Groundwater levels, climate and anthropogenic factors affect the hydrology and water quality of an intermittent and a regulated subtropical stream

Stream hydrology and water quality are highly interconnected and impacted by climate, land use and geology. We examined this connection using monitoring data from 2000 to 2019 for two streams with contrasting hydrological regimes—intermittent and regulated perennial—in subtropical Queensland, Australia. Our main objective was to evaluate relationships between groundwater levels, climate and flow regulation on the hydrology and water quality of an intermittent and a regulated subtropical stream. In intermittently flowing Lockyer Creek, flow was highly dependent on groundwater levels and occurred when the aquifer was recharged to elevations exceeding the upper 90‐percentile value. With 9.4% of the catchment area in irrigated horticulture, flow in Lockyer Creek was also likely to be reduced by drawdown of the aquifer for irrigation, with no flow for 30% to 81% of days over the observation period for stations in Lockyer Creek. In contrast, flow in the mid‐Brisbane River was continuous, regulated by discharge from a large upstream dam. Nutrient and suspended sediment concentrations in Lockyer Creek were generally higher than in the mid‐Brisbane River, likely associated with runoff from agricultural areas adjacent to the stream, while the upstream dam likely reduced the concentration and variability of nutrients and suspended sediment in the mid‐Brisbane River. During periods of low flow in the mid‐Brisbane River, longitudinal changes in nutrient and suspended sediment concentrations occurred, notably a significant decrease in total and dissolved inorganic nitrogen concentrations downstream (p < 0.05), indicating a possible effect of in‐stream algal uptake and denitrification. This study highlights the impact of human modifications on stream hydrology and water quality in the face of climate change. The findings can inform decision‐making on groundwater irrigation or dam release control for water security.

Increasing severity of floods and droughts, exacerbated by climate change, is a major environmental issue affecting water availability and quality, damaging critical infrastructure and threatening drinking water supplies (Nema et al., 2012).For flowing waters, human interventions, such as dam construction, water extraction for irrigation and changes in channel morphology modulate climate impacts (Essaid & Caldwell, 2017;Magilligan & Nislow, 2005).Understanding the combined effects of climate variability and human impacts that affect the natural or reference condition of rivers and streams is essential to inform management (Chang et al., 2016).Good monitoring programs across all parts of the hydrological system are essential to support state of the environment assessments and increase confidence in water quality models designed to provide predictions under future global change.
Subtropical streams are characterized by highly variable streamflow due to seasonally skewed distributions of rainfall (Eccles et al., 2019).Seasonal changes interact with the El Niño-Southern Oscillation (ENSO) cycle, which has teleconnections with other climatic cycles, to produce large interannual variations in rainfall and streamflow (Chung & Power, 2017;Miralles et al., 2013).Recent severe weather events globally, many in tropical and subtropical systems, have focused attention on building catchment resilience to droughts and floods, as well as adapting to their impacts (Geris et al., 2015).Human stressors provide another overlay of complexity in subtropical regions, affecting whether flow is intermittent or perennial (Goodrich et al., 2018;von Schiller et al., 2008).Groundwater extraction, for example, aggravates the intermittency of streamflow and can act synergistically with climate variability to affect surface water availability (Cui et al., 2017).
Streamflow in subtropical Southeast Queensland, Australia, is highly variable and linked to climate cycles, including severe drought in the 2000s (van Dijk et al., 2013) and a major flood in 2011 during a strong La Niña cycle (Cai & van Rensch, 2012).The interactions of climate and streamflow provided the motivation for the current study, which compares the hydrology and water quality of two streams within the same basin but different subcatchments in the Southeast Queensland region.These streams have different human pressures as a result of groundwater extraction, flow regulation associated with a large dam, and diffuse sources of nutrients from areas of irrigated agriculture around and within the floodplain of the streams.
Flow regulation changes the nutrient contribution balance for sediment and nutrient sources, that is, external terrestrial sources and internal sources from in-stream processing (Ford et al., 2019).Hydraulic structures, such as dams and weirs trap pollutants (Ounissi & Bouchareb, 2013) and extend periods of continuously regulated flow in downstream receiving waters.They are important regulators of biogeochemical processes (Cunha et al., 2018).Flow regulation also causes longitudinal changes in water quality along stream reaches, and temporal changes due to altered in-stream residence times (Kaushal et al., 2014).Water residence time affects the fraction of sediment and particulate nutrients trapped in the streambed (Briggs et al., 2014) and influences the potential for eutrophication as periphyton and phytoplankton biomass build up when not flushed out of the system (Bae & Seo, 2021).Conversely, nitrogen removal via denitrification may be increased with reduced flushing as a result of an increase in the presence of anoxic environments, but also depending on the supply of nitrate and carbon (Sharp et al., 2021).Flow is therefore a 'master factor', regulating channel morphology, trophic state and spatio-temporal variability of streams, particularly in subtropical regions where there is highly variable flow (Biggs et al., 2005).
Phytoplankton growth is often limited by the availability of nitrogen (N) and phosphorus (P).The ratio of these nutrients can be important for nutrient limitation and is affected by climate and land use (Hayes et al., 2015), as well as internal processing (i.e., nutrient transformations), with dependence on the system residence time (Cunha et al., 2018).To assess potential nutrient limitation status, the Redfield atomic ratio (total nitrogen [TN]: total phosphorus [TP] = 16:1) (Redfield, 1958) is widely used, while Ptacnik et al. (2010) has suggested that the dissolved inorganic nitrogen (DIN):TP atomic ratio (N limitation below 2:1 and P limitation above 5:1 by atoms) is a superior indicator as DIN is a better indicator of the nutritional demands of phytoplankton and the occurrence of denitrification.
The sites in this study were two streams in subtropical Southeast Queensland, Australia.One of the streams was naturally intermittent.
The other was perennial due to perturbation of the natural flow by a large upstream dam.Using these study sites, the objectives of this study were to (1) evaluate how climate variability affects flow regimes through changes in groundwater levels in the intermittent stream catchment and water releases from the dam in the perennial stream, (2) examine temporal and spatial variations in water quality in the two streams.We tested the relative contributions of external and internal (in-stream) processes on the biogeochemical properties of each stream using a 20-year dataset that included two severe droughts interspersed with periods of severe flooding.

