The selected cases studies are the Lachlan River catchment (Case Study 1) and the Hawkesbury–Nepean River catchment (Case Study 2). The catchments share a boundary along the Great Dividing Range, with the Lachlan River draining inland to the west, and the Hawkesbury–Nepean River draining east to the Pacific Ocean (Figure 1).

The Lachlan River catchment (the Lachlan) is within the Murray–Darling Basin, which is sparsely populated (ca. 2 million people in 1.06 million km²), accounts for approximately 40% of agricultural production in Australia, and crosses the borders of five states and territories (MDBA, 2010). Major water reforms over the last 30 years culminated in the implementation of the 2012 Murray–Darling Basin Plan, intended to return water over-allocated for irrigation back to the environment (Colloff and Pittock, 2022). The major water storage in the Lachlan is Wyangala Dam, constructed in 1935 and expanded in 1971 to a capacity of 1,220 GL (WaterNSW, 2023). The headwaters of the Lachlan River are in the mountainous eastern parts of the catchment where average rainfall is the highest, with climate conditions becoming increasingly arid downstream, where evaporation generally exceeds rainfall throughout the year (Bureau of Meteorology, 2024a). Wyangala Dam supplies water primarily for irrigation, but other downstream uses are town water supply, stock and domestic, mining and major floodplain wetlands, which provide habitat for a number of threatened species (NSW Department of Planning and Environment, 2020). Water management in the Lachlan is the responsibility of the Murray–Darling Basin Authority and the Commonwealth Environmental Water Holder (for environmental water management) at the federal level, WaterNSW and the NSW Department of Department of Climate Change, Energy, the Environment and Water at the state level, and urban water supply is the responsibility of local government.

The Hawkesbury–Nepean catchment (the Hawkesbury) is on the east coast of New South Wales, and includes Australia’s largest city, Sydney, which has a population of approximately five million (Australian Bureau of Statistics, 2022). The primary water storage is Warragamba Dam, completed in 1960 with a storage capacity of 2,065 GL. The highest average annual rainfall in the catchment is on the coast (1,500 mm) and decreases towards the west where Warragamba Dam is located. The primary use of water from Warragamba Dam is for urban water supply, contributing up to 80% of Sydney’s water needs (Sydney Water, 2024). Operation of the dam is the responsibility of WaterNSW at the state level and Sydney Water, a state-owned corporation, is responsible for water distribution to consumers. Downstream of the dam, considerable floodplain development has occurred, and the floodplain of the Hawkesbury has the highest flood exposure in Australia (Infrastructure NSW, 2019).

In this study we hypothesise the dominance of irrigated agriculture or urbanisation, and the different climate characteristics and catchment water balance, will contribute to the emergence of different sociohydrological phenomena. The Lachlan was selected to examine the role of irrigated agriculture and water scarcity in shaping water management challenges. The Hawkesbury was chosen to investigate urban water use behaviours in response to drought, and how relatively abundant water availability and low competition and contestation affects management decisions. Additional considerations in case study selection was their proximity (Figure 1), similar administrative bodies, and the presence of a large dam constructed in the last century.

We combined quantitative modelling using a system dynamics model with qualitative analysis of the historical hydroclimate, water management decisions, and socioeconomic context to explore the presence of emergent phenomena. Qualitative data was derived from scientific articles, reports by government agencies, water authorities, local governments, media reports and online sources such as the Bureau of Meteorology website. This mixed methods case study approach (Harwell, 2011) is intended to generate generalizable understandings about specific human-water system features which drive specific emergent phenomena (Mostert, 2018).

The system dynamic model in this study contains five modelling sub-routines (Figure 2), namely:

Sources of data for model inputs and observations used to test model performance are detailed in Supplementary Table S2. The equations for the dam, drought, flood, and population systems are applied as by Mazzoleni et al. (2021) and are provided in the Supplementary Table S3 together with initial conditions and parameter values (Supplementary Tables S4, S5).

