Predicting Suspended Sediment Transport in Urbanised Streams: A Case Study of Dry Creek, South Australia

Abstract

Sediment transport in urban streams is a critical environmental issue, with significant implications for water quality, ecosystem health, and infrastructure management. Accurately estimating suspended sediment concentration (SSC) is essential for effective long-term environmental management. This study investigates the relationships between streamflow, turbidity, and SSC in Dry Creek, South Australia, to understand sediment transport dynamics in urbanised catchments. We collected grab samples from the field and analysed them in the laboratory. We employed statistical modelling to develop a sediment rating curve (SRC) that provides insights into the sediment transport dynamics in the urban stream. The grab sample measurements showed variations in SSC between 3.2 and 431.8 mg/L, with a median value of 77.3 mg/L. The analysis revealed a strong linear relationship between streamflow and SSC, while turbidity exhibited a two-regime linear relationship, in which the low-turbidity regime demonstrated a stronger linear relationship compared to the high-turbidity regime. This is attributed to the urbanised nature of the catchment, which contributes to a first-flush effect in turbidity. This contributes to sediment hysteresis, resulting in non-proportional turbidity and SSC responses to streamflow changes. The findings demonstrate the effectiveness of a streamflow-based SRC for accurately predicting sediment discharge, explaining 97% of the variability in sediment discharge. The sediment discharge predicted using the SRC indicated a sediment load of 341.8 tonnes per year along the creek. The developed sediment rating curve provides a valuable tool for long-term sediment management in Dry Creek, enabling the assessment of downstream environmental risks. By addressing data limitations, this study contributes to a deeper understanding of sediment transport dynamics in urbanized environments, offering insights for informed decision-making and effective sediment management strategies.

1. Introduction

Erosion and sedimentation pose significant environmental challenges worldwide [1]. These processes can lead to a cascade of negative consequences, including reduced agricultural productivity [2], diminished reservoir capacity [3], degraded water quality [4], and impaired channel function [5,6]. Sediment originates from various sources, including soil erosion from land-use changes, deforestation, construction activities, and channel scouring [7]. Increased sediment loads from urban areas significantly impact coastal ecosystems and water quality [8,9]. Understanding the transport of suspended sediment in streams is crucial for assessing its implications on ecological and environmental factors [10]. Studies like the Adelaide Coastal Waters Study [11] highlight the significant impact of sediment loads on coastal ecosystems, with approximately 67% of discharged material affecting coastal aquatic life. Turbidity, a water quality parameter influenced by suspended sediment concentration (SSC), is often elevated in urban streams due to factors like increased stormwater runoff and resuspension of pollutants [12,13,14]. Urbanisation also alters flow patterns and sediment transport, particularly during construction phases [15]. Given the significant environmental consequences, accurate SSC estimation is essential for effective water and catchment management [16].

Modelling sediment transport in rivers relies heavily on observed data for calibration and validation [17]. However, many catchments lack adequate hydrological and sediment data due to cost and complexity [12,18]. Direct measurement of SSC, while highly accurate, is time-consuming and expensive [19,20]. In contrast, indirect methods like turbidity measurement, Acoustic Doppler Current Profilers (ADCPs), remote sensing, and sediment rating curves (SRCs) offer cost-effective alternatives, despite potential limitations in precision [21,22,23,24]. The choice of method depends on accuracy requirements, data availability, resources, and river characteristics. Estimating suspended sediment through indirect methods is vital for water resource management, especially with access to long-term streamflow data, turbidity measurements, and/or operational sensors [25,26]. This scenario allows for the development of SRCs, which can be used in conjunction with advanced sediment transport models to achieve accurate predictions of river sediment loads [27,28]. SRC is a widely used method that requires site-specific calibration due to the dynamic nature of soil erosion and transport. Several factors, including rainfall, discharge, catchment area, and sampling frequency, influence their accuracy. Despite these limitations, SRCs remain a valuable tool for estimating SSC, understanding sediment dynamics and erosion processes, and analysing changes in sediment load over time [29,30,31,32].

