ALBERT

Description: Abstract Optimizing the remediation of light nonaqueous phase liquids (LNAPLs) to achieve an acceptable endpoint status for a site is not trivial. Recently, Sookhak Lari, Johnston, et al. (2018, https://doi.org/10.1016/j.jhazmat.2017.11.006), Sookhak Lari, Rayner, and Davis (2018, https://doi.org/10.1016/j.jenvman.2018.07.041) conducted three‐dimensional multiphase, multicomponent simulations to address LNAPL remediation endpoints for a single recovery well. However, optimized LNAPL remediation for multiple wells is not addressed in the literature. In the first part of this paper, we establish a matrix of 10 simulation scenarios to show the sensitivity of the remedial endpoint (i.e., what is feasibly achieved) to several parameters including viscosity and partitioning attributes of the LNAPL, heterogeneity of the formation, and the location and number of the recovery wells. While this addresses the variability of LNAPL removal from the subsurface and is valuable in its own right, it does not address the optimal removal of LNAPL. We address this in the second part of the paper by linking a genetic algorithm to TMVOC‐MP to allow, for example, the assessment of the optimal number and location of LNAPL recovery wells in a field‐scale problem. Using supercomputing facilities and within 49 genetic algorithm generations, each including 150 members, highly optimized answers to different objective functions were obtained. For the first time, such a multiphase, multicomponent optimization tool promises the possibility of optimizing LNAPL remediation at field scale to achieve practicable endpoint conditions, within computational affordability.

Description: Abstract Understanding streamflow generation and its dependence on catchment characteristics requires large spatial datasets and is often limited by convoluted effects of multiple variables. Here we address this knowledge gap using data‐informed physics‐based hydrologic modelling in two catchments with similar vegetation and climate but different lithology (Shale Hills, SH, Shale, 0.08 km2 and Garner Run, GR, Sandstone, 1.34 km2), which influences catchment topography and soil properties. The sandstone catchment, Garner Run, is characterized by lower drainage density, extensive valley fill, and boulder soils. We tested the hypothesis that the influence of topographic characteristics is more significant than that of soil properties and catchment size. Transferring calibration coefficients from the previously‐calibrated SH model to GR cannot reproduce monthly discharge until after incorporating measured boulder distribution at GR. Model calibration underscored the importance of soil properties (porosity, van Genuchten parameters, and boulder characteristics) in reproducing daily discharge. Virtual experiments were used to swap topography, soil properties, and catchment size one at a time to disentangle their influence. They showed that clayey SH soils led to high nonlinearity and threshold behavior. With the same soil and topography, changing from SH to GR size consistently increased dynamic water storage (Sd) from ~ 0.12 m to ~ 0.17 m. All analyses accentuated the predominant control of soil properties, therefore rejecting the hypothesis. The results illustrate the use of physics‐based modelling for illuminating mechanisms and underscore the importance and challenges for subsurface characterization as we move toward hydrological Prediction in Ungauged Basins (PUB).

Description: In this study, we automate tree species classification and mapping using field-based training data, high spatial resolution airborne hyperspectral imagery, and a convolutional neural network classifier (CNN). We tested our methods by identifying seven dominant trees species as well as dead standing trees in a mixed-conifer forest in the Southern Sierra Nevada Mountains, CA (USA) using training, validation, and testing datasets composed of spatially-explicit transects and plots sampled across a single strip of imaging spectroscopy. We also used a three-band ‘Red-Green-Blue’ pseudo true-color subset of the hyperspectral imagery strip to test the classification accuracy of a CNN model without the additional non-visible spectral data provided in the hyperspectral imagery. Our classifier is pixel-based rather than object based, although we use three-dimensional structural information from airborne Light Detection and Ranging (LiDAR) to identify trees (points 〉 5 m above the ground) and the classifier was applied to image pixels that were thus identified as tree crowns. By training a CNN classifier using field data and hyperspectral imagery, we were able to accurately identify tree species and predict their distribution, as well as the distribution of tree mortality, across the landscape. Using a window size of 15 pixels and eight hidden convolutional layers, a CNN model classified the correct species of 713 individual trees from hyperspectral imagery with an average F-score of 0.87 and F-scores ranging from 0.67–0.95 depending on species. The CNN classification model performance increased from a combined F-score of 0.64 for the Red-Green-Blue model to a combined F-score of 0.87 for the hyperspectral model. The hyperspectral CNN model captures the species composition changes across ~700 meters (1935 to 2630 m) of elevation from a lower-elevation mixed oak conifer forest to a higher-elevation fir-dominated coniferous forest. High resolution tree species maps can support forest ecosystem monitoring and management, and identifying dead trees aids landscape assessment of forest mortality resulting from drought, insects and pathogens. We publicly provide our code to apply deep learning classifiers to tree species identification from geospatial imagery and field training data.

