Introduction
Droughts are often long-lasting phenomena, without a distinct start or end and with impacts cascading across sectors and systems, creating long-term legacies. Nevertheless, our current perceptions and management of droughts and their impacts are often event-based, which can limit the effective assessment of drought risks and reduction of drought impacts. Here, we advocate for changing this perspective and viewing drought as a hydrological–ecological–social continuum. We take a systems theory perspective and focus on how “memory” causes feedback and interactions between parts of the interconnected systems at different timescales.
We first discuss the characteristics of the drought continuum with a focus on the hydrological, ecological, and social systems separately, and then we study the system of systems. Our analysis is based on a review of the literature and a study of five cases: Chile, the Colorado River basin in the USA, northeast Brazil, Kenya, and the Rhine River basin in northwest Europe. We find that the memories of past dry and wet periods, carried by both bio-physical (e.g. groundwater, vegetation) and social systems (e.g. people, governance), influence how future drought risk manifests. We identify four archetypes of drought dynamics: impact and recovery, slow resilience building, gradual collapse, and high resilience–big shock. The interactions between the hydrological, ecological, and social systems result in systems shifting between these types, which plays out differently in the five case studies.
We call for more research on drought preconditions and recovery in different systems, on dynamics cascading between systems and triggering system changes, and on dynamic vulnerability and maladaptation. Additionally, we advocate for more continuous monitoring of drought hazards and impacts, modelling tools that better incorporate memories and adaptation responses, and management strategies that increase societal and institutional memory. This will help us to better deal with the complex hydrological–ecological–social drought continuum and identify effective pathways to adaptation and mitigation.
Conceptual framework: Viewing drought as a system of systems
In this paper we build on the concepts of complex systems and systems thinking to conceptualise drought as a hydrological–ecological–social system and to draw on elements from social–ecological systems, socio-hydrology, and earth system science. Our specific focus is the dynamic aspects of these systems interacting over time as they are affected by and create system memory.
The field of systems thinking defines complex systems as composed of a set of elements (which can be systems themselves) that have connections between each other (Jackson, 2019; Shaked and Schechter, 2017). The interactions between these interconnected elements can lead to unexpected emergent results (Westra and Zscheischler, 2023). Elements can interact and feedback at different scales, creating a multidimensional complex adaptive system (Rammel et al., 2007). Systems theory is applied to, for example, agriculture, natural resource management (Ison et al., 1997), and disaster recovery (Bahmani and Zhang, 2021).
Social–ecological systems (SESs) are examples of complex adaptive systems characterised by integrated bio-physical and socio-cultural processes (Ahmed and Abdalla, 2005; Delgado-Serrano et al., 2015; Ostrom, 2009; Tellman et al., 2018). Socio-hydrology or hydrosocial systems can be seen as a specific type of SES revolving around the interactions between people and water (Konar et al., 2019; Sivapalan et al., 2012; Wesselink et al., 2017). Many studies, for example, use socio-hydrology to understand and model the complex dynamics of flood risk resulting from the interplay between floods and people (Di Baldassarre et al., 2013; Vanelli et al., 2022).
Earth system science (ESS) focuses on the complex adaptive components of the earth system and their interactions (Steffen et al., 2020). ESS is strongly based in the natural sciences (meteorology, climate physics, environmental science) but has more recently recognised the important role of humans as agents of change of the earth system (Alessa and Chapin, 2008). One difference between SESs and ESS is the scale at which they are studied, with ESS focusing on the planetary scale (Steffen et al., 2020).
Within complex social–ecological or earth systems, the interactions between the elements or subsystems happen across both spatial and temporal scales (Konar et al., 2019; Vanelli et al., 2022). In this paper, we are interested in temporal aspects. Naylor et al. (2020) state that to understand complex systems and their emergent properties, it is necessary to examine the changes in relationships between system elements over time.
The concept of time is studied extensively in the separate systems – the hydrological system (Koutsoyiannis, 2013), ecosystem (Jackson et al., 2021), and social system (Peixoto and Rosvall, 2017) – despite common features between them. Aspects like antecedent conditions, response times to disturbances, and recovery to the original state (or transition to a new state) jointly shape the response of a system to external drivers. These factors determine whether the system changes quickly or slowly, depending on the system’s memory.
The memory of a system refers to its ability to retain information about past states, conditions, and experiences, which influences its current behaviour and response to future events. Memory is often manifested through legacies and responses. Legacies are the lasting effects of past conditions that might continue to influence the system’s structure, function, and behaviour over time. Responses are the manifestation of how quickly the system reacts to disturbances and adapts to changes. A system with a long memory retains past influences for a longer period, leading to slower responses and longer legacies, while a system with a short memory quickly responds to disturbance and has short legacies (Gunderson and Holling, 2002; Kchouk et al., 2023; Redman and Kinzig, 2003).
