In situ and satellite observations of the visible and infrared albedo of sea ice during spring melt

Abstract

Using in situ and satellite data, this thesis identifies and explains the spatial and temporal characteristics of sea ice visible (0.15-0.70 um) and infrared albedo (0.7-4.0 um) during the annual spring melt. Specifically, spectral albedo and complementary physical measurements were collected over first-year and multiyear ice located in the Barrow St./Lancaster Sound area of the Canadian Archipelago in the spring of 1993. Concurrently, clear-sky Advanced Very High Resolution Radiometer (AVHRR) satellite data from NOAA 12 were recorded for analysis. Of specific interest was the utility of the satellite's visible (0.58-0.68 um) and near-infrared (0.73-1.1 um) channels for estimating visible and infrared sea ice surface albedo during the dynamic spring transition period. Surface albedos were inverted from satellite top-of-atmosphere radiances after correcting for sensor calibration drift, scene anisotropy and atmospheric attenuation. In order to determine the uncertainty in the inversion method, a complete error budget for the satellite-derived surface albedos was performed. The sensitivity of the method to uncertainties in input parameters was also examined.

A significant drift (8-25%) was detected in the calibrations of the visible and near-infrared satellite channels. Updated calibration coefficients were derived and are presented herein. Scene anisotropy was corrected using scene models derived from Nimbus-7 ERBE data. The high variability in the anisotropic reflectance factors for individual view configurations limits the precision of the albedo inversion method presented here. A statistical approach was used to remove atmospheric attenuation effects by incorporating linear coefficients derived from radiative transfer simulations. During winter and early spring in the arctic, increased atmospheric turbidity warrants careful specification of aerosol optical depth within the atmospheric correction procedure. Simulated albedo surfaces were used to derive linear coefficients used to convert narrowband albedos to broadband visible and infrared albedos. The satellite-derived surface albedos compared well to in situ albedo measurements (0.05-0.1). However the inversion method suffers from uncertainty regarding the employed anisotropic reflectance factors. The in situ albedo observations were used along with the satellite-derived albedo data to identify and characterize the spatial and temporal variability of sea ice albedo over the experiment region. Pairwise image comparison techniques were used to examine change between image dates and differences between visible and infrared albedo. Principal component analysis was used to identify the major and minor scale variance structures in the multitemporal satellite albedo dataset.

The satellite and in situ albedo measurements indicated that sea ice albedo is high and relatively homogenous over the study scene for winter and early spring conditions. Variability in visible albedo for this part of the season was found to be linked to the visibility of the darker underlying ice surface. Due to their thick snow volumes, visible albedo of multiyear ice melt pond areas stayed constant for the winter and the onset of melt periods. Alternatively, thinner volumes on first year ice resulted in visible albedo decreasing with the onset of melt conditions. The consistently lower infrared albedo decreased in response to the seasonal enlargement of snow grains in the top layer of the snow volumes. The saturation of the snow volumes resulted in decreases in both the visible and infrared albedo. The subsequent development of surface melt ponds resulted in significant decreases in both visible and infrared albedo, as well as an increase in the scene variability. The higher ponding on ice that consolidated in late 1992 compared to that ice which consolidated in late winter 1993 illuminated the importance of snow depth and ice macromorphological characteristics in determining late season patterns of albedo. Breakup and dispersal of ice in the study scene resulted in a significant lowering of the magnitude and variability of both visible and infrared albedo. High frequency fluctuations in albedo were found due to melt water drainage and ice convergence and divergence. Feedback diagrams describing the relationships between both seasonal visible and infrared albedo and relevant environmental factors are presented herein.