Climate change, water quality, and water diversion are associated with the shifts of aquatic vegetation structure and phenology in five temperate regulating lakes of eastern China.

Abstract

Aquatic plants are vital for lake ecosystem functioning and water-quality stability, yet their community dynamics and phenology rhythms remain insufficiently understood, due to the lack of effective strategies for fine-scale species mapping and phenology extraction. In this study, based on Sentinel-2 MSI imagery, we developed an integrated framework combining machine learning and a priori ecological knowledge to quantify the spatiotemporal changes in aquatic plant distribution, species composition and phenological dynamics for eight dominant species in five regulating lakes along the Eastern Route of the South-to-North Water Diversion Project in China. Results showed that the proposed framework enabled accurate aquatic plant identification, achieving an overall classification accuracy of 96.16% and over 90% accuracy for each species. Since 2016, aquatic vegetation coverage has substantially declined in most lakes, mainly due to the retreat of submerged vegetation. Community structure has shifted from submerged-plant dominance to emergent and floating-leaved dominance in two of them. Phenologically, we found that most aquatic vegetation exhibited a longer growing season, characterized by earlier growth onset (-0.28 days/year) and peak timing (-0.67 days/year) and delayed senescence (0.54 days/year). Correlation analysis indicated that aquatic vegetation dynamics was associated with climate variation, nutrient enrichment, turbidity, and water diversion, with warming and solar radiation likely promoting the growth of some emergent species, while nutrient enrichment and turbidity could be linked with submerged vegetation decline and earlier phenological shifts. Overall, this study provides an effective framework for species-level mapping and phenological monitoring of aquatic vegetation, offering valuable support for the management and conservation of lake ecosystems.

Plant phenotyping relevance

Sentinel-2画像と機械学習等を統合した、植物種分布・構成・フェノロジーを抽出する手法を開発し、精度検証と大規模適用を行っているため、植物フェノタイピング手法が中心である。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.