Graduation Date
Spring 2021
Document Type
Thesis
Program
Master of Science degree with a major in Natural Resources: option Environmental Science and Management
Committee Chair Name
Dr. James Graham
Committee Chair Affiliation
HSU Faculty or Staff
Second Committee Member Name
Dr. David Gwenzi
Second Committee Member Affiliation
HSU Faculty or Staff
Third Committee Member Name
Dr. Frank Shaughnessy
Third Committee Member Affiliation
HSU Faculty or Staff
Keywords
GIS, Natural resources, Modeling, LULC, Mapping, Landcover, Remote sensing, Geospatial, Aquatic, Terrestrial
Subject Categories
Environmental Science and Management
Abstract
Land use and land cover (LULC) mapping plays a vital role in understanding the state of the world, showing us a visual representation of the natural and anthropogenic features covering our planet. Northern California in the United States is home to many critical habitats that provide for a variety of endemic and some threatened and engendered species, making it an area of particular concern to better understand and monitor. There is a greater need to identify specific methods for vegetation modeling in Northern California due to its unique species; to do this we examined two case studies with the following objectives: 1) Determine whether unmanned aerial system (UAS) image analysis can provide similar estimates of eelgrass biometrics, such as percent coverage, to those obtained in situ using traditional field survey methods; 2) To develop a GIS data fusion workflow for high-resolution habitat classification in the Napa Watershed of central California with a focus on oak savanna habitat. UAS Imagery for two eelgrass sites were collected during June, 2019 using a DJI Matrice 100 equipped with MicaSense RedEdge Multispectral sensor (5-band). Following UAS image collection, ground survey data were collected at three tidal elevation transects per site, with 20 quadrats stationed randomly along each transect. Eelgrass percent coverage was measured for each quadrat and then compared to eelgrass classification models derived from UAS derived imagery. In the Napa watershed, we examined methods necessary to accurately incorporate ancillary geospatial spatial datasets into a remote sensing land cover classification. By doing so, I developed a habitat distribution dataset that may better analyze interactions of wildlife, humans, and the endemic habitat types of the Napa watershed in California. UAVs provided a means to obtain high resolution remote sensing imagery of eelgrass at a resolution of 3.46 – 3.70 cm per pixel or greater at specific tidal periods, providing a useful methodology that allowed for percent coverage estimates with an R2 value of 0.6496 compared to in situ measurements. While developing a land cover classification workflow for the Napa watershed, I found that by incorporating ancillary geospatial data, remotely sensed data, and threshold classification, I could obtain a LULC model that more accurately depicts the endemic land use and land cover features of the Napa watershed. With an overall accuracy of 70.20% and a kappa statistic of 0.6140, this modeling method proved more accurate than traditional image classification methods. With ground sampled reference data and remotely sensed data gathered at the same temporal and spatial scales these classification methods would be robust and replicable for future analyses.
Citation Style
APA
Recommended Citation
Corro, Lucila, "A look at land cover classification methods in Northern California with the use of high spatial resolution geospatial data" (2021). Cal Poly Humboldt theses and projects. 488.
https://digitalcommons.humboldt.edu/etd/488