Error analysis of spectral reflectance derived from imaging spectrometer data

As the field of remote imaging spectrometry grows, interest increases in applications that require the data be "corrected" to surface reflectance. However, the utility of the data for these applications will be limited by the accuracy to which the correction can be performed. Two approaches to atmospheric compensation are reviewed and applied to airborne spectrometer data to study their error characteristics. An end-to-end analysis model is used to extend the study to examine the effects of individual sources of error. Results indicate that random errors of 1 to 2% reflectance units and bias errors of 1 to 4% are achievable in atmospheric window regions, with considerably higher errors in atmospheric absorption bands.