Swin Transformer for Complex Coastal Wetland Classification Using the Integration of Sentinel-1 and Sentinel-2 Imagery

The emergence of deep learning techniques has revolutionized the use of machine learning algorithms to classify complicated environments, notably in remote sensing. Convolutional Neural Networks (CNNs) have shown considerable promise in classifying challenging high-dimensional remote sensing data, particularly in the classification of wetlands. State-of-the-art Natural Language Processing (NLP) algorithms, on the other hand, are transformers. Despite the fact that transformers have been utilized for a few remote sensing applications, they have not been compared to other well-known CNN networks in complex wetland classification. As such, for the classification of complex coastal wetlands in the study area of Saint John city, located in New Brunswick, Canada, we modified and employed the Swin Transformer algorithm. Moreover, the developed transformer classifier results were compared with two well-known deep CNNs of AlexNet and VGG-16. In terms of average accuracy, the proposed Swin Transformer algorithm outperformed the AlexNet and VGG-16 techniques by 14.3% and 44.28%, respectively. The proposed Swin Transformer classifier obtained F-1 scores of 0.65, 0.71, 0.73, 0.78, 0.82, 0.84, and 0.84 for the recognition of coastal marsh, shrub, bog, fen, aquatic bed, forested wetland, and freshwater marsh, respectively. The results achieved in this study suggest the high capability of transformers over very deep CNN networks for the classification of complex landscapes in remote sensing.

Salinity reduces coastal marsh respiration more than photosynthesis

A paradigm in carbon cycling science predicts that sea-level rise will enhance carbon accumulation in an apparent negative carbon-climate feedback1,2. However, ecosystems exposed to combinations of stressors and subsidies – such as saltwater intrusion and sea-level rise – may adapt, transition to an alternative state, or experience a decline in functions, such as carbon storage, thereby altering their response trajectories to environmental changes3,4. Climate change is increasing salinity in coastal ecosystems worldwide yet the effects on ecosystem metabolism remain uncertain4-8. Here, we synthesized gross ecosystem productivity (GEP), ecosystem respiration [CO2 and CH4 (ERCO2 and ERCH4)], and net ecosystem productivity (NEP) from diverse coastal marshes exposed to experimental additions and observational gradients in salinity. Increases in salinity generally caused decreases in median GEP, ERCO2, and ERCH4 but increases in GEP and NEP from ~5 to 10 ppt. Increased saltwater intrusion can stimulate or stress wetlands based on relative exposure and acclimation to increased salinities, and we detected positive NEP where salinity increases had greater negative effects on ERCO2 and ERCH4 than GEP. Although increases in NEP are detectable at low salinities, saltwater intrusion and climate-driven disturbances may reduce carbon storage capacity of coastal ecosystems as productivity declines toward higher salinities.

Understanding the Fates and Chemical Compositions of Weathered Oil in Coastal Marshes since the 2010 Deepwater Horizon Oil Spill

Coastal marshes were heavily impacted by the Deepwater Horizon (DWH) oil spill in 2010, with approximately 90% of shoreline impacts occurring in Louisiana's coastal wetlands. Spilled crude oils impact an environment through four major mechanisms: ecosystem exposure to reactive and toxic aromatic compounds; covering and smothering that hinders normal plant and animal physiology; depletion of dissolved oxygen; and disruption of the aquatic food web. Crude oil's ability to cause environmental harm depends upon its composition, which is a very complex mixture of many thousands of reduced carbon compounds made from the degradation of plant material deposited deep underground. This study reviews the results from the chemical characterization of petroleum hydrocarbons, at various weathering stages, in >2000 marsh surface sediments and select sediment cores samples collected from various sampling locations in Terrebonne Bay, Grand isle, and northern Barataria Bay from 2010 to 2018. The sediment samples were analyzed for target saturated alkanes, polycyclic aromatic compounds, and the forensic biomarker (hopane and sterane) compounds. The chemical characterization of the compositional changes of target compounds in DWH oil, from its pre-stranding stage just offshore in the Louisiana Bight, through stranding on marshy shorelines and through its degradation and weathering over eight years has given insights into the complexity of oil residues and potential for impacts in these varying environmental conditions. Stranded oil initially had two prominent fates: settling on surface sediment/soils of the marshes, and subsurface deposition primarily by means of settling into fiddler crab burrows. Both initial fates affected shorelines and 10–20 meters inward. Over time, surface oil residues were spread beyond initially impacted areas by Tropical Storm Isaac in 2012 and other weather events, and oil residues were quickly degraded. Subsurface stranded oil was degraded much more slowly under anaerobic conditions and some was re-released as fairly fresh oil during the coastal erosions caused by DWH surface oiling damage to the marsh plants. However, these re-releases were relatively slow and were quickly aerobically degraded once the stranded oil reached marsh surfaces. There was also evidence of anaerobic degradation of heavily weathered surface oil residues during the 2015 to 2018 timeframe. This eight-year study establishes a very complex narrative between the physical and chemical properties of stranded oil and its interactions with coastal marsh environments.

Quantifying interior coastal marsh degradation at two North Carolina marshes

Marshes along the coast of North Carolina are currently at risk due to ongoing land loss, and as they are highly productive waterways, understanding the processes driving land loss is critical. By focusing on two marshes adjacent to waterways — Roanoke Marsh, south of Manns Harbor, and Mackay Island National Wildlife Refuge — we created a dataset of land loss rates from 1983 to 2016, both within the marsh interior (due to expanding ponds) and on the outer edge of the marsh (coastal retreat). We investigated the hypothesized primary driver behind the interior pond expansion (wind-driven waves in the pond interior) and the coastal edge retreat of the marsh (wind-driven waves within Currituck Sound). The total land area lost over the 34- year study period was 1.49 km2 and 2.32 km2 on Roanoke Marsh and Mackay Island, respectively. The percentage of total land lost due to internal pond expansion was 60% in Roanoke Marsh and 44% in Mackay Island. Internal pond expansion is at least of equal importance to outer coastal retreat for net land loss in these coastal marshes. The local wind has a dominant direction from the north-northwest with more energetic winds during the winter. However, the wind directions and direction of pond expansion do not appear to be correlated. This may be because the winds are bimodal and drive expansion in multiple directions. In addition, there is subsidence in this portion of North Carolina that may be an additional factor contributing to the pond area expansion.