Abstract
The Upper East Region (UER) of Ghana is characterised by unreliable rainfall, rising temperatures, and a dry sub-humid climate. It is also prone to land degradation (e.g., erosion). Despite its low precipitation, the region floods annually due to rainfall and the opening of the Bagre Dam in Burkina Faso. A large proportion of its workforce is engaged in agriculture, and the region contributes a substantial portion to Ghana’s annual crop production. Notwithstanding, how the region’s climate and annual floods influence land deformation has not yet been studied. Land deformation is a component of land degradation, thus, understanding the spatial and temporal variability of ground deformation and its causes can help mitigate land degradation in this region. However, in situ data to study ground deformation, water resources, and climate change and variability are not readily available in the UER. Thus, in this research, the relationship between land surface deformation and groundwater storage variations was analysed by combining freely available global data and advanced computational techniques. The three research objectives were thoroughly addressed in three analysis chapters. The key findings of the analysis chapters are summarised as follows:1. The terrestrial water budget (TWB) was leveraged to evaluate the reliability of data from the Gravity Recovery and Climate Experiment (GRACE) mission (from April 2002 to June 2017) for water resources monitoring in the UER. TWB describes water availability as the difference between water influx (precipitation) and outflux (evapotranspiration and runoff). To improve the reliability and accuracy of
the TWB analysis, multiple solutions of precipitation, evapotranspiration, and total water storage anomalies (TWSA) were evaluated using the three-cornered hat method to select the most accurate data sets for the UER. The selected data sets were then used in the TWB analysis to determine the reliability of GRACE data, and to also analyse, in conjunction with additional meteorological variables such as temperature, humidity and wind speed, the relationship between meteorological conditions and water resources in the region. The water balance analysis indicated that GRACE data has a high signal-to-noise ratio (SNR) in the small, data-poor UER, while water storage from the Global Land Data Assimilation System (GLDAS)-Noah cannot be used for total water storage analysis in the region because GLDAS-Noah does not simulate subsurface water storage. The results also revealed a lagged relationship between rainfall and TWSA, with the former leading the latter. Overall, meteorological conditions appeared to support a dry climate, but there was also a moderate to high relationship between evapotranspiration and rainfall, suggesting that high levels of evapotranspiration corresponded to high levels of rainfall. From 2002 to 2017, water storage in the region was characterised by a significant increasing linear trend (9.9 ± 1.8 mm year–1), with the region generally having gained more water than it lost.
2. Although GRACE has a high SNR in the UER, its spatial resolution limits a detailed analysis of the spatial variability of the relationship between land surface deformation and groundwater changes. To improve the resolution of GRACE, this research compared multiple linear regression (MLR), Random Forest (RF), and geographically weighted regression (GWR) for large scale downscaling of GRACE data. For improved downscaling results, it is critical to account for spatial dependence. Accounting for spatial structure in the measurements exploits the property that near measurements are more similar than those farther apart, which can improve model accuracy and preserve spatial patterns in the data. For this reason, I used the variogram to analyse the residuals of the three regression models. However, on the sphere, the Euclidean distance does not correctly characterise the horizontal separation between two points. This poses a challenge for variography as inaccurate distances may improperly characterise the spatial relationship between data points. Therefore, a key contribution of this research is testing the use of the chordal length as the distance metric in variography to provide weights for ordinary kriging of the regression residuals. The results showed GWR to be the most accurate regression method for downscaling GRACE data over the study extent as it accounts for spatial structure in the data. Combining the regression models with kriging of their residuals substantially improved the prediction accuracies of MLR and RF. The prediction accuracy of GWR increased by more than 30% in terms of mean absolute error (MAE) and root mean squared error (RMSE) when it was combined with kriging of its residuals. These results demonstrate that large scale regression-kriging of GRACE data using the chordal distance as the distance metric in variography improves modelling of GRACE data and prediction accuracy.
3. Techniques for a detailed spatial analysis of the relationship between groundwater storage anomalies (GWSA) and land surface deformation in the UER were introduced. The primary goal was to evaluate the use of satellite-only and reanalysis data to study the impact of groundwater loading and unloading on surface deformation. Since in situ groundwater measurements were unavailable at the time of conducting the experiments, GWSA was obtained by estimating the difference between GRACE TWSA and GLDAS-Noah total water content anomaly (TWCA); and downscaled by combining GWR and ordinary kriging of the residuals. The resulting fine-resolution GWSA data were then analysed in conjunction with ground displacement data that were obtained from a multi-temporal Interferometric Synthetic Aperture Radar (InSAR) analysis to determine their relationship using methods such as regression and correlation analyses and the Granger causality test. The results of the correlation analysis revealed important information governing the relationship between displacement and groundwater, including: (1) the relationship
between the two may be obscured by short-term variations in both time series and (2) the noise in the time series may be strongly contributing to the weak relationships between the two time series. The analyses revealed a strong long-term
positive linear relationship between GWSA and displacement, indicating that the two variables change in the same direction. The Granger causality analysis indicated a bi-directional relationship between the two, highlighting the influence
of the complex mechanisms of groundwater recharge on land surface uplift and vice versa in the dry sub-humid UER. Additionally, forward modelling GWSA and displacement revealed that the region is experiencing both elastic and poroelastic deformation; with poroelastic deformation responsible for the observed uplift at five selected locations. These results demonstrate that satellite-only products can provide a high-level to detailed insights into the relationship between ground displacement and groundwater changes in the UER. The results also highlight the potential to assimilate InSAR observations into GWSA to improve groundwater estimates. However, augmenting space-borne measurements with in situ data can substantially improve our understanding of the impact of changing hydrology on ground deformation in the UER.
The research significantly advances our understanding of the relationship between changing hydrology and ground deformation in the UER. It makes substantial contributions to the field of hydrogeodesy by developing techniques for improved integration of GRACE and InSAR data for hydrological analysis in data-poor regions. The three key contributions of this research are: (1) evaluating the reliability of GRACE data, (2) enhancing the resolution of GRACE data for detailed spatial analysis alongside InSAR, and (3) providing a comprehensive approach to analysing groundwater and land deformation dynamics. These advancements offer new tools and methodologies for the scientific community, providing valuable insights into the complex interactions between hydrological changes and ground deformation. By enabling more detailed and accurate analysis of groundwater storage anomalies and land deformation, this research has profound implications for mitigating land degradation, supporting sustainable agricultural practices, and improving water resource management in the UER and similar regions. Overall, this research represents a significant step forward in monitoring and understanding the impacts of changing hydrology on ground deformation, particularly
in data-poor regions, holding promise for future efforts to preserve vital water resources and maintain land integrity.
| Date of Award | Oct 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Nicholas Hamm (Supervisor), Stephen Grebby (Supervisor), Stuart Marsh (Supervisor) & Craig M. Hancock (Supervisor) |