Description: Abstract Groundwater dependent ecosystems are often defined by the presence of deeply‐rooted phreatophytic plants. When connected to groundwater, phreatophytes in arid regions decouple ecosystem net primary productivity from precipitation, underscoring a disproportionately high biodiversity and exchange of resources relative to surrounding areas. However, groundwater dependent ecosystems are widely threatened due to the effects of water diversions, groundwater abstraction, and higher frequencies of episodic drought and heat waves. The resilience of these ecosystems to shifting ecohydrological‐climatological conditions will depend largely on the capacity of dominant, phreatophytic plants to cope with dramatic reductions in water availability and increases in atmospheric water demand. This paper disentangles the broad range of hydraulic traits expressed by phreatophytic vegetation to better understand their capacity to survive, or even thrive under shifting ecohydrological conditions. We focus on three elements of plant water relations: 1) hydraulic architecture (including root area to leaf area ratios and rooting depth), 2) xylem structure and function, and 3) stomatal regulation. We place the expression of these traits across a continuum of phreatophytic habits from obligate to semi‐obligate to semi‐facultative to facultative. Although many species occupy multiple phreatophytic niches depending on access to groundwater, we anticipate that populations are largely locally adapted to a narrow range of ecohydrological conditions regardless of gene flow across ecohydrological gradients. Consequently, we hypothesize that reductions in available groundwater and increases in atmospheric water demand will result in either 1) stand replacement of obligate phreatophytic species with more facultative species as a function of wide‐spread mortality in highly groundwater dependent populations, or 2) directional selection in semi‐obligate and semi‐facultative phreatophytes towards the expression of traits associated with highly facultative phreatophytes in the absence of species replacement. Anticipated shifts in the expression of hydraulic traits may have profound impacts on water cycling processes, species assemblages and habitat structure of groundwater dependent woodlands and riparian forests.

Description: The construction technology market is competitive and complicated, due to the high-risk of digital technology utilisation in construction projects and the conservative character of construction companies. This complexity affects the process of job-site technology dissemination and adoption in which construction companies make decisions to purchase and utilise the new technology. The complexity is one of the reasons that many new remote technologies, positioning and locating systems, lasers and drones, 3D printing, and robots are not widely adopted in the short term, despite vendors making determined efforts to overcome this. Three objectives are investigated in this paper: (i) to define criteria for examining patterns of vendors’ strategies to support technology adoption; (ii) to present fact-based evidence of different vendors’ demonstration methods; and (iii) to present examples of different technology groups based on their required strategies. This paper presents the results of a longitudinal investigation of the construction technology market, including patterns of technology demonstration and a conceptual model of classifying vendors and their technologies in construction market places. The model involves the three most important factors that distinguish technology exhibitors: Physical appearance, Interpersonal relationship and Technology demonstration. Data was collected from technology exhibitions, involving randomly selected vendors. This data was analysed using hierarchical and c-means clustering techniques. The hard-clustering techniques resulted in vendors being placed in five classes based on the elements of the PIT framework. Fuzzy analysis shows how these classes fit into an underlying strategy spectrum. Understanding the strategies used in each class enables new vendors to select their own dissemination strategies based on their own particular circumstances. The practical implication of this study is to present a set of dissemination strategies to new technology stakeholders involved in Industry 4.0. The identified patterns of technology vendor strategies and the novel conceptual model contribute to the body of knowledge in technology diffusion.

Description: In this study, we automate tree species classification and mapping using field-based training data, high spatial resolution airborne hyperspectral imagery, and a convolutional neural network classifier (CNN). We tested our methods by identifying seven dominant trees species as well as dead standing trees in a mixed-conifer forest in the Southern Sierra Nevada Mountains, CA (USA) using training, validation, and testing datasets composed of spatially-explicit transects and plots sampled across a single strip of imaging spectroscopy. We also used a three-band ‘Red-Green-Blue’ pseudo true-color subset of the hyperspectral imagery strip to test the classification accuracy of a CNN model without the additional non-visible spectral data provided in the hyperspectral imagery. Our classifier is pixel-based rather than object based, although we use three-dimensional structural information from airborne Light Detection and Ranging (LiDAR) to identify trees (points 〉 5 m above the ground) and the classifier was applied to image pixels that were thus identified as tree crowns. By training a CNN classifier using field data and hyperspectral imagery, we were able to accurately identify tree species and predict their distribution, as well as the distribution of tree mortality, across the landscape. Using a window size of 15 pixels and eight hidden convolutional layers, a CNN model classified the correct species of 713 individual trees from hyperspectral imagery with an average F-score of 0.87 and F-scores ranging from 0.67–0.95 depending on species. The CNN classification model performance increased from a combined F-score of 0.64 for the Red-Green-Blue model to a combined F-score of 0.87 for the hyperspectral model. The hyperspectral CNN model captures the species composition changes across ~700 meters (1935 to 2630 m) of elevation from a lower-elevation mixed oak conifer forest to a higher-elevation fir-dominated coniferous forest. High resolution tree species maps can support forest ecosystem monitoring and management, and identifying dead trees aids landscape assessment of forest mortality resulting from drought, insects and pathogens. We publicly provide our code to apply deep learning classifiers to tree species identification from geospatial imagery and field training data.