森林土壤呼吸监测与评价方法研究

发布时间：2018-01-26 09:42

本文关键词： 土壤呼吸分布式监测土壤温度监测点布置贝叶斯最大熵　出处：《北京林业大学》2016年博士论文　论文类型：学位论文

【摘要】：土壤是一个巨大的碳库,总储存量达到1394PgC,土壤呼吸占整个陆地生态系统呼吸的600%-90%,其微小的变化都可能引起大气CO2浓度较大的改变,是导致全球气候变化的关键过程。森林是陆地生态系统的主体,其土壤中的碳贮量占全球土壤碳库的73%。故探索森林土壤呼吸的监测手段和评价方法,不仅对于碳汇计量、碳交易、缓解气候变化具有重要的战略意义,而且对森林碳储量的经营管理具有重要的指导价值。论文首先研究土壤呼吸监测平台,然后研究土壤温度空间分布,以土壤温度指导监测点布置理论,并把土壤温度作为软数据结合贝叶斯最大熵方法提高土壤呼吸预测的可靠性,最后研发土壤呼吸分析与评价软件,为研究者分析与评价土壤呼吸提供便利的方法。本论文主要形成以下结论：(1)研究了基于气体扩散定律的土壤呼吸仪,在仪器内设置多层传感器采集数据,建立CO2在仪器内的扩散模型计算得到土壤碳通量值,此土壤呼吸监测仪平均误差为10%左右。然后通过AD-HOC与GPRS无线通信技术,组建土壤呼吸分布式监测网络,实现多点、同时监测。研究表明在实验区域内仅用一个采样点进行土壤呼吸监测平均误差为42%左右,说明单一点监测会产生较大误差,用分布式监测网络可以减小监测误差。(2)研究了基于BP神经网络的土壤温度空间分布方法。针对土壤温度分布具有不确定性和非线性等特征,提出分组训练、组合优化的改进BP神经网络算法,此算法隐含层节点数量设为10,学习率设为0.75,算法达到较好的性能。并以真实值与预测值进行对比,用R2作为评价标准,样条插值算法,克里金算法,BP神经网络算法的R2分别为0.59,0.73和0.92,而改进BP神经网络算法R2为0.97,说明此算法能有效提高土壤温度空间分布的可靠性。(3)研究了基于土壤温度梯度与图像分割算法的土壤呼吸监测点优化布置方法,以土壤温度为依据,确定实验区土壤呼吸监测点数量设置5个比较合理,并利用区域生长图像分割算法,确定这5个监测点位置。然后以平均误差率为评价标准,均匀网格法确定位置和随机布置方法确定位置的平均误差分别为24%和29%,而以区域生长图像分割算法平均误差为15%,说明此方法能提高土壤呼吸监测点布置的合理性。(4)研究了以土壤温度为辅助信息的贝叶斯最大熵算法,利用土壤温度和土壤呼吸之间的函数关系,把土壤温度作为软数据融入到土壤呼吸估计中,可以提高土壤呼吸空间插值算法的准确性。贝叶斯最大熵算法(BME)的平均CR为0.745,明显高于协同克里金算法(Co-OK)方法的0.505以及普通克里金算法(OK)方法的0.267;BME的平均RMSE结果为0.568,均低于Co-OK方法的0.851以及OK方法的1.511；BME方法的Bias绝对值平均值为0.192,相比较Co-OK方法的0.508以及OK方法的1.143,实验表明BME方法的性能比Co-OK方法和OK方法好。同时,利用BME算法还能减少监测点数量。(5)利用SQL Server数据库存储传感器数据,利用ArcGIS建立土壤呼吸相关图层,利用Visual Studio 2008研发土壤呼吸分析与评价软件,为用户进行土壤呼吸研究提供便利。研究结果为监测与评价森林土壤呼吸提供一套优化的方法,为进一步研究土壤碳的动态变化以及指导碳汇林和森林资源的经营管理作必要的技术支撑。
[Abstract]:The soil is a huge carbon pool, the total storage capacity reached 1394PgC, soil respiration accounted for the entire land ecosystem respiration 600%-90%, the tiny change can cause the change of atmospheric CO2 concentration is larger, which is a critical process of global climate change. The forest is the main terrestrial ecosystem, monitoring and evaluation methods of carbon the storage of soil accounted for the global soil carbon pool 73%. therefore exploring the forest soil respiration, not only for carbon accounting, carbon trading, has an important strategic significance for the mitigation of climate change, but also has an important guiding value for the management of forest carbon storage. The paper firstly studies the soil respiration monitoring platform, and then study the soil temperature the soil temperature distribution, to guide the layout of monitoring points and the theory of soil temperature as the reliability of soft data based on Bayesian maximum entropy method to improve the prediction of soil respiration, finally. Analysis and evaluation of soil respiration software, provide a convenient method for the research of analysis and evaluation of soil respiration. The main conclusions as following: (1) study of soil respiration instrument gas diffusion law based on multilayer data acquisition sensor is arranged on the instrument, the diffusion model is established in CO2 instrument in the calculation of soil carbon flux the value of soil respiration monitor, the average error is about 10%. And then through the AD-HOC and GPRS wireless communication technology, the formation of soil respiration in distributed monitoring network, realize multi point simultaneous monitoring. The results show that only one sampling point of