| Study area
Two streams, Lockyer Creek and the mid-Brisbane River, located west of Brisbane in Southeast Queensland, Australia (Figure 1), were selected for this study.Both streams are in agricultural catchments and prone to flooding.The two streams have highly variable streamflow due to episodic subtropical rainfall patterns.Lockyer Creek is an intermittent stream and has a catchment with a large proportion of irrigated horticulture (9.4% of the catchment) irrigated by groundwater, which influences groundwater levels and streamflow (Cui et al., 2017).The rich alluvial soils where agricultural production is focused in the Lockyer Creek catchment are readily eroded during storms and contribute massive loads of sediment and nutrients to the mid-Brisbane River (Grinham et al., 2021).Flow in the mid-Brisbane River is regulated by dam releases, resulting in nearly constant baseflow that is elevated above natural conditions (Kemp et al., 2015).The mid-Brisbane River is 61 km long from the outflow of Wivenhoe Dam to the Mt Crosby Weir drinking water intake and has several smaller tributaries including Lockyer Creek, which enters 2.5 km downstream of the dam outflow (Figure 1).Below Mt Crosby Weir, the mid-Brisbane River flows through Brisbane and discharges to the ecologically important Moreton Bay, a Ramsar wetland site (Kemp et al., 2015).Most of the catchment area of both streams is bushland and rural production, including pastoral grazing and cropping, with some peri-urban development (Garzon-Garcia et al., 2017;The State of Queensland, 2009).groundwater levels from four stations were downloaded from the Water Monitoring Information Portal managed by DRDMW (The State of Queensland, 2022).The water quality data from the other 29 stations included high frequency monitoring of storm events from one Seqwater station and four DES stations (Figure 1 and Table S1).

| Data description and analysis
The water quality parameters were analysed at DES Chemistry Centre When more than one reading was available at a station on any day (i.e., samples taken at high frequency during storm events), a median value was calculated.The resulting daily median value was used to calculate monthly median values for annual and seasonal trend analysis.For streamflow and groundwater level data, a daily mean value was calculated from hourly readings.Due to irregularities in storm event data, median values were preferred for water quality analysis, while mean values were considered for uniformly monitored data like streamflow and groundwater levels.The processed data were analysed statistically using R software (R Core Team, 2020).For statistical analyses, we utilized both parametric tests (t-test and ANOVA) for streamflow and nonparametric tests (Mann-Whitney U test and Kruskal-Wallis test) for water quality.The t-test and Mann-Whitney U test were used when comparing two variables, while the other two tests were applied for comparisons involving more than three variables.We conducted Spearman correlation tests between variables, including rainfall, streamflow and water quality parameters on a yearly and monthly basis.shown effectiveness in previous studies (Lim et al., 2010;Raffensperger et al., 2017;Stoelzle et al., 2020), but we found, for example, that the Web-based Hydrograph Analysis Tool (Lim et al., 2005) tended to categorize some high streamflow values (e.g., > 76 m 3 s À1 at St. 81 and >1870 m 3 s À1 at St. 94) as baseflow (i.e., low flow), suggesting these methods may not be well adapted to the highly variable flows occurring in subtropical regions.Instead, we adopted a segmented regression method, implementing it with the 'segmented' package in R (Muggeo, 2003).This package identifies breakpoints, indicating structural changes in data and delineation of different regression models (Muggeo, 2003).This functionality has been validated in subsequent studies (Muggeo, 2017;Muggeo et al., 2014).To apply the segmented regression method, we followed these steps: (1) calculation of daily mean streamflow values, (2) exclusion of streamflow of <0.001 m 3 s À1 , which was classified as no-flow,

| Analysis for stream hydrology
(3) application of a linear regression model to the remaining streamflow data which excluded no-flow cases and (4) utilization of the 'segmented' package to analyse the linear regression results and obtain the breakpoint to separate the high and low flow.

| Relationships of streamflow, groundwater level and rainfall for Lockyer Creek
We conducted a comparative analysis of streamflow, groundwater levels and rainfall data at each of the four stations: St. L-22, L-42, L-52 and L-81 for streamflow; St. L-22-G, L-42-G, L-52-G and L-81-G for groundwater level; and St. L-22-R, L-42-R, L-52-R and L-81-R for rainfall (see locations in Figure 1) within the Lockyer Creek catchment to examine the drivers of streamflow occurrence.We used daily mean streamflow and groundwater level data to calculate the percentages of streamflow (categorized as no-flow, low flow and high flow) observed on days with both high and low groundwater levels.Here, the 'high level' category refers to values exceeding the 99th percentile of the observed groundwater levels, while 'low' encompasses the remaining data.To analyse the impact of rainfall on streamflow occurrence, we considered the daily mean streamflow and 2-day rainfall over the 3 days leading up to and including the day of observation for the categories of high and low groundwater levels.Comparisons were made between time-series of monthly Southern Oscillation Index (SOI), annual rainfall (St.L-R, representing the Lockyer Creek catchment) and groundwater level (the four stations listed above).Values of SOI indicate three phases: La Niña (sustained positive values >7), El Niño (sustained negative values <À7) and neutral (À7 to 7) (Commonwealth of Australia, 2022a).

| Groundwater use for irrigation in the Lockyer Creek catchment
To gauge the scale of annual water usage from groundwater irrigation, which can be compared with the volume of Lockyer Creek flow, we collected irrigation data focusing on Lockyer Creek catchment area from reports and online sources managed by government agencies.
While some data for metered entitlement areas was available, data for areas with unmetered (unregulated) entitlement was either unavailable or based on estimated values.The compiled data include the area where groundwater irrigation is performed, the collected period of irrigation, the amount of water used from groundwater irrigation, and its sources.

| The percentage of dam release and Lockyer Creek, as major inflows of the mid-Brisbane River
To assess the impact of dam discharge on the mid-Brisbane River flow, we obtained dam release data from Seqwater.We compared this data with our collected flow data at St. M-94 (mid-point of the mid-Brisbane River) and St. L-81 (downstream point of Lockyer Creek) to evaluate the influence of the two major inflows on the river.
The comparison period is from 2012 to 2019.With the analysis of the total period, we divided it into wet years (2012)(2013)(2014)(2015) and dry years (2016)(2017)(2018)(2019), as well as wet season (October-March) and dry season (April-September).This yearly and seasonal comparison was conducted to account for potential variations in dam operation, which may affect flow patterns both annually and seasonally.between the two streams and between low flow and high flow of each stream were also tested using Mann-Whitney U test.We used the atomic ratios in the creek to assess the balance between N and P, aiming to understand how in-stream nutrient processing may affect ratios.It is important to note that atomic N:P ratios may not be good indicators of nutrient limitation in streams due to possible measurement issues for soluble reactive phosphorus (Dodds, 2003), and other factors, such as light availability, flow velocity and impact of benthic autotrophs (Keck & Lepori, 2012;King et al., 2014).

