所有生物量表现出较高的相关性决定系数R2=0.79,且其建立模型的预测值与实测值之间的关系相比CI(R2=0.76,均方根误差RMSE=954,平均相对误差RE=52.0%)和RSI(R2=0.80,RMSE=926,RE=35.8%)具有最小的偏差(R2=0.77,RMSE=756,RE=37.1％),故选取其建立的非线性二次回归方程y=23288.0x2+16116.0x+422.2为估算本实验生物量的最优生物量模型。(3)NDSI更适合不同草地类型、不同生长时期、不同冠层盖度等复杂环境草地生物量的估算,它提高了光谱指数与生物量之间的相关性,对生物量变化更敏感,不存在饱和问题,所建模型能够用于估算复杂环境下的草地生物量,这不仅为本试验区根据遥感估算草地生物量提供了准确的生物量估算模型,也为遥感植被指数的理论研究提供了更深入的理论基础。
[Abstract]:The development of remote sensing technology provides the possibility for rapid and non-destructive dynamic monitoring of grassland. This paper analyzes the biomass and canopy spectral data of natural grassland and artificial grassland in Inner Mongolia from 2013 to 2014. The spectral bands that are sensitive to biomass change are extracted by the conceptual model and the normalized algorithm and simple ratio algorithm from all possible combinations of two bands at 350 nm to 1150 nm. The biomass estimation model was established by linear and nonlinear regression analysis. The robustness of spectral indices calculated by different optimization algorithms in estimating grassland biomass was discussed. The optimized spectral index can significantly improve the prediction ability of biomass, and there is no saturation problem based on the red position vegetation index. And the band combination of the optimized spectral index is different with the grassland type and forage variety. The band combination includes blue light and red light. Red edge position and near infrared position. The correlation between the existing vegetation index and grassland biomass changes with the change of grassland type and growth period. The optimized spectral index overcomes the saturation problem of the spectral index based on red light under the condition of high biomass. Chlorophyll index.Optimum spectral index NDSI with blue and red edge positions. Normalized difference spectral index and simple ratio of blue light and red edge vegetation index (RSI). Simple spectral index showed a better correlation with biomass than the existing vegetation index (R2P 0.36-0.80). The correlation coefficient between CI and RSI was higher than that of R696-R652 / R696R652 (R _ (696-R _ (652) / R _ (696) R _ (652))). The coefficient of correlation between CI and RSI was 0.79. The relationship between the predicted value and the measured value of the model is compared with that of CIR _ 2N _ (0.76) and the root mean square error (RMSE=954). The mean relative error (RE52.0) and RSI R2N 0.80 RMSE 926RE35.8) have the minimum deviation of R2 / 0.77. RMSE 756 and RMSE 37.1). Therefore, the nonlinear quadratic regression equation yang23288.0x216116.0x422.2 was chosen as the optimal biomass model for estimating the experimental biomass. NDSI is more suitable for different grassland types. The estimation of grassland biomass in different growth period and different canopy coverage increased the correlation between spectral index and biomass, and was more sensitive to biomass change and had no saturation problem. The model can be used to estimate grassland biomass in complex environment, which not only provides an accurate model for estimating grassland biomass based on remote sensing. It also provides a deeper theoretical basis for the theoretical study of remote sensing vegetation index.
【学位授予单位】：内蒙古农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S812

【参考文献】