欢迎访问《应用生态学报》官方网站,今天是 分享到:

应用生态学报 ›› 2023, Vol. 34 ›› Issue (8): 2123-2132.doi: 10.13287/j.1001-9332.202308.019

• • 上一篇    下一篇

基于神经网络优化模型的丝绵木瞬态液流模拟

周鹏1, 韩磊2,3,4*, 彭苓1, 柳利利2,3,4, 王娜娜2,3,4, 马军1, 马云蕾2,3,4   

  1. 1宁夏大学农学院, 银川 750021;
    2宁夏大学地理科学与规划学院, 银川 750021;
    3中阿旱区特色资源与环境治理国际合作联合实验室, 银川 750021;
    4宁夏旱区资源评价与环境调控重点实验室, 银川 750021
  • 收稿日期:2023-02-21 接受日期:2023-05-25 出版日期:2023-08-15 发布日期:2024-02-15
  • 通讯作者: *E-mail: layhan@163.com
  • 作者简介:周 鹏, 男, 1998年生, 硕士研究生。主要从事干旱区生态水文过程研究。E-mail: zhoupeng0235@163.com
  • 基金资助:
    宁夏自然科学基金项目(2022AAC03094)和国家自然科学基金项目(31760236)

Instantaneous sap flow velocity simulation of Euonymus bungeanus based on neural network optimization model

ZHOU Peng1, HAN Lei2,3,4*, PENG Ling1, LIU Lili2,3,4, WANG Nana2,3,4, MA Jun1, MA Yunlei2,3,4   

  1. 1College of Agriculture, Ningxia University, Yinchuan 750021, China;
    2School of Geography and Planning, Ningxia University, Yinchuan 750021, China;
    3China-Arab Joint International Research Laboratory for Featured Resources and Environmental Governance in Arid Regions, Yinchuan 750021, China;
    4Ningxia Key Laboratory of Resource Environmental Regulation in Arid Region, Yinchuan 750021, China
  • Received:2023-02-21 Accepted:2023-05-25 Online:2023-08-15 Published:2024-02-15

摘要: 树木的液流规律是复杂的,难以用多元线性或经验模型表达,在理解林木树干液流规律的基础上,寻找一种简易可行的方法模拟林木树干液流对环境因子的响应过程,对定量分析森林生态水文过程及区域生态需水量尤为重要。本研究以宁夏河东沙区防护林树种丝绵木为对象,采用热扩散茎流计连续测定树干液流速率,分析环境因子对丝绵木树干液流的影响,并构建基于粒子群算法(PSO)和麻雀搜索算法(SSA)优化的神经网络模型对丝绵木液流速率进行预测。结果表明: 影响丝棉木树干液流的主要因素为太阳辐射、饱和水汽压差、气温和相对湿度,重要度依次为32.5%、25.3%、22.0%和16.1%,其响应过程均呈现时滞回环关系。采用优化后的BP、Elman和ELM神经网络模型模拟瞬态液流,综合评价指标(GPI)分别提高1.5%、30.0%和5.3%。但与PSO-Elman和SSA-ELM优化模型相比,SSA-BP优化模型预测结果最佳,GPI分别提高1.0%和23.2%。基于麻雀搜索算法优化的BP神经网络模型可以作为预测丝绵木树干瞬态液流速率的首选模型。

关键词: 神经网络, 优化算法, 瞬态液流, 模拟, 预测, 丝绵木

Abstract: The sap flow of trees is complex and difficult to express with multivariate linear or empirical models. A simple and feasible method on the basis of understanding sap flow variation to simulate its variation with environmental factors is of special importance for quantitatively analyzing forest ecohydrological processes and regional water demand. In this study, with one of the shelter forest species Euonymus bungeanus in the east sandy land of Yellow River in Ningxia as the research object, we continuously measured the trunk sap flow velocity by thermal diffusion sap flow meter, and analyzed the effects of environmental factors on stem sap flow. We used the particle swarm optimization (PSO) and sparrow search algorithm (SSA) optimized neural network model to predict sap flow velocity of E. bungeanus. Results showed that the main environmental factors influencing sap flow were solar radiation, vapor pressure deficit, air temperature, and relative humidity, with the influencing importance of 32.5%, 25.3%, 22.0% and 16.1%, respectively. The response process between sap flow and environmental factors presented a hysteresis loop relationship. The optimized BP, Elman and ELM neural network models improved the comprehensive evaluation index (GPI) by 1.5%, 30.0% and 5.3%, respectively. Compared with the PSO-Elman and SSA-ELM optimization models, the SSA-BP optimization model had the best prediction results with an improvement of 1.0% and 23.2% in GPI, respectively. Therefore, the prediction results of the BP neural network model based on the sparrow search algorithm could be used as an optimal model for predicting instantaneous sap flow velocity of E. bungeanus.

Key words: neural network, optimization algorithm, instantaneous sap flow, simulation, prediction, Euonymus bungeanus.