Sediment transport in streams is a complex process influenced by flow energy, which dictates the movement and concentration of sediment particles [33]. Analysing the relationships between SSC, discharge (Q), and turbidity (T) sheds light on these dynamics [34]. Studies have revealed the prevalence of hysteresis phenomena in these relationships, characterised by non-linear, loop-like behaviour during flow events [35]. Sediment hysteresis is a time lag between peak flow and sediment transport rates. This arises from the disparity in SSC between the rising and falling limbs of the hydrograph, caused by temporal variations in discharge and SSC [36,37]. Several factors influence hysteresis, including the magnitude and sequence of flood events, sediment particle size distribution, basin size, land use, and sediment source availability [36,38,39]. Additionally, alterations in channel morphology (planform, gradient, and bed material) due to natural or anthropogenic factors can contribute to SSC-Q hysteresis [40]. Sediment hysteresis methods, such as indexes and statistical analyses, quantify these relationships and provide valuable insights into sediment transport dynamics [41,42]. Understanding these dynamics informs engineering and environmental management decisions related to flood risk management, morphological adjustments, infrastructure design, and environmental impact assessment [35,43,44].

Although extensive research exists on riverine sediment transport, the accurate estimation of SSC in urbanised catchments remains a significant challenge due to factors such as limited monitoring stations, data gaps, and the dynamic nature of urban environments. This research presents a framework for developing effective sediment management strategies in urbanised stream environments. Using the upstream portion of Dry Creek, South Australia, as a case study, we aimed to (1) develop a sediment rating curve to address data limitations and enable long-term modelling of sediment transport, (2) gain insights into sediment dynamics within Dry Creek, and (3) inform future modelling efforts and sustainable management strategies. This study helps to address data limitations and provides a valuable tool for understanding and mitigating sediment transport challenges in urbanised stream environments. The findings can be applied to similar settings, contributing to a broader understanding of sediment dynamics, and informing future research, policy, and sustainable management practices.

2. Materials and Methods

2.1. Study Catchment

This study focuses on the Dry Creek catchment in the northern metropolitan area of Adelaide, South Australia, covering an area of 115 km². As its name suggests, Dry Creek is a seasonal surface water system that is dry most of the year, with intermittent flow that depends heavily on rainfall. There are several gauging stations along Dry Creek, but only the upstream station at Conway Crescent Valley View has long-term flow records and turbidity measurements that can be used for this study. The catchment area upstream of the Conway Crescent Valley View gauge covers 44 km². The catchment area is largely urbanised, with the majority of it covered by impervious surfaces (Figure 1). The elevation in the catchment area upstream of the gauging station ranges from 84 to 424 m AHD (Australian Height Datum), and the slope within the catchment varies from flat to very steep. However, while certain sections of the upper regions are situated on steep and highly steep slopes, most of the catchment is positioned on gently sloping, moderately gentle, and very gentle slopes. The spatial mean annual rainfall of the upstream catchment ranges from 532 mm at the downstream (west) point to 713 mm at the uppermost (east) point with high variability during dry seasons. Due to erosion and sediment transport processes occurring in the upland areas and throughout Dry Creek, the downstream sections of the creek have undergone channel aggradation [45].

2.2. Data Collection

Water level, streamflow, turbidity, and rainfall data for the Dry Creek catchment were obtained from an online database of flow and water quality records maintained by a private consulting firm (Water Data Services Pty Ltd., Edwardstown, Australia) that compiles data collected on behalf of South Australian Government departments in the Adelaide region (https://greenadelaide.waterdata.com.au/Amlr.aspx?station=A5041051&scroll=0 accessed on 15 December 2023). Data for the Conway Crescent Valley View gauging station was collected for the period from August 2022 to September 2023. Grab samples (n = 30) were collected from the centroid of the stream cross-section just downstream of the Conway Crescent gauging station during or immediately following rainfall events to supplement the continuous flow data sets with manually collected turbidity and SSC data. A 1-litre plastic container was used for sample collection.

