基于近地高光谱遥感与作物生长模型的冬小麦品质监测预报研究
发布时间：2018-08-19 10:46
【摘要】：生产优质专用、蛋白质含量稳定的小麦,是解决我国优质专用小麦缺乏的必经之路,作物模型和遥感技术的出现,为科学、无损、高效监测预报作物品质提供了新的思路。作物模型的机理性、时间扩展性,以及与遥感数据的空间性、实时性相耦合,成为一种可以提高作物品质监测预报精度的新手段。本文以2012-2013、2013-2014年陕西省杨凌区及其周边地区的冬小麦为研究对象,选择DSSAT作物生长模型,结合近地面高光谱遥感数据和粒子群同化算法(Part icle Swarm Optimization,PSO)开展遥感信息与作物模型结合的冬小麦品质监测预报研究。首先,对DSSAT模型的适应性进行评价,利用2013-2014年田间实验数据对DSSAT作物模型进行参数本地化研究,并用2012-2013年的试验数据进行验证。然后,针对2013-2014年的杨凌揉谷试验区,利用全生育期地面高光谱数据反演冬小麦的生物量,解决了单生育期反演精度低的问题,间接计算冠层氮素累积量,将遥感反演的冠层氮素累积量作为粒子群同化变量对作物模型模拟的冠层氮素累积量进行优化,提高籽粒蛋白质含量的模拟精度。结果表明:参数本地化后的DSSAT作物模型模拟的2012-2013、2013-2014年冬小麦生长的全生育期误差均小于1天,各生育期误差均小于2天,2013-2014年、2012-2013年的生物量的实测值和模拟值的RMSE分别为0.996t/ha和1.678t/ha,产量的实测值与模拟值的绝对误差为-59kg/ha和-26kg/ha,籽粒蛋白质含量的绝对误差分别为-1.2%和-1.55%,研究证明了参数本地化后的DSSA T模型在杨凌地区有一定的适用性。利用粒子群同化算法对大田环境下每个单点的模拟结果进行优化,整个区域20个地块的籽粒蛋白质监测误差RMSE由同化前的2.39%降低为2.26%,表明了遥感信息与作物模型相结合监测预报作物品质是有效的,并且利用粒子群同化算法对两者进行耦合,可以提高模型模拟的精度。
[Abstract]:The production of wheat with high quality and stable protein content is the only way to solve the shortage of high quality special wheat in China. The emergence of crop models and remote sensing technology provides a new idea for scientific, non-destructive and efficient monitoring and prediction of crop quality. The mechanism, time expansibility, spatial and real-time coupling of crop models with remote sensing data have become a new method to improve the precision of crop quality monitoring and prediction. In this paper, the winter wheat in Yangling region and its surrounding areas of Shaanxi Province from 2012-2013 to 2014 was selected as the research object, and the DSSAT crop growth model was selected. Based on the near surface hyperspectral remote sensing data and particle swarm optimization (Part icle Swarm), a study on winter wheat quality monitoring and forecasting based on remote sensing information and crop model was carried out. Firstly, the adaptability of DSSAT model was evaluated, and the field experiment data from 2013 to 2014 were used to localize the parameters of DSSAT crop model. Then, the biomass of winter wheat was retrieved from the surface hyperspectral data of the whole growth period in the Yang Ling kneading valley experimental area from 2013 to 2014, which solved the problem of low inversion accuracy in the single growth period and indirectly calculated the accumulation of nitrogen in the canopy. In order to improve the simulation accuracy of grain protein content, the nitrogen accumulation in the canopy was optimized as a variable of particle swarm assimilation (PSO). The results showed that the errors of the whole growth period of winter wheat from 2012-2013 to 2013-2014 were all less than 1 day by the DSSAT crop model after localization of parameters. The RMSE of biomass measured and simulated values in 2013-2014 and 2012-2013 were 0.996t/ha and 1.678 t / ha, respectively. The absolute errors of yield and simulated values were -59 kg / ha and -26 kg / ha, and the absolute errors of grain protein content were -1.2%, respectively. And -1.55, it is proved that the DSSA T model after parameter localization is applicable in Yang Ling region. The particle swarm optimization algorithm is used to optimize the simulation results of each single point in the field environment. The grain protein monitoring error (RMSE) of 20 plots in the whole region was reduced from 2.39% before assimilation to 2.26, which indicated that the combination of remote sensing information and crop model was effective in monitoring and forecasting crop quality, and the particle swarm assimilation algorithm was used to couple them. The accuracy of model simulation can be improved.
【学位授予单位】：山东农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S512.11;S127
本文编号：2191433
[Abstract]:The production of wheat with high quality and stable protein content is the only way to solve the shortage of high quality special wheat in China. The emergence of crop models and remote sensing technology provides a new idea for scientific, non-destructive and efficient monitoring and prediction of crop quality. The mechanism, time expansibility, spatial and real-time coupling of crop models with remote sensing data have become a new method to improve the precision of crop quality monitoring and prediction. In this paper, the winter wheat in Yangling region and its surrounding areas of Shaanxi Province from 2012-2013 to 2014 was selected as the research object, and the DSSAT crop growth model was selected. Based on the near surface hyperspectral remote sensing data and particle swarm optimization (Part icle Swarm), a study on winter wheat quality monitoring and forecasting based on remote sensing information and crop model was carried out. Firstly, the adaptability of DSSAT model was evaluated, and the field experiment data from 2013 to 2014 were used to localize the parameters of DSSAT crop model. Then, the biomass of winter wheat was retrieved from the surface hyperspectral data of the whole growth period in the Yang Ling kneading valley experimental area from 2013 to 2014, which solved the problem of low inversion accuracy in the single growth period and indirectly calculated the accumulation of nitrogen in the canopy. In order to improve the simulation accuracy of grain protein content, the nitrogen accumulation in the canopy was optimized as a variable of particle swarm assimilation (PSO). The results showed that the errors of the whole growth period of winter wheat from 2012-2013 to 2013-2014 were all less than 1 day by the DSSAT crop model after localization of parameters. The RMSE of biomass measured and simulated values in 2013-2014 and 2012-2013 were 0.996t/ha and 1.678 t / ha, respectively. The absolute errors of yield and simulated values were -59 kg / ha and -26 kg / ha, and the absolute errors of grain protein content were -1.2%, respectively. And -1.55, it is proved that the DSSA T model after parameter localization is applicable in Yang Ling region. The particle swarm optimization algorithm is used to optimize the simulation results of each single point in the field environment. The grain protein monitoring error (RMSE) of 20 plots in the whole region was reduced from 2.39% before assimilation to 2.26, which indicated that the combination of remote sensing information and crop model was effective in monitoring and forecasting crop quality, and the particle swarm assimilation algorithm was used to couple them. The accuracy of model simulation can be improved.
【学位授予单位】：山东农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S512.11;S127
【参考文献】
相关博士学位论文 前2条
1 刘兰翠;我国二氧化碳减排问题的政策建模与实证研究[D];中国科学技术大学;2006年
2 姜志伟;区域冬小麦估产的遥感数据同化技术研究[D];中国农业科学院;2012年
相关硕士学位论文 前1条
1 项艳;AquaCrop模型在华北地区夏玉米生产中的应用研究[D];山东农业大学;2009年
，本文编号：2191433
本文链接：https://www.wllwen.com/kejilunwen/nykj/2191433.html
