Previous simulations have shown that wind farms have an impact on the near-surface atmospheric boundary layer (ABL) as turbulent wakes generated by the turbines enhance vertical mixing of momentum, heat and moisture. These changes alter downstream atmospheric properties. With the exception of a few observational data sets that focus on the impact to near-surface temperature within wind farms, little to no observational evidence exists with respect to vertical mixing. These few experimental studies also lack high spatial resolution due to their use of a limited number of meteorological sensors or remote sensing techniques. This study utilizes an instrumented small unmanned aerial system (sUAS) to gather high resolution in-situ field measurements from two state-of-the-art Midwest wind farms in order to differentially map downstream changes to relative humidity. These measurements are complemented by numerical experiments conducted using large eddy simulation (LES). Observations and numerical predictions are in good general agreement around a single wind turbine and show that downstream relative humidity is altered in the vertical, lateral, and downstream directions. A suite of LES is then performed to determine the effect of a turbine array on the relative humidity distribution in compounding wakes. In stable and neutral conditions, and in the presence of a positive relative humidity lapse rate, it is found that the humidity decreases below the turbine hub height and increases above the hub height. As the array is transitioned, the magnitude of change increases, differentially grows on the left-hand and right-hand side of the wake, and move slightly upward with downstream distance. In unstable conditions, the magnitude of near-surface decrease in relative humidity is a full order of magnitude smaller than that observed in a stable atmospheric regime.

<p dir="ltr">Efficient management of plant diseases and their spread within fields requires a system capable of early and accurate disease identification and its severity estimation. Many plant diseases have distinct visual symptoms, which can be used to correctly identify, classify, and manage them. Recent technological advancements have led to increased adoption of deep neural networks (DNN) for developing deep learning (DL)-based computer vision systems. An accurate disease identification and severity estimation system using a DL-based computer vision framework is critical for efficiently managing corn diseases under field conditions and further restricting the spread of disease. Image processing and machine learning methods for disease identification and classification have been employed in the last two decades using high-cost sensors that need frequent calibration. Researchers have used low-cost red, green, and blue (RGB) sensors to mostly identify single diseases affecting crops, whereas, in real-world applications, a single leaf can be affected by multiple diseases. This research identifies gaps in knowledge of DL applications to field crops by reviewing 70 research articles published between 1983 and 2022. It creates a much-needed disease database for corn grown under field conditions by adding custom-acquired image data to other publicly available image repositories. The image data was used to train and evaluate the performance of commonly used DL-based image classification models for differentiating single diseases on individual corn leaves under field conditions. However, many disease lesions of different shapes and sizes can simultaneously develop on infected leaves. The performance of DL-based image classification and object detection models was evaluated to accurately identify multiple simultaneous diseases with varying symptoms. Disease identification under field conditions is necessary to implement an effective disease management system. However, recent work has demonstrated poor generalization accuracies of DL models trained on lab-acquired imagery for identifying diseases in the field. Therefore, after achieving promising results for disease identification, DL generalization performance was assessed and improved using different dataset combinations with varying backgrounds. A novel neural network architecture using a hierarchical structure was also proposed, which resulted in improved generalization performance. Additionally, disease severity must be estimated to implement an effective management response. DL models were evaluated to estimate the severity of multiple corn diseases under field conditions using aerial and ground-based platforms to identify specific lesions from above and below the canopy. A progressive web application was designed to empower end users with disease recognition capabilities. Overall, this research reports findings of the performance of deep learning image processing, object detection, and segmentation models for identifying single/multiple diseases on field corn and the development of tools that can potentially be a component of production-ready disease diagnosis systems for implementing effective management practices.</p>

No description available.

Images have become a ubiquitous and efficient data form to record information. Use of this option for data capture has largely increased due to the widespread availability of image sensors and sensor platforms (e.g., smartphones and drones), the simplicity of this approach for broad groups of users, and our pervasive access to the internet as one class of infrastructure in itself. Such data contains abundant visual information that can be exploited to automate asset assessment and management tasks that traditionally are manually conducted for engineering systems. Automation of the data collection, extraction and analytics is however, key to realizing the use of these data for decision-making. Despite recent advances in computer vision and machine learning techniques extracting information from an image, automation of these real-world tasks has been limited thus far. This is partly due to the variety of data and the fundamental challenges associated with each domain. Due to the societal demands for access to and steady operation of our infrastructure systems, this class of systems represents an ideal application where automation can have high impact. Extensive human involvement is required at this time to perform everyday procedures such as organizing, filtering, and ranking of the data before executing analysis techniques, consequently, discouraging engineers from even collecting large volumes of data. To break down these barriers, methods must be developed and validated to speed up the analysis and management of data over the lifecycle of infrastructure systems. In this dissertation, big visual data collection and analysis methods are developed with the goal of reducing the burden associated with human manual procedures. The automated capabilities developed herein are focused on applications in lifecycle visual assessment and are intended to exploit large volumes of data collected periodically over time. To demonstrate the methods, various classes of infrastructure, commonly located in our communities, are chosen for validating this work because they: (i) provide commodities and service essential to enable, sustain, or enhance our lives; and (ii) require a lifecycle structural assessment in a high priority. To validate those capabilities, applications of infrastructure assessment are developed to achieve multiple approaches of big visual data such as region-of-interest extraction, orthophoto generation, image localization, object detection, and image organization using convolution neural networks (CNNs), depending on the domain of lifecycle assessment needed in the target infrastructure. However, this research can be adapted to many other applications where monitoring and maintenance are required over their lifecycle.