The main driving equation in the model is the dam storage water balance equation, which influences all other systems (Equation 1):

Where: V t is storage volume at time t, Q t I Δt is the monthly total inflow to the dam, Q t E Δt is the total monthly required environmental flow volume, Q t Bf Δt is the minimum monthly outflow required to maintain river flow, Q t S Δt are spills or controlled releases from the dam to maintain flood storage capacity, as required for dam safety, and W t Δt is the total of monthly withdrawals from the dam for consumptive purposes, including irrigation (Mazzoleni et al., 2021).

The irrigation component was newly developed for this study and affects the dam system by determining the volume of irrigation withdrawal, and the population system by limiting population growth when drought awareness is high. It is affected by dam storage volumes and actual dam withdrawals, simulating water shortages (when withdrawals cannot meet demand) and reductions in consumption when water availability is low, approximating the operational water sharing rules.

Irrigation water demands are calculated annually to reflect annual decision making on plantings and are based on the demand function of Gonzales and Ajami (2017), which captures rapid recovery of water demand after periods of drought. The maximum possible demand is limited by the dam operational full supply volume (OFSV) and the coefficient ( c Dmax ) representing the proportion of storage assumed to be accessible for irrigation on a monthly basis (Equation 2).

Modelled irrigation demand ( D t I ) (Equation 3) declines towards the minimum observed demand ( D min I ) if drought awareness is increased over the past year, and recovers if drought awareness has decreased. The rate at which demand recovers or declines is proportional to the drought awareness ( A t DI ), and the sensitivity of demand to drought memory ( α DI ), decaying over time at rate which reflects increases in water use efficiency ( β I ).

Irrigators experience drought losses ( L t DI ) when demands exceed withdrawals (Equation 4).

Irrigators also experience profitability losses when demand is reduced in response to low water availability ( L t P ). This approximates water sharing rules which reduce water available for irrigation as a priority when dam storage is low. Accordingly, irrigators make decisions about on-farm water use to manage reduced allocations. In the model, irrigation losses related to reduced profitability during periods of low water availability are calculated when dam storage volumes are below a threshold ( τ I ) (Equation 5).

Profitability losses are a reliable drought indicator, and can therefore be understood to raise drought awareness (Hughes et al., 2022). Drought awareness of irrigators ( A t DI ) is computed by considering the maximum of water shortage and water demand losses (Equation 6).

The model was parameterised to fit observation data where available or to reflect operational rules and physical characteristics of each case study. Parameter values were restricted to the range tested in the sensitivity analysis conducted by Mazzoleni et al. (2021). Model parameterisation values and methods are provided in more detail in the Supplementary Table S5.

A sensitivity analysis of key model variables to parameter values was performed to test the model robustness to parameter assumptions (details are included in Supplementary material). The sensitivity analysis showed that all variables of interest but particularly drought losses are sensitive to parameters which influence irrigation water demands. Where irrigation water demands are absent, drought losses are instead most sensitive to the fractional efficiency adoption rate (the rate at which water demands can be reduced).

Model scenarios were designed to test the effects of major water management interventions in each case study. In Case Study 1, Lachlan Catchment, the goal is to test the role of the dam enlargement for enabling the growth of irrigated agriculture. Therefore, the following scenarios were modelled: Baseline, which approximates the real historical condition, including the construction of Wyangala Dam in 1935, the raise in the height of the dam in 1971, and where translucency flow rules are introduced in 1998; and No dam raise (NDR), whereby the dam is not raised in 1971 and remains at its initial capacity of 375 GL.

In Case Study 2, Hawkesbury–Nepean Catchment, the goal is to assess the effectiveness of public drought awareness programs and water restrictions in encouraging lower per-capita water use during droughts. Therefore, the following scenarios were modelled: Baseline, which approximates the current operating rules, which do not allow for dam operation for flood mitigation and include water restrictions during droughts; and Water exploitation (WE), where drought awareness is only raised when communities experience water shortages.

