High-Resolution Land Cover Mapping
During my tenure at the Iowa Geological Survey, I played a pivotal role in the statewide Land Use/Land Cover (LULC) classification project. Using ERDAS IMAGINE, I conducted unsupervised classification on high-resolution aerial imagery, segmenting Iowa’s diverse landscapes into 15 distinct land cover classes. This process incorporated imagery from multiple timeframes, including 2008 and 2009 NAIP and four-band spring captures, enabling a comprehensive spectral analysis across the entire state.
Following classification, I was deeply involved in the post-processing phase, where I refined the outputs for accuracy and consistency. This included reclassification and merging of land cover types to correct misclassifications and improve thematic coherence. I implemented robust quality assurance and quality control (QA/QC) procedures to ensure the reliability of results.
Working in collaboration with a multidisciplinary team, I trained team members in best practices for classification and data refinement. My end-to-end involvement, from image classification through to final output, was instrumental in delivering a statewide, high-resolution LULC dataset that now supports environmental planning and natural resource management efforts in Iowa.
Webmap sourced from: https://programs.iowadnr.gov/geospatial/rest/services/LandCover/LC_2009_1m/MapServer
Closeup view of the classification
Segment showing distinct classification of buildings, roads, tree canopy, and grassy areas.
Segment showing distinct classification of buildings, roads, tree canopy, and grassy areas.
Land Use and Land Cover Classification
Preliminary raster output showing six land cover classes generated through object-based classification in eCognition Developer; later refined through QA/QC and manual editing.
Mosaicked DEM Classification Output: Skagit County
This output was generated through a combination of supervised and unsupervised classification techniques using eCognition Developer, applied to satellite and aerial imagery for generating Land Use/Land Cover (LULC) raster outputs with six distinct classes. The process began with unsupervised classification and progressed into rule-based supervised classification using object-based analysis, where vector objects were created and assigned classes based on defined spatial and contextual rules. Additional datasets,such as OSM road buffers, building layers, and water bodies, were incorporated to improve classification precision and contextual accuracy.
A robust QA/QC process ensured the outputs aligned with source imagery, and further refinement was done through manual editing using ERDAS Imagine, ArcMap, and NAIP imagery. I supervised this critical phase, introducing an efficient method of using the original object polygons to guide cell reclassification, implemented through a custom-built in-house ArcMap tool.
Beyond technical execution, I played a key role in workflow design, team training, and documentation of editing procedures, equipping the team with the tools and knowledge needed to correct misclassifications efficiently. Under my guidance, we successfully delivered a seamlessly mosaicked and validated raster product as the final output.
This website uses cookies.
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.