Taylor Holm

Date of Award

January 2020

Document Type


Degree Name

Master of Science (MS)



First Advisor

Robert Newman


Monitoring land use and land cover (LULC) change and its effect on habitat availability and connectivity is crucial for species conservation and management. This is most efficiently accomplished at a sufficiently broad scale using remote sensing technology. Current approaches benefit from freely available data sources such as National Land Cover Database (NLCD), derived from 30 m Landsat imagery. However, temporal frequency and spatial resolution of NLCD may be insufficient for some applications. Specifically, 30 m resolution data may not detect small or narrow landscape features that may be important habitat for small animals such as amphibians. Other available data sources, such as the higher resolution (0.6-1 m) imagery obtained by the National Agriculture Imagery Program (NAIP), may be more suitable for detecting fine-scale features, but requires users to estimate their own landscape classifications to make the imagery useful for ecological analysis.

The purpose of my study was to create LULC maps from NAIP imagery and estimate LULC change over a ~10-year period of a 24.2 × 21.7 km portion of North Dakota’s Prairie Pothole Region (PPR). I also used the classified NAIP maps to estimate changes in potential functional connectivity for amphibians based on hypothesized effects of land cover on their movement across the landscape. For both objectives, I compared estimates obtained from NAIP imagery to those obtained from lower resolution NLCD maps. I found that object-based image analysis (OBIA) of NAIP images (overall accuracy 88.4-93.2%; Kappa 0.840-0.901) classified the landscape more accurately than two pixel-based classification methods (overall accuracy 70.9-77.3%; Kappa 0.603-0.677) and produced more accurate maps than NLCD (overall accuracy 69.7-75.0%; Kappa 0.563-0.658) in the study area. From 2009 to 2018, cropland and wetland cover increased while grassland/pasture/hay and tree/forest cover decreased within the classified NAIP images. The increase in wetland cover corresponded with an increase in smaller, more closely positioned wetlands. However, wetland patches also appeared to become functionally less connected for amphibians, based on assumed relationships between land cover and constraints on movement. In contrast, NLCD maps estimated different overall LULC proportions and the opposite trend in potential connectivity. My research suggests that higher resolution data may be more suitable for assessing the landscape patterns and trends for small-bodied animals with limited mobility, such as amphibians.