The memory of the systems within a complex system strongly determines the emerging properties, such as (i) self-organisation and emergence, (ii) non-linear behaviour and tipping points, (iii) state shifts and feedback loops, and (iv) resilience and adaptation (Carmichael and Hadžikadić, 2019; Preiser et al., 2018). Such properties are particularly evident when examining the co-evolution of human and water systems across time. For example, Srinivasan et al. (2012) introduced the concept of “syndromes” to conceptualise and describe the evolving nature of human–water interactions over time. These syndromes represent specific patterns of water use, reflecting the dynamic state of the system as it changes and adapts with time. Similarly, Roobavannan et al. (2017) modelled a “pendulum swing” in the management of the Murrumbidgee Basin in Australia, which is in fact a shift from agricultural to environmental water allocation. This shift reflects the “memory properties” of systems as it was shaped by accumulated experiences, past policies, and societal values, showing how historical experiences influence current practices.
Time is an important element in the development of drought and drought impacts, as recognised by previous studies (Hall and Leng, 2019; Tijdeman et al., 2022; Wilhite and Glantz, 1985; WMO, 2021), and time characteristics have been studied empirically in the separate systems (see some examples in Table 1). In the next sections, we explore and discuss the concept of memory shaping drought over time from different perspectives: hydrology, ecology, and social science. Next, we analyse potential temporal interactions across these systems to understand how they impact the broader drought system across time.
Table 1. Examples of drought as a continuum in hydrological systems, ecosystems, and social systems based on specific studies.
| Hydrological systems | Ecosystems | Social systems |
|---|---|---|
| – Catchment memory modulating drought severity, duration, and recovery (Alvarez-Garreton et al., 2021) | – Drought legacies affecting ecosystem structure, function, and resilience (Kannenberg et al., 2020) | – Dynamic vulnerability to drought influenced by past experiences and adaptation (de Ruiter and van Loon, 2022) |
| – Groundwater depletion leading to persistent low-flow states (Peterson et al., 2021) | – Compounding effects of consecutive droughts on vegetation and soil processes (de Vries et al., 2018) | – Perception of drought risk and management shaped by institutional memory (Nohrstedt, 2022) |
| – Shift from snow- to rainfall-dominated hydrological regimes altering drought response (Arheimer and Lindström, 2015) | – Drought-induced shifts in plant community composition and ecosystem functioning (Crausbay et al., 2020) | – Maladaptation through over-reliance on water storage infrastructure (Di Baldassarre et al., 2018) |
The drought continuum in hydrological, ecological, and social systems
Hydrological systems
The emphasis on drought as a hydrological extreme event has led to drought detection and definition using indices specified over defined timescales (McKee et al., 1993; Mishra and Singh, 2010) or considering a limited range of lagged hydro-meteorological variables (Mishra and Singh, 2011). However, it is increasingly recognised that hydrological droughts result from complex interactions between multiple bio-physical processes and human influences (Van Loon et al., 2016). This implies that hydrological droughts occur not as singular events but rather as a result of the continuous evolution of multiple hydrological fluxes and states. Therefore, we cannot fully characterise droughts without considering the (wet and dry) hydrological conditions that either precede or follow what is considered a drought event, as well as how baseline conditions may be shifting over time due to climate change.
The duration for which these hydrological conditions need to be considered to understand the evolution of drought and subsequent recovery primarily depends on the processes that contribute to catchment memory (Stoelzle et al., 2020). Catchment memory, in the context of drought, modulates the cumulative effects of anomalous meteorological and hydrological conditions and their persistence over time and thus the severity, duration, and recovery of droughts (Alvarez-Garreton et al., 2021). This memory depends on the heterogeneous and spatially distributed characteristics of the catchment, such as topography, land cover, soil types, storage properties, and variability in hydro-climatic conditions (Cranko Page et al., 2023; Fowler et al., 2020; De Lavenne et al., 2022).
For instance, catchment memory in surface-water-dominated catchments may be quite short, depending on soil moisture and vegetation memory (Ghajarnia et al., 2020; Gu et al., 2023; Fig. 1a, dark-blue line). By contrast, in groundwater-dominated catchments, catchment memory may typically be longer, as groundwater acts as a storage reservoir that buffers short-lived rainfall anomalies and sustains baseflow in rivers and streams (Sutanto and Van Lanen, 2022). Such a long memory will, however, lead to slower recovery, particularly if groundwater levels have been significantly depleted due to more persistent rainfall deficits (Fig. 1b, light-blue line). This was found during the 2018–2022 drought in groundwater-dominated systems in the eastern part of the Netherlands, which showed minimal or no recovery despite the drought being interspersed with relatively wet conditions in the winter of 2019–2020 (Brakkee et al., 2022; see the Rhine River basin case study, Sect. S5).
