Date of Award

12-2014

Document Type

Thesis

Degree Name

Master of Science (MS)

Legacy Department

Plant and Environmental Science

Advisor

Dr. Patrick McMillan

Committee Member

Dr. David White

Committee Member

Dr. Dara Park

Abstract

This study was conducted to better predict and assess damage to high-value small-spatial scale landscapes from storm water. Storm water damage in the form of rill formation across the South Carolina Botanic Gardens (SCBG) Natural Heritage Garden Trail has been modelled as a function of contributing area using D8 and D-infinity flow direction algorithms on a preprocessed LiDAR-derived elevation raster. D8 and D-infinity algorithms were also applied over a set of stochastic Monte Carlo simulations (n=1,000) representing elevation error. The contributing area was calculated using each of the four methods for each 5'x5' cell along the trail. The output was then filtered using a moving kernel calculating a value for each cell according to the maximum value within specified radii of neighboring cells. Observed storm water damage along the trail was geo-referenced as a validation dataset for the model. The receiver operating characteristic (ROC) curves of the three contributing area estimates filtered at various filter radii were graphed by comparison with geo-referenced rills. Results indicate that high resolution LiDAR elevation data can be used to localize storm water damage risks. The D-8 and D-infinity algorithms performed equivalently, and the Monte Carlo procedure improved the performance of both. These models should prove effective in predicting and preventing damage in high-value public landscapes.

HydroDEM.tif (7830 kB)
supplement

MCsim.py (3 kB)
supplement

MCagg.py (1 kB)
supplement

MCsimDinf.py (2 kB)
supplement

Share

COinS