Integrated hydrological modelling of streamflow and groundwater level through time series modelling
Loading...
Citations
Altmetric:
Altmetric::
Download
Associated Records
Authors
File Licence
Editors
Date
2023-03
Source Title
Journal Title
Journal ISSN
Volume Title
Publisher
Melbourne Authors
Affiliation
Infrastructure Engineering
Document Type
PhD thesis
Subject Code
Citation
Access Status
Open Access
DOI
Access URL
Open Access URL
Open Access at PMC
Linked Resource URL
NHMRC Grant Code
ARC Grant Code
Description
© 2023 Giancarlo Bonotto
Abstract
Agriculture is the largest consumer of water, and it depends on the availability of cheap and reliable sources of surface and groundwater (e.g., rivers and aquifers). Although the integrated management of aquifers and rivers is of paramount importance for future food, water, and environmental security, the quantification of the interactions between groundwater (GW) and streamflow (SF) is complex and remains challenging, particularly in the low-income world where hydrological monitoring is limited, both spatially and temporally. In order to contribute to this challenge, this research investigates and develops methods that aim to use commonly available hydrological data to generate better insights into the interactions between aquifers and rivers at the catchment scale. After an introduction and systematic literature review, the results of this thesis are presented in the format of three research papers: (i) application and investigation of a causal inference method to identify the strength and direction of interactions between groundwater and streamflow, spatially and temporally; (ii) development and application of a new integrated groundwater-streamflow time series model; (iii) sensitivity of the modelling performance to the selection of representative bore and model structure.
The first results chapter (Chapter 3) demonstrated the potential of using convergent cross-mapping (CCM) to identify causal interactions between SF and GW head. Moreover, this chapter shows that CCM can provide valuable causal information from common hydrological time series, which is relevant to a wide range of applications in hydrology. Applied to three upland catchments across different hydrogeologic settings and climatic conditions in Victoria, Australia, the CCM results indicated causal links that largely agreed with an a priori perceptual understanding of the study areas and offered additional insights about hydrological processes. For instance, it suggested weaker SF-GW interactions during and after the Millennium Drought than in the previous wet periods. At the same time, this paper demonstrated that limitations of the method around seasonality, data sampling frequency, and long-term trends impact the physical interpretability of the causal links and highlight the need for future research.
Having identified different levels of SF-GW interactions at the study areas, the second results chapter (Chapter 4) investigated the potential of expanding the HydroSight GW time series model to better represent both SF and GW responses by trialling different formulations for the unsaturated zone and baseflow conceptualisation in conjunction with the existing saturated zone formulation. Applied to two of the study catchments using a joint calibration to SF and GW head, the model captured both SF and GW dynamics (including multiyear trends) with little calibration trade-off. The results also offered insights into the dominating hydrological processes, suggesting that baseflow is important in Brucknell Creek but negligible in Sunday Creek, where flow ceased during the Millennium Drought. Additionally, the joint SF-GW calibration improved the realism of the simulations by avoiding unrealistic fluxes in the unsaturated and saturated zone, which is important for the understanding of catchment processes and the management of water resources. This is an important improvement in the joint time series simulation of groundwater and streamflow.
Having developed and applied the new GW-SF time series model, the third results chapter (Chapter 5) investigated the sensitivity of modelling performance to the selection of representative bore. To investigate this, different structural configurations of the model were jointly calibrated to the SF and GW head from multiple bores in Brucknell Creek and Sunday Creek. The results suggested that the bore selection affects simulation performance more than model structure, particularly in terms of GW simulation but also to a lesser degree regarding SF simulation. Moreover, the results demonstrated that rainfall partition is affected by both model structure and bore selection, which further highlights the need to use multicriteria calibration and carefully select representative bores and model structures. At the same time, it was not possible to achieve accurate simulations of the intermittent streamflow of Sunday Creek, possibly due to the monotonic gaining river formulation of the model and the need for improved representation of feedback between saturated and unsaturated zones.
Overall, the results of Chapter 3 and Chapter 5 indicate CCM causal analysis as a potential means of bore selection for modelling studies that use conceptual models and representative bores to investigate SF-GW interactions. The new GW-SF time series model developed in Chapter 4 contributes to more realistic simulations of both groundwater and streamflow components (i.e., GW recharge, baseflow, interflow & runoff) using the limited monitoring data that is available in the developing world. Therefore, the results from this thesis give crucial contributions to the sustainable integrated management of aquifers and rivers when and where it will be most needed, i.e., during the next two decades in the developing world.
In particular, this thesis demonstrated, for the first time with real streamflow and groundwater monitoring data, that convergent cross-mapping (CCM) seems widely applicable to hydrology but also that the method has limitations and thus should be used and interpreted with care. This thesis developed a new integrated GW-SF model to better represent both streamflow and groundwater responses. The application of this new model demonstrated its ability to capture both streamflow and groundwater dynamics but also demonstrated the importance of joint SF-GW calibration to improve the realism of hydrological simulations. This thesis contributed to the very limited literature on the sensitivity of model performance to the selection of representative bore by demonstrating that integrated GW-SF modelling can be more sensitive to bore selection than model structure. Finally, it presented a new methodology to choose the best representative bore for a particular catchment based on the best joint SF-GW modelling performance and presented indications that convergent cross-mapping can be used as a new method to select representative bores that produce the best joint SF-GW simulation.
Notes
Terms of Use
Keywords
hydrology, catchment, surface-groundwater interactions, nonlinear causal inference, causality analysis, Convergent Cross-Mapping, integrated hydrological model, multi-objective calibration, model evaluation, sensitivity test, multiple working hypotheses, HydroSight