Turbidity and SSC analysis of the collected water samples was conducted at the water chemistry and civil engineering laboratories of the University of South Australia. Samples were immediately transported to the laboratory and the turbidity was analysed on the same day using a portable Hach^® model 2100P turbidimeter. The turbidimeter was calibrated according to American Public Health Association (APHA) [46] protocols before each measurement. Analysis of SSC followed the APHA Standard Methods, Method 2540D. To measure SSC, whole samples were mixed thoroughly before measurement, then filtered through ashless Whatman filter paper (grade 42, 125 mm diameter, 100 g/m² base weight). Following filtration, samples were oven-dried at 105 °C for 24 h. The mass of dried suspended sediment was determined using a digital mass balance with a precision of 0.01 mg.

2.3. Suspended Sediment Concentration (SSC) Estimation

The collected grab samples were used to calculate SSC for each sample using Equation (1) and to develop a relationship between turbidity, streamflow, and SSC by utilising Equations (2) and (3). Available streamflow and turbidity data were used to estimate SSC using regression analysis [47]. Based on exploratory data analysis, a linear relationship was initially proposed to estimate SSC using streamflow and turbidity. At the start, we observed data following more than one distribution or regime. Further investigations revealed a distinct bimodal distribution, suggesting two distinct regimes. To account for this, separate SSC estimation models were developed for low-turbidity and high-turbidity regimes. To avoid biased estimates during zero-flow periods and account for the seasonal nature of flow in Dry Creek, a zero-intercept model was used in the regressions.

$$SSC={\displaystyle \frac{M}{V}}$$

(1)

$$SSC \left(Q\right)=a\ast Q$$

(2)

$$SSC \left(T\right)=b\ast T$$

(3)

where M is the mass of oven-dried sediment sample in milligrams (mg), V is the volume of the water sample in litres (L), SSC is the observed suspended sediment concentration in (mg/L), Q is the streamflow (L/s), T is turbidity in NTU, SSC (Q) and SSC (T) are the estimated suspended sediment concentration derived from streamflow and turbidity, respectively, and a and b are the coefficient derived from regressions of the observed SSC data with Q and T.

2.4. Development of Sediment Rating Curves

SRC development facilitates the long-term estimation of sediment discharge based on continuous measurements of streamflow. The SRC enables long-term SSC estimation based on continuous turbidity and streamflow measurements. Following common practice [48,49], a power equation was employed to establish the rating relationship between streamflow (Q) and suspended sediment discharge (Q_s). Equations (4) to (6) were used to calculate sediment discharge and sediment loads.

$$Q_s=SSC\ast Q$$

(4)

$$Q_s=a\ast Q^b$$

(5)

$$S_L=\sum Q_s$$

(6)

where Q_s is suspended sediment discharge (mg/s), S_L is sediment load over time (tonnes), and a and b are the coefficient and exponent, respectively, for developing a sediment discharge rating curve.

2.5. Model Selection and Performance Evaluation

The optimal model to predict SSC and sediment load transport based on measured streamflow was selected using several criteria: higher coefficients of determination (R²), Nash–Sutcliffe efficiency (NSE), minimum residual sum of squares (RSS) values, Akaike’s information criterion (AIC), and the Bayesian information criterion (BIC) [50,51,52]. Model performance was evaluated using these indicators based on Equations (7) to (10):

$$AIC=n\ast \mathit{ln}\left({\displaystyle \frac{RSS}{n}}\right)+2(k+1)$$

(7)

$$BIC=n\ast \mathit{ln}\left({\displaystyle \frac{RSS}{n}}\right)+\mathit{ln}\left(n\right)\ast (k+1)$$

(8)

$$R^2=1-{\displaystyle \frac{RSS}{TSS}}$$

(9)

$$NSE=1-{\displaystyle \frac{\sum _{n=1}^n{(Q_s-Q_{s\_p})}^2}{\sum _{n=1}^n{(Q_s-{Q_s\_}_{av})}^2}}$$

(10)

where n is the number of samples, k is the number of independent variables, RSS and TSS represent the total residual sum of squares and the total sum of squares, respectively, for the measured ($Q_s$) and predicted ($Q_{s\_p}$) sediment discharge values, and $Q_{\_av}$ is the average measured sediment discharge.

