Search

Accurately mapping urban above-ground vegetation carbon density presents challenges due to fragmented landscapes, mixed pixels, and shadows induced by buildings and mountains. To address these issues, a novel methodological framework is introduced, utilizing a linear spectral unmixing analysis (LSUA) for shadow removal and vegetation information extraction from mixed pixels. Parametric and nonparametric models, incorporating LSUA-derived vegetation fraction, are compared, including linear stepwise regression, logistic model-based stepwise regression, k-Nearest Neighbors, Decision Trees, and Random Forests. Applied in Shenzhen, China, the framework integrates Landsat 8, Pleiades 1A & 1B, DEM, and field measurements. Among the key findings, the shadow removal algorithm is effective in mountainous areas, while LSUA-enhanced models improve urban vegetation carbon density mapping, albeit with marginal gains. Integrating kNN and RF with LSUA reduces errors, and Decision Trees, especially when integrated with LSUA, outperform other models. This study underscores the potential of the proposed framework, particularly the integration of Decision Trees with LSUA, for advancing the accuracy of urban vegetation carbon density mapping.

Our understanding of broad-scale forest disturbances under climatic extremes remains incomplete. Drought, as a typical extreme event, is a key driver of forest mortality but there have been no reports on continental-scale quantification of its impact on forest mortality or how it compares to other natural or anthropogenic drivers. Thus, our ability to understand and predict broad-scale carbon cycling in response to changing climate and extreme events is limited. In this study, we applied an attribution approach based on different sources of data to quantify the area and potential carbon loss/transfer in continental U.S. (CONUS) from four types of disturbance: (1) anthropogenic (especially timber harvest); (2) fire; (3) drought-associated; and (4) other from 2000 to 2014. Our results showed that anthropogenic disturbances, fire, drought-associated disturbances, and other disturbances accounted for 54.3, 10.7, 12.7, and 22.3% of total canopy area loss, respectively. Anthropogenic disturbance was the most important driver contributing to 58.1% potential carbon loss/transfer in CONUS for 2000–2014. The potential carbon loss/transfer from natural disturbances (fire, drought, and other) for the same study period accounted for approximately 41.9% of the total loss/transfer from all agents, suggesting that natural disturbances also played a very important role in forest carbon turnover. Potential carbon loss/transfer associated with drought accounted for approximately 11.6% of the total loss/transfer in CONUS, which was of similar magnitude to potential carbon loss/transfer from fire (∼11.0%). The other natural disturbance accounted for 19.3% of potential carbon loss/transfer. Our results demonstrated the importance of the impacts of various disturbances on forest carbon stocks at the continental scale, and the drought-associated carbon loss/transfer data developed here could be used for evaluating the performance of predictive models of tree mortality under droughts.

Wetlands serve many important ecosystem services, yet the United States lacks up-to-date, high-resolution wetland inventories. New, automated techniques for developing wetland segmentation maps from high-resolution aerial imagery can improve our understanding of the location and amount of wetlands. We assembled training and testing data sets (patch sizes of 28 × 28 m2 and 56 × 56 m2) of high-resolution aerial imagery of wetlands using Illinois Natural History Survey wetland location data and National Agricultural Imagery Project data. Each patch was labeled as wetland or non-wetland. To augment these data sets with additional information, we incorporated digital surface and digital terrain models and topographic wetness index data in the same two patch sizes. Subsequently, we evaluated convolutional neural network (CNN) and Gaussian process-based machine learning methods to produce wetland segmentation maps. We developed the best performing method into a new CNN algorithm, WetSegNet. It exhibited an area under the curve of 98% when used with 56 × 56 m2 patch sizes. WetSegNet developed reliable wetland segmentation maps in test cases in which wetlands would have gone undetected using only the National Land Cover Database. The development of WetSegNet exemplifies the types of data sets and methods that are needed to accelerate the use of high-resolution aerial imagery towards an improved understanding of wetlands. This algorithm could be used by state and federal agencies or other groups to identify wetlands with higher accuracy and at a finer scale than previously possible.