Evaluation of an airborne optical remote sensing system and operational pre-processing methods

For remote sensing (RS) to be a truly useful technique in environmental science, the remotely sensed data need to be of known quality and reliability. The various steps known collectively as ‘pre-processing’ are vitally important as error and uncertainty at this stage has the potential to cause large errors in the final data product. Two important aspects of pre-processing were investigated in this research. First, the radiometric conversion by which acquired signals are converted into physically meaningful values, and which allows the comparison of datasets from different sensors or from different dates with the same sensor was studied. Second, the process by which the influence of the atmosphere is removed from the remotely sensed signal was investigated, focusing upon practical methods to correct data collected by CASI-2, an imaging spectrometer produced by Itres Research. The radiometric (and other) characteristics of any sensor are normally obtained by laboratory calibration. Several experiments were conducted to investigate and enhance knowledge of the performance of the CASI-2. The results suggested that wavelength calibration could reveal systematic optical distortion due to the ‘optical smile’ effect, and that the uncertainty of the wavelength calibration could be reduced if this was taken into account. Another possible source of error in the sensor calibration was identified and traced to a spatially non-uniform radiance standard, improvement of which could greatly reduce systematic error across the image. In addition to investigation of the conventional sensor calibration, several studies were conducted to acquire a new information. For example, the spectral response function of the CASI-2 was determined independently of the manufacturer for the first time, using an innovative iterative procedure. In addition to the research conducted on assessing the laboratory calibration procedure of the CASI, and investigating its performance, a series of laboratory investigations were undertaken to characterise the CASI Incident Light Sensor (ILS). Generally, the performance of the ILS, such as its angular response radiometric linearity were acceptable for the purpose for which it was designed. However, signals in the short wavelength region seemed to suffer from relatively low signal-to-noise ratio. In the interest of operational aspects of airborne multispectral RS, the contribution of atmospheric variation to remotely sensed data was reviewed, focusing in particular, upon numerical models developed to characterise the sky radiance distribution. Atmospheric effects on remotely sensed data were reviewed and the effects of atmospheric variability studied in terms of how this influences the remotely sensed signal. Temporal variations in atmospheric clarity (and by extension, spatial variations typical of RS image data), were shown to cause errors which also affected multispectral ratio-based analysis. Two novel practical methods of atmospheric correction were developed following a series of practical experiments and theoretical studies.