| Relationships of streamflow, groundwater level and rainfall in Lockyer Creek
Streamflow in Lockyer Creek was highly variable and dependent on groundwater levels.Figure 3  exhibited streamflow when groundwater levels were above the 99th percentile ('high levels').Conversely, these stations rarely had streamflow under 'low groundwater level' (Figure 3).In contrast, stations L-52 and L-81 had relatively high percentages of time with streamflow when groundwater levels fell below the 99th percentile ('low level').Streamflow at St. L-52 and L-81 mostly occurred when groundwater levels were above the 40th percentile.Station L-52 had a smaller range between the maximum and minimum groundwater levels compared with St. L-22 and L-42 (see Figure 3).Station L-81, lower in the catchment, receives multiple inflows, including those from three stations and adjacent inflows.
Figure 4 illustrates the relationship between rainfall and streamflow over the monitoring period from 2012 to 2019.Across all stations, many days were observed when streamflow did not occur despite there being over 50 mm of rainfall over a 2-day period under low groundwater level conditions.Furthermore, when streamflow did occur under low groundwater levels, the duration of streamflow was significantly reduced compared with periods with high groundwater levels.Stations L-22 and L-42 showed rapid transitions from flow to no-flow within 1-4 days.In contrast, streamflow was more sustained at stations L-52 and L-82, often exceeding 20 days following rain events.The 2-day antecedent rainfall explained more than 50% of streamflow only at St. L-52.Time-series of groundwater level showed different patterns among stations in the Lockyer Creek catchment (Figure 5).The temporal changes of groundwater levels were smaller but more frequent at St. L-52 than the other three stations.Groundwater     3).In the mid-Brisbane River, annual median concentrations of all water quality parameters were significantly higher from 2010 to 2013 than the other years (Kruskal-Wallis test, p < 0.05; Figure 6), when streamflow was also elevated.Monthly median concentrations of TSS, TN, TP and PO 4 -P in the mid-Brisbane River exhibited peaks in February, followed by April (Figure 7).In the statistical analysis, the monthly median concentrations of TP, and PO 4 -P were significantly higher in February than in July to September (Kruskal-Wallis test, p < 0.05; Figure 7).These concentrations correlated with observed monthly rainfall and streamflow, while the monthly concentrations of the other parameters (TSS, TN, NH 4 -N and NO x -N) were not related to monthly rainfall or streamflow (Table 3).4).For all stations, the concentrations of TSS, TN, TP and

| Proportions of nutrient forms and ratios of TN:TP and DIN:TP in Lockyer Creek and the mid-Brisbane River
The Kruskal-Wallis test and Mann-Whitney U test were used to assess whether medians and atomic ratios of selected nutrients differed as a function of flow (high or low) (see Table 5).The dominant species of nitrogen was DON in Lockyer Creek and the mid-Brisbane River, followed by PN in Lockyer Creek and NO x -N in the mid-Brisbane River (Kruskal-Wallis test, p < 0.05; Table 5).The dominant forms of phosphorus were PP and PO 4 -P for both streams (Kruskal-Wallis test, p < 0.05).In Lockyer Creek, the proportion of PP/TP was higher than PO 4 -P/TP during high flows, whereas under low flows, it was reversed (Kruskal-Wallis test, p < 0.05).Median atomic ratios of TN:TP and DIN:TP ratios were significantly higher in the mid-Brisbane River than Lockyer Creek (Mann-Whitney U test, p < 0.05; Table 5).
The TN:TP ratios were higher under low flow than high flow for both streams (Mann-Whitney U test, p < 0.05).

| DISCUSSION
In this study, two subtropical streams were used to understand the effects of human influence on stream hydrology and water quality.ensuring water security and effectively removing pollutants (Arthington & Pusey, 2003;Janes et al., 2017).Failure to do so may result in significant time and budget expenditures for flow restoration efforts in the future.

| Unbalanced nutrient exports from the Lockyer Creek catchment
In the study region, N is recognized as a primary element causing degradation of water quality (Bunn et al., 2007), related to the higher exports of P compared to N, associated with agricultural development in the catchment.Elevated P inputs from the Lockyer Creek catchment during storm events contribute to decreasing N:P ratios in downstream waters.This may result in N limitation for algal productivity in Moreton Bay (Glibert et al., 2006).Seasonal variations in P concentration, influenced by storm events, also impact cyanobacteria productivity in Moreton Bay by regulating N fixation rates (Elmetri & Bell, 2004).The high P export is likely due to erodible fertile soils that contain higher P than N, and the application of sulphur phosphate fertilizers on horticultural land in the Lockyer Creek catchment (Dickson & Asher, 1974;Sarker et al., 2008).Our data indicate that the relative increase in TN concentrations between low flow and high flow was less than for TP, as the high-flow events contributed sediment from the catchment that was enriched in phosphorus compared with nitrogen (see Lu et al., 2023).A study by Alvarez-Cobelas et al.In-stream processes, such as nutrient processing (e.g., uptake and denitrification), can significantly affect water quality under low flow (Cunha et al., 2018;Hensley et al., 2014).However, these processes are likely limited during high flow conditions due to short residence times.During periods of low flow in the downstream part of the mid-Brisbane River, the decrease in NH 4 -N and NOx-N, along with the increase in PN, might be linked to in-stream processes occurring due to longer residence times.Losses of nitrate from denitrification can also be expected downstream.However, algal blooms are more likely to occur in coastal receiving waters where light conditions are favourable for growth and phytoplankton are retained in coastal inlets or bays.For example, blooms of cyanobacteria, for example, Lyngbya majuscula, occur in Moreton Bay (Albert et al., 2005).Further studies are needed to clarify DIN input and denitrification in the mid-Brisbane River and how these processes could be managed to impact algal growth in the receiving waters of Moreton Bay.Previous research (Eyre & McKee, 2002;Mineau et al., 2015;Wollheim et al., 2008) confirms that N processing, including denitrification under warm and low flows, can influence DIN removal in rivers.Thus, managing DIN in the river could be critical to reducing the risk of algal blooms in the mid-Brisbane River and Moreton Bay.Investigation of conditions to promote denitrification potential could be also beneficial in reducing algal growth in this N-limited area (Adame et al., 2021;Newcomer Johnson et al., 2016).

