File(s) under permanent embargo
Reason: Restricted by author. A copy can be supplied under Section 51(2) of the Australian Copyright Act 1968 by submitting a document delivery request through your library or by emailing email@example.com
Joint active passive microwave soil moisture retrieval
thesisposted on 23.02.2017, 23:15 by Gao, Ying
Surface soil moisture is of great importance to the disciplines of agriculture, hydrology and meteorology. Over the past three decades, researchers have made significant advances in developing the algorithms and techniques for retrieving soil moisture by remote sensing, a technique which measures the emitted, reflected and/or scattered electromagnetic radiation from the land surfaces. A large number of remote sensing approaches have been developed and tested to measure soil moisture. Among them, passive microwave remote sensing (at L-band) has been demonstrated as the most promising tool for global soil moisture estimation. However, passive microwave soil moisture retrieval is highly dependent on the availability of ancillary surface parameters such as vegetation water content and surface roughness. It is difficult to characterise these information at the scale of L-band radiometer footprints (40 km) globally by ground measurement. Nevertheless, global information on vegetation water content can potentially be obtained from optical sensing technologies, while surface roughness can potentially be characterised by active microwave sensors, because of the high sensitivity to water absorption and surface roughness respectively. Up to now there has been research about retrieving soil moisture using passive or active microwave observations individually. However, no research has incorporated active-derived roughness into the passive retrieval model, in order to improve the passive soil moisture retrieval accuracy. Therefore, this research aimed to characterise surface roughness from active measurements, and then apply these information to passive soil moisture retrieval accuracy improvement. This research is mostly based on field data collected from the Soil Moisture Active Passive Experiments (SMAPEx) as part of this PhD. First, estimation of the vegetation water content needed in the passive microwave emission model for calculation of the vegetation optical depth was explored. This information was retrieved from MODIS (Moderate Resolution Imaging Spectroradiometer)-derived vegetation indices, using empirical formulations developed from historical field and satellite data sets. Subsequently, the Tau-Omega Model, which is the most frequently used passive emission model for vegetated surfaces, was evaluated using the SMAPEx airborne brightness temperature and ground soil moisture data sets, together with the vegetation water content developed from the previous step. This provided a baseline soil moisture map for the entire study area. Moreover, results showed that the default model parameters provided by SMAP ATBD (the Algorithm Theoretical Basis Documents of the Soil Moisture Active Passive mission) provided a soil moisture accuracy of 0.11 m³/m³ for cropland and 0.06 m³/m³ for grassland. After calibration with ground soil moisture data, the results could be improved to 0.06 m³/m³ for cropland and 0.05 m³/m³ for grassland. Last, the active microwave retrieval of surface roughness and its usefulness in passive microwave retrieval of soil moisture was explored. In order to improve the passive soil moisture retrieval accuracy through roughness, the relationship between the passive roughness parameter, HR, and the active roughness parameter standard deviation of surface height, SD, was clarified. Existing relationships have only focused on ground measured SD. However, no research has related HR to SD retrieved from active microwave measurements. Therefore, a new formulation was developed to relate SD to HR using remotely sensed and field data. An iterative algorithm combining an active microwave (Oh) model and a passive microwave (Tau-Omega) model has been developed to retrieve soil moisture and surface roughness simultaneously. The new roughness formulation developed in the previous step was then applied here to relate surface roughness in the active and passive models. Results showed that the iterative algorithm could achieve a soil moisture accuracy of 0.085 m³/m³ for cropland and 0.05 m³/m³ for grassland, without relying on any model calibration. This result outperformed the retrieval accuracy when using default HR from the SMAP ATBD by 0.02 m³/m³ for cropland and 0.01 m³/m³ for grassland, suggesting that use of active microwave data for surface roughness estimation can lead to more accurate near-surface soil moisture mapping globally.