Mapping Melt Pond Bathymetry on Arctic Sea Ice by Means of Optical Remote Sensing
The last decades saw an increasingly rapid decrease of Arctic sea ice with associated consequences for the climate, ecosystems and human activities on local, regional and global scales. The magnitude of sea ice loss observed exceeded predictions of sea ice and climate models. This deficiency and disagreements between different model outputs indicate that some sea ice processes are not yet entirely understood. Improving our understanding of how the Arctic sea ice changes requires regular, intercomparable large-scale observations of essential sea ice parameters, which are only possible by means of remote sensing. One of the main drivers of Arctic sea ice loss is the ice-albedo-feedback mechanism which is strongly influenced by the presence and evolution of melt ponds. Contemporary remote sensing products quantify melt ponds almost exclusively in terms of coverage. A comprehensive understanding of the spatio-temporal evolution of melt ponds, however, also requires monitoring of pond depth, which is an important parameter in sea ice models. This thesis presents a new method to derive the depth of melt ponds on Arctic sea ice by means of optical remote sensing and is based on a uniquely comprehensive dataset of field based observations and remote sensing data. Firstly, a new method to retrieve melt pond depth from optical field data with unprecedented accuracy is presented. The newly developed method rests upon the log-scaled remote sensing reflectance at 710 nm as a function of depth, is largely independent from bottom albedo, and addresses varying sun zenith angles, making it particularly suitable for remote sensing applications. Secondly, the new method is applied to atmospherically corrected hyperspectral airborne imagery to map two-dimensional melt pond bathymetry. Retrieved bathymetries are in good agreement with field observations and confirm the method's large independence from bottom albedo but results indicate that precise atmospheric correction is a challenge in the Arctic sea ice environment. Thirdly, five different atmospheric correction algorithms for satellite imagery are compared for their applicability over Arctic sea ice using the example of Sentinel-2 data. Comparisons with field data question the suitability of most algorithms for application in this environment in their current state but we found potential for further development. Findings of the dissertation indicate that the new method enables accurately mapping the bathymetry of melt ponds on Arctic sea ice from hyperspectral remote sensing data with subject to precise atmospheric correction. This thesis may serve as a first step towards an operational remote sensing-based monitoring of melt pond depth on Arctic sea ice.