Impacts of lossy compression on hyperspectral products for forestry

Hyperspectral data from satellites are voluminous. Satellite data from a hyperspectral sensor can be transmitted at either Ka-band rates without compression or at X-band rates with lossy compression through links to ground stations. If lossy compression is used, there can be a 10:1 reduction in the amount of data to be transmitted. The Canadian Space Agency (CSA) has developed algorithms for lossy compression. For Canada´s Hyperspectral satellite HERO (Hyperspectral Earth Resource Observer), consideration is being given to lossy compression of the data prior to data transmission. Experiments have been conducted with Hyperion and AVIRIS data to assess the impacts of compression on forest information products. These assessments have included forest classification products for forest inventory. This paper presents the results of these experiments with multiple analysis methods. The results indicate that 10:1 lossy compression produces too large a loss in information content in hyperspectral imagery for forest information products. Comparisons are given with other lossy compression methods