PS 53-178 - Detecting terrestrial cyanobacteria using Landsat imagery and the Phycocyanin Content algorithm

Wednesday, August 10, 2011
Exhibit Hall 3, Austin Convention Center
Enrique Gomezdelcampo, School of Earth, Environment, and Society, Bowling Green State University, Bowling Green, OH and Lee M. Bartholomew, Department of Geology, Bowling Green State University, Bowling Green, OH
Background/Question/Methods

The Phycocyanin Content algorithm (PYC) was originally developed to quantify blooms of the toxic cyanobacteria, Microcystis aeruginosa, in large freshwater bodies, particularly in Lake Erie, where the data for the development of the algorithm was obtained . However, when Landsat data was processed with PYC for a lake it also showed some highlighted sites on land, depending on season, and land cover vegetation. Theoretically, because of the wide distribution of cyanobacteria on land, the PYC may be applicable in many different types of terrestrial environments, including farmland, sand dunes, deserts, alpine tundra, and areas affected by recent wildfires.

Can the Phycocyanin Content algorithm (PCY) be used to quantify cyanobacteria pigments on land? The Algodones Wilderness Area in the Imperial Sand Dunes in California was chosen as the study site based on the known influences of the algorithm, and the need for a site which was likely to be cloud free and in an overlap area between two paths that the L5 and L7 Landsat satellites travel.

The PYC algorithm, when applied on land, was negatively affected by the averaging effect on each pixel of vegetation and minerals. Filters to reduce this image noise on a terrestrial setting were developed by using a conceptual idealized dataset to establish the hypothesized relationship between chlorophyll and phycocyanin. The ratio of Landsat TM bands 4/3 was used as a vegetation filter and the 3/2 ratio was used to lessen the effects of iron oxides on the application on land of the PYC.

Results/Conclusions

The PYC performed successfully in areas of very low vegetation as the vegetation and mineral filters worked as designed. However, further calibration is necessary for this algorithm to function as a quantification tool. The mineral filter was based on the iron oxide mineral as it is the prevalent mineral in the region. Unfortunately, the iron oxide filter also detected areas of senescent vegetation and inversely indicated areas with green vegetation, complicating the performance of PYC in these areas.

Copyright © . All rights reserved.
Banner photo by Flickr user greg westfall.