As the fastest growing trend in big data analysis, deep learning technology has proven to be both an unprecedented breakthrough and a powerful tool in many fields, particularly for image segmentation tasks. Nevertheless, most achievements depend on high-quality pre-labeled training samples, which are labor-intensive and time-consuming. Furthermore, different from conventional natural images, coastal remote sensing ones generally carry far more complicated and considerable land cover information, making it difficult to produce pre-labeled references for supervised image segmentation. In our research, motivated by this observation, we take an in-depth investigation on the utilization of neural networks for unsupervised learning and propose a novel method, namely conditional co-training (CCT), specifically for truly unsupervised remote sensing image segmentation in coastal areas. In our idea, a multi-model framework consisting of two parallel data streams, which are superpixel-based over-segmentation and pixel-level semantic segmentation, is proposed to simultaneously perform the pixel-level classification. The former processes the input image into multiple over-segments, providing self-constrained guidance for model training. Meanwhile, with this guidance, the latter continuously processes the input image into multi-channel response maps until the model converges. Incentivized by multiple conditional constraints, our framework learns to extract high-level semantic knowledge and produce full-resolution segmentation maps without pre-labeled ground truths. Compared to the black-box solutions in conventional supervised learning manners, this method is of stronger explainability and transparency for its specific architecture and mechanism. The experimental results on two representative real-world coastal remote sensing datasets of image segmentation and the comparison with other state-of-the-art truly unsupervised methods validate the plausible performance and excellent efficiency of our proposed CCT.
The Amundsen Sea (AS) sector in West Antarctica accounts for a significant proportion of Earth's ice losses and is the largest contributor of Antarctica's mass loss. To evaluate its contribution to global sea‐level rise, we reconstruct the long‐term continuous surface elevation changes (CSEC) record of the AS sector by an improved least‐squares plane fitting method (ILSPFM), which merged the relative surface elevation change (SEC) series instead of height from Envisat, ICESat, CryoSat‐2, and ICESat‐2 missions during 2003–2021. The accuracy of CSEC is improved by 25.9% using ILSPFM. The average rate of CSEC in the AS sector was −24.25 ± 0.48 cm yr−1 during 2003–2021. The largest signals of SEC are found over Pine Island, Thwaites, and Pope Glaciers, with the largest decline of SEC over Pope Glacier with a total SEC of −82.44 ± 7.21 m and an annual change rate of −4.34 ± 0.38 m yr−1. The ridge between Pine Island and Thwaites Glaciers is found in the AS sector, indicating that the change of ice sheet is dynamic thinning and closely related to the topography and the distance from the grounding line. Compared with meteorological data sets, we find that the codirectional fluctuation in CSEC is delayed by 3 months with surface temperature, and the precipitation leading SEC series as the phase arrow points straight down from the cross wavelet transform. Our new record shows that the AS sector thinned rapidly from 2003 to 2021 but decelerated from 2019 to 2021, and it was clearly correlated to the surface temperature, precipitation, and local terrain.
Satellite altimeters have been used to monitor Arctic sea ice (ASI) thickness for several decades, but whether the different altimeter missions (such as radar and laser altimeters) are in agreement with each other and suitable for long-term research needs to be investigated. To analyze the spatiotemporal characteristics of ASI, continuous long-term first-year ice, and multi-year ice of ASI freeboard, thickness, and volume from 2002 to 2021 using the gridded nadirization method from Envisat, CryoSat-2, and ICESat-2, altimeter data are comprehensively constructed and assessed. The influences of sea surface temperature (SST) and sea surface wind field (SSW) on ASI are also discussed. The freeboard/thickness and extent/area of ASI all varied seasonally and reached their maximum and minimum in April and October, March and September, respectively. From 2002 to 2021, the freeboard, thickness, extent, and area of ASI all consistently showed downward trends, and sea ice volume decreased by 5437 km3/month. SST in the Arctic rose by 0.003 degrees C/month, and the sea ice changes lagged behind this temperature variation by one month between 2002 and 2021. The meridional winds blowing from the central Arctic region along the eastern coast of Greenland to the North Atlantic each month are consistent with changes in the freeboard and thickness of ASI. SST and SSW are two of the most critical factors driving sea ice changes. This study provides new data and technical support for monitoring ASI and exploring its response mechanisms to climate change.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.