In the Lachlan catchment, major sociohydrological changes are related to the prioritisation of water resources for irrigation agriculture purposes, which has driven the economic development of the region. Following colonial settlement in the region in the 1830s, river regulation and the establishment of European-style agriculture led to hydrological change and widespread environmental impacts (Leblanc et al., 2012). Agricultural expansion was possible only through the expropriation of land and water resources from First Nations peoples, the legacy of which continues to shape water justice issues in the Basin (Hartwig et al., 2020). Water reforms are typically triggered by water crises (Pittock, 2019), and have included the establishment of water markets, the capping of extractions, the re-allocation of water from irrigation to the environment through government ‘buy backs’ of water entitlements, and the introduction of environmental flow requirements.

In the Hawkesbury–Nepean Catchment, changes in water management have been driven by the need to provide secure and safe water to the population of Sydney (Davies and Wright, 2014). Floodplain development has occurred downstream of Warragamba Dam, exposing approximately 140,000 people to flood waters (NSW Government, 2021). Major water management decisions have been the construction of Warragamba Dam to support ongoing population growth, and the enforcement of water restrictions during drought periods to reduce per-capita water demands. More recently, the potential for the dam to be operated for flood mitigation has been investigated (e.g., Devanand et al., 2023).

A timeline of major changes in socio-economics, hydroclimate, and water management is presented for each case study (Figure 3). A more detailed review is provided in the Supplementary Table S1.

For Case Study 1, Lachlan Catchment, the modelled storage volume of Wyangala Dam shows a good correlation with historical observations (correlation coefficient: 0.88), and captures significant drought-related drawdown periods in the early 1980s, the Millennium Drought (2000–2010) and the severe Tinderbox Drought (2017–2019) (Figure 4a). Modelled irrigation water demand approximates longer-term trends but does not reproduce the drop in demand observed in 1990 (Figure 4b). The comparison of modelled to observed growth in irrigation demand between 1987 and 2000 supports the model assumption that growth in irrigation demand is related to the catchment water storage capacity. Reductions in demand for irrigation water in wet years (1990, 2016 and 2021) are not captured by the model, which does not simulate soil moisture or on-farm water storage. The effect of the Millennium Drought on population growth is captured, although the model does not accurately simulate the steep growth in population after approximately 1980, and instead simulates a slower but more steady population increase (Figure 4c). The model generally captures flood events prior to the Millennium Drought, with some discrepancies in flood flow volume, but the model does not capture the drought-breaking flood events of 2010–2012 and 2016 (Figure 4d). Overall convergence of model results with observations indicates the model structure and assumptions realistically capture aspects of the dynamic relationships between dam operations, drought awareness, and irrigation water use.

Model results for Case Study 2, Hawkesbury–Nepean Catchment, indicate the model effectively represents trends in dam storage volumes (correlation coefficient: 0.74) (Figure 5a). Trends in population growth are captured by the model, except for the effects of the COVID-19 pandemic in 2020 (Figure 5b). Modelled per-capita water demands reproduce the reductions associated with water restrictions in the 1990s and during the Millennium Drought (Figure 5c). Modelled total water demands reflect the balance between per-capita water demand changes and population growth (Figure 5d). There is limited data available for water consumption trends prior to 1990, so these are verified by convergence of dam storage volumes with observations. Where data is available, the model overestimates water consumption prior to 1990. Downstream flooding regime is adequately represented, with critical flood events captured, though with some discrepancies in flood water level (Figure 5e).

The Nash-Sutcliffe Efficiency (NSE) was calculated to provide a quantitative comparison between modelled and observed values (Table 2) for variables with relevant observational data. NSE is a measure of deviation of modelled results from observations, considering the variability of observations from the mean. The calculated NSE shows that the model adequately approximates observations, particularly reservoir volumes, population and flood water level.

