Menu
Research Questions/Objectives:
The overarching aim of this project is to improve the understanding of rainfall runoff processes in an alluvial gully catchment (I.e., the Burdekin catchment in Northern Queensland). Specifically, the first objective is to set up a hydrological model, with the second being to calibrate and validate that model. Finally, the study will incorporate the assessment of scenarios by testing different rainfall events and evaluating changes in vegetation extent, rainfall volume and soil infiltration capacity. Questions revolving this include:
Brief Description of the Project: This study aims to utilise hydrological modelling to investigate rainfall-runoff relationships within an alluvial gully complex at Spyglass Cattle Station. The study will integrate the use of in situ data (e.g., rain gauges and water samplers) and computer generated modelling to understand the effect of rainfall under different environmental regimes. With few studies approaching this in detail, rainfall and runoff activity is a primary instigator and driver of the formation and progression of alluvial gullies. Not only do these erosional features influence the state of water quality both in situ and in downstream/base level habitats (i.e., the Great Barrier Reef), but they have deleterious effects on production within important rangelands as such.
Background and Significance of the Research Question to drought risk, vulnerability, preparedness, or resilience: Land degradation is a process in which the value of the biological and physical environment is affected by a combination of anthropogenically-induced and natural processes acting upon the land (Conacher et al., 1995). Following such changes, the land is perceived to exist in an undesirable state; one that both directly and indirectly affects the state of the environment in situ and in adjacent areas. For example, land degradation processes occurring in the headwaters of coastal catchments (e.g., erosion or land clearing) may have deleterious effects on downstream ecosystems and marine environments, particularly during stochastic events such as flooding or cyclonic storms (Shellberg, 2021). Moreover, land degradation directly impacts agricultural and horticultural production, particularly in Australia’s rangelands where soil is vulnerable to dispersion and subsequent erosion. Land degradation through processes such as soil erosion removes fertile soil that is imperative to agricultural processes. The loss of soil as such results in lower yields and higher production cost. Erosion often causes the formation of rills and gullies, of which leave the cultivation of paddocks almost impossible and unsafe (Dregne, 1982).
Gullies are a major land degradation issue on a global scale, confronting the capabilities of land and water resource management. Gullies occur where water is concentrated in narrow flows, incising, widening and elongating as the gully is further exposed to soil loss and moving water. Gullies eventually become so large that normal management techniques such as tillage cannot destroy them (Le Roux and Sumner, 2012). Gullies are of particular concern in Northern Queensland, where they provide a bountiful sediment supply to downstream catchment areas and are a source of sediment on the Great Barrier Reef (GBR). Likely the result of introduced grazing post-European settlement in the mid 1800’s, gully development has increased in erosion rates, with increasing evidence that soil derived from the rangelands is directly affecting nearshore marine environments such as seagrass beds and coral reefs (Koci et al., 2021). As of 2020, there were 87,000km of gullies documented in rangeland tributaries to the Great Barrer Reef World Heritage Area in Queensland (Koci, 2020). This is the primary source of sediment and nutrients such as nitrogen and phosphorus to the GBR. The vast and increasing area dominated by gully complexes in Northern Queensland has placed an emphasis on their remediation, as well as the need to expand on our understanding of the erosional processes that contribute to gully formation and evolution.
There are two main forms of gullies; hillslope and alluvial gullies. Whilst the former is more commonly described throughout the literature, alluvial gullies are complex in morphologies and are influenced by a suite of intricate processes. Despite their extremely challenging behaviour, alluvial gullies are seldom studied or represented in hydrology and water resource articles when compared to their hillslope counterparts. With their damaging and problematic nature, as well as a lack of focused research, alluvial gullies are the main focus of this project.
Connected to drainage systems such as the Burdekin River (one of Australia’s largest rivers and exporters of catchment derived sediment to the GBR), sediment has little opportunity for deposition before being transported to streams and events out of the catchment and into the GBR lagoon. In an area that experiences high annual rainfall and storm events, it is critical to understand how rainfall translates into runoff at particular sites, in order to recognise landscape condition and erosion susceptibility. Moreover, it is an apt proxy for measuring the response of alluvial gullies to remediation attempts. Very few studies investigate rainfall runoff relationships in alluvial gullies in detail, of which this study will address through hydrological modelling.
