Search

Nebraska's eastern saline wetlands are globally unique and highly vulnerable inland salt marsh ecosystems. This research aims to evaluate the status of the saline wetlands in eastern Nebraska to discover the conditions of saline wetland hydrology, hydrophytes, and hydraulic soil. The research adopts machine learning and Google Earth Engine to classify Sentinel-2 imagery for water and vegetation classification and the National Agriculture Imagery Program imagery for salinity conditions. Six machine learning models are applied in water, soil, and vegetation detection in the study area. The optimal model (linear kernel SVM) generates an overall accuracy of 99.95% for water classification. For saline vegetation classification, the optimal model is the gradient tree boost with an overall accuracy of 94.07%. The overall accuracy values of saline soil classification using the optimal model (linear kernel SVM) varied among different years. The results of this study show the possibility of an observation approach for continuously monitoring Nebraska's eastern saline wetlands. The water classification results show that the saline wetlands in this area all have a similar temporal water cover pattern within each year. For saline vegetation, the peak season in this area is between June and July. The years 2019 (19.00%) and 2018 (17.69%) had higher saline vegetation cover rates than 2017 (10.54%). The saline soil classification shows that the saline soil area is highly variable in response to changes in the water and vegetation conditions. The research findings provide solid scientific evidence for conservation decision-making in these saline wetland areas. [ABSTRACT FROM AUTHOR]

The combination of Unmanned/Unoccupied Aerial Vehicle (UAV) data and deep learning, especially convolutional neural networks (CNNs), offers robust new tools for precision land cover mapping. However, its successful application is highly dependent on local experiences that are rarely documented, resulting in practical limitations during implementation. Cost-effective deep learning frameworks for fast deployment are required. This study presents a deep learning adaptation framework, named Auto-UNet++, trying to streamline wetland mapping tasks (including training data labeling and organizing). The framework treats mapping tasks as an intact semantic segmentation pipeline and then integrates automatic strategies into each step to reduce human intervention. These automatic strategies are achieved by standard computer vision techniques, including multi-view (MV) imaging—highly overlapped UAV images over an area (for labeling/voting), unsupervised clustering (for labeling), multi-scale CNN (for feature extraction), and attention mechanism—a CNN design used to select informative features from input (for feature exploration/selection). The framework was tested on playa wetland mapping in the Rainwater Basin, Nebraska, USA, with multispectral UAV datasets. Generally, the multi-scale CNN mapping task achieved a high of 87% overall accuracy and over 90% accuracy in water delineation. The results indicate that the multi-view and attention strategies have the potential to improve segmentation performance, and together with unsupervised learning, save considerable labor/expertise. Interestingly, evidence shows that the band/scale attention (weight) is adaptively associated with the land cover percentages per input image, indicating spatial contexts captured. This finding highlights the potential usages of the attention rule in automatic feature exploration, selection, and model interpretation. The framework illustrating a highly automated deep learning deployment on small MV datasets facilitates cost-effective wetland cover mapping. Although limitations exist, the study demonstrated the possibility of where/how conventional segmentation pipelines can be improved in typical UAV wetland mapping tasks. The framework and findings are useful for similar applications (including non-UAV studies) that only have limited time, labor, and expertise to implement sophisticated semantic segmentation models. • A highly fast deep learning adaptation framework for wetland mapping. • First deep learning initiative trying to streamline the wetland mapping pipeline. • The unsupervised/MV labeling strategies increase data organization efficiency. • The attention rule show potentials for model interpreting and feature selection. • The framework can be adjusted to fit needs in different applications. [ABSTRACT FROM AUTHOR]