在LSTM网络中,即使有不完整的时空数据,也能提供有效的蓝藻细菌繁殖预测
Claudia Fournier1, Raúl Fernandez-Fernandez2, Samuel Cirés1
1Departamento de Biología, Universidad Autónoma de Madrid, 28049 Madrid, Spain.
Water research
|October 10, 2024
概括
一个有效的早期预警系统 (EWS) 预测了使用 phycocyanin (PC) 数据的蓝藻细菌繁殖. 多变量长期短期 (LSTM) 神经网络模型准确地预测到28天前的开花.
科学领域:
- 环境科学 环境科学
- 生态生态学 生态生态学
- 水质管理水质管理
背景情况:
- 由于人类活动和气候变化,内陆水域的蓝藻细菌繁殖量正在增加.
- 这些开花威胁到生态系统的健康和水质,特别是产生毒素的菌株.
- 早期预警系统 (EWS) 对于及时管理蓝藻细菌的开花至关重要.
研究的目的:
- 开发和评估一个有效的EWS来预测蓝菌花朵的发展.
- 用时空数据比较六种不同的预测模型的性能.
- 通过使用混合评估系统,在多个时间范围内评估预测准确性.
主要方法:
- 利用来自多参数探测器的6年不完整的高频时空数据,专注于植物 (PC) 光.
- 开发了一种探测无关的方法,用于数据预处理和时间序列生成,用于花预测.
- 使用回归,分类和技能指标,比较了六种预测模型 (线性回归,随机森林,LSTM - 自动回归和多变量).
主要成果:
- 多变量长期短期 (LSTM) 神经网络在所有预测视野和指标中显示出最佳和最一致的性能.
- 在预测拟议的PC报警水平 (10微克PC/L) 时,LSTM达到高达90%的准确性.
- 积极的技能值证实了LSTM在提前16至28天预测蓝藻细菌开花方面的有效性.
结论:
- 拟议的EWS,特别是使用多变量LSTM,对于预测蓝藻细菌开花非常有效.
- 该系统提供了显著的预警,使主动管理策略成为可能.
- 该方法为监测和预测内陆水域有害藻类繁殖提供了可复制的方法.
更多相关视频
05:21Operation of Laboratory Photobioreactors with Online Growth Measurements and Customizable Light Regimes
Published on: October 28, 2021
2.1K
05:44Assembly and Quantification of Co-Cultures Combining Heterotrophic Yeast with Phototrophic Sugar-Secreting Cyanobacteria
Published on: December 27, 2024
669
相关概念视频
01:25Microbial Mats
Microbial communities forming biofilms and mats represent complex, spatially structured ecosystems where metabolic processes are stratified according to light, oxygen, and nutrient gradients. Biofilms are initial colonization stages, only a few millimeters thick, while mature microbial mats can reach centimeter-scale thickness and display intricate vertical organization. Their structural and functional heterogeneity allows microorganisms to occupy distinct ecological niches within a few...
01:24Freshwater Microbial Ecology
Freshwater systems such as streams, rivers, and lakes exhibit distinct physical and biological characteristics that influence their microbial communities. These environments are broadly categorized into lotic systems—those with flowing waters like streams and most rivers—and lentic systems, which include still or slow-moving waters such as lakes, ponds, and marshes.In lentic systems, phytoplankton drive primary production, generating autochthonous organic carbon. In contrast, lotic systems...