Academic and research experience relevant to the honours project: A Master of Science (Professional) student majoring in marine biology, I began my academic journey at Deakin University, Melbourne, with a Bachelor of Environmental Science majoring in conservation and wildlife biology. Since commencing at JCU in 2021, I have found my interests straying further inland than originally expected, and subsequently have found myself with a strong interest in alluvial systems. Specifically, I feel passionate about addressing the processes that have a powerful influence on the state of human livelihoods, as well as downstream ecosystems such as our beloved Great Barrier Reef.
After a successful semester studying under my lecturer Dr Jack Koci in the subject ‘Coastal and Catchment Geomorphology’, I have become intimately familiar and concerned with environmental processes. This is particularly relevant for ones that have broader impacts on both a temporal and spatial scale and that underpin the condition of entire ecosystems. I have found that this interest is strongest in coastal catchments. Concurrently, I have gained a strong understanding of the financial, emotional and social stress of decreased production in rangeland areas of environmental degradation. I believe that using my knowledge to address both the environmental and anthropogenic burdens of erosional processes is the best and most fulfilling way to implement my passions in a real world scenario.
I have already conducted fieldwork on two occasions at Spyglass Station, installing methods of and conducting data collection. This enabled me to gain a first-person perspective on the extent to which alluvial gully complexes, and erosional processes generally, have impacted important grazing lands in the area. To me, the attempt to address the issues associated demonstrate the resilience and strength of the Australian spirit, and it inspires me to understand these systems better. With an apt understanding in landscape processes and management, years of experience in environmental volunteer positions both on a domestic and international scale, and hardworking team-oriented attitude, I am excited to undertake my Master’s minor project in a field as such.
Principal Supervisor’s skills and experience in relation to this project topic: Jack Koci is an integral component of the alluvial gully research being developed and implemented at Spyglass Cattle Station. A such, he is directly affiliated with the Drought Hub through projects such as this and others including the bare ground erosion project based in McKinlay in Outback Queensland. His PhD largely focused on hillslope gully erosion in savanna rangelands tributary to the Great Barrier Reef, investigating of hydrogeomorphic processes, sediment and nutrient yields. Moreover, his research was centred locationally in an area adjacent to Spyglass Station, also within the Burdekin catchment.
An earth and environmental scientist with an interest in catchment geomorphology and water quality, Jack currently conducts research and teaching within the School of Earth and Environmental Science at James Cook University, Queensland. Jack’s current research primarily addresses a lack of understanding and treating of the causes, impacts and management of soil erosion in tropical landscapes. Jack also utilises novel spatial analysis tools such as drone-based remote sensing techniques in both the research and modelling of tropical landscapes.
Jack has direct experience in working in the agricultural sector and shows a passion toward the rehabilitation of degraded agricultural land both on a domestic and international level. As such, he has worked with Seqwater and the Australian Centre for International Agricultural Research (ACIAR). The latter involved working for Researchers in Agriculture for International Development (RAID); an Australian-based network of researchers involved in international agricultural research for development. RAID does the following:
Prior to his position at James Cook University, Jack worked as a Research Fellow at the University of the Sunshine Coast.
References:
CONACHER, A., CONACHER, J., DRAGOVICH, D. & MAUDE, A. 1995. Rural land degradation in Australia, Oxford University Press.
DREGNE, H. E. 1982. Impact of land degradation on future world food production, US Department of Agriculture, Economic Research Service.
KOCI, J. 2020. Hillslope gully erosion in savanna rangelands tributary to the Great Barrier Reef: Investigation of hydrogeomorphic processes, sediment and nutrient yields. University of the Sunshine Coast.
KOCI, J., WILKINSON, S. N., HAWDON, A. A., KINSEY‐HENDERSON, A. E., BARTLEY, R. & GOODWIN, N. R. 2021. Rehabilitation effects on gully sediment yields and vegetation in a savanna rangeland. Earth Surface Processes and Landforms, 46, 1007-1025.
LE ROUX, J. J. & SUMNER, P. 2012. Factors controlling gully development: comparing continuous and discontinuous gullies. Land Degradation & Development, 23, 440-449.
SHELLBERG, J. G. 2021. Agricultural development risks increasing gully erosion and cumulative sediment yields from headwater streams in Great Barrier Reef catchments. Land Degradation & Development, 32, 1555-1569.