3. Results

3.1. Patterns of Turbidity, Streamflow, and SSC

Q, T, and SSC were checked for normality using the Anderson–Darling test. The results indicated a non-normal distribution, as the p-values were below 0.05. The SSC values of grab water samples varied from 3.2 to 431.8 mg/L, with a median of 77.3 mg/L (

Table 1. Statistical summary of measured streamflow, turbidity, and SSC parameters.

Variable	Min	Mean	Median	Max	StDev	CV (%)	p-Value
Streamflow (L/s)	11.0	1703.0	516.0	5889.0	1935.0	114	<0.005
Turbidity (NTU)	3.4	131.8	105.0	348.0	131.1	99	<0.005
SSC (mg/L)	3.2	119.9	77.3	431.8	122.2	102	0.007

Min = minimum, Max = Maximum, StDev = Standard deviation, CV = Coefficient of variation.

). The median streamflow rate and turbidity values were 516 L/s and 105 NTU, respectively. It was noted that higher SSC values corresponded to higher streamflow.

While the telemetry provides data records at 10 min intervals, only the data corresponding to the grab sampling times were used for comparison. Both grab sampling and telemetry data collection techniques exhibited a consistent trend in terms of turbidity measurements. However, some discrepancies were evident between the two methods. Following the exclusion of three outliers (Figure 2a) identified through Interquartile Range (IQR) and visual inspection, a linear regression analysis (grab sample turbidity = 0.71 * telemetry turbidity) indicated that telemetry turbidity measurements explained 68% of the variance observed in grab sample turbidity measurements (Figure 2b). Interestingly, telemetry turbidity measurements generally produced higher values compared to grab samples.

Figure 3 illustrates the dynamic relationship between streamflow and turbidity during selected rainfall events in Dry Creek. The observed patterns highlight the complex interplay between streamflow and turbidity in urban environments. Figure 3a demonstrates the classic “first flush effect” where a rapid increase in turbidity occurs at the onset of high-flow conditions. This is likely due to the rapid mobilisation of accumulated pollutants from urban surfaces at the onset of high-intensity rainfall. Figure 3b highlights a more complex pattern, with multiple turbidity peaks occurring both before and after the peak flow. This suggests a delayed response of sediment mobilisation, possibly due to the release of stored sediments from urban surfaces or channel banks. In contrast, Figure 3c exhibits a distinct counterclockwise flow-turbidity relationship. Turbidity levels remain elevated for an extended period following the peak flow, likely due to the resuspension of settled sediments or continuous input from upstream sources.

3.2. Hysteresis in Sediment Transport Dynamics of Dry Creek

The analysis of streamflow, turbidity, and suspended sediment concentration at the Dry Creek case study location revealed a complex interplay between these variables (

Table 2. Summary statistics for streamflow (Q), turbidity (T), suspended sediment concentration (SSC), and derived ratios for different phases of the hydrograph in Dry Creek.

Hydrograph	Statistics	Streamflow (L/s)	Turbidity (NTU)	SSC (mg/L)	SSC/Q	SSC/T
Rising limb	Min	65.0	3.4	4.7	0.04	0.31
	Mean	1223.3	169.2	66.9	0.15	2.77
	Median	1198.8	162.8	68.8	0.05	0.87
	Max	2430.7	348.0	125.3	0.47	9.05
	Stdev	1323.9	191.7	58.5	0.21	4.21
	CV (%)	108.2	113.3	87.3	134.8	151.9
Peak	Min	4178.8	105.0	232.4	0.06	2.09
	Mean	4816.0	127.6	299.4	0.06	2.32
	Median	4476.1	125.5	262.1	0.06	2.21
	Max	5889.5	157.0	431.8	0.07	2.75
	Stdev	711.8	20.3	81.6	0.01	0.26
	CV (%)	14.8	15.9	27.2	11.6	11.3
Falling limb	Min	258.7	17.7	3.2	0.01	0.18
	Mean	294.6	18.9	11.7	0.04	0.63
	Median	263.5	18.3	8.4	0.03	0.43
	Max	347.9	21.6	30.3	0.09	1.71
	Stdev	44.9	1.6	10.6	0.03	0.61
	CV (%)	15.3	8.2	90.7	76.78	96.51

). While streamflow exhibited fluctuations, particularly during the rising limb of the hydrograph, turbidity and SSC closely followed these trends, indicating a strong relationship. However, the SSC/Q and SSC/T ratios reveal a more complex dynamic, with variations in sediment mobilisation and the influence of particle characteristics.

