Integrating SAR and Optical Remote Sensing for Conservation-Targeted Wetlands Mapping

The Prairie Pothole Region (PPR) contains numerous depressional wetlands known as potholes that provide habitats for waterfowl and other wetland-dependent species. Mapping these wetlands is essential for identifying viable waterfowl habitat and conservation planning scenarios, yet it is a challenging task due to the small size of the potholes, and the presence of emergent vegetation. This study develops an open-source process within the Google Earth Engine platform for mapping the spatial distribution of wetlands through the integration of Sentinel-1 C-band SAR (synthetic aperture radar) data with high-resolution (10-m) Sentinel-2 bands. We used two machine-learning algorithms (random forest (RF) and support vector machine (SVM)) to identify wetlands across the study area through supervised classification of the multisensor composite. We trained the algorithms with ground truth data provided through field studies and aerial photography. The accuracy was assessed by comparing the predicted and actual wetland and non-wetland classes using statistical coefficients (overall accuracy, Kappa, sensitivity, and specificity). For this purpose, we used four different out-of-sample test subsets, including the same year, next year, small vegetated, and small non-vegetated test sets to evaluate the methods on different spatial and temporal scales. The results were also compared to Landsat-derived JRC surface water products, and the Sentinel-2-derived normalized difference water index (NDWI). The wetlands derived from the RF model (overall accuracy 0.76 to 0.95) yielded favorable results, and outperformed the SVM, NDWI, and JRC products in all four testing subsets. To provide a further characterization of the potholes, the water bodies were stratified based on the presence of emergent vegetation using Sentinel-2-derived NDVI, and, after excluding permanent water bodies, using the JRC surface water product. The algorithm presented in the study is scalable and can be adopted for identifying wetlands in other regions of the world.

Redefining Waters of the US: A Case Study from the Edge of the Okefenokee Swamp

Defining the upslope extent of Federal Clean Water Act jurisdiction over wetlands and streams has been contentious since the passage of the Act but has large effects on the type, number, and area of wetlands that are protected by legislation. Federal guidance in the US has changed and evolved in response to scientific knowledge, Supreme Court decisions, and policy goals of Presidential Administrations. In 2020, the Trump administration replaced the Obama administration Clean Water Rule with the Navigable Waters Protection Rule with the goal of reducing jurisdiction over so-called isolated depressional wetlands and small streams. Here we use a case study of a titanium sands mining proposal on Trail Ridge southeast of Okefenokee Swamp to illustrate the large reduction in wetland and stream protection engendered by this policy change. Under the Navigable Waters Protection Rule, all seven wetlands within the 232 ha mining area, totaling 131 ha or 56% of the project area, were deemed non-jurisdictional and thus the project required no federal review or permitting. Under an earlier mining application under the Clean Water Rule, all of these wetlands were declared jurisdictional. Trail Ridge is located on the Atlantic Coastal Plain, an ecological province rich in depressional wetlands and ill-defined surface drainages. This case study shows that in such environments, the Navigable Water Protection Rule will allow destruction of large numbers and areas of ecologically significant wetlands.

Limited Evidence For Indicator Species of High Floristic Quality Wetlands In The US Southern Plains

Floristic Quality Assessment requires compiling a full list of vascular plant species for the wetland. Practitioners may lack the time and taxonomic skills for full-community vegetation surveys, especially when wetlands are large and complex. In this paper we broadly ask whether floristic quality indicator species may exist for wetlands, specifically evaluating indicator species potential for high floristic quality wetlands in the US southern plains region. Indicators were identified for a broader context (wetlands in Oklahoma prairie ecoregions) and narrower context (depressional wetlands in the northern Central Great Plains ecoregion of Oklahoma) based on indicator value, indicator validity, hydrophytic status, and ecological conservatism. No candidate indicators satisfied all criteria for high floristic quality. Indicator values improved with increasing spatial-environmental context, but many candidates occurred too frequently in non-high quality sites or too infrequently in high quality sites, relative to predicted rates. The best performing indicator (Eleocharis compressa) lacked validity in the broader context and showed high false-positive rates in the narrower context. Combining E. compressa with select other candidates (Amorpha fruticosa, Juncus torreyi, Leersia oryzoides, Schoenoplectus pungens) may compensate for weaknesses but the combinations may rarely be found across the region. Overall, these results do not support relying on indicator species to rapidly identify or verify high floristic quality wetlands in the US southern plains. We recommend similar studies in other regions and testing other quality levels (low, moderate) before broadly concluding that floristic quality indicator species do not exist for wetlands.

