欢迎访问林业科学,今天是

林业科学 ›› 2022, Vol. 58 ›› Issue (2): 42-48.doi: 10.11707/j.1001-7488.20220205

• 前沿与重点: 森林碳汇专题 • 上一篇    下一篇

基于GA-BP神经网络的雷竹林CO2浓度反演

侯志康1,曾松伟1,*,莫路锋1,周宇峰2   

  1. 1. 浙江农林大学信息工程学院 杭州 311300
    2. 浙江农林大学环境与资源学院 杭州 311300
  • 收稿日期:2020-07-02 出版日期:2022-02-25 发布日期:2022-04-26
  • 通讯作者: 曾松伟
  • 基金资助:
    国家自然科学基金两化融合重点项目(U1809208);浙江省自然基金公益项目(LGN18C200017)

CO2 Concentration in Phyllostachys praecox Stand Inversion Based on GA-BP Neural Network

Zhikang Hou1,Songwei Zeng1,*,Lufeng Mo1,Yufeng Zhou2   

  1. 1. College of Information Engineering, Zhejiang A & F University Hangzhou 311300
    2. College of Environment and Resources Zhejiang A & F University Hangzhou 311300
  • Received:2020-07-02 Online:2022-02-25 Published:2022-04-26
  • Contact: Songwei Zeng

摘要:

目的: 研发竹林气象因子采集系统, 分析雷竹林CO2浓度与温湿度等气象因子之间的关系, 探讨基于GA-BP神经网络的雷竹林CO2浓度反演模型(简称GA-BP模型), 为竹林碳储量、竹林增汇、竹林固碳能力等研究提供基础数据。方法: 根据微气象学相关原理、方法及森林碳通量动态感知的需求, 设计基于嵌入式的森林碳通量数据远程实时监测系统, 该监测系统以成熟雷竹林为监测对象, 进行为期2个月(2019年10—11月)的气象数据监测; 在此基础上, 提出GA-BP模型。结果: 根据GA-BP模型和BP模型反演的结果可知: GA-BP模型反演结果的决定系数R2为0.86, 比BP模型的R2(0.79)提高了0.07; 平均绝对误差为8.12 mg·m-3, 比BP神经网络下降2.79 mg·m-3。GA-BP模型相较于BP网络具有更稳定的反演性能和更高的反演精度。结论: 可以利用竹林气象因子采集系统获取相关气象数据; 基于CO2浓度与温湿度等气象因子之间的相关性, 本研究提出的基于GA-BP神经网络的CO2浓度反演模型能够有效反演研究区的CO2浓度。

关键词: 生态系统, 碳通量, GA-BP, 碳储量, 雷竹林

Abstract:

Objective: The purpose of this work is to develop meteorological factor acquisition system of Phyllostachys praecox stand, to obtain the relationship between CO2 concentration and meteorological factors (temperature and humidity, etc.), to discusses the CO2 concentration inversion model based on GA-BP neural network (abbreviated as GA-BP model), and to provide fundamental data for carbon storage, carbon sinks and carbon sequestration capacity of P. praecox stand. Method: According to the relevant principles and methods of micrometeorology and the requirements of dynamic sensing of forest carbon flux, a remote and real-time monitoring system of forest carbon flux data based on embedded system is designed. Taking the mature P. praecox stand, stand as monitored object, this system monitored the environmental data for two months (October ~ November 2019). After analyzing these data, a CO2 concentration inversion model based on genetic classification optimization neural network is proposed. Results: According to the inversion results of GA-BP and BP inversion model, the determinative coefficient R2 of the inversion results of GA-BP inversion model is 0.86, which is 7 percentage points higher than that of BP inversion model. The mean absolute error is 8.12 mg·m-3, which is 2.79 mg·m-3 lower than that of BP inversion model. Compared with the BP inversion model, the GA-BP inversion model has more stable inversion performance and higher inversion accuracy. Conclusion: The P. praecox stand meteorological factor acquisition system can be used to obtain relevant meteorological data. Based on the correlation between CO2 concentration and meteorological factors (temperature and humidity, etc.), the CO2 concentration inversion model based on GA-BP neural network can effectively invert the CO2 concentration data in the survey region.

Key words: ecosystem, carbon flux, GA-BP, carbon storage, Phyllostachys praecox stand

中图分类号: