**4. System dynamics and its application to water management**

The water crisis is not an exception to the crisis of perception stated in Section 2. In this context, system dynamics represents an interesting method, as argued next. The approach of system dynamics was proposed by Forrester in 1956 for industrial applications. Forrester combined applied control theory, information theory, decision theory, and other relevant theories and methods relevant to this purpose [40] and argued that an event-oriented vision and a lineal-causal thought are not useful to address complex problems [41]. His books Industrial Dynamics, Urban Dynamics, and Principles of Systems were the starting point of system dynamics.

Some years later, Forrester and the Club of Rome, headed by Aurelio Peccei, collaborated to bring the system dynamics approach to a global scale and to a time horizon of more than 100 years. In effect, in Meadows et al. [42] there is an assessment of which would be the behavior of flow and storage of natural resources in the future and describe how they may mean limits to growth. The title of this report, Limits to Growth, became the motto of the global environmental movement [43]. Likewise, Forrester published individually his book World Dynamics. Later, Forrester and his collaborators developed a programming language called DYNAMO which made the Limits to Growth model possible. The textbook of Richardson and Paugh published in 1981 is based on this language.

Next, the ideas of system dynamics were brought to corporate and citizen practice in the book The Fifth Discipline [44]. The five disciplines described are personal mastery, mental models, shared vision, team learning, and systems thinking. This book created a new era in business and management because it inspiringly presented a framework for the learning organization [45]. During this decade, many companies, consultations firms, and governmental organizations began to use system dynamics to address critical issues [46]. In effect, Sterman published his book Business Dynamics in 2000, which collects concepts, techniques, and tools for dynamic simulation. Eventually, these developments permeated in IWRM, as described next.

Currently, the four interdisciplinary approximations to water management are IWRM, Water-Energy-Food Nexus (WEF Nexus), Nature-Based Solutions (NbS), and socio-hydrology [47]. In an extensive bibliometric and network analysis, the authors identified the main academic communities dedicated to studying the WEF nexus in the urban environment. There is no doubt that the current paradigm of water management is interdisciplinary. Given the complexity of water problems, the adequate approach to address them is a flexible one, that includes "hard" and "soft" approximations, that is, that it considers the construction of hydraulic infrastructures like dams, aqueducts, pipelines, and centralized treatment plants but at the same

time, looks to increase water use productivity using economic tools and public policy [48]. For that matter, assessment approaches are crucial.

Assessment approaches are classified into four categories: water urban systems, which encompasses integrated modeling of water systems; Urban Metabolism, which reunites territorial and mass balance; consumption methods like Life Cycle Analysis, water footprint, and input-output analysis; and complex systems (ecological network analysis and system dynamics) [49]. Similarly, in Hassan and Garg [48], there is a description of the different modeling techniques that can boost the flexible approach and conclude that system dynamics allow to model it while in Ghalehkhondabi et al. [50] there is a review on the soft computing techniques that allow to include nonlinearity, stochasticity, and imprecision of water demand data and mentions that the potential of these techniques and especially that of system dynamics has not been availed.

The history of dynamic simulation for water management has been classified previously in Winz et al. [51]. For stages are distinguished: limitation in adaptation, during the 60s and 70s decade; the surge of integrated water resources management, at the end of the 80s decade; the 90s decade; and the recent developments, from 2000 onwards. Next, the advances in the application of dynamic modeling to water management are described through their historical development.

A few years after Forrester's developments, the first application of system dynamics to hydrology can be found by Crawford and Linsley in 1966. In effect, in Crawford and Burges [52] it is documented that the model was unique in its times because it reunited knowledge and theory of the different hydrologic processes but considered their interdependences and incorporate them in a model whose objective was to describe hydrologic balance continuously on time. In Winz et al. [51] it is mentioned that later Hamilton (1969) developed a model that besides flow and storage, included socioeconomic factors that looked to explain the systemic causes of the crisis in the American watershed of Susquehanna river. The perspective of these authors preceded the surge of integrated water resources management concept, so in their time it was visionary.

However, during the 70s decade, there was a limit in the adaptation of these models, attributable to two factors pointed out by Winz et al. [51]: the indifference for complexity and uncertainty in the daily operation of water systems and negative publicity of the findings of Forrester and Meadows. The adoption of the concept of IWRM and its inherent interdisciplinary feature, along with the soaring of computing power and the increase in the complexity of water problems constitute the paradigm of complexity in water management [8], under which proliferated a high quantity of studies using system dynamics approach. There have been several review articles published that will be discussed further (see [45, 46, 53–56]).

Dynamic simulation aids our capacity to make predictions of future states and to improve our understanding of the problem as a necessary step towards affecting sustainable and effective change [51]. The former activity is critical to inform public policy or projects related to water management but also to involve relevant stakeholders (including, of course, policy-makers and authorities that materialize these policies and projects) because it allows describing the behavior of the system under different scenarios.

Dynamic simulations models have been classified previously. The first classification of dynamic simulation models for water management is reported in Winz et al. [51] and it was realized based on the problem that they address: regional analysis and river basin planning, urban water, flooding, irrigation, and pure process models. A second

## *The Usefulness of System Dynamics for Groundwater Management DOI: http://dx.doi.org/10.5772/intechopen.105162*

classification, proposed in Chen and Wei [40] divides the models into three categories: flooding control and disaster mitigation, water resources security, and water environment security. The first one encompasses management and forecast of floodings; the second reunites water resources management, water resources carrying capacity, water resources planning, and sustainability; the third one consists of pollution control, water quality management, pollution early warning systems, and water ecology.

However, the classification that has been complemented further in literature can be found in Mirchi et al. [41], which is based on the model's objective. This way, models are classified as predictive simulation models, descriptive integrated models, and participatory and shared vision models. Combined system dynamics models can be added to this classification [57]. In the first category, the processes governing particular subsystems within a broader water resources system are simulated with a quantitative approach to defining tactics, while in the second, the approach is holistic and strive to identify and characterize the main feedback loops among two or more disparate subsystems and finally, the third involves stakeholders and decision-makers to foster the understanding of the system [41]. In Zomorodian et al. [58], the three kinds of descriptive integrated models are reported: hydro-economic, socio-hydrologic, and integrated water resources management models.

Previously, the modeling process in the classic literature of system dynamics was reported in Luna-Reyes and Andersen [59]. A comprehensive explanation of the methodology for dynamic simulation in the context of water management can be found in Simonovic [8]. In **Table 1**, there is a summary of the methodology for dynamic simulation in the context of water management and it is possible to appreciate that essentially, there are six processes to be conducted. Firstly, the problem articulation shapes the entire modeling [60] because involves identifying the key variables and their past and possible future behavior, as well as defining the time horizon [46].


#### **Table 1.**

*Processes for dynamic simulation in water management.*

Next, there is conceptual modeling where qualitative tools can be used, such as causal relationships, where positive or negative relationships between the variables are identified; causal loop diagrams (CLD) and feedback loops; stock and flow diagrams (SFD); reference modes [4] and system archetypes [61]. To accomplish a successful system dynamics application, extensive computer simulations should be performed only after a clear picture of the integrated water resources system has been established through reasonably simplified conceptual models [61]. A result of this preliminary sketch is the development of developing a dynamic hypothesis, which plays a significant role in complexity reduction [60] and describes the problematic behavior of the system [46]. This process is appealing to involve stakeholders to identify the key variables that affect the system, as well as to propose and refine the dynamic hypothesis that describe the problematic behavior. It is important to note that relevant stakeholders are to be mapped for particular local conditions.

Following, the simulation model is formulated. The simulation model is the refinement and closure of a set of dynamic hypotheses to an explicit set of mathematical relationships, where the experimentation process is iterative and flexible [8]. It is important to underscore that the connection between qualitative and quantitative modeling is the use of the dynamic hypothesis and how it can be modified and adjusted, according to the findings made during the simulation process. The mathematical equations used in the simulation models describe detail complexity, and frequently the modeler will be tempted to include more math into the model to increase accuracy. However, it must be considered that, at the strategic level, emphasis should be placed on trend identification and pattern recognition rather than exact quantitative predictions of dynamic variables [61]. This claim leads to the following step of the modeling process.