soil respiration monitoring the average error is about 42% in the experimental area, the single point monitoring will produce large error, distributed monitoring network monitoring can reduce the error. (2) studied the spatial distribution of soil temperature method based on BP neural network. According to the soil temperature distribution is The uncertainty and nonlinear characteristics, put forward the team training, improved BP neural network algorithm in combinatorial optimization, the algorithm of hidden layer node number is set to 10, the learning rate is set to 0.75 algorithms to achieve better performance. And the true value and predictive value compared with R2 as the evaluation standard, spline interpolation. Kriging algorithm, BP neural network algorithm R2 0.59,0.73 and 0.92 respectively, and the improved BP neural network algorithm R2 is 0.97, indicating that this algorithm can effectively improve the reliability of the soil temperature distribution. (3) studied the soil temperature gradient image segmentation algorithm and the optimization of monitoring points of soil respiration to soil temperature based method according to determine the number of experimentation area soil respiration monitoring point setting 5 is reasonable, and the use of regional growth image segmentation algorithm, determine which of the 5 monitoring points. Then as the evaluation standard to the average error rate, uniform grid Method to determine the location and random layout method to determine the average error of location is 24% and 29% respectively, while in the region growing image segmentation algorithm the average error is 15%, indicating that this method can improve the rationality of the arrangement of soil respiration monitoring. (4) studied the Bias maximum entropy algorithm of the auxiliary information to the soil temperature, the relationship between soil the soil temperature and the soil temperature between breath, as soft data into the estimation of soil respiration, can improve the accuracy of the interpolation algorithm of breathing space soil. Bias maximum entropy algorithm (BME) and the average CR was 0.745, significantly higher than the collaborative Krikin algorithm (Co-OK) method and 0.505 ordinary Krikin algorithm (OK) for 0.267; the average RMSE BME was 0.568, 1.511 and 0.851 were lower than that of Co-OK method OK method; BME method Bias the absolute value of the average value is 0.192, compared to 0.508 Co-OK method 1.143 and OK method, experiments show that the performance of BME method than the Co-OK method and OK method. At the same time, using the BME algorithm can reduce the number of monitoring points. (5) the use of sensor data storage SQL Server database, using ArcGIS to establish soil respiration related layer, using Visual Studio 2008 software R & D analysis and evaluation of soil respiration, provide convenience study of soil respiration for users. The results provide a set of optimization methods for forest soil respiration monitoring and evaluation, the management for further research on dynamic changes of soil carbon and carbon sequestration forest and forest resources as necessary technical support.

【学位授予单位】：北京林业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：S714.3

【相似文献】