Active-Passive Microwave Detection of Shallow Ground Freeze/Thaw Dynamics
Jan Hofste is a PhD student in the Department of Water Resources. (Co)Promotors are prof.dr. Z. Su and dr.ir. C. van der Tol from the Faculty ITC.
Land surface models (LSMs) rely on accurate values of soil liquid water content (mvl(z,t)) and soil temperature (Ts(z,t)) at the skin- (z = 0) and near-surface (top few cm’s) soil layers to correctly model near-surface diurnal freeze/thaw processes. In-situ buried mvl(z,t)- and Ts(z,t) sensors, however, are vulnerable to disturbances by weather, plants and animals. Furthermore, LSMs struggle to properly model skin- and near-surface mvl(z,t) and frozen soil water content (mvi(z,t)) due to inaccurate values of soil thermal properties.
This thesis proposes a novel remote sensing procedure involving active- and passive microwave signatures of freezing/thawing soil – measured at high-temporal frequency: 1x, 2x per hour – that can potentially solve both issues. The procedure uses many microwave channels to infer the skin- and near-surface mvl(z,t) and mvi(z,t): L-band (fc = 1.4 GHz) brightness temperature (TB) at vertical- (v) and horizontal (h) polarization at incidence angles of 40°, 55° and 70° and backscattered power (Prx) across L-, S-, C-, and X-band (fc = 1.625, 2.75, 4.75, 9.5 GHz) at vv- and hh polarization.
Measured Tskin(t) and Ts(z = 5 cm, t) form a profile Ts(z, t) satisfying the heat equation using an effective thermal diffusivity (Deff). Combined with near-surface total water content (mv(z)), a parametric freeze/thaw algorithm generates many scenarios for mvl(z, t) and mvi(z, t). These are converted to compound soil dielectric profiles (ϵc(z, t)) using a four-phase dielectric mixing model. For each scenario, TB(θi) and Prx are simulated using a stratified reflectivity model combined with the advanced integral equation method (AIEM) model for noncoherent scattering. Comparison between simulations and measurements identify the most plausible near-surface freeze/thaw profile scenarios.
Besides aforementioned procedure, this thesis describes the design, calibration, and operation procedure of the above-mentioned scatterometer that was installed on an alpine meadow site (Maqu site) on the Tibetan Plateau next to an (already present) ELBARA-III radiometer.
Furthermore, the influence of vegetation on the microwave signatures of the soil (scattering, absorption) was quantified by the Tor Vergata (TVG) microwave vegetation scattering model which was extended to include canopy heterogeneity over height. The extended TVG model showed no significant improvement over the default model concerning L- to X-band backscatter of the Maqu site grass cover during summer. However, simulations predicted significant differences for bistatic scattering at shorter wavelengths (C-, X-band), relevant for microwave emission modelling and GNSS-R. Crucially, the winter/early spring litter layer caused only small-to-moderate attenuation at shorter wavelengths, easily correctable in bare soil backscatter models.
The procedure with the freeze/thaw algorithm was applied to two 6-day periods: in April 2018 during the thawing phase and in January 2018 during the frozen period. With the thawing period, simulated TB(t) and Prx(t) matched satisfactory with most of the measured channels, demonstrating the suitability of both the microwave model and freeze/thaw algorithm. The best scenario’s mvl(z = 5 cm, t) matched fairly-well with the in-situ measurement. Results for the January period were less successful, attributed to the limited parameter sets used in the freeze/thaw algorithm. Manual adjustment with an alternative mvl(z,t)/ mvi(z,t) scenario yielded a better.
Analysis of results further revealed: interference effects in emission and backscatter around noon due to the formation (and subsequent afternoon thickening) of a thin liquid water layer (≈ λ/4 thickness), that addition of a Surface Impedance Matching Structure (SIMS) to the emission model for roughness correction on the coherent component proved ineffective, and that the TBmismatch at 70° v-pol. suggests an abrupt transition between frozen soil overlying non-frozen soil within the near-surface layer.
We conclude that microwave remote sensing provides a valuable means to infer near-surface mvl and mvi, overcoming limitations of unreliable shallow in-situ sensors. It also offers a pathway for LSMs to infer near-surface soil thermal properties, which are otherwise extremely laborious to quantify over large areas. The scheme proposed in this thesis can be integrated with deeper layer Ts(z,t)- and mv(z,t) measurements to provide LSMs with accurate initialization- and validation data for modelling Tibetan Plateau energy- and water cycles. High-temporal frequency microwave signature measurements are essential for this procedure, as are the highest possible number of channels. Currently this is more readily provided with ground-based systems than with airborne or spaceborne sensors.
Future work should extend the procedure by incorporating a time-dependent compound thermal diffusivity profile Dc(z, t) instead of the constant Deff.