| CONCLUSIONS
Our study aimed to examine the impacts of human activities and climate variability on stream hydrology and water quality in a subtropical agricultural catchment.Specifically, we investigated the effects of changes in groundwater levels and regulated dam release on the flow patterns and spatio-temporal changes in water quality in two streams: Lockyer Creek and the mid-Brisbane River.Our findings imply that groundwater irrigation can exacerbate the effect of climate driven variability on the intermittency of streamflow.Managing groundwater extraction in a systematic manner, including considering how to reduce agriculturally driven no-flow periods, could improve water security.For streams with regulated dam outflows, dam releases should be carefully managed for flood risk reduction but also for downstream water quality through flushing and dilution at key times.
Flow regulation by the dam facilitates in-stream water quality changes, promoting nutrient processing.For example, under low flow conditions, in-stream processes, such as denitrification can be further promoted to manage limiting nutrients, particularly nitrogen, reducing the risk of algal blooms in downstream (Brisbane River) and inshore (Moreton Bay) areas.Further research is necessary to quantify specific in stream processes for N removal, to effectively manage nitrogen in these receiving waters.
We used monitoring data from 2000 to 2019 in the Lockyer Creek catchment and the mid-Brisbane River catchment, including water quality, streamflow, groundwater level and rainfall.The water quality monitoring data were from 39 stations and provided by three Government departments and authorities; the Queensland Department of Regional Development, Manufacturing and Water (DRDMW), Seqwater, and the Queensland Department of Environment and Science (DES).Monitoring locations are shown in Figure 1 and details of monitoring, including frequency and parameters measured, are listed in Table S1.The water quality and streamflow data from 10 stations and F I G U R E 1 (a) Location of Southeast Queensland within Australia, (b) location of the study area within Southeast Queensland, (c) location of water quality (WQ) and streamflow monitoring sites of Department of Regional Development, Manufacturing and Water (DRDMW), Seqwater, and Department of Environment and Science (DES) and location of rain and groundwater level gauges within the catchments of Lockyer Creek (LOC) and the mid-Brisbane River (MBR).

and
Seqwater using standard methods (American Public Health Association, 2012).Water quality measurements included total suspended solids (TSS), total Kjeldahl nitrogen (TKN) and total Kjeldahl phosphorus (TKP), with additional samples filtered (0.45 μm syringe membrane filters, Minisart, Germany) prior to analysis for dissolved Kjeldahl nitrogen (DKN), dissolved Kjeldahl phosphorus (DKP), ammonium (NH 4 -N), nitrate/nitrite (NO x -N) and dissolved reactive phosphorus (PO 4 -P).Rainfall data were sourced from the Scientific Information for Land Owners climate database (The State of Queensland, 2021b) for six stations (Figure 1).Total nitrogen (TN = TKN + NO x -N), total phosphorus (TP = TKP), particulate nitrogen (PN = TKN À DKN), particulate phosphorus (PP = TKP À DKP), dissolved organic nitrogen (DON = DKN À NH 4 -N), dissolved organic phosphorus (DOP = DKP À PO 4 -P) and dissolved inorganic nitrogen (DIN = NH 4 -N + NO x -N) were calculated from the measured nutrient parameters.A minor proportion of censored nutrient concentrations below the detection limits (0.002 mg L À1 for NH 4 -N and NO x -N, and 0.001 mg L À1 for PO 4 -P, comprising 16%, 6% and 2% of the samples, respectively) were addressed by a tailored function.This function considered both the nutrient type and the measurement timestep, generating a replacement estimate through a uniformly distributed random number (0.01-0.99) multiplied by the relevant detection limit.
2.3.1 | Streamflow classification: No-flow, low flow and high flow Streamflow was divided into no-flow, low flow and high flow at eight monitoring stations (St.L-11, L-14, L-22, L-42, L-52, L-55, and L-81 in Lockyer Creek and St. M-94 in the mid-Brisbane River) using daily mean data for 20 years from 2000 to 2019 (calendar year).To classify streamflow into low flow and high flow, we explored various flow separation methods, including local minimum methods, digital filter methods and breakpoint analysis.Several different methods have The identified breakpoints were used to serve as indicators of the flow threshold differentiating low flow and high flow for each station in Lockyer Creek and the mid-Brisbane River, individually.This data were employed for water quality analysis partitioned as low flow and high flow.The percentage of time with each streamflow condition (no-flow, low flow and high flow) during the monitoring period was calculated for each station.
We assessed annual and monthly trends in water quality, streamflow and rainfall in Lockyer Creek and the mid-Brisbane River from 2007 to 2019 (calendar year).The annual analysis included the year 2007 to illustrate the low annual rainfall and no-flow condition in Lockyer Creek that year, while the monthly analysis was conducted for the period from 2008 to 2019.For annual and monthly comparisons, boxplots of streamflow used daily mean values at St. L-81 and M-94, whereas plots of water quality (TSS, TN, TP, NH 4 -N, NO x -N, and PO 4 -P) used monthly median values at each of the monitoring stations due to irregularity of sampling days, which included event-based monitoring within each month.Rainfall data were compared annually for the wet and dry seasons.Two rain gauges (St.L-R and St. M-R) located at the centre of the Lockyer Creek catchment and the closest gauge to Wivenhoe dam outflow were used to represent rainfall for the Lockyer Creek catchment and the mid-Brisbane River catchment, respectively.The annual and monthly variations of streamflow and water quality for each stream were used to test statistical differences among different years and months (treatments) using one-way ANOVA with post hoc Tukey's HSD test for streamflow and the Kruskal-Wallis test for water quality.We conducted a Spearman correlation analysis between variables (rainfall, streamflow and water quality parameters [TSS, TN and TP]) on a yearly and monthly basis.2.4.2 | Comparison of water quality between low flow and high flow and among multiple stations Water quality parameters (TSS, TN, TP, NH 4 -N, NO x -N and PO 4 -P) were examined at five major stations under low flow and high flow.Stations L-14, L-42 and L-55 have a similar sub-catchment size but different land use.Station L-84 is located further downstream from station L-81 in Lockyer Creek, and station M-91 is directly upstream in the mid-Brisbane River, adjacent to Wivenhoe Dam.We used land use information and a 25 m digital elevation model (DEM) data both obtained from The State of Queensland (2021a) to calculate subcatchment area and the percentages of land use, using QGIS software (QGIS Development Team, 2019).Land use was classified according to the Australian Land Use and Management Classification (Commonwealth of Australia, 2016) (Table 1).We calculated the median and 10th and 90th percentile values for concentrations of TSS and nutrients.We then conducted a Mann-Whitney U test to compare the median values between low flow and high flow conditions and a Kruskal-Wallis test to compare the median values across the five stations for each streamflow condition.2.4.3 | Changes in longitudinal water quality, N:P ratios and proportions of nutrient forms Longitudinal changes in the concentration of TSS and nutrients and nutrient ratios in the mid-Brisbane River from upstream (dam release point) to downstream (Mt Crosby Weir; Figure 1) were compared under low flow and high flow.Monthly data (TSS, TN, PN, DON, NH 4 -N, NO x -N, TP, PP and PO 4 -P) sampled on the same day from 2011 to 2019 at seven stations (St.M-91 to M-97) were tested for longitudinal concentration changes, N:P ratios (TN:TP, DIN:TP), and proportions (NH 4 -N/TN, NO x -N/TN, PN/TN, PO 4 -P/TP and PP/TP) of nutrients using boxplots.Differences in mean values of the nutrient ratios and proportions among the seven stations under low flow and high flow conditions, respectively, were tested using Kruskal-Wallis test.Differences of the atomic N:P ratios and proportions of nutrients | Streamflow (no-flow, low flow and high flow) in Lockyer Creek and the mid-Brisbane River Most of creeks in the Lockyer Creek catchment, except St. L-11 located in the uppermost part of the catchment, exhibited a high percentage of no-flow days, ranging from 30% to 81% during the observation period (Figure 2).The downstream station L-81 did not flow for 51% of the time, despite a catchment area of 2490 km 2 (The State of Queensland, 2022).Station M-94, located below the dam outflow in the mid-Brisbane River, flowed continuously.The breakpoint for discriminating between high and low flows in the mid-Brisbane River (St.M-94) was 38 times higher than that in the downstream station in Lockyer Creek (L-81), as shown in Figure 2.This discrepancy becomes pronounced when considering the four-fold difference in catchment size between the two stations: 10 170 km 2 for M-94 and 2490 km 2 for L-81 (The State of Queensland, 2022).
indicates the percentages of flow conditions (no-flow, low flow and high flow) relative to groundwater levels (low and high) at four stations (St.L-22, St. L-42, St. L-52 and St. L-81).Between 2012 and 2018, stations L-22 and L-42 usually levels at stations L-22-G and L-42-G peaked in 2011 after high rainfall years (>1000 mm year À1 ) in 2008, 2010 and 2011.These years also had sustained positive SOI values (>7, corresponding to strong La Niña).Groundwater levels gradually decreased until the end of 2015, corresponding to years with reduced rainfall (<700 mm year À1 ).F I G U R E 3 The relationship between daily mean groundwater (GW) levels and streamflow at four groundwater stations (St.L-22, L-42, L-52 and L-81) within the Lockyer Creek catchment from 2012 to 2019 (total 2922 days).Any sporadic missing groundwater level data were supplemented using an interpolation method.The red dashed lines indicate the 99th percentile value of the GW level between the maximum and minimum levels at each station when the stream has no-flow (streamflow <0.001 m 3 s À1 was replaced with 0.001 m 3 s À1 ).

