Land Change Extrapolation for Virginia Coast Reserve with TerrSet’s Land Change Modeler
Course: GIS and Land Change Models
Date: December 2019
Authors: Rachel Corcoran-Adams, Julia Moore, Elizabeth Lohr, and Jess Sztrempko
Background: As the final project for "GIS & Land Change Models," my group and I used land cover classification data provided by the National Science Foundation's Long Term Ecological Research Network to simulate transitions between Water, Marsh, and Upland in the Virginia Coast Reserve. This research uses the TerrSet’s Land Change Modeler (LCM) to extrapolate land change during 1957-2013. This method used maps of change from 1949-1957 to calibrate and then utilized a Markov matrix to determine the quantity of extrapolated change, employing maps of distance to edge to determine the spatial allocation of extrapolated change.
Data:
Methods: We ran LCM two times, each using a particular set of distance to edge driver maps. The first run used distance to edge of land cover change 1949-1957, while the second run used distance to edge of land cover patches at 1949 and 1957. Validation compared the reference change to the extrapolated change during 1957-2013.
Outcomes: Results for both runs show that the extrapolated change is 3.7 times greater than the reference change, because reference change during the calibration time interval is faster than reference change during the validation time interval. Error due to an incorrect quantity is greater than error due to incorrect allocation of simulated change. Distance to edge of land cover patches created a marginally more accurate allocation than distance to edge of land cover change
Date: December 2019
Authors: Rachel Corcoran-Adams, Julia Moore, Elizabeth Lohr, and Jess Sztrempko
Background: As the final project for "GIS & Land Change Models," my group and I used land cover classification data provided by the National Science Foundation's Long Term Ecological Research Network to simulate transitions between Water, Marsh, and Upland in the Virginia Coast Reserve. This research uses the TerrSet’s Land Change Modeler (LCM) to extrapolate land change during 1957-2013. This method used maps of change from 1949-1957 to calibrate and then utilized a Markov matrix to determine the quantity of extrapolated change, employing maps of distance to edge to determine the spatial allocation of extrapolated change.
Data:
- PIE Land Cover Classifications for 1949, 1957, and 2013 were provided by our project adviser, Dr. Robert Gilmore Pontius. The images were classified into three land cover categories: water, marsh, and upland.
- Edges of land cover types in 1949 and 1957 were used as a naive variables to simulate land cover transitions. This data was created by our group.
Methods: We ran LCM two times, each using a particular set of distance to edge driver maps. The first run used distance to edge of land cover change 1949-1957, while the second run used distance to edge of land cover patches at 1949 and 1957. Validation compared the reference change to the extrapolated change during 1957-2013.
Outcomes: Results for both runs show that the extrapolated change is 3.7 times greater than the reference change, because reference change during the calibration time interval is faster than reference change during the validation time interval. Error due to an incorrect quantity is greater than error due to incorrect allocation of simulated change. Distance to edge of land cover patches created a marginally more accurate allocation than distance to edge of land cover change