Assessment of Depressional Wetland Degradation, Spatial Distribution, and Geological Aspects in Southern Brazil

This paper presents a procedure to study depressional wetlands in southern Brazil and focuses on the mechanisms controlling water dynamics and environmental degradation due to anthropogenic interference. The study is based on an inventory of wetlands, a digital elevation model, the geological and geotechnical characteristics of geological materials, a multitemporal analysis of satellite images, the distribution of land use types, and onsite monitoring of water level and rainfall data. One hundred and twelve depressional wetlands were identified with a total area of 902 ha and a catchment area of 5456.8 ha. These wetlands were grouped into two classes with different hydrological control mechanisms. From the water level monitoring, the wetlands were found to present different hydrological conditions. Before rainy periods, the wetlands were almost dry or had little water; after rainy periods, over half of the wetlands were still dry or had groundwater levels below the surface, and the water levels of the other wetlands increased. The multitemporal analysis showed a reduction in the wetland water surface area from 270 ha in 1991 to 60 ha in 2019, which confirms the monitoring result that the amount of stored water is decreasing because of anthropogenic activities. Anthropogenic activities affect wetland water dynamics because of changes in the landscape and soil characteristics of the catchment area, and drainage of wetland areas by ditches for agricultural water supply; more than 50% of wetlands showed a high degree of change (environmental degradation), with conditions that make restoration or remediation very difficult.

Modeling the Supporting Ecosystem Services of Depressional Wetlands in Agricultural Landscapes

We explored how a geographic information system modeling approach could be used to quantify supporting ecosystem services related to the type, abundance, and distribution of landscape components. Specifically, we use the Integrated Valuation of Ecosystem Services and Tradeoffs model to quantify habitats that support amphibians and birds, floral resources that support pollinators, native-plant communities that support regional biodiversity, and above- and below-ground carbon stores in the Des Moines Lobe ecoregion of the U.S. We quantified services under two scenarios, one that represented the 2012 Des Moines Lobe landscape, and one that simulated the conversion to crop production of wetlands and surrounding uplands conserved under the USDA Agricultural Conservation Easement Program (ACEP). While ACEP easements only covered 0.35% of the ecoregion, preserved wetlands and grasslands provided for 19,020 ha of amphibian habitat, 21,462 ha of grassland-bird habitat, 18,798 ha of high-quality native wetland plants, and 27,882 ha of floral resources for pollinators. Additionally, ACEP protected lands stored 257,722 t of carbon that, if released, would result in costs in excess of 45-million USD. An integrated approach using results from a GIS-based model in combination with process-based model quantifications will facilitate more informed decisions related to ecosystem service tradeoffs.

Mapping Forested Wetland Inundation in the Delmarva Peninsula, USA Using Deep Convolutional Neural Networks

The Delmarva Peninsula in the eastern United States is partially characterized by thousands of small, forested, depressional wetlands that are highly sensitive to weather variability and climate change, but provide critical ecosystem services. Due to the relatively small size of these depressional wetlands and their occurrence under forest canopy cover, it is very challenging to map their inundation status based on existing remote sensing data and traditional classification approaches. In this study, we applied a state-of-the-art U-Net semantic segmentation network to map forested wetland inundation in the Delmarva area by integrating leaf-off WorldView-3 (WV3) multispectral data with fine spatial resolution light detection and ranging (lidar) intensity and topographic data, including a digital elevation model (DEM) and topographic wetness index (TWI). Wetland inundation labels generated from lidar intensity were used for model training and validation. The wetland inundation map results were also validated using field data, and compared to the U.S. Fish and Wildlife Service National Wetlands Inventory (NWI) geospatial dataset and a random forest output from a previous study. Our results demonstrate that our deep learning model can accurately determine inundation status with an overall accuracy of 95% (Kappa = 0.90) compared to field data and high overlap (IoU = 70%) with lidar intensity-derived inundation labels. The integration of topographic metrics in deep learning models can improve the classification accuracy for depressional wetlands. This study highlights the great potential of deep learning models to improve the accuracy of wetland inundation maps through use of high-resolution optical and lidar remote sensing datasets.