The observed clockwise hysteresis, indicated by higher SSC/Q ratios during the rising limb compared to the falling limb, likely results from a combination of factors. The urbanised nature of the catchment, coupled with the availability of readily erodible sediments, contributes to the first-flush phenomenon and subsequent hysteresis patterns. The upper sections of the catchment, characterised by steep slopes and vegetation cover, are particularly susceptible to erosion. While vegetation can mitigate erosion, the steep slopes may amplify the force of water runoff, leading to increased soil loss and sediment transport. However, the spatial distance between these erosion-prone areas and the gauging station may introduce a time delay, contributing to additional peaks or hysteresis in subsequent monitoring cycles.

The observed hysteresis patterns indicate a non-proportional response of SSC and T to changes in discharge, particularly during the rising and falling limbs of the hydrograph. During the rising limb, the system mobilises the readily available coarse sediments. The higher coefficient of variation (CV) observed during the rising limb suggests increased variability in Q, T, and SSC with each flow increment. Additionally, the lower SSC/T ratio during falling limbs indicates the presence of finer sediment particles that remain suspended for longer periods as the streamflow rate decreases, leading to increased turbidity despite the decrease in SSC.

The SSC/Q ratio generally remained low, suggesting a relatively slow sediment mobilisation compared to the increase in streamflow. This could be influenced by factors such as the availability of sediment sources and the complex dynamics of sediment inflow and movement within the channel. The combined effects of increasing discharge scouring existing sediment deposits and increased erosion of channel banks during rising water levels may contribute to the observed patterns. However, the SSC/Q ratio may remain relatively constant during peak flows, as various sediment sources can influence peak flow concentrations without significantly affecting the overall ratio.

3.3. Relationship Between Streamflow, Turbidity, and SSC

The selection of appropriate methods for estimating SSC depends on the availability of measured streamflow and turbidity records. The performance metrics in

Table 3. Established relationships and performance metrics to estimate the SSC from Q and T.

Function	Model	R2	NSE	p-Value	RSS	AIC	BIC
SSC = f (Q)	SSC = 0.064Q	0.92	0.85	<0.0001	45,248	165	167
SSC = f (T); T < 250 NTU	SSC = 2.378T	0.97	0.94	<0.0001	15,149	142	144
SSC = f (T); T > 250 NTU	SSC = 0.316T	0.73	0.13	0.014	22,434	150	152

and the graphical representations of the parameters in Figure 4 demonstrate strong relationships between SSC, Q, and T. This indicates a more accurate model fit with stronger explanatory power for SSC variations. Streamflow-based SSC estimation, using a linear regression model utilising streamflow as the sole predictor for SSC estimation, demonstrated a statistically significant positive association between the two variables (p-value < 0.0001). This indicates that increasing streamflow corresponds with higher suspended sediment concentrations at the Dry Creek case study site. The model explained 92% of the variance in SSC, suggesting a strong linear relationship between streamflow and SSC (Figure 4b and Table 3).

The initial turbidity-based SSC estimation (SSC = 0.57T; R² = 0.39) revealed a weak association between turbidity and SSC using a simple linear regression model. This explained only 39% of the variance in SSC, suggesting limited predictive power for turbidity. However, observing the patterns of turbidity vs SSC graphs, the data were divided into two groups, namely a low-turbidity regime (T < 250 NTU) and a high-turbidity regime (T > 250 NTU). This showed a significant improvement with the low-turbidity regime explaining 97% of the variance, while the high-turbidity regime explained 73% of the variance in SSC (see Figure 4a). The strong positive correlation with a high R² value suggests a reliable relationship between turbidity and SSC during low-turbidity events in Dry Creek, indicating that turbidity is a reliable proxy for SSC when T and SSC values were considered in lower T conditions. Conversely, the high-turbidity model exhibited a relatively low predictive capability, suggesting that factors other than turbidity influence SSC during high turbidity periods. Figure 5 graphically compares measured and predicted SSC values, demonstrating the models’ performance in predicting SSC.

