A genetic programming model assimilates SAR images and geoenvironmental parameters to assess soil moisture at the watershed scale.A genetic programming (GP)-based, nonlinear modeling structure relates soil moisture with synthetic-apertureradar (SAR) images to present representative soil moisture estimates at the watershed scale. Surface soil moisture measurement is difficult to obtain over a large area due to a variety of soil permeability values and soil textures. Point measurements can be used on a smallscale area, but it is impossible to acquire such information effectively in largescale watersheds. This model exhibits the capacity to assimilate SAR images and relevant geoenvironmental parameters to measure soil moisture. In the past, spaceborne radar imaging satellites used all-weather observation, but estimation methods of soil moisture based on active or passive satellite images remains uncertain. Estimation of soil moisture based on SAR measurement was made possible by developing linear regression models and nonlinear regression models in a single land use/land cover from several hundred square meters to several square kilometers, based on traditional statistical regression theory. This GP-based artificial intelligence mode uses an evolutionary computational approach to estimate soil moisture with a variety of land use/land cover patterns.
The function derived in the evolutionary
computation links a series of crucial
topographical and geographical features
including slope, aspect, vegetation
cover, and soil permeability with well-calibrated
SAR data. Research findings
indicate that this development and
application of the GP model has proved
useful for generating a highly nonlinear
structure in regression regimes, which
exhibit strong statistical correlations
between the model estimates and the
ground truth measurements (volumetric
water content), based on unseen
Using this model, science missions
would be capable of handling large-scale
moisture estimation using spaceborne
satellite images, and could generate
multi-temporal soil moisture maps over
seasons. The GP-model is ultimately
extensible and interoperable for any
river basin of interest, though the
impact of landscape complexity needs to
be studied further.
This work was done by Ni-Bin Chang of Texas A&M University for Stennis Space Center.
Inquiries concerning rights for its commercial
use should be addressed to:
Texas A&M University
332 Wisenbacker Eng. Research Center
College Station, TX 77843-3000
Phone No.: (407) 823-1375
Refer to SSC-00256, volume and number of this NASA Tech Briefs issue, and the page number.