University of Twente Student Theses


Spectral merging of MODIS / MERIS Ocean Colour Data to improve monitoring of coastal water processes

Freitas, Fernanda Henderikx (2009) Spectral merging of MODIS / MERIS Ocean Colour Data to improve monitoring of coastal water processes.

[img] PDF
Abstract:MODIS and MERIS are considered potential sensors for observing coastal oceanographic processes due to their appropriate spectral and spatial resolutions. In order to make use of the full spectral range provided by both sensors, a spectral merging technique was proposed. The method was tested for fifteen sampling points in the Santa Barbara Channel (SBC), California, for which field data was available (absorption coefficients and water-leaving reflectances). Because merging at any level requires that the data involved be as comparable and as free of errors as possible, Level 1B data from both sensors were pre-processed and atmospherically corrected utilising the same assumptions and algorithms. Since absorption and backscattering coefficients for the sampling units were either provided or calculated, a bio-optical model in the forward mode was used to estimate remote sensing reflectance for the complete range of wavelengths covered by MODIS and MERIS. Once all discrete wavelengths of one sensor had a common pair in the other sensor, a wavelength-based multi-linear regression was performed, having each sensor as an independent variable, and the field data (also modelled for the same wavelengths) as the dependent variable. The regression coefficients generated were tested for seven validation points. The merged spectra resulted in more accurate estimations of the field observations, if compared to the accuracy of the sensor’s estimations alone. That was considered a positive result, as only a limited number of validation points were available, and the SBC region itself is highly variable in time and space. The bio-optical model was able to predict the original data at reference wavelength around 443 nm (<1% error), and the results obtained are highly dependent on the model’s limitations and assumptions. In order to know if in practice the merged spectra can contribute with depicting information that the sensors alone cannot recognise, the technique should be applied for the whole region, rather than pixelbased, so that the processes could be directly observed. Important to notice that this work represents a first attempt in dealing with the complexity of merging radiometric units of sensors that are not commonly studied together because of their inherent differences. Despite the many limitations observed, it is believed that applying the technique for a larger number of sampling units or in a less complex environment could greatly improve the results obtained.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page