Understanding the Effects of Spatial and Temporal Scale on State and Transition Models Based Upon Remotely Sensed Data
Abstract
The grassland-forest ecotone is a highly diverse and complex region that encompasses a multitude of grasslands, savannas, and forests. Ecological and anthropogenic pressures are causing dramatic changes in the health of the landscapes within this ecotone. A new methodology was created using State and Transition Models and Remote Sensing to understand and assess the landscape health. This methodology used six spectral indices (MTVI, NDSVI, NDVI, NDWI, SATVI, and SWIR32) created from times series Landsat 4-5TM. These spectral indices and LiDAR were used to characterize the spatial, spectral, and temporal properties of vegetation states and substates across 20 sites within the grassland-forest ecotone region of North Dakota and Minnesota. This suite of characteristics was used to create spectral keys. These spectral keys were then used to identify states and substates of Landsat-scaled State and Transition Models within Sheyenne National Grasslands. The effectiveness of 6 State and Transition Models was tested using the metrics of kappa, producer's accuracy, user's accuracy, and overall accuracy. This methodology was successful in identifying Tallgrass, Mixed, and Sand Prairies states. The entirety of the Tallgrass Prairie State and Transition Model met with the highest overall accuracy of over 80%. Spectral mixing was one of the main causes for low overall accuracy within the State and Transition Models. The characteristics of these landscape State and Transition Models were then applied to create a baseline for determining landscape health. The properties of the vegetation states, substates, and transitions were then applied to calculate values for landscape health, and its' four indicators (Vigor, Organization, Resilience, and Ecosystem Services). An example was then used to illustrate how these values for indicators of landscape health could be used to help identify problems or improve the landscape health. The metrics used to determine landscape health, and landscape State and Transition Models with remote sensing are an important step to monitoring, understanding, and researching the Landscape Health of the highly diverse grassland-forest ecotones of this world. The supplementary materials contain additional over time and phenology graphs of ecological sites and their managed lands for each study site.
- Publication:
-
Ph.D. Thesis
- Pub Date:
- 2018
- Bibcode:
- 2018PhDT........44L
- Keywords:
-
- Environmental science;Remote sensing;Range management