F
I G U R E 2 Daily mean streamflow (minimum to maximum) between 2000 and 2019 (7305 days) at eight observation stations; upper subcatchment of Lockyer Creek (St.L-11, L-14, L-22, L-42, L-52 and L-55), downstream of Lockyer Creek (St.L-81), and the mid-Brisbane River (St.M-94).The red dashed line indicates the calculated breakpoint (threshold between low flow and high flow).The number of days and percentages on which no-flow was observed is stated for each station.3.3 | Groundwater use in Lockyer Creek catchment and percentage of dam release magnitude in the mid-Brisbane River flow Groundwater irrigation in the Lockyer Creek catchment is both metered and unmetered, with high inter-annual variation reported.Seqwater (2019) reported annual total water usage and water allocations for groundwater irrigation across all water sources in the Central Lockyer Valley region between 2003 and 2019.Through calculations involving water usage and groundwater irrigation allocations between 2003 and 2019, the metered groundwater irrigation ranges between approximately 1140 and 5700 ML per year.Earlier Wolf and Moore (2011) estimated total groundwater withdrawal (metered and unmetered) for cultivation in the 1980s of 46 500 ML per year.This groundwater irrigation represents a significant amount, whether metered or unmetered, particularly when compared with the mean annual streamflow volumes (76 175 ML per year) at the downstream point of Lockyer Creek (St.L-81), even during extended dry periods.The contribution of dam releases to mid-Brisbane River flows is substantial, accounting for 86% of the total flow from 2012 to 2019, while Lockyer Creek contributes about 11% (Table 2).Particularly during the dry years (2016-2019), the dam's contribution was more significant (the total may exceed 100% due to the potential time differences caused by dam release control).The dam releases attenuated the relative streamflow variation between dry and wet seasons compared with that in the Lockyer Valley.

