- Title
- Estimating soil moisture at a high spatial resolution with remote sensing
- Creator
- Senanayake, Indishe Prabath
- Relation
- University of Newcastle Research Higher Degree Thesis
- Resource Type
- thesis
- Date
- 2020
- Description
- Research Doctorate - Doctor of Philosophy (PhD)
- Description
- High spatial resolution soil moisture information is important for hydrologic, climatic and agricultural applications. The point-scale in-situ observations and coarse spatial resolution (10s km) satellite retrievals are unable to satisfy the resolution requirement of many applications. This makes hydrologists and climatologists to call high spatial resolution soil moisture as the ‘missing piece of the puzzle’ in their modelling approaches. With the advancement of the remote sensing technologies and satellite missions dedicated to measure soil moisture, researchers realized that downscaling L-band passive satellite soil moisture retrievals as a viable method to achieve near-surface (~top 5 cm) soil moisture information at a high spatial resolution. The available downscaling models have shown varying performance over regions with heterogeneous land surface properties and climatic conditions. The thermal data based downscaling techniques appear to be robust methods over arid and semi-arid regions with a high atmospheric evaporative demand such as most of Australia. Among them, soil thermal inertia based downscaling methods have shown encouraging results over semi-arid catchments in the United States, therefore has a good potential over Australian agricultural landscapes. These methods have not yet been tested in Australia, by taking the land surface properties the agricultural catchments of the country. In this work, downscaling algorithms were developed based on the soil thermal inertia relationship between diurnal soil temperature difference (ΔT) and the daily mean soil moisture (μSM) over two Australian agricultural catchments to estimate soil moisture at a 1 km spatial resolution. Factors affecting ΔT- μSM relationship such as vegetation and soil texture were included in these algorithms to improve the models. Here, regression trees and machine learning models (Gaussian process regression, Levenberg-Marquardt and Bayesian regularization backpropagation algorithms) were developed with the near-surface soil moisture and soil temperature information obtained from (i) long-term in-situ observations and (ii) 25 km gridded Global Land Data Assimilation System (GLDAS) outputs. The machine learning models were tested by considering their potential to capture the complex, non-linear relationships between μSM, ΔT and factors such as normalized difference vegetation index (NDVI) and soil texture. The 1 km spatial resolution ΔT, and NDVI values extracted from the Moderate Resolution Imaging Spectroradiometer (MODIS) products and other modulating factors at 1 km spatial resolution were entered to the models to estimate soil moisture at 1 km spatial resolution. SMAP 36 km, SMAP-E 9 km and SMOS 25 km gridded soil moisture products along with airborne soil moisture products aggregated over coarse resolution pixels were downscaled. The downscaled products were validated against in-situ soil moisture observations and high spatial resolution airborne soil moisture retrievals to evaluate the model performance. The soil thermal inertia based models tested here showed promising results over the study areas, with strengths and weaknesses associated with each model. Both long-term in-situ databased and GLDAS datasets were able to capture the catchment scale spatial variability of soil moisture, given that the datasets are representative of the land surface characteristics of the area. The results provide insights for developing a long–term time series of soil moisture over agricultural landscapes in Australia and other semi-arid regions. Combining these models and including factors such as topography and surface albedo can provide a robust approach in developing such a time-record of soil moisture over larger extents for hydrologic, climatic and agricultural applications.
- Subject
- soil moisture; remote sensing; downscaling; SMAP; SMOS; MODIS; thesis by publication
- Identifier
- http://hdl.handle.net/1959.13/1433479
- Identifier
- uon:39261
- Rights
- Copyright 2020 Indishe Prabath Senanayake
- Language
- eng
- Hits: 703
- Visitors: 686
- Downloads: 0