Evaluating soil moisture variability using Synthetic Aperture RADAR and LiDAR-derived wetness indices

Thumbnail Image
Powell, Kathryn A.
Journal Title
Journal ISSN
Volume Title
University of Guelph

Soil moisture spatial and temporal variability is influenced by precipitation patterns, local topography, soil texture and vegetation. This research aims to assess whether the ln('As'/tan[beta]) (Beven and Kirkby, 1979) wetness index (WI) derived from a 1 m LiDAR digital elevation model can be used as a surrogate for soil moisture spatial patterns through time by comparing Synthetic Aperture RADAR (SAR) backscatter to the WI under variable moisture regimes and using different sensor parameters. Over multiple dates spanning Fall 2009 and Spring/Fall 2010, fine quad-polarimetric mode RADARSAT-2 imagery and coincident 'in situ' surface parameter data are acquired over a small agricultural watershed in southwestern Ontario. Soil moisture maps are derived using several backscatter models and compared to 'in situ' soil moisture using goodness-of-fit statistics. The spatial pattern of soil moisture was observed to be most variable under moderate moisture regimes through evaluation of 'in situ' soil moisture data and semivariogram analysis. Both SAR-derived soil moisture and SAR linear intensity channels were compared to the WI using the Spearman rank correlation coefficient ('rs'). SAR soil moisture derived using the Oh et al. (1992) model showed the highest correlation to the WI, achieving significant but weak positive correlations at both high and low incidence angles. The cross-polarized intensity channel (HV) correlated more strongly with WI than either co-polarized channel (HH or VV), achieving significant weak to moderate 'rs' values. Standardized anomalies of soil moisture values were calculated for all SAR image acquisitions and compared to the WI using 'rs'. Discontinuous but comparable time stable points were identified using all backscatter models, although only those derived using the Oh et al. (1992) model correlated significantly to the WI. For all model outputs, most points were time stable within one standard deviation of the mean.

RADAR, synthetic aperture, LiDAR, wetness index, soil moisture