3. 4 |
Annual and monthly variation in streamflow and water quality of Lockyer Creek and the mid-Brisbane River At both annual (Figure6) and monthly (Figure7) time scales, differences were observed in the streamflow between Lockyer Creek and the mid-Brisbane River.Flow in Lockyer Creek was highly variable, remaining dry between 2007 and early 2008, followed by a sustained flow through 2015.The flow continuity during this periodF I G U R E 4The relationship between 2-day rainfall (leading up to and including the day of observation) and daily mean streamflow at the four groundwater stations.The coloured groups are based on different groundwater levels ('high' represented in blue and 'low' in orange), differentiated by 99-percentile values at each station).The blue lines indicate linear models fitted to the data only when groundwater levels were 'high', and the corresponding R-squared values were calculated.F I G U R E 5 Time-series variation of climate and groundwater (GW) levels.Monthly Southern Oscillation Index (SOI) calculated by the method of the Australian Bureau of Meteorology, annual rainfall volume at St. L-R (the centre of the Lockyer Creek catchment), and groundwater level at the four groundwater stations within the Lockyer Creek catchment from 2000 to 2019.The red dashed lines on the first graph indicate three phases: La Niña (>7), El Niño (<À7), and neutral (À7-7).corresponded to years of high rainfall over 1000 mm year À1 , observed in 2008, 2010 and 2011.In Lockyer Creek (St.L-81), the two highest annual mean streamflow years, 2011 and 2013 (ANOVA, p < 0.05), occurred during the period of sustained flow.Streamflow ceased in early 2016 due to annual rainfall consistently below 700 mm year À1 from 2014.The mean streamflow in the mid-Brisbane River for the monitoring period was almost 10 times higher than that in Lockyer Creek (22.6 m 3 s À1 vs. 2.4 m 3 s À1 , t-test, p < 0.05), despite only about a four-fold difference in catchment size (10 170 km 2 vs. 2490 km 2 , The State of Queensland, 2022).Table 2 shows that from 2008 to 2019, excluding 2011, the mean flow at St. M-91 (immediately below the dam) is seven times higher than the streamflow at St. L-81 (downstream point of Lockyer Creek) (t-test, p < 0.05).Annual mean streamflow in the mid-Brisbane River at St. M-94 was significantly higher in four successive years (2010-2013) compared with the other years between 2007 and 2019 (ANOVA, p < 0.05) when there were large outflows from Wivenhoe Dam.Monthly variations in rainfall and streamflow tended to be similarly proportionately higher in the months from December to March than the other months in bothLockyer Creek and the mid-Brisbane River.Monthly mean streamflow in Lockyer Creek was larger during December to March (9.9 m 3 s À1 in mean) than during the other months (0.7 m 3 s À1 ).Streamflow in the mid-Brisbane River was also larger during December to March (75 m 3 s À1 ) than the other months (9.6 m 3 s À1 ) (see Figure7).Median concentrations of water quality parameters differed in the streams at annual and monthly time scales.Annual mean TSS values in Lockyer Creek were positively correlated with annual streamflow (Spearman's rho, p < 0.05; Table3), with peaks in 2011 and 2013 followed by a decrease after 2013 (Figure6).The higher median concentrations of TN, TP and PO 4 -P were observed in 2008, 2011 and 2017 compared with the other years (Kruskal-Wallis test, p < 0.05; Figure 6).The peaks in concentrations of those parameters in 2008 and 2017 coincided with antecedent periods of no-flow.Annual median concentrations of NH 4 -N and NO x -N were elevated from 2011 to 2014, similar to annual patterns of TSS.Monthly median concentrations of TSS, TP, TN, NH 4 -N and NOx-N tended to correlate more strongly with streamflow than rainfall in Lockyer Creek, confirmed by higher Spearman's rho values (see Table Median concentrations of water quality parameters (TSS, TN, TP, NH 4 -N, NO x -N and PO 4 -P) at multiple stations (St.L-14, L-42, L-55 and L-84 in Lockyer Creek and M-91 in the mid-Brisbane River) were compared between low flow and high flow and among the stations (Table PO 4 -P were higher under high flow than low flow (Mann-Whitney U test, p < 0.05).However, concentrations of NH 4 -N at St. L-42, L-55 and L-84, and NO x -N at St. L-42 did not show significant differences (Mann-Whitney U test, p ≥ 0.05).During both low and high flow conditions, St. L-55 had the highest median concentrations of all water quality parameters, except for TSS in high flow conditions, which had a higher median concentration at St. L-42 (Kruskal-Wallis test, p < 0.05).Median concentrations of water quality parameters were generally lowest at the uppermost Lockyer Creek station L-14 or the mid-Brisbane upstream station M-91, with no difference in TN and TP concentrations under high flow for stations L-14, L-84 and M-91.3.6 | Longitudinal water quality changes under low flow and high flow in the mid-Brisbane River

F
I G U R E 6 Annual total rainfall (at St. L-R and M-R), monthly mean streamflow (at St. L-81 and M-94), and monthly median concentrations of TSS, TN, TP, NH 4 -N, NO x -N and PO 4 -P (at St. L-81 to L-84 and M-91 to M-97) in the downstream sector of Lockyer Creek and the mid-Brisbane River between 2007 and 2019.This data comparison is based on monthly median concentrations to mitigate potential data distortion resulting from irregular sampling frequency during storm events.The centre, upper, and lower lines of the boxplots represent the median and upper quartile and lower quartile of the data, and red 'Â' marks on streamflow represent the mean values.The data sample numbers are twelve per year for all treatments (years) unless numbers are shown on the graphs (some missing data of Lockyer Creek is due to no-flow condition and no measurements of some parameters).significant differences in median concentrations between stations were found under low flow (for water quality parameters except for DON and PO 4 -P) but not high flow (for all parameters) (Figure 8).The small sample size of the high flow category may have contributed to being unable to statistically detect differences.Under low flow, median concentrations of TSS, TN and TP increased significantlyF I G U R E 7 Monthly total rainfall (at St. L-R and M-R), monthly mean streamflow (at St. L-81 and M-94), and monthly median concentrations of TSS, TN, TP, NH 4 -N, NO x -N and PO 4 -P (at St. L-11 to L-84 and M-91 to M-97) in Lockyer Creek and the mid-Brisbane River for 12 years between 2008 and 2019.This data comparison is based on monthly median concentrations to mitigate potential data distortion resulting from irregular sampling frequency during storm events.The centre, upper, and lower lines of the boxplots represent the median and upper quartile and lower quartile of the data, and red 'Â' marks on streamflow represent the mean values.The number of data samples for all treatments (months) are twelve per month unless the numbers are shown on the graphs.between St. M-91 and M-92, that is, before and after the Lockyer Creek inflow (Kruskal-Wallis test, p < 0.05; Figure 8), whereas the other parameters (PN, DON, NH 4 -N, NO x -N and PO 4 -P) did not show significant differences (Kruskal-Wallis test, p ≥ 0.05).In the downstream region of the mid-Brisbane River (St.M-96 and M-97), median concentrations of NH 4 -N and NO x -N decreased significantly (Kruskal-Wallis test, p < 0.05).At St. M-97, NH 4 -N and NO x -N concentrations fell below the limit of detection in 75% and 54% of the samples during low flow, respectively.Changes in atomic ratios of TN:TP and DIN:TP and proportions of dissolved and particulate nutrients from stations M-91 to M-97 were used along with concentrations to indicate potential for nutrient limitation.The atomic ratios were compared against the balanced nutrient ratios from Redfield (1958) and Ptacnik et al. (2010) (see Section 1).During low-flow conditions, both the N:P ratios and the proportions of N and P in particulate and dissolved forms showed significant differences among stations (Figure9).Median TN:TP ratios peaked at St. M-91 and exhibited a decreasing trend downstream (Kruskal-Wallis test, p < 0.05), consistently remaining higher than the Redfield ratio of 16:1.Median DIN:TP ratios significantly decreased at St. M-96 and further declined at St. M-97 (downstream stations) (Kruskal-Wallis test, p < 0.05).At St. M-97, DIN/TN ratio decreased significantly, whereas PN/TN ratio increased (Kruskal-Wallis test, p < 0.05; Figure9).Proportions of other nutrients (DON/TN, PP/TP, DOP/TP and PO 4 -P/TP) remained stable along the river (Kruskal-Wallis test, p ≥ 0.05).
The two rivers are in close proximity but have different human influences.The stations within the Lockyer Creek catchment experience intermittent flows that depend on groundwater levels, which are in turn affected primarily by rainfall and irrigation for cropping, while the mid-Brisbane River is a regulated perennial stream influenced mostly by upstream dam releases.These differences generate different spatio-temporal patterns of streamflow and water quality.4.1 | Two contrasting hydrological systemsMost studies (see literature review by Eccles et al., 2019) project increased extreme riverine flooding from future climate change in both tropical and subtropical regions.Eccles et al. (2019) suggest that these changes are complicated by impacts of anthropogenic activities, such as hydraulic structures and land use change.Our results indicate that human impacts, such as groundwater extraction and flow control by a dam, may significantly alter the nature of streamflow in various ways.Vu et al. (2018) reported that groundwater levels have a significant impact on surface water hydrology in both tropical and subtropical regions, but this impact may be critical in subtropical agricultural catchments where extraction is required to sustain productivity during dry periods.The percentage of no-flow days in Lockyer Creek (St.L-14 to L-81) ranged from 30% to 81%.Drier conditions in the creek are influenced by various factors, including irrigate cropping area, groundwater storage levels, rainfall pattern and the location of the monitoring stations in the catchment.Particularly, areas adjacent to

