Estimating subcanopy soil moisture with radar

• Published in: Journal of Geophysical Research: Atmospheres 2000-06, Vol.105 (D11), p.14899-14911
• Main Authors: Moghaddam, Mahta, Saatchi, Sasan, Cuenca, Richard H.
• Format: Article
• Language: English
• Subjects: Earth, ocean, space; Exact sciences and technology; External geophysics; Geophysics. Techniques, methods, instrumentation and models
• ISSN: 0148-0227; 2156-2202
• DOI: 10.1029/2000JD900058

The subcanopy soil moisture of a boreal old jack pine forest stand is estimated using polarimetric L and P band airborne synthetic aperture radar (AIRSAR) data. Model simulations have shown that for this stand the principal scattering mechanism responsible for radar backscatter is the double‐bounce mechanism between the tree trunks and the ground. The data to be used here were acquired during five flights from June to September 1994 as part of the Boreal Ecosystem‐Atmosphere Study (BOREAS) project. The dielectric constants, or equivalently moisture contents, of the trunks and soil can change significantly during this period. To estimate these dynamic unknowns, parametric models of radar backscatter for the double‐bounce mechanism are developed using a series of simulations of a numerical forest scattering model. A nonlinear optimization procedure is used to estimate the dielectric constants. Ground measurements of soil and trunk moisture content are used to validate the results. The trunk moisture content measurements are used to gain confidence that the respective estimation results are accurate enough not to corrupt the soil moisture estimation, which is the main focus of this paper. After conversion of the trunk moisture measurements to dielectric constants it is found that the estimated values are within 14% of the measurements. Owing to possible calibration uncertainties in the soil moisture measurements on the ground as well as in AIRSAR data, the variations rather than the absolute levels of the estimated soil moisture are considered. The results indicate that the estimated variations closely track the measurements. The worst case average estimated change differs by