Contributors
Abstract
Subsurface drainage is known to adversely impact the water quality and contribute to the formation of harmful algal blooms (HABs). In early August of 2014, a HAB developed in the western Lake Erie Basin and resulted in over 400,000 people being unable to drink their tap water. HAB development is aided by excess nutrients from agricultural fields, which are transported through subsurface tile and enter the watershed. Compounding the issue, the trend has been to increase the installation of tile drains in both total extent and density. Due to the immense area of drained fields, it is necessary to establish a cost-effective technique to monitor tile installations and their associated impacts.
This thesis aimed at developing an automated method in order to identify subsurface tile locations from high resolution aerial imagery by applying an object based image analysis (OBIA) approach utilizing eCognition. This process was accomplished through a set of algorithms and image filters, which segment and classify image objects by their spectral and geometric characteristics. The algorithms utilized were based on the relative location of image objects and pixels in order to maximize the robustness and transferability of the final rule-set. These algorithms were coupled with convolution and histogram image iv filters to generate results for a 10km² study area located within Clay Township in Ottawa County, Ohio.
The eCognition results were compared to previously collected tile locations from a concurrent project that applied heads-up digitizing. The heads-up digitized locations were used as a baseline for the accuracy assessment. The accuracy assessment generated a range of agreement values from 67.20% – 71.20%, and an average agreement of 69.76%. The confusion matrices calculated a range of kappa values from 0.273 – 0.416 with an overall K value of 0.382, considered fair in strength of agreement. This thesis provides a step forward in the ability to automatically identify and extract tile drains and will assist future research in subsurface drainage modeling.
