Detection of Agricultural Drainage Infrastructure in Intensively-managed Landscape
Description
Agricultural drainage infrastructure is widely implemented in intensively managed landscapes such as the U.S. Midwest. As a key agricultural water management practice, it commonly comprises components including subsurface drainage pipes and surface drainage ditches, which helps maintain optimal soil moisture, enhancing crop yields and boosting farm income. These agricultural water practices transport not only water and sediments but also chemicals, leading to hydrological and environmental issues. Explicit representation of agricultural drainage infrastructure is vital to better understand their impacts on hydrologic processes to improve hydrological and water quality prediction. However, existing studies often simplified the effects of agricultural drainage infrastructure due to insufficient explicit representations of these agricultural practices. Therefore, there is a pressing need to improve the representations of agricultural drainage infrastructure. The dissertation aims to detect both subsurface and surface components of drainage infrastructure in agricultural landscapes to enhance their explicit representations in hydrological models. The first part tackles the lack of specific methods to guide survey times or filter historical remote sensing imageries for subsurface drainage detection. A simulation-guided approach is developed to determine the optimal time window to detect subsurface drainage pipes. The second part focuses on overcoming challenges in preserving surface drainage features while eliminating noise in high-resolution LiDAR DEMs. A smoothing algorithm that accounts for both gradient magnitudes and orientations is introduced, enabling effective noise reduction while preserving key features. This algorithm, combined with a detection method based on the principal slope, is applied to accurately delineate surface drainage networks at the watershed scale in the Midwest. To further enhance detection efficiency, the third part leverages the identified drainage data in the second part to label data and train a UNet model aiming at detecting comprehensive watershed-scale agricultural surface drainages from LiDAR data. The UNet model is successfully applied to identify surface drainage networks and cross-sectional geometries of drainage channels across multiple watersheds in the Midwest. By achieving the explicit representation of these infrastructures, this study will lead to a more accurate understanding of their roles in hydrologic processes, ultimately enhancing agricultural water management, boosting crop production and promoting environmental sustainability.
Details
Contributors
- Yao, Chuncheng (Author)
- Zeng, Ruijie (Thesis advisor)
- Vivoni, Enrique R. (Committee member)
- Kumar, Saurav (Committee member)
- Xu, Tianfang (Committee member)
- Wang, Zhihua (Committee member)
- Arizona State University (Publisher)
Date Created
The date the item was original created (prior to any relationship with the ASU Digital Repositories.)
2024
Topical Subject
Resource Type
Language
- eng
Note
- Partial requirement for: Ph.D., Arizona State University, 2024
- Field of study: Civil, Environmental and Sustainable Engineering
Additional Information
English
Extent
- 151 pages