Spatio-temporal variability of X-band radar backscatter and coherence over the Lena River Delta, Siberia.

Abstract

Time series of TerraSAR-X backscatter intensity and 11-day interferometric coherence with high temporal resolution have been used to interpret major seasonal land surface changes in a variety of tundra environments, namely an area of wet polygonal tundra, a drier Ice Complex upland area, a recently drained well-vegetated lake basin, a partly well-vegetated floodplain, a bare sandbank, and a very dry area of rocky outcrops. Seasonal variations in intensity and coherence were evaluated in the context of meteorological conditions such as air temperatures, precipitation and snow cover status. The TSX signal appeared to have very limited penetration through vegetation and the observed variations in backscatter and coherence were therefore mainly attributed to processes in the upper layer of vegetation. Variations in the TSX backscatter intensities were mostly moderate throughout the annual cycle. Backscatter was found to be insensitive to ground freezing and thawing as well as being generally insensitive to precipitation, but it was sensitive to (i) an individual rain event at the time of SAR acquisition, (ii) an individual snow shower coinciding with unusually high air temperature, and (iii) the spring melt of the snowpack (likely with a refrozen icy crust on the surface). Flooding of the sandbank was clearly detectable from extremely low backscatter values. The selected regions of interest (ROIs) demonstrated generally good separability on the basis of differences in their backscatter intensities: rough and very sparsely vegetated rocky outcrops yielded the highest backscatter and the smooth barren sandbank yielded the lowest backscatter. The backscatter from the vegetated ROIs yielding intermediate values, with the less vegetated ROIs returning lower backscatter. Interferometric coherence comprises both amplitude and phase signal components and should therefore be more sensitive to surface changes than backscatter intensity alone, especially at the X-band frequency, an assumption that is strongly supported by the results of our investigations. The coherence decreased dramatically with the onset of snow cover in all of the landscape types. The snow melt period was also clearly identified by another reduction in coherence. The snow shower that affected the backscatter also caused a reduction in coherence. January and February yielded the highest coherence values for all of the ROIs (with mean values of up to 0.9 for the rocky outcrops).