Abstract
Soil moisture is an important aspect of the Earths hydrological system. Accuratemonitoring of soil moisture is essential for advancing our knowledge of its influence
on Earth system processes. Cosmic-Ray Neutron Sensing (CRNS) offers an opportunity
to fill the knowledge gap between point-scale sensors and large-scale sensors (i.e.,
satellite remote sensing) by capturing field scale, root-zone soil moisture. However,
increasing global deployment of CRNS in regional networks has led to disparate
processing methods across the sensors, hindering the use of these sensors in broad
scale global studies.
This thesis explores the opportunities presented in utilising CRNS stations from across
the globe as a harmonized global network of sensors. Firstly, an open-source python
processing tool was developed to facilitate a much-needed harmonization of CRNS
data from 163 stations from across multiple networks. Using this tool, we demonstrate the problems that can come from a non-harmonized set of sensors in global studies. Utilizing this harmonized dataset, we conducted a comparative study against a satellite-derived soil moisture product (ESA-CCI) and a reanalysis product (ERA5-Land). Our analysis reveals residual biases between these products and CRNS values, which notably increases at the extremes of wet and dry conditions, whilst correlation differences increase under moderate conditions. Lastly, machine learning models were used to evaluate the role of soil moisture spatial representation in predictions of land surface fluxes of water (evapotranspiration) and photosynthesis (gross primary productivity). Our findings indicate that in-situ soil moisture data is particularly important for accurate predictions of evapotranspiration in water stressed regions when compared to indirect estimates from empirical models or satellite remote sensing. Unlike evapotranspiration, we observe that the contribution of deeper soil moisture, in the form of soil moisture memory, plays a more significant role in
predicting photosynthesis, pointing to the importance of identifying distinct mechanisms driving water and carbon fluxes at the land-atmosphere interface.
Overall, this thesis demonstrates the value of CRNS being treated as a harmonized and global network, reveals the importance of soil moisture spatial representation in modelling of land-atmosphere processes, and highlights where future soil moisture sensor deployment can be most beneficial.
| Date of Award | 23 Jan 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Rafael Rosolem (Supervisor) & Miguel A Rico-Ramirez (Supervisor) |