L-Band Multi-Polarization Radar Scatterometry over Global Forests: Modelling, Analysis, and Applications

Citation

Lim, Yu Xian (2020) L-Band Multi-Polarization Radar Scatterometry over Global Forests: Modelling, Analysis, and Applications. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/7Y4D-JD17. https://resolver.caltech.edu/CaltechTHESIS:07252019-221240608

Abstract

Spaceborne L-band radars have the ability to penetrate vegetation canopies over forested areas, suggesting a potential for regular and frequent global monitoring of both the vegetation state and the subcanopy soil moisture. However, L-band radar’s sensitivity to both vegetation and ground also complicates the relationship between the radar observations and the ecological and geophysical parameters. Accurate yet parsimonious forward models of the radar backscatter are valuable to building an understanding of these relationships. In the first part of this thesis, a model of L-band multi-polarization radar backscatter from forests, intended for use at regional to global spatial scales, is presented. Novel developments in the model include the consideration of multiple scattering within the dense vegetation canopy, and the application of a general model of plant allometry to mitigate the need for much intensive field data for training or over-tuning towards specific sites and tree species.

Aided by our model, in the remainder and majority of the thesis, a detailed analysis and interpretation of L-band backscatter over global forests is performed, using data from the Aquarius and SMAP missions. Quantitative differences in backscatter predicted by our model due to freeze/thaw states, branch orientation, and flooding are partially verified against the data, and fitted values of aboveground-biomass and microwave vegetation optical depths are comparable to independent estimates in the literature. Polarization information is used to help distinguish vegetation and ground effects on spatial and temporal variations. We show that neither vegetation nor ground effects alone can explain spatial variations within the same land cover class. For temporal variations during unfrozen periods, soil moisture is found to often be an important factor at timescales of a week to several months, although vegetation changes remain a non-negligible factor. We report the observation of significant differences in backscatter depending on beam azimuthal angle, possibly due to plant phototropism.

We also investigated diurnal variations, which have the potential to reveal signals related to plant transpiration. SMAP data from May-July 2015 showed that globally, co-polarized backscatter was generally higher at 6PM compared to 6AM over boreal forests, which is not what one might expect based on previous studies. Based on our modelling, increased canopy extinction at 6AM is a possible cause, but this is unproven and its true underlying physical cause undetermined.

Finally, by making simplifying approximations on our forward model, we propose and explore algorithms for soil moisture retrieval under forest canopies using L-band scatterometry, with preliminary evaluations suggesting improved performance over existing algorithms.

Thesis Files