It is worth recalling that a model is useful when addresses the right problem at the right scale and scope, and represents the system response correctly [51]. In that sense, for models to be useful as decision support tools for water resources planning and management, it is necessary to verify the model structure to ensure that mathematical equations and interrelationships between subsystems follow logical explanations and are not spurious or erroneous [61]. One method to do this is through calibration, which is concerned with adjusting the model variables in such a way that the model outputs can fit the collected data [60]; however, due to limited data or lack of appropriate methods to quantify particular subsystems [61] often makes this process not possible, which gives way to the concept of model validation [3]. This process is conducted using structural tests, sensitivity analysis [60], behavior tests (which encompasses calibration and comparison with reference modes), and implications tests [51].

Finally, depending on the models objective it can either be used for policy formulation and evaluation, where it is investigated how specific change in a parameter in a system dynamics model affects the system's response [60], or theory building, where the outcome is a theory that is a structured, explanatory, abstract and coherent set of statement about a partial reality [62]. In addition, in Randers [63] it is argued that besides some tests, implementation deals with the identification of potential users, translation of study insights to an accessible form, and diffusion of study insights. This can be done through the building of a decision support system (DSS) (see for example, Martínez-Valderrama et al. [53]), the organization of technology transfer workshops with decision-makers, and/or divulgation activities with the community.

Dynamic models are useful because they are multidisciplinary, cross-scalar, modular object-oriented, transparent, adaptable; in addition, they support a variety

## *The Usefulness of System Dynamics for Groundwater Management DOI: http://dx.doi.org/10.5772/intechopen.105162*

of goals, allow users to quickly become familiar, and have methods for consensus building and team learning [51]. When the complexity of water management issues is considered, these are valuable features. In effect, water management problems are multiscalar because besides involving the biophysical borders of watersheds, they happen in political-administrative divisions. Also, they can be described using different subsystems according to water users, as described in the next section. Finally, they require the involvement of different stakeholders: several authorities, local communities, cities, and different kinds of users and each one of them has a distinct mental model of what are the root causes of problems, as well as their approach to solutions. These arguments support why system dynamics are useful for water management. Nevertheless, there are several opportunities for system dynamics application to water management, and they have been identified by the authors that have undertaken reviews of the literature on the topic (see [45, 46, 53–56]).

In Chen and Wei [40], a synthesis of the insights of 35 studies is reported. Firstly, they indicate the necessity of integrating studies of different water subsystems: flooding control, disaster mitigation, security in water supply, and environmental supply, for example. Secondly, to consider climate change and the intensification of human activities. Thirdly, to advance in uncertainty theories that limit applications of system dynamics. Finally, to integrate system dynamics with other methods like geographic information systems (GIS), remote sensing (RS), hydrologic models, and water quality to broaden their spatial scope.

In Zarghami et al. [60], a revision focused on applications of dynamic simulation to water supply and distribution networks was made. The authors identify 32 records oriented towards water supply and four to sewage and conclude that this approach is useful to be applied in water distribution networks because their performance indicators have a dynamic behavior and are subject to internal and external variables of the system. However, they suggest that they must be complemented to reinforce their capacities, with reliability methods, for example.

After the revision and synthesis reported in Zomorodian et al. [58], the authors coincide with the ideas in Chen and Wei [40] and Mashaly and Fernald [57]: it is necessary to integrate water subsystems to have a better perspective of the behavior of the entire system and to combine system dynamics with other methods and tools to strengthen the modeling framework. A relevant contribution of these authors is the identification of the necessity of adding qualitative variables to modeling, especially in the moment of incorporating social and political processes to modeling.

Recently, in Mashaly and Fernald [57] it was reported that since 1996, 199 articles have been published using system dynamics methods to solve agricultural water management problems. Besides, they found that 64% are integrated descriptive models, 16% are predictive simulation models, 8% are shared vision models, and 12% are combined. The authors advocate for preventing the creation of simplistic models through the involvement of specialists that have worked with the system and the use of validation and/or calibration methods.

Finally, in Cerecedo Arroyo and Martínez Austria [3] the first systematic review of dynamic modeling reported in the literature can be found. It is distinguished from previous work because of its transparency in search protocols. The authors reported 27 studies applied to watersheds and aquifers and indicate that there are relevant opportunities for the validation of these models as mentioned in Mashaly and Fernald [57], as well as the incorporation of water quality aspects.