3.4. SRC and Sediment Discharge Dynamics

Following the lower predictive capability of turbidity for SSC estimation, we used streamflow-based SSC prediction and developed an SRC to analyse the sediment discharge dynamics of the creek. Analysis of the relationship between streamflow (Q) and sediment discharge (Qs) in Dry Creek at the case study site revealed that a power function provided the most suitable model. The calculated exponent of the power function (1.88) signifies the sensitivity of sediment discharge to changes in streamflow (Figure 6a). With an R² of 0.97, the SRC accurately predicts sediment discharge in Dry Creek based on streamflow data, suggesting its reliability for long-term estimation. The scatterplot (Figure 6b) confirms the strong agreement between predicted and measured sediment discharge, as shown by an NSE of 0.95 and R² of 0.97. These estimates are valuable for rapidly assessing potential downstream environmental impacts of sediment transport, such as channel morphology changes. This relationship can aid in understanding sediment transport processes and support the development of SRCs for urban stream management.

Using the sediment rating curve (SRC) and measured streamflow data from the Dry Creek case study site, the estimated annual sediment discharge was approximately 13.6 g/s (or 341.8 tonnes per year). This value highlights the sediment load in the upper reaches of the stream. Figure 7 shows how the monthly sediment discharge varies with changes in streamflow. A strong positive correlation is evident, with higher sediment discharge coinciding with periods of increased streamflow, particularly during the winter months. The figure also reveals several distinct peaks in sediment discharge, likely associated with specific storm events or periods of intense rainfall. Observations from May to September revealed a period of elevated streamflow (exceeding 122 L/s on average) and sediment discharge (surpassing 13.6 g/s on average). This seasonal trend coincides with the typical wet winter season in the region and aligns with the expected increase in sediment mobilisation during high-flow events.

Figure 8 illustrates the temporal variability of annual rainfall, streamflow (Q), and sediment discharge (Qs) in Dry Creek from 2001 to 2022. The figure reveals distinct patterns in these hydrological variables, with notable fluctuations over the study period. The annual variability in sediment discharge was significant, with a coefficient of variation of 67%. This highlights the sediment discharge sensitivity to variations in environmental and contributing factors. Notably, the annual variability of rainfall (19%) and streamflow (34%) was found to be lower compared to sediment discharge. The estimated annual sediment discharge can be used to calibrate and improve sediment transport models, leading to a better understanding of sediment dynamics in urban streams.

4. Discussion

The variation in streamflow during sample collection led to a higher coefficient of variation (CV) for all parameters (Q, T, and SSC), resulting in greater fluctuations in both turbidity and SSC. These findings are consistent with previous research highlighting the influence of flow variability on water quality parameters in streams [53,54]. Factors such as seasonal variations, periods of high discharge, and anthropogenic activities significantly impact turbidity and SSC levels [55]. For instance, Gibson and Hancock [12] reported significant variation in SSC within an ungauged agricultural catchment in southeast Australia, with monitoring data ranging from 30 to 350 mg/L. Similarly, a study conducted in the Yellow River within the Atlanta metropolitan area of the United States documented a peak SSC value of 648 mg/L, which was attributed primarily to the impacts of urbanisation and associated development activities [56]. Additionally, Fortesa, et al. [57] emphasise that the majority of annual sediment yield typically occurs during high-intensity flood events in agriculturally dominated streams. This highlights the importance of considering the temporal distribution of flow regimes when assessing sediment transport dynamics in streams.