T
A B L E 4 Quantile and median concentration values of TSS, TN, TP, NH 4 -N, NO x -N and PO 4 -P under low flow and high flow between 2008 and 2019 at five stations; upper sub-catchments of Lockyer Creek (St.L-14, L-42 and L-55) and downstream of Lockyer Creek (St.L-84), and upstream point of the mid-Brisbane River (St.M-91).The superscript letters following the median values, arranged in alphabetical order from 'a' (highest to lowest, p < 0.05), represent statistical comparisons among stations from St. L-14 to M-91.Comparisons were conducted separately for low flow and high flow conditions, using Wilcoxon Signed-Rank test.Except for NH extensive irrigated horticulture, where groundwater extraction likely intensified the natural intermittency of streamflow resulting from seasonally and annually variable rainfall.The substantial magnitude of groundwater irrigation occurring in the Lockyer Creek catchment could significantly impact both aquifer recharge and streamflow dynamics.

F
I G U R E 8 Longitudinal water quality concentration (TSS, TN, PN, DON, NH 4 -N, NO x -N, TP and PO 4 -P) gradients in the mid-Brisbane River from upstream to downstream (St.M-91-M-97) under low flow and high flow.The monthly data collected on the same date for all stations between 2011 and 2019 was used.The data sample numbers per station are stated.The superscript letters on the boxplot show the significant difference in median values of the water quality parameters among stations, with 'a' representing the highest value.Stations with same letters indicate no significant difference (Kruskal-Wallis test, p ≥ 0.05).Test for high flow was excluded due to the small sample size that may make it difficult to detect differences in median concentrations.Groundwater recharge and depletion exhibited multi-year variability, which was closely linked to annual rainfall patterns affected by SOI variations.However, the slow pace of groundwater recharge and depletion during transition between dry and wet years contributed to a time lag in the occurrence of streamflow.During the Millennium drought from 1997 to 2009, groundwater levels remained low, but began to recharge following heavy rainfall in 2008 (999 mm).High rainfall in 2010 and 2011 corresponded to a strong La Niña F I G U R E 9 Longitudinal changes in atomic ratio values of TN:TP (the first row) and DIN:TP (the second row) and proportions of DIN/TN and PN/TN (the third line) and PO 4 -P/TP and PP/TP (the fourth line) in the mid-Brisbane River under low flow and high flow.The monthly data collected on the same date for all stations between 2011 and 2019 was used.Statistical comparison results are based on Kruskal-Wallis test, and test for high flow was excluded due to the small sample size.The red dashed lines in the first row indicate the Redfield N:P atomic ratio of 16:1 and those in the second row indicate the DIN:TP ratio, suggested by Ptacnik et al. (2010) (N limitation below 2:1 and P limitation above 5:1).T A B L E 5 Median values of proportions of nutrient parameters (PN, DON, NH 4 -N and NO x -N in TN and PP, DOP and PO 4 -P in TP) and N:P atomic ratios (TN:TP and DIN:TP) between 2008 and 2019 in Lockyer Creek (LOC) and in the mid-Brisbane River (MBR).
phase and recharged groundwater to its maximum recorded level (99th upper percentile) from January 2011 to January 2016.Major flooding occurred at this time January 2011) throughout Southeast Queensland, encompassing all the monitored wells.Also, in January and February 2013, massive flood occurred at Laidley in the Lockyer Creek catchment (Commonwealth of Australia, 2013), with subsequent high groundwater levels and continuous flow until late 2015.A subsequent El Niño or neutral phase through 2015 (Commonwealth of Australia, 2022b) coincided with the continuous decline in groundwater levels after 2015 and intermittency and drying of streams.Despite higher annual rainfall in Lockyer Creek in 2008, the duration and magnitude of streamflow was significantly lower than in 2013 due to overall lower groundwater levels.Groundwater levels are closely linked to flooding occurrences and can be used as an early warning for periods of increased risk of floods during storms.Dam release significantly offsets the impact of variable rainfall on the flow dynamics in the mid-Brisbane River.Prior to dam construction in the mid-1980s, the mid-Brisbane River flowed continuously, but flow varied more by season compared with the current regime (The State of Queensland, 2022).Dam releases damp the variation of flow in the mid-Brisbane River, increasing the magnitude of the lowest flows and flattening the flow recession, thereby reducing flood peaks.It is probable that high flow extremes decreased while there was increased frequency of high flows of lower magnitude.However, significant increases in flow were observed between late 2010 and early 2011, which was one of the strongest La Niña events on record (Commonwealth of Australia, 2022c).Throughout this period, extensive dam releases, combined with floodwater from the saturated Lockyer Creek catchment, caused the highest magnitude and longest duration of high flows recorded during the monitoring period, leading to widespread damage downstream of the dam (van den Honert & McAneney, 2011).Quantifying the combined impact of flow variations from the uncontrolled Lockyer Creek catchment and releases from the Wivenhoe Dam is critical to improving flood preparedness.4.2 | Water quality changes under different hydrological systemsThe different hydrological regimes of Lockyer Creek and the mid-Brisbane River had significant impacts on sediment and nutrient concentrations in each system.In Lockyer Creek, where streamflow is intermittent and flow regulation is limited, water quality changes were primarily driven by rainfall and likely also by land use, with stations in the lower Lockyer Creek located adjacent to intensive horticulture.The intermittency corresponded to rapid transitions between short periods of high throughflow with limited in-stream processing and extended periods of no-flow.Conversely, Wivenhoe Dam traps sediment and particulate nutrients, resulting in less variability in water quality in the mid-Brisbane River, except for years of floods(2011,   2013).The residence time within the dam was otherwise likely to exert strong control on sediment and nutrients in the upper mid-Brisbane River and under constant low flow, there was adequate time for in-stream nutrient processing, resulting in obvious longitudinal changes in nutrient concentrations.Compared with Lockyer Creek, the mid-Brisbane River exhibits low variability of TSS, TN and TP, with the lowest concentrations observed at the station immediately below the dam, indicating how the dam acts in net retention of sediment and nutrient.Moreover, the dam appeared to promote instream processing downstream of the dam, leading to a decrease in DIN.To enhance river water quality management in this region, further studies are required to better understand the specific processes and environmental factors contributing to DIN reduction.Simultaneously, it is crucial to implement water management strategies that maintain a balance between preserving pre-existing natural stream communities,