In this study we identified emergent sociohydrological phenomena in two catchments that can be attributed to the distinct characteristics of each sociohydrological system (summarised in Table 3).

The levee effect is apparent in the Hawkesbury, where major floodplain development has occurred due to pressures on space for residential development and urbanisation. Options of water managers to reduce flood exposure are constrained, as the costs of relocating people are prohibitive (Infrastructure NSW, 2019). Water managers have more options to deal with the levee effect in the Lachlan, where space is less constrained, population sparser and costs of relocating infrastructure are lower. The extent of floodplain development in the Lachlan in relation to flood risk is not well documented. Further research could explore the relationship between flood risk and irrigated agriculture.

In the Hawkesbury, adaptive management of Warragamba Dam to manage both droughts and floods is constrained by the need to secure urban water supply. Therefore, sequence effects and lock-in behaviour are consequences of the reservoir effect. Strategies to diversify supply and reduce reliance on the dam for water supply create the opportunity for greater adaptive management. In the Lachlan, we identified the prospect of sequence effects via multiple pathways for which there is growing general understanding (Barendrecht et al., 2024). Droughts and floods tend to be managed separately in the Murray–Darling Basin and their interactions are typically not considered. The potential for sequence effects indicates water managers should consider interacting effects of their interventions and that floods and droughts be treated as two ends of a continuous hydrological spectrum rather than as separate, unrelated events (Ward et al., 2020).

Rebound effects are expected where there are strong economic incentives to consume more water. In the Lachlan, this is driven by water efficiency improvements in irrigated agriculture (Wheeler et al., 2020), and changes from annual to perennial crops with higher water requirements (Loch and Adamson, 2015). In the Hawkesbury, rebound effects have not been apparent historically, though may emerge following the introduction of new water supplies and efficiency improvements (Sharifi et al., 2024). Rebound effects are considered more likely in catchments with water-intensive industries, such as irrigated agriculture, and less likely in urban catchments.

Reservoir effects and lock-in behaviours were identified in both case studies, suggesting these are inherent risks in management of large dams. The driving characteristic of these phenomena is the presence of the dam itself. In the Hawkesbury, specific human-water system features present options to address reservoir effects through diversification of water sources to reduce drought vulnerability. For example, Sydney has access to seawater for desalination, providing additional supply (Sydney Desalination Plant, 2025). High local rainfall and domestic rainwater harvesting can reduce demand pressures on Warragamba Dam (Sountharajah et al., 2017). Decentralisation of urban water systems also has potential to relieve lock-in effects (Rabaey et al., 2020), and options to operate the dam more dynamically may arise as flood forecasting improves (Devanand et al., 2023).

By comparison, options to address reservoir effects and lock-in behaviours in the Lachlan are constrained by low water availability, leaving managers few options to increase supply without effects to downstream users and wetlands. For example, a recent proposal to expand storage of Wyangala Dam was rejected due to potential adverse environmental impacts (NSW Minister for Water, 2023). Unlike the Hawkesbury, diversifying supply via decentralisation and private infrastructure presents management challenges to water modelling and accounting, which form the basis for decisions on water allocations to irrigators and between consumptive and environmental uses (Pittock et al., 2023). In this context, decentralisation of water sources is unlikely to be a viable option to relieve demand for scarce water resources, and reduce reservoir effects. This is further complicated by the emergence of supply–demand cycles and rebound effects, with the high dependency of irrigators on the operation of Wyangala Dam interpreted as both a product of, and a constraint on, water supply options in the catchment. This indicates that water management strategies aiming to reduce scarcity by increasing supply or water-use efficiency are likely to instead stimulate demand, and would have limited effects on improving water security. Instead, attention should focus on strategies aiming to control or reduce demand (Di Baldassarre et al., 2019). The recent rejection of the NSW government proposal to expand the storage capacity of Wyangala Dam by 650 GL because of a lack of return on investment and adverse downstream environmental impacts suggests a slight shift in approach in this direction (NSW Minister for Water, 2023). However, the failure of the pendulum swing at the policy level to produce real change at the level of implementation suggests more transformative changes to water management are needed.

