Search

This study presents an automated methodology to generate training data for surface water mapping from a single Sentinel-2 granule at 10 m (4 band, VIS/NIR) or 20 m (9 band, VIS/NIR/SWIR) resolution without the need for ancillary training data layers. The 20 m method incorporates an ensemble of three spectral indexes with optimal band thresholds, whereas the 10 m method achieves similar results using fewer bands and a single spectral index. A spectrally balanced and randomly generated set of training data based on the index values and optimal thresholds is used to fit machine learning classifiers. Statistical validation compares the 20 m ensemble-only method to the 20 m ensemble method with a random forest classifier. Results show the 20 m ensemble-only method had an overall accuracy of 89.5% (±1.7%), whereas the ensemble method combined with the random forest classifier performed better, with a ~4.8% higher overall accuracy: 20 m method (94.3% (±1.3%)) with optimal spectral index and SWIR thresholds of −0.03 and 800, respectively, and 10 m method (93.4% (±1.5%)) with optimal spectral index and NIR thresholds of −0.01 and 800, respectively. Comparison of other supervised classifiers trained automatically with the framework typically resulted in less than 1% accuracy improvement compared with the random forest, suggesting that training data quality is more important than classifier type. This straightforward framework enables accurate surface water classification across diverse geographies, making it ideal for development into a decision support tool for water resource managers.

Tree cover maps derived from satellite and aerial imagery directly support civil and military operations. However, distinguishing tree cover from other vegetative land covers is an analytical challenge. While the commonly used Normalized Difference Vegetation Index (NDVI) can identify vegetative cover, it does not consistently distinguish between tree and low-stature vegetation. The Forest Cover Index (FCI) algorithm was developed to take the multiplicative product of the red and near infrared bands and apply a threshold to separate tree cover from non-tree cover in multispectral imagery (MSI). Previous testing focused on one study site using 2-m resolution commercial MSI from WorldView-2 and 30-m resolution imagery from Landsat-7. New testing in this work used 3-m imagery from PlanetScope and 10-m imagery from Sentinel-2 in imagery in sites across 12 biomes in South and Central America and North Korea. Overall accuracy ranged between 23% and 97% for Sentinel-2 imagery and between 51% and 98% for PlanetScope imagery. Future research will focus on automating the identification of the threshold that separates tree from other land covers, exploring use of the output for machine learning applications, and incorporating ancillary data such as digital surface models and existing tree cover maps.

Feature extraction algorithms are routinely leveraged to extract building footprints and road networks into vector format. When used in conjunction with high resolution remotely sensed imagery, machine learning enables the automation of such feature extraction workflows. However, many of the feature extraction algorithms currently available have not been thoroughly evaluated in a scientific manner within complex terrain such as the cities of developing countries. This report details the performance of three automated feature extraction (AFE) datasets: Ecopia, Tier 1, and Tier 2, at extracting building footprints and roads from high resolution satellite imagery as compared to manual digitization of the same areas. To avoid environmental bias, this assessment was done in two different regions of the world: Maracay, Venezuela and Niamey, Niger. High, medium, and low urban density sites are compared between regions. We quantify the accuracy of the data and time needed to correct the three AFE datasets against hand digitized reference data across ninety tiles in each city, selected by stratified random sampling. Within each tile, the reference data was compared against the three AFE datasets, both before and after analyst editing, using the accuracy assessment metrics of Intersection over Union and F1 Score for buildings and roads, as well as Average Path Length Similarity (APLS) to measure road network connectivity. It was found that of the three AFE tested, the Ecopia data most frequently outperformed the other AFE in accuracy and reduced the time needed for editing.

