基于集成学习的内陆水体叶绿素a浓度反演
Chlorophyll-a Concentration Retrieval in Inland Water Based on Ensemble Learning
孟黎 1孟静2
作者信息
- 1. 山东城市建设职业学院,山东 济南 250103
- 2. 山东省国土测绘院,山东 济南 250102
- 折叠
摘要
利用卫星数据监测内陆或水质状态,对生态决策具有重要意义.基于具有高时空分辨率的哨兵二号卫星数据,联合2种集成学习算法反演山东省南四湖叶绿素a(Chla)浓度,结果表明经遥感反射率校正后的哨兵二号数据更加适用于水质反演.XGBoost模型在五折交叉验证反演结果上表现最优(R2=0.732 5,RMSE=9.168 1 μg/L),反演结果更符合实际情况.因此,使用该模型反演南四湖叶绿素a浓度,能较好地掌握其时空变化情况,对其他区域类似研究可提供一定参考.
Abstract
Using satellite data to monitor inland or water quality status is of great significance for ecological decision-making.The concentration of Chlorophyll-a(Chla)in Nansi lake,Shandong Province is retrieved by combining two ensemble learning algorithms,based on Sentinel-2 satellite data with high spatiotemporal resolution.The results show that Sentinel-2 data corrected for remote sensing reflectance are more suitable for water quality inversion.The XGBoost model performs optimally on the 5-fold cross-validation inversion results(R2=0.732 5,RMSE=9.168 1 μg/L),making the inversion results more realistic.Therefore,using this model to invert the Chla concentration in the Nansi lake can provide a better understanding of its spatiotemporal variability,and the conclusions of this paper can provide some reference for similar studies in other regions.
关键词
哨兵二号数据/南四湖/叶绿素a/集成学习Key words
sentinel-2 data/Nansi Lake/Chlorophyll-a/ensemble learning引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
维普
万方数据