Background
Agribusinesses in Tropical North Queensland are familiar with the adversities brought by drought. With effects extending beyond farmers and producers, the wider community and local economy also experience the impacts of drought conditions. Efforts to promote drought and ecosystem resilience rely on an in depth understanding of core issues and hydrological processes, which can be used to equip agricultural businesses with the knowledge necessary to implement potential solutions. This is particularly important in catchments such as the Upper Burdekin, where extensive gullying is present and high amounts of terrestrial runoff from intermittent rainfall is being channelled into major systems such as the Burdekin River. Not only is excessive runoff contributing to a reduction in downstream ecosystem health, but it is also a cause of active erosion on productive soils, reducing landscape productivity. Capturing current runoff rates via hydrological modelling has shown promise for simulating and quantifying rainfall runoff processes, however, is seldom seen at the small scale of individual gullies. With little known about the complex nature of the relationship between intermittent yet high intensity rainfall and the physical environment (i.e., vegetation, soil and complex morphology), remediation efforts may be limited by a lack of baseline studies and information on how the system is behaving.
Aim
This study provides vital information on prospects and challenges of using hydrological modelling to understand rainfall runoff processes at the scale of individual gullies for potential use in the future. Specifically, the overarching aim of this study was to set-up, calibrate and validate a hydrological model for predicting runoff from rainfall in a small-scale savanna rangeland catchment dissected by an alluvial gully.
Method
The study was conducted within a singular gully unit at Spyglass Research Station in the Upper Burdekin catchment, over one season spanning between January and April of 2023. The open source hydrological modelling system HEC HMS was chosen as the most suitable modelling tool for calculating rainfall runoff in the gully, named JJ2. Following catchment delineation of JJ2 in ArcGIS, the model required several inputs. Vegetation and soil parameters were included as indirect measurements of soil infiltration capacity and antecedent moisture content. Rainfall data was collected using a tipping bucket rain gauge programmed to log rainfall amount every minute during an event. Terrain data was also collected using a real-time kinematic drone (DJI Phantom) with a standard red, green, blue camera, which was flown at 50m with 80% forward and side overlap between images. This data was geographically corrected using 9 ground control points which were surveyed using a real time kinematic Navigation Satellite System (RTK GNSS). The images were processes in Agisoft Metashape 2.0 to extract true ground points. Lastly, observed discharge during a rainfall event was calculated manually, using depth data (SBLT depth gauge), channel width data, surveyed using an RTK GNSS and an assumed velocity rate. Water infiltration and rainfall-runoff lag estimates were adjusted to model the total runoff volume as closely to the observed runoff volume as possible. The study used both a visual and statistical (percentage bias) assessment of the goodness-of-fit between the modelled and observed discharge.
Results
Key result 1: Model performance
Our study has shown the model can reliably estimate runoff volume for 50% of rainfall events, with a total volume difference of 1.5mm from observed data across these events. The model was most reliable when simulating ‘Event 11’ in mid-April, of which had a 1.19mm difference between the observed and modelled runoff volume. According to a previously outlined performance criteria, these events were deemed as ‘satisfactory’ to ‘very good’. The model had difficulty in simulating the other 50% of events, in which the percentage difference between modelled and simulated runoff volume varied between 35.3% and 48.4%. Error margins as such are likely explained by scale, where in small catchments slight differences exacerbate error margins as they account for a high percentage of total runoff volume.
Key result 2: Rainfall runoff patterns and process understanding
The model recognises the main patterns of runoff throughout an event. Illustrated on a hydrograph, discharge at JJ2 tends to increase quickly at the start of an event followed by a gentler decrease toward the end. There is also a short lag time between the time of rainfall and peak discharge. Both patterns suggest that discharge increases rapidly over a short period of time and responds very quickly to rainfall, characterising JJ2 as a ‘flashy’ hydrological system. The average rainfall runoff ratio was 42.55% across modelled events, indicating that almost half of the input rainfall data is being lost as discharge. Physical factors such as rainfall location in relation to the rain gauge, gully morphology, soil infiltration capacity and vegetation presence are likely contributing to these results.
Results
Key result 1: Model performance
Our study has shown the model can reliably estimate runoff volume for 50% of rainfall events, with a total volume difference of 1.5mm from observed data across these events. The model was most reliable when simulating ‘Event 11’ in mid-April, of which had a 1.19mm difference between the observed and modelled runoff volume. According to a previously outlined performance criteria, these events were deemed as ‘satisfactory’ to ‘very good’. The model had difficulty in simulating the other 50% of events, in which the percentage difference between modelled and simulated runoff volume varied between 35.3% and 48.4%. Error margins as such are likely explained by scale, where in small catchments slight differences exacerbate error margins as they account for a high percentage of total runoff volume.