The variations in turbidity between grab sample measurements and telemetry in this study could potentially be attributed to sensor calibration for telemetry turbidity measurements and the presence of stagnant water near the sensor location [58,59]. To ensure the accuracy and reliability of telemetry turbidity data, regular calibration and monitoring of the field and laboratory sensors are crucial [60].The dynamics of urban environments, such as the combination of impervious surfaces, stormwater runoff, and channel modifications, contribute to the variability between streamflow and turbidity. The initial flush removes contaminants from surfaces, leading to higher turbidity levels. As the storm progresses, water quality gradually improves, resulting in a decrease in turbidity [61,62]. On the other hand, the resuspension of settled sediments post-peak flow causes turbidity peaks [63].

The observed clockwise hysteresis in the relationship among Q, T, and SSC likely arises from a combination of factors. These factors may include but are not limited to; spatial variability of land cover in the catchment, channel erosion, and urban pollutants (such as oil and grease increasing the turbidity of water). Smith and Dragovich [34] reported a similar clockwise SSC-Q hysteresis due to rapid bank erosion and sediment remobilisation during peak flow events in an upland headwater catchment within the Lachlan River system, a tributary in the Murray–Darling Basin. Clockwise hysteresis patterns in SSC vs Q relationships typically indicate readily depletable sediment sources near the measurement point [43]. The dynamic SSC behaviour is attributed to the depletion of readily available sediment sources on the rising limb, leading to higher SSC than the falling limb [1,34,64]. In contrast, in-channel processes and distant sediment sources can lead to counter-clockwise hysteresis [44,64]. Understanding hysteresis dynamics is crucial for accurate sediment transport modelling and management [57]. Furthermore, stream complexity, catchment characteristics (size, land use, rainfall distribution), and particle properties (size, shape) influence the relationships between SSC, T, and Q [33,56,57]. Urbanisation leads to surface erosion and increases suspended particles during development [65,66]. Following development, urban surfaces still contribute sediment, oil, grease, and other wastes that cause fluctuations in turbidity [65,66,67].

This study revealed a strong linear relationship between Q and SSC. However, T exhibited non-linear relationships with SSC, suggesting a more complex process influenced by particle characteristics such as size, shape, and colour [68,69]. These characteristics can be temporally dynamic, potentially biasing turbidity measurements and hindering a perfect correlation with SSC [20,56]. Regime-based analysis was utilised to account for its non-linearity. Q-based models, typically employing a linear relationship, have been widely used due to the established link between high-flow conditions and increased SSC [70,71]. Kang, et al. [72] established relationships between T and SSC using power and polynomial methods in a 50 m long circulating channel in the Andong River Experiment Centre of the Korea Institute of Civil Engineering and Building Technology l in Korea, achieving satisfactory results (R² = 0.7). However, T-based models are susceptible to variations in particle characteristics, as evidenced by Sherriff, et al. [18] who observed higher correlations for lower turbidity levels in agricultural catchments. While the relationship between streamflow and SSC is often linear [73], reflecting the direct influence of discharge on sediment mobilisation, the relationship between turbidity and SSC can exhibit non-linearity [74]. This non-linearity can be attributed to factors such as variations in particle size distribution or the presence of organic matter, which can influence light scattering properties and turbidity measurements. While turbidity-based sediment measurement is widely utilised, its effectiveness varies based on specific environmental conditions. Studies have demonstrated a strong correlation between turbidity and SSC in rural catchments, especially when calibrated through simultaneous measurements during flood events [75]. Conversely, in urban areas, turbidity readings can be significantly impacted by water quality issues such as pollution, leading to less dependable estimates of SSC [53].

To account for the non-linearity and temporal variability inherent in the Q–T relationship, Wang and Steinschneider [33] employed dynamic linear models to understand flow-turbidity dynamics, particularly in situations with fluctuating particle characteristics. In the context of the case study of Dry Creek, the predictive capability of T for estimating SSC was constrained by the dynamic influence of the sediment discharge, perhaps related to the urban surroundings. Our recommendation is to employ streamflow as a means to estimate SSC, and we advise careful consideration when incorporating turbidity data.

