Introduction: Marine pollution can have a significant impact on the blue carbon, which finally affect the ocean’s ability to sequester carbon and contribute to achieving carbon neutrality. Marine pollution is a complex problem that requires a great deal of time and effort to measure. Existing machine learning algorithms cannot effectively solve the detection time problem and provide limited accuracy. Moreover, marine pollution can come from a variety of sources. However, most of the existing research focused on a single ocean indicator to analyze marine pollution. In this study, two indicators, marine organisms and debris, are used to create a more complete picture of the extent and impact of pollution in the ocean.Methods: To effectively recognize different marine objects in the complex marine environment, we propose an integrated data fusion approach where deep convolutional neural networks (CNNs) are combined to conduct underwater object recognition. Through this multi-source data fusion approach, the accuracy of object recognition is significantly improved. After feature extraction, four machine and deep learning classifiers’ performances are used to train on features extracted with deep CNNs.Results: The results show that VGG-16 achieves better performance than other feature extractors when detecting marine organisms. When detecting marine debris, AlexNet outperforms other deep CNNs. The results also show that the LSTM classifier with VGG-16 for detecting marine organisms outperforms other deep learning models.Discussion: For detecting marine debris, the best performance was observed with the AlexNet extractor, which obtained the best classification result with an LSTM. This information can be used to develop policies and practices aimed at reducing pollution and protecting marine environments for future generations.

This study evaluated the simulated cloud radiative feedbacks (CRF) during the El Nino-Southern Oscillation (ENSO) cycle in the latest version of the Chinese Academy of Meteorological Sciences climate system model (CAMS-CSM). We conducted two experimental model simulations: the Atmospheric Model Intercomparison Project (AMIP), forced by the observed sea surface temperature (SST); and the preindustrial control (PIcontrol), a coupled run without flux correction. We found that both the experiments generally reproduced the observed features of the shortwave and longwave cloud radiative forcing (SWCRF and LWCRF) feedbacks. The AMIP run exhibited better simulation performance in the magnitude and spatial distribution than the PIcontrol run. Furthermore, the simulation biases in SWCRF and LWCRF feedbacks were linked to the biases in the representation of the corresponding total cloud cover and precipitation feedbacks. It is interesting to further find that the simulation bias originating in the atmospheric component was amplified in the PIcontrol run, indicating that the coupling aggravated the simulation bias. Since the PIcontrol run exhibited an apparent mean SST cold bias over the cold tongue, the precipitation response to the SST anomaly (SSTA) changes during the ENSO cycle occurred towards the relatively warmer western equatorial Pacific. Thus, the corresponding cloud cover and CRF shifted westward and showed a weaker magnitude in the PI-control run versus observational data. In contrast, the AMIP run was forced by the observational SST, hence representing a more realistic CRF. Our results demonstrate the challenges of simulating CRF in coupled models. This study also underscores the necessity of realistically representing the climatological mean state when simulating CRF during the ENSO cycle.