School of Geographical Sciences, China West Normal University, Nanchong 637001, China; Sichuan Provincial Engineering Laboratory of Monitoring and Control for Soil Erosion in Dry Valleys, China West Normal University, Nanchong 637001, China.
School of Geoscience and Technology, Southwest Petroleum University, Chengdu 610500, China; Xizang Autonomous Region Key Laboratory of Satellite Remote Sensing and Application, Lhasa 851400, China.
Sci Total Environ. 2024 Oct 1;945:174076. doi: 10.1016/j.scitotenv.2024.174076. Epub 2024 Jun 21.
Chlorophyll-a (Chl-a) is a crucial pigment in algae and macrophytes, which makes the concentration of total Chl-a in the water column (total Chl-a) an essential indicator for estimating the primary productivity and carbon cycle of the ocean. Integrating the Chl-a concentration at different depths (Chl-a profile) is an important way to obtain the total Chl-a. However, due to limited cost and technology, it is difficult to measure Chl-a profiles directly in a spatially continuous and high-resolution way. In this study, we proposed an integrated strategy model that combines three different machine learning methods (PSO-BP, random forest and gradient boosting) to predict the Chl-a profile in the Mediterranean by using several sea surface variables (photosynthetically active radiation, spectral irradiance, sea surface temperature, wind speed, euphotic depth and KD490) and subsurface variables (mixed layer depth) observed by or estimated from satellite and BGC-Argo float observations. After accuracy estimation, the integrated model was utilized to generate the time series total Chl-a in the Mediterranean from 2003 to 2021. By analysing the time series results, it was found that seasonal fluctuation contributed the most to the variation in total Chl-a. In addition, there was an overall decreasing trend in the Mediterranean phytoplankton biomass, with the total Chl- decreasing at a rate of 0.048 mg/m per year, which was inferred to be related to global warming and precipitation reduction based on comprehensive analysis with sea surface temperature and precipitation data.
叶绿素 a（Chl-a）是藻类和大型水生植物中的关键色素，使得水柱中总 Chl-a（total Chl-a）的浓度成为估算海洋初级生产力和碳循环的重要指标。整合不同深度的 Chl-a 浓度（Chl-a 剖面）是获取总 Chl-a 的重要方法。然而，由于成本和技术的限制，很难以空间连续和高分辨率的方式直接测量 Chl-a 剖面。在本研究中，我们提出了一种综合策略模型，该模型结合了三种不同的机器学习方法（PSO-BP、随机森林和梯度提升），利用卫星和 BGC-Argo 浮标观测到的或估算的几个海面变量（光合有效辐射、光谱辐照度、海面温度、风速、透光深度和 KD490）和次表层变量（混合层深度）来预测地中海的 Chl-a 剖面。在进行精度估计后，该综合模型被用于生成 2003 年至 2021 年地中海的时间序列总 Chl-a。通过分析时间序列结果，发现季节性波动对总 Chl-a 的变化贡献最大。此外，地中海浮游植物生物量呈总体下降趋势，总 Chl-a 每年减少 0.048 mg/m，根据与海面温度和降水数据的综合分析，这被推断与全球变暖以及降水减少有关。