Although reservoir effects and lock-in behaviours are inherent risks for large dams, they will be lessened in catchments with options to diversify water sources and where there is alignment of water management objectives among users. In catchments with high competition for scarce water resources and economic incentives to increase irrigation demand, reservoir effects, supply–demand cycles, lock-in behaviours and rebound effects become interlinked and mutually reinforcing. The extent to which demand can be managed by farm-scale adaptation to less water remains to be seen.

In the Lachlan, water managers are trying to satisfy multiple, often conflicting objectives. Interventions to address some phenomena could exacerbate effects of others. Competition for resources drives trade-offs among management objectives, which shift together with societal values of water and changing economic and environmental conditions, producing pendulum swing effects (van Emmerik et al., 2014). By comparison, water management objectives in the Hawkesbury appear to have high coherence, and management options to diversify supply have potential to reduce risks associated with other phenomena. We consider the pendulum swing effect is more likely in catchments with vulnerable downstream wetlands and high competition for water, as in the Lachlan.

Confidence in the findings of this study would be improved by inclusion of more case studies from a range of catchments. Contextual analysis and the development of model scenarios could have benefitted from input from water managers. Despite these shortcomings, we consider the model framework as sufficiently generalisable to be adapted and tailored to a range of contexts and issues water managers might be required to investigate.

Model representation of the safe-development paradox and sequence effects in the Lachlan is limited by gaps in knowledge of drought-flood interactions and how irrigation water demand responds to extreme wet conditions. Drought awareness in each case study was modified to represent different mechanisms by which it is developed, highlighting the complexity of drought awareness-raising processes. Relating awareness only to experience of water shortages is too simplistic. In Australia, media reporting increasingly covers drought crises (Wei et al., 2015), and the effects of drought in irrigation areas have flow-on effects to the wider community, for example though increases in food prices. Therefore, it is likely that droughts in the Murray–Darling Basin affect drought-aware behaviours in Australian cities. This situation raises the question of whether the catchment is the most relevant scale to address droughts and their impacts (Fischer et al., 2021).

Further, the model was able to endogenously represent changes in urban and irrigation water demand as a result of variable water availability. More significant changes in water management, for example reservoir storage expansion, were instead modelled using scenarios. In both catchments, recent droughts have prompted calls to increase reservoir storage which were ultimately dismissed on social and environmental grounds. This suggests that the management response to drought in both catchments has changed over time, where social and environmental factors now serve to counteract pressure to increase supply. These feedback mechanisms are not included in this model, highlighting the value of qualitative data to complement quantitative modelling when understanding emergent behaviour in human-water systems.

An issue that warrants attention is the role of colonial processes in sociohydrological phenomena in Australia. Colonial processes enabled expansion of irrigation in the Murray–Darling Basin and served to exclude First Nations Peoples from participation in agricultural livelihoods (Hartwig et al., 2022). Historical and continuing injustices over water rights and access has led to major power imbalances: those who influence and benefit most from use of water resources tend to drive the emergence of sociohydrological phenomena to the detriment of those with least social power (Savelli et al., 2021). Power dynamics and how they play out in the management of water resources is an important sociohydrological issue which would benefit from further research (Konar et al., 2019).

We compared two case studies to determine the characteristics of human-water systems that contribute to the development of emergent sociohydrological phenomena. The generalised model we present here can be used by water managers and decision-makers as a framework to consider and address behavioural responses to water policy and management interventions. This framework can be used to question current norms in water management practice, whether they are achieving objectives and explore options for change. A nuanced interpretation of the modelling requires a detailed contextual understanding of each catchment, which water managers are well-placed to contribute to and engage with. The identification of specific emergent phenomena and their attribution to the characteristics of the human-water system enables water managers to determine which phenomena are most important to consider, based on their knowledge and experience.