Characterizing remotely sensed soil moisture over an agricultural catchment

Garett William Pignotti, Purdue University

Abstract

As global population and subsequent demand on the provisioning services of natural systems continue to rise, it is increasingly important to identify creative, pragmatic solutions to meet these needs while minimizing environmental degradation. Of particular importance are agricultural ecosystems, among the most widespread anthropogenically driven and managed systems. Soil moisture is critical to characterizing many atmospheric, hydrologic, and land-surface interactions relevant to monitoring and modeling such systems. However, soil moisture is spatially and temporally highly variable, and is thus challenging to measure in a meaningful, representative manner with point based in situ sensors. Alternatively, remote sensing in the microwave region of the electromagnetic spectrum is capable of providing all weather, day/night gridded global soil moisture measurements. Unfortunately, the spatial resolution of standard passive microwave based products is too low (∼40 km) for most hydrologic applications (1–10 km). One approach to overcome this limitation is to combine soil moisture maps with fine scale ancillary remote sensing data to downscale soil moisture data products. This research focused on evaluating two variants of a downscaling algorithm applied to soil moisture products derived from data acquired by the SMOS (Soil Moisture and Ocean Salinity) satellite over a sensor network in the Upper Cedar Creek Watershed in northeastern Indiana. Results in this region indicated that while the accuracy of standard SMOS products does not meet the stated mission objectives (0.068 m3/m3 vs. 0.04 m3/m3), the derived downscaled data did not degrade either the root mean square difference or correlation with in situ measurements. Potential deficiencies in the downscaling models were identified and discussed. Overfitting was detected in one approach, and both models were inadequate for tracking temporal variations associated with vegetation data in the Midwest. A qualitative evaluation of within network variability and auxiliary soil moisture data sources was also performed. Comparison of remotely sensed data to ground networks provided insight into sensor derived response over a heterogeneous area and potential to exploit such relationships. By refining downscaling approaches, it should be possible to better resolve the scale at which soil moisture processes act. Specific to agricultural applications, accurate soil moisture estimation can improve yield prediction, forecasting of drought and floods, and water management practices.

Degree

M.S.E.

Advisors

Crawford, Purdue University.

Subject Area

Environmental engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS