The performance of vegetation indices when applied to early Landsat data for irrigation system detection in Cambodia

Authors: Corrine Coakley*, Kent State Geography Dept., James Tyner, Kent State University, Mandy Munro-Stasiuk, Kent State University, Sokvisal Kimsroy, Kent State University
Topics: Remote Sensing, Coupled Human and Natural Systems, Historical Geography
Keywords: Vegetation Index, Cambodia, Hydrology, Khmer Rouge
Session Type: Poster
Day: 4/4/2019
Start / End Time: 9:55 AM / 11:35 AM
Room: Lincoln 2, Marriott, Exhibition Level
Presentation File: Download


Landsats 1-3 still have much information to provide the remote sensing scientist, especially when applied to areas of the globe where ground surface data from that time period is scarce. Using Landsat data from 1973 and 1979, we created a database of irrigation canals and reservoirs built during the timeframe when the Khmer Rouge occupied Cambodia (1975-1979). By applying multiple vegetation indices to early Landsat data, our investigation has revealed that over 9,000 kilometers of canals and over 400 reservoirs that were built during this time. Many of the structures built during the time of the Khmer Rouge are known to have failed during or soon after their construction, and the remnants are barely visible even using high resolution modern imagery. This study compares the results of NDVI, CTVI, and TTVI methods at two locations in Banteay Meanchey Province, Cambodia. Linear features at these sites appear very similar in the satellite imagery, and in false color infrared it is easy to confuse canals and roads. While all three methods proved adequate at isolating linear features, the NDVI output often overemphasized water pixels, making it hard to distinguish surface water areas from vegetation. The TTVI results confused clouds and surface water, resulting in overcounting of water pixels. The CTVI algorithm was most successful in detecting canal features and eliminating roads and most successfully documented the building of the canals.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login