Key result 2: Rainfall runoff patterns and process understanding
The model recognises the main patterns of runoff throughout an event. Illustrated on a hydrograph, discharge at JJ2 tends to increase quickly at the start of an event followed by a gentler decrease toward the end. There is also a short lag time between the time of rainfall and peak discharge. Both patterns suggest that discharge increases rapidly over a short period of time and responds very quickly to rainfall, characterising JJ2 as a ‘flashy’ hydrological system. The average rainfall runoff ratio was 42.55% across modelled events, indicating that almost half of the input rainfall data is being lost as discharge. Physical factors such as rainfall location in relation to the rain gauge, gully morphology, soil infiltration capacity and vegetation presence are likely contributing to these results.
Key result 3: Challenges in fine-scale hydrological modelling
Although no model is explicitly correct, addressing challenges may aid in the applicability of a model to describe a system.
Reliable terrain data
DTMs are often considered the most crucial component of a model, showing the characteristics of the earth’s surface as opposed to all objects on the ground (e.g., vegetation). An inaccurate DTM can impact the reliability of catchment boundaries and stream networks, as well as flow patterns and terrain slope and roughness analysis. This, ultimately influencing how much runoff a model may calculate to be reaching a gauging point.
Rainfall spatial variability
Other challenges surround the high spatial variability of rainfall at JJ2, of which is difficult to capture with a single tipping bucket rain gage with a small coverage area (200mm catch dimension). This was evident in our study when six rainfall events were excluded from reaching the modelling stage in HEC HMS following highly improbable rainfall runoff ratios (over 100%). With several runoff events not associated with any preceding rainfall data, gullies were in flow, yet no rainfall data was captured. Spatially variable rainfall likely explains missing data, where rainfall is occurring in a part of the catchment that does not reach the rain gauge yet is still contributing to runoff. If data cannot be collected to accurately represent the amount of rainfall contributing to recorded runoff, then the event data is unreliable.
Reliable velocity measurements
An elementary issue in the calibration of the model presented here is the estimation of average velocity, as a state variable that expresses a system dynamic. To test this, one of the worst performing events, was re-modelled using an average velocity estimate of 0.2 m/s, replacing the previous estimation of 0.3 m/s. Preliminary results show that changing the velocity accounted for a significant improvement in the PBIAS value of over 50% (-47% to -20.49%).
In summary, this model shows the great promise that hydrological modelling has for describing the behaviour of hydrology in a system. Whilst further improvements can be made to strengthen the applicability of the model, even a preliminary version suggests that hydrological modelling could be used to evaluate remediation techniques, both prospectively and retrospectively.
Background and project overview
In the savanna rangelands of north-eastern Queensland extensive gullying is present. Excess runoff following rainfall events diminishes the landscapes’ ability to capture and store water. Decreasing the amount of water available for plant uptake, excess runoff contributes to the active erosion of productive soils, reducing landscape productivity and negatively impacting downstream ecosystems. Management strategies looking to promote ecosystem health and drought resilience in water limited systems used for grazing rely on available information on hydrological processes. Current information on the challenges of quantifying hydrological processes at the gully scale is severely lacking, and little is known about the complex relationships between rainfall and features such as vegetation, soil and complex topography at this scale. Hence, our study aims to provide important information on the prospects and challenges of implementing hydrological modelling at small scales to understand rainfall runoff processes for potential use in the future.
To achieve this, a singular gully complex located in the Upper Burdekin Catchment at Spyglass Research Station (named ‘JJ2’) was equipped with instruments to measure rainfall and runoff over one wet season spanning between January and April of 2023. These included a tipping bucket rain gauge and a depth monitor. Terrain data was collected using a real time kinematic (RTK) drone, for the extraction of the overall catchment size and morphology, and an RTK global positioning system, for gully channel morphology. The latter, in combination with an assumed velocity speed, was used to manually calculate observed discharge during a rainfall event. All data was used to calibrate and validate a hydrological model within the open source modelling system HEC HMS to test the ability of the model to reliably predict runoff for a given rainfall event, in which there were six. Water infiltration and rainfall-runoff lag estimates were adjusted to model the total runoff volume as closely to the observed runoff as possible. The study used both a visual and statistical assessment of the goodness-of-fit between the modelled and observed discharge.
