Arctic Ecology Laboratory

Land Cover Classification and Modeling of Ecosystem Carbon Flux in the Barrow Environmental Observatory Using IKONOS Satellite Imagery

Fig. 4 - Flow chart illustrating model development and allocation of tasks.

• A map of land cover is fundamental to most terrestrial research programs and sound land management practice. Until now, no high resolution land cover map existed for the Barrow area.

• An automated classification of pan-sharpened IKONOS imagery offers an opportunity to generate a high resolution land cover map for the BEO and neighbouring lands.

• A land cover classification scheme with decision rules was developed based on plot measurements that the Webber group has collected over recent years throughout the Barrow area. This classification is hierarchical and cross walks with other classifications formerly used in the Barrow area4. Our aim was to as closely as possible link classifications detectable at the plot level to the landscape level.

• Prior to classification, the 4-meter multispectral image was pan- sharpened with the 1-meter panchromatic band, resulting in a multispectral image with 1 meter resolution (Fig. 3). Data including vegetation type and GPS location were taken from 225 sites and used for training classification algorithms. Water and clouds were removed from the image prior to land cover classification to reduce overall spectral variation. Supervised classification algorithms were run on the pan-sharpened imagery in ERDAS Imagine 8.4. Initially, the classification was limited to areas that fell within one standard deviation of the training sites. Areas that fell outside of this range were removed and classified separately using a maximum likelihood match with the same training sites. A flow chart describing the classification process and resulting map is shown to the right along with areal cover of each land cover type classified.

• An accuracy assessment of the classification will be completed in summer 2002 when a real-time Differential Global Positioning System (DGPS) will be available in Barrow.

Continue browsing this poster:

1. Beginning
2. Introduction
3. IKONOS Satellite Imagery
4. Land Cover Classification
5. Modeling of Carbon Flux (GEE)
6. Conclusions
7. References