Our analysis revealed that the developed rating curve at the case study site effectively explained 97% of the variability in sediment discharge based on streamflow. This finding underlines the importance of streamflow in governing sediment transport dynamics within an urban creek. In this study, a higher exponent (close to two) indicates a disproportionately larger increase in sediment discharge with increasing streamflow [76,77,78]. This finding is consistent with the understanding that higher streamflow possesses greater erosive power, mobilising and transporting larger quantities of sediment [79]. However, it is crucial to acknowledge the limitations associated with relying solely on SRCs for suspended sediment load estimation [80]. These limitations include inconsistencies in sampling techniques, inadequate data quantity or quality, and a limited flow range for which the curve is valid. Studies conducted in large catchments of Australia by Smith and Croke [80] in the Murrumbidgee River catchment and Joo, et al. [81] in the Fitzroy Catchment highlight these limitations. Climate change is another factor that can influence sediment discharge, as demonstrated by Kido, et al. [82] in the Pekerebetsu River basin, Hokkaido, Japan. Their study revealed a disproportionate increase in sediment discharge compared to precipitation and flow discharge, suggesting that factors beyond increased streamflow contribute to elevated sediment loads. This can be attributed to changes in rainfall patterns, which lead to more frequent slope failures and an increased sediment transport capacity of mountainous channels due to the supply of fine-grained sediment from eroded slopes.

While this study provides valuable insights into sediment transport dynamics at the case study site, it is important to acknowledge certain limitations as a first study of sediment discharge in the creek. The use of grab samples for SSC measurements could be refined through the adoption of flow-weighted sampling to improve the accuracy of event-mean concentration estimates. Additionally, expanding the spatial and temporal coverage of sampling sites and increasing the sampling frequency would enhance our understanding of sediment transport variability. Characterising sediment properties, such as particle size and organic matter content, would further enrich the analysis. Furthermore, investigating the influence of channel morphology and bank stability on sediment transport dynamics would provide a more comprehensive understanding of the complex factors influencing sediment transport in Dry Creek under various flow conditions.

5. Conclusions

This study examined sediment transport dynamics for a case study urban creek site, located in Dry Creek, South Australia, by analysing the relationships between streamflow (Q), turbidity (T), and suspended sediment concentration (SSC). This study involved fieldwork collecting grab water samples, laboratory analysis, and statistical modelling to establish a relationship and rating curve to predict SSC and sediment loads. The findings highlight the complex interplay between Q, T and SSC. In Dry Creek, turbidity levels exhibited a characteristic first-flush phenomenon during the onset of rainfall events mainly after a long antecedent dry period, reflecting the mobilisation of easily eroded sediments and pollutants from urban surfaces. Subsequent decreases in turbidity are likely due to the depletion of these readily available sources. The initial surge in flow aids the removal of pollutants from urban surfaces, contributing to elevated turbidity levels during the early stages of the hydrograph. A strong linear relationship was found between streamflow and SSC, with 92% of the variability in SSC being explained by the streamflow. However, the linear relationship with turbidity was more variable. To account for this, we employed a regime-based analysis. This approach revealed two distinct patterns: a stronger linear relationship between flow and turbidity in low-turbidity regimes compared to high-turbidity regimes. The explanatory power of the model was found to be 97% and 73% for low-turbidity and high-turbidity regimes, respectively. The observed hysteresis in the relationship indicates a non-proportional response of SSC and turbidity to changes in streamflow, emphasising the complex and dynamic nature of sediment transport processes. Given the limited correlation between turbidity and SSC during periods of high turbidity, we developed a sediment rating curve utilising streamflow using a power function. This approach demonstrated exceptional performance, with R² and NSE values of 0.97 and 0.95, respectively. These findings offer valuable insights into the long-term dynamics of sediment transport, facilitating a better understanding of sediment load in the urban stream, and informing the development of effective sediment management strategies. Based on the findings of this study, future research could benefit from establishing continuous monitoring programs to track changes in sediment transport dynamics over time. Additionally, increasing the number of sampling sites would allow for a more comprehensive understanding of spatial and seasonal variations in sediment transport within the catchment.