Overall, the model showed to reliably model 50% of rainfall events, with a total volume difference of 1.5mm between observed and simulated runoff. The other 50% of events were less accurately modelled than the others. The scale of runoff is a likely explanation for these events, in which smaller catchments tend to accentuate error margins as they account for a higher ratio of total runoff volume than larger catchments. However, the model recognises main patterns in runoff from a rainfall event, in which the model shows the ‘flashiness’ of JJ2. This meaning discharge increases quickly at the start of an event and tails off slower toward the end. The short lag time between rainfall and runoff shows discharge responds very quickly to the introduction of rainfall in the system with almost half of the rainfall input being lost as discharge. Physical factors such as rainfall location, gully shape and terrain, soil infiltration capacity and the percentage of vegetation accounts for this. The study also highlighted several areas acting as challenges and areas to improve upon for future modelling attempts. These include obtaining reliable terrain data, the impact of spatially variable rainfall in the region, and the importance of reliable velocity measurements.
Practical applications
Practical application 1: Remediation planning and evaluation
Hydrological modelling is an essential tool for planning and undertaking improvements to landscape condition and increasing landscape and community resilience to drought. A well calibrated hydrological model can demonstrate the effects of remediation works both prospectively and retrospectively. By changing the input values of a model to represent changes in the environment (e.g., soil infiltration rates), land managers may be able to forecast the effects of proposed solutions, such as introducing gypsum into the soil with the goal to improve soil structure, for example. Concurrently, a hydrological model may lay bare the deficiencies and areas for improvement when evaluating an already implemented remediation strategy. Proxies of determining whether landscape condition has improved from remediation action includes lower total runoff volume and peak runoff, lower ratios of rainfall to runoff, as well as a longer lag time between rainfall and the creation of discharge. Both, resulting in a less ‘flashy’ system.
With a significant amount of funding being invested into understanding and managing alluvial gully erosion, it is important that these investments are underpinned by a good understanding of local drivers of gully formation, as well as the spatial variability in hydrological processes. These, pertaining to where and why runoff is occurring in different parts of a catchment so that management bodies can prioritise area during remediation. By expanding the number of gullies being modelled in an area, they can be prioritised based on severity of landscape degradation or a site’s likelihood of effective remediation.
Practical application 2: Sediment yield analysis and erosion monitoring
Sediment yield analysis combined with the identification of local mechanisms of soil erosion is essential for developing effective management approaches and enhancing water conservation methods. Small scale hydrological modelling can assist in sediment yield estimations and the identification of sediment-producing hotspot areas. There are already rising stage samplers and an automatic pump sampler implemented at JJ2, commonly used to collect water samples from flashy, intermittent streams. These samples can be used to calculate how much sediment is being suspended during a rainfall/flow event of a particular magnitude. As landscape condition improves and repeat surveys are conducted over time, it can be expected that less sediment will be entrained during a rainfall or flow event of the same magnitude. By implementing these sampling methods at other sites, we can begin to understand how erosion is varying across the landscape, prioritising areas for remediation efforts.
Practical application 3: Addressing challenges and improving workflow
One of the main objectives of this study was to highlight the challenges and limitations in modelling hydrology at the small scale of individual gullies. Firstly, accessibility is one of the primary restraints in obtaining enough accurate field data for modelling hydrology. One of the main impacts of inaccessible areas is the inability to survey true ground points for terrain data. This is especially important in extracting flow pathways and networks, which in turn, determines how much water is reaching an outflow point, or a gauging station, in a hydrological model. It is also important for obtaining characteristics of gully morphology for use in discharge calculations and catchment extraction. Although topographic data can be derived from Ligh Detection and Ranging (LiDAR) surveys, they are expensive and require expertise to function. One solution being investigated is the use of low-cost drone flights to generate digital terrain models, using different camera angles to improve accuracy. An apt next step is to test this data in a hydrological model, to see if the restraint of accessibility can be alleviated using low-cost drones for terrain data.
Moreover, the phenomenon of equifinality in systems theory suggests that any value can be assigned to a model parameter (i.e., soil infiltration or lag time) to eventually reach a calibrated state. Especially with the use of indirect measurements of landscape condition (i.e., infiltration rather than actual soil and vegetation characteristics), a set of model parameters that have no physical relevance to a system can be used to calibrate a model. Hence, obtaining more field data (i.e., vegetation and soil characteristics) for use in measured parameters may address this problem. Also related to the accuracy of field data, it was seen here that velocity measurements have a large impact on the calibration of a model. After re-running a poorly performing model after adjusting the assumed velocity, the statistical value for model performance improved by over 50%. Future studies should prioritise equipment related to calculating observed discharge so as to avoid this issue (i.e., flow monitors). Whilst this list of challenges is not finite, they are appropriate places to start with improving the modelling of hydrology at small scales.
