Recently, Researcher He Xianqiang and his collaborators have published a research paper titled Satellite-estimation of the Global Ocean Primary Productivity via BGC-Argo Measurements in the renowned international marine journal Journal of Geophysical Research: Oceans. The first author is Zhang Yinxue, a joint-cultivated doctoral student trained by our laboratory and Hohai University, and Researcher He Xianqiang is the corresponding author. Co-authors include Researcher Bai Yan, Associate Researcher Li Teng, Researcher Wang Difeng, Senior Engineer Gong Fang, Professor-level Senior Engineer Zhu Qiankun from our institute, and Associate Professor Wang Guifen from Hohai University.
Although the biomass of marine phytoplankton is far lower than that of terrestrial vegetation, it contributes nearly 50% of global net primary productivity (NPP) and plays a vital role in the global carbon cycle. Decades of in-situ observations have laid a basic understanding of the spatial distribution pattern of global marine NPP. Nevertheless, field measurement of marine primary productivity involves complicated procedures and extremely low sampling efficiency, resulting in scarce measured samples and limited spatiotemporal coverage.
Since the launch of the world’s first ocean color satellite sensor in 1978, developing remote sensing retrieval algorithms to realize long-term monitoring of global marine NPP has been a core goal of satellite ocean color observation. However, the shortage of field data and the limitation that satellite remote sensing can only capture surface-layer information lead to high uncertainties in existing marine NPP retrieval models. Previous international intercomparison projects of marine primary productivity remote sensing algorithms revealed that the estimated annual total global marine NPP derived from different models can differ by up to 30 Pg C, which greatly hinders the accurate recognition of long-term variation characteristics and driving mechanisms of global marine NPP.
Massive in-situ vertical bio-optical observation data obtained from the Biogeochemical Argo (BGC-Argo) program can effectively make up for the deficiency of field samples and the limitation of satellite surface observation. BGC-Argo is capable of observing various biogeochemical and optical parameters including chlorophyll-a (Chl-a) and particulate backscattering coefficient at 700 nm (bbp(700)). Theoretically, vertical profile data from BGC-Argo can be adopted to calculate NPP, and such in-situ derived NPP data can provide abundant training samples for remote sensing model construction.
Following this technical route, this study established a large-scale quasi-in-situ NPP dataset by integrating BGC-Argo profile measurements and the improved Satellite_CbPM model. This dataset effectively compensates for the lack of global in-situ NPP observations and expands the quantity and spatial coverage of available NPP samples. On the basis of the quasi-in-situ dataset, a high-precision marine NPP remote sensing retrieval model named XGBoost_CbPM was constructed using the XGBoost machine learning algorithm, which supports long-term global ocean NPP estimation based on satellite remote sensing data.
Independent verification results show that the model achieves a coefficient of determination of 0.87 and a mean absolute deviation of 12.52%, proving its outstanding retrieval accuracy. Validated against two long-term observation stations (BATS and HOT), the XGBoost_CbPM model outperforms the traditional Satellite_CbPM model in retrieving both magnitude and temporal variation trends of water column integrated NPP. It also presents favorable performance in vertical NPP profile inversion, with a determination coefficient of 0.57 for NPP profile retrieval at the HOT station, and can precisely capture seasonal variation features of vertical NPP distribution.
In general, this study leverages BGC-Argo profile data to fill the gap of scarce in-situ NPP measurements and greatly improve the accuracy of global marine NPP remote sensing retrieval. The NPP products generated by this method will facilitate further exploration on the evolutionary characteristics and mechanisms of the global marine carbon cycle.
This research is jointly funded by the National Natural Science Foundation of China, Zhejiang Provincial Natural Science Foundation of China, Zhejiang Provincial Pioneer R&D Program and other research projects.
Fig.1 Spatial distribution of global quasi-in-situ primary productivity samples in (a) spring, (b) summer, (c) autumn and (d) winter. The numbers in brackets represent sample sizes in different seasons. (e) Frequency distribution histogram of the quasi-in-situ dataset
Fig.2 Spatial distribution of monthly mean ocean primary productivity in 2021 retrieved by the XGBoost_CbPM model
Fig.3 Magnitude comparison between model-retrieved and in-situ observed vertical NPP profiles at (a) BATS station and (b) HOT station
Citation
Zhang, Y., He, X., Bai, Y., et al. (2025). Satellite-Estimation of the Global Ocean Primary Productivity via BGC-Argo Measurements.
Journal of Geophysical Research: Oceans, 130(4), e2024JC021163.
https://doi.org/10.1029/2024JC021163