Adaptation is now firmly embedded in the societal discourse regarding the management of climate risk. In this discourse, adaptation planning and implementation at the local level are seen as particularly important for developing robust responses to climate change. However, it is not clear whether the mantra that adaptation is local holds true given the multi-level nature of climate risk governance. Using a multi-method approach, this paper examines the extent to which adaptation should be framed as a local issue and, specifically, the role of local government in adaptation relative to other actors. In so doing, the paper first explores the extent to which the local framing of adaptation is embedded in the international adaptation literature. This is followed by a specific case study from Southeast Queensland, Australia, which focuses on the critical examination of the processes of responsibility shifting and taking among actors involved in coastal adaptation planning. Results indicate the assumption that adaptation is local remains widely held in adaptation science, although counter arguments can be readily identified. Interviews with adaptation actors revealed unclear divisions of responsibility for climate change adaptation as a significant constraint on actors’ willingness to implement adaptation. Furthermore, attributing responsibility for adaptation to local actors might not necessarily be a robust strategy, due to the existence of particularly strong constraints and value conflicts at local levels of governance. Greater appreciation by researchers and practitioners for the interactions between local actors and those at higher levels of governance in shaping response capacity may contribute to more equitable and effective allocations of responsibilities for adaptation action.

The adaptation science enterprise has expanded rapidly in recent years, presumably in response to growth in demand for knowledge that can facilitate adaptation policy and practice. However, evidence suggests such investments in adaptation science have not necessarily translated into adaptation implementation. One potential constraint on adaptation may be the underlying heuristics that are used as the foundation for both adaptation research and practice. Here, we explore the adaptation academic literature with the objective of identifying adaptation heuristics, assessing the extent to which they have become entrenched within the adaptation discourse, and discussing potential weaknesses in their framing that could undermine adaptation efforts. This investigation is supported by a multi-method analysis that includes both a quantitative content analysis of the adaptation literature that evidences the use of adaptation heuristics and a qualitative analysis of the implications of such heuristics for enhancing or hindering the implementation of adaptation. Results demonstrate that a number of heuristic devices are commonly used in both the peer-reviewed adaptation literature as well as within grey literature designed to inform adaptation practitioners. Furthermore, the apparent lack of critical reflection upon the robustness of these heuristics for diverse contexts may contribute to potential cognitive bias with respect to the framing of adaptation by both researchers and practitioners. We discuss this phenomenon by drawing upon heuristic-analytic theory, which has explanatory utility in understanding both the origins of such heuristics as well as the measures that can be pursued toward the co-generation of more robust approaches to adaptation problem-solving.

The consequences of future sea-level rise for coastal communities are a priority concern arising from anthropogenic climate change. Here, previously published methods are scaled up in order to undertake a first pass assessment of exposure to hurricane storm surge and sea-level rise for the U.S. Gulf of Mexico and Atlantic coasts. Sea-level rise scenarios ranging from +0.50 to +0.82 m by 2100 increased estimates of the area exposed to inundation by 4–13% and 7–20%, respectively, among different Saffir-Simpson hurricane intensity categories. Potential applications of these hazard layers for vulnerability assessment are demonstrated with two contrasting case studies: potential exposure of current energy infrastructure in the U.S. Southeast and exposure of current and future housing along both the Gulf and Atlantic Coasts. Estimates of the number of Southeast electricity generation facilities potentially exposed to hurricane storm surge ranged from 69 to 291 for category 1 and category 5 storms, respectively. Sea-level rise increased the number of exposed facilities by 6–60%, depending on the sea-level rise scenario and the intensity of the hurricane under consideration. Meanwhile, estimates of the number of housing units currently exposed to hurricane storm surge ranged from 4.1 to 9.4 million for category 1 and category 4 storms, respectively, while exposure for category 5 storms was estimated at 7.1 million due to the absence of landfalling category 5 hurricanes in the New England region. Housing exposure was projected to increase 83–230% by 2100 among different sea-level rise and housing scenarios, with the majority of this increase attributed to future housing development. These case studies highlight the utility of geospatial hazard information for national-scale coastal exposure or vulnerability assessment as well as the importance of future socioeconomic development in the assessment of coastal vulnerability.