(
2008), which compiled an extensive dataset of 946 systems from the literature, also demonstrated that N export is only weakly related to runoff and catchment area.Thus, monitoring potential N inputs from diffuse sources associated with a single land use or point sources like wastewater can be beneficial in understanding the sources of N, which can vary in their timing and location, and therefore improving water quality management strategies.4.4 | Longitudinal changes in water quality in the mid-Brisbane RiverLongitudinal changes in water quality reflect the balance between external inputs and nutrient processing within the river(Hensley et al., 2014).In the mid-Brisbane River, dam releases are the primary determinant of dilution effects and residence times, which have different impacts on water quality.Our data indicate that Lockyer Creek inflows are likely to be significant contributors the increase in concentrations of TSS, TN and TP during low flows in the mid-Brisbane River.Other tributary inflows may obscure obvious longitudinal patterns.For example, although not statistically significant, there were small increases in NH 4 -N and NO x -N between stations M-92 and M-94 under low flow were potentially due to wastewater discharges located between these stations.The volume of these discharges ($0.017-0.051m 3 s À1 ), less than 1% of the mid-Brisbane River low flow, is small but among all the stations in the river, areas receiving wastewater discharges showed clear increases in the mean TN and DIN concentrations, although these increases were not statistically significant.TN and TP concentrations are elevated relative to the main tributary.Therefore, nutrient management for tributaries, such as the Lockyer Creek and discharges from wastewater treatment plants is crucial to prevent potential water quality degradation in the mid-Brisbane River under low flow conditions, while dam release control may help to dilute or flush water when there are high external inputs.
T A B L E 1 Catchment area and percentage of land use at major stations; St. L-14 (Upper Lockyer Creek catchment), St. L-22 (Flagstone Creek catchment), St. L-42 (Tenthill Creek catchment), St. L-52 (mid-point of Laidley Creek catchment), St. L-55 (Laidley Creek catchment), St. L-84 (The Lockyer Creek catchment), St. M-91 (Lake Wivenhoe catchment) and St. M-97 (the sum of the Lockyer Creek catchment, the mid-Brisbane River catchment and Lake Wivenhoe catchments).Note: Catchment boundaries are shown in Figure 1.The catchments of the stations were delineated from 25 m Digital Elevation Mapping data and calculations of area and land use were done in QGIS.Land use was classified according to the Australian Land Use and Management Classification (Commonwealth of Australia, 2016).
The period from 2012 to 2015 is categorized as wet years (when groundwater levels were high), and the years 2016-2019 are classified as dry years.In addition, the wet season is defined as the months from October to March, and the dry season is from April to September.The dam release data at St. M-91, used for this comparison, was provided from Seqwater and calculated based on daily mean data, consistent with the calculation methodology for other dataset analysis.
In the mid-Brisbane River, longitudinal variations in median concentrations from upstream (St.M-91) to downstream (St.M-97) were statistically tested using Kruskal-Wallis test, and all the tests for statistical significance are based on a significance level of 0.05.Statistically T A B L E 3 Correlations (Spearman's rho) between treatments: rainfall versus streamflow, rainfall versus water quality concentrations, and streamflow versus water quality concentrations for Lockyer Creek (LOC) and the mid-Brisbane River (MBR) between 2008 and 2019.The water quality parameters include TSS, TN, TP, and NH 4 -N, NO x -N and PO 4 -P.Data comparison is based on yearly and monthly data of total rainfall, mean streamflow and median concentrations of water quality parameters.An asterisk (*) indicates a p-value of <0.05, denoting statistical significance (in bold).
Note: Kruskal-Wallis test was used to compare the proportion values for each N and P parameter within the categories (LOC overall, MBR overall, low flow in LOC, high flow in LOC, low flow in MBR and high flow in MBR).Mann-Whitney U test was used to compare the ratio values between Lockyer Creek and the mid-Brisbane River, as well as between low flow (L) and high flow (H) for each stream.The superscript letters following median values, arranged from highest to lowest with 'a' being the higher, represent significant differences between two treatments.Treatments that share the same letter for each comparison are not significantly different (Kruskal-Wallis test or t-test, p > 0.05).The sum of proportions for each N and P parameter may not total 100% as they are derived from medians.