Characterization of land surface thermal structure from NOAA-AVHRR data over a northern ecosystem

Abstract

This paper focuses on the estimation and analysis of surface thermal parameters (emissivities and surface temperatures) in a nordic environment (Québec, Canada). The land cover in this region varies from boreal forest in the south to tundra in the north. The thermal parameters are estimated from two variants of a new model that combines the radiances of the short wave infrared (SWIR) spectral band [advanced very high resolution radiometer (AVHRR) channel 3: 3.55–3.93 μm] and the thermal bands (AVHRR channel 4: 10.5–11.5 μm; and AVHRR channel 5: 11.5–12.5 μm). The study, carried out for images acquired on different dates, reveals that, in most situations, the two approaches allow the separation of emissivities and surface temperatures. Analysis of the variations of the estimated emissivities in relation to surface patterns shows that they are slightly variable in spectral bands 4, and 5, with values generally greater than 0.95. Variations are more important in the SWIR channel, where values less than 0.90 appear, especially in urban areas. In general, surface emissivities increase with the density of the vegetation cover. Moreover, for densely vegetated areas, SWIR surface reflectivities, which can be derived from emissivities, appear to be well correlated with the reflectivities of the AVHRR visible channel. As with emissivities, variations of the estimated land surface temperatures (LST) in relation to vegetation density, characterized by the normalized difference vegetation index (NDVI), were considered. The relations between the two parameters (LST and NDVI) show essentially two opposite directions of linear variations (positive and negative correlations). In the light of the main results obtained, the synergistic use of the different spectral regions (visible, near, mid-, and thermal infrareds) could be very useful in the parameterization of boreal ecosystems.