Temporal Characteristics of Boreal Forest Radar Measurements

Doktorsavhandling, 2020

Radar observations of forests are sensitive to seasonal changes, meteorological variables and variations in soil and tree water content. These phenomena cause temporal variations in radar measurements, limiting the accuracy of tree height and biomass estimates using radar data. The temporal characteristics of radar measurements of forests, especially boreal forests, are not well understood. To fill this knowledge gap, a tower-based radar experiment was established for studying temporal variations in radar measurements of a boreal forest site in southern Sweden.
The work in this thesis involves the design and implementation of the experiment and the analysis of data acquired. The instrument allowed radar signatures from the forest to be monitored over timescales ranging from less than a second to years. A purpose-built, 50 m high tower was equipped with 30 antennas for tomographic imaging at microwave frequencies of P-band (420-450 MHz), L-band (1240-1375 MHz) and C-band (5250-5570 MHz) for multiple polarisation combinations. Parallel measurements using a 20-port vector network analyser resulted in significantly shorter measurement times and better tomographic image quality than previous tower-based radars. A new method was developed for suppressing mutual antenna coupling without affecting the range resolution. Algorithms were developed for compensating for phase errors using an array radar and for correcting for pixel-variant impulse responses in tomographic images.
Time series results showed large freeze/thaw backscatter variations due to freezing moisture in trees. P-band canopy backscatter variations of up to 10 dB occurred near instantaneously as the air temperature crossed 0⁰C, with ground backscatter responding over longer timescales. During nonfrozen conditions, the canopy backscatter was very stable with time. Evidence of backscatter variations due to tree water content were observed during hot summer periods only. A high vapour pressure deficit and strong winds increased the rate of transpiration fast enough to reduce the tree water content, which was visible as 0.5-2 dB backscatter drops during the day. Ground backscatter for cross-polarised observations increased during strong winds due to bending tree stems. Significant temporal decorrelation was only seen at P-band during freezing, thawing and strong winds. Suitable conditions for repeat-pass L-band interferometry were only seen during the summer. C-band temporal coherence was high over timescales of seconds and occasionally for several hours for night-time observations during the summer. Decorrelation coinciding with high transpiration rates was observed at L- and C-band, suggesting sensitivity to tree water dynamics.
The observations from this experiment are important for understanding, modelling and mitigating temporal variations in radar observables in forest parameter estimation algorithms. The results also are also useful in the design of spaceborne synthetic aperture radar missions with interferometric and tomographic capabilities. The results motivate the implementation of single-pass interferometric synthetic aperture radars for forest applications at P-, L- and C-band.