High meteorological variability can dissipate catchment memory given rapid and frequent changes in hydrological states, especially in systems with shallow groundwater tables where excess water cannot be stored and the system is reset during wet periods (Appels et al., 2017; van der Velde et al., 2009). However, large subsurface storage with deep groundwater levels tends to attenuate the effects of variability in precipitation and evapotranspiration on the hydrological system. This then contributes to the accumulation of drought deficits and the lagging and pooling of meteorological drought events, thus extending the recovery process (Sutanto and Van Lanen, 2022). Other forms of storage can also contribute to long catchment memory, such as extensive wetlands and lakes (Gu et al., 2023). Furthermore, human-made storage reservoirs can increase catchment memory and buffer drought (Ribeiro Neto et al., 2022), though only up to certain critical thresholds such as when the reservoir falls empty (Rangecroft et al., 2019; Fig. 1a, orange line). Recovery in such reservoir-influenced catchments may be slower (Margariti et al., 2019; see also the northeast Brazil case study, Sect. S3).
Catchment memory varies across different climate types. In arid and semi-arid climates, propagation from meteorological to hydrological drought may be slower than in wet–tropical climates (Gevaert et al., 2018; Odongo et al., 2023). This may be exacerbated by land–atmosphere interactions, which can lead to the self-propagation of droughts, thus extending them in space or time (Miralles et al., 2019; Schumacher et al., 2022). Catchment memory also varies in climates with distinct seasonality, such as tropical savannas, snow-dominated catchments, or Mediterranean-type climates (Gevaert et al., 2018; Seager et al., 2019) where drought propagation has a strong intra-seasonal (De Lavenne et al., 2022) or even multiannual timescale (Gevaert et al., 2018). For example, in the Andes Cordillera, snow deficits lead to streamflow deficits not only during the summer melting season but also in the following autumn season (Alvarez-Garreton et al., 2021; see the Chile case study, Sect. S1). Similarly, winter snow droughts in the snow-dominated catchments of the Alps affect summer discharges of the river Rhine (Ionita and Nagavciuc, 2020; Khanal et al., 2019), while in the winter of 2022–2023, unprecedented dry and warmer-than-normal conditions over the Italian Alps caused critical hydrological conditions in the Po and Adige rivers in the ensuing spring (Colombo et al., 2023). Another example is the Mediterranean region, where precipitation is highly seasonal due to winter storms. The weakening of the storm systems combined with long–dry summers leads to a precipitation deficit and, thus, increased drought risk in the region (Cook et al., 2014; Ionita and Nagavciuc, 2021).
Catchment memory can, therefore, connect climate and hydrological anomalies across different temporal scales. Precipitation anomalies occurring at a particular time of the year can be compounded and lead to long-memory streamflow anomalies later in the year (Mudelsee, 2007). Figure 1 schematically shows how the superposition of different drought signals and hydrological states with long and short memory may result in either amplifying or dampening the duration and severity of hydrological droughts. However, the interaction of these signals is not always linear, as witnessed by the unexpectedly quick recovery in groundwater systems in Germany (Tijdeman et al., 2022: see the Rhine River basin case study, Sect. S5).
It is worth noting that the processes that constitute catchment memory are not stationary. Unprecedented climatic conditions such as multi-year droughts may alter how a catchment responds to precipitation and/or how surface and groundwater systems interact (Fuchs et al., 2019; Fig. 1b, dashed line). This can lead to persistent shifts in rainfall–runoff relationships (Eltahir and Yeh, 1999; Kleine et al., 2021) and less runoff than expected (Alvarez-Garreton et al., 2021; Fowler et al., 2020; Saft et al., 2015). Further, catchments may not always fully recover and return to their original states after protracted droughts end, leading to new persistently low-flow states due to changes in the dominant hydrological processes and catchment memory (Peterson et al., 2021; Fig. 1b, yellow line).
In addition to the effects of protracted dry conditions, climate change can also lead to non-stationary catchment responses through aridification as a result of greater atmospheric water demand, increased evaporation, and lower soil moisture (Boisier et al., 2018; Overpeck and Udall, 2020), as experienced in the Colorado River basin in the southwestern USA (Colorado case study, Sect. S2). Climate warming may lead to a modified drought response by shifting the hydrological regime of a basin from snow-dominated to rainfall-dominated. A shift from snow to rain in winter may reduce catch
