Towards Resilient Cities: Systematic Review of the Literature on the Use of AI to Optimize Water Harvesting and Mitigate Scarcity

Abstract

This article develops a systematic literature review with a focus on the optimization of water harvesting through the use of artificial intelligence (AI) applications. These are framed in the search for sustainable solutions to the growing problem of water scarcity in urban environments. The analysis is oriented towards urban resilience and smart water management, incorporating interdisciplinary approaches such as systems thinking to understand the complex dynamics involved in water governance. The results indicate a growing trend in the utilisation of AI in various domains, including demand forecasting, leak detection, and catchment infrastructure optimization. Additionally, the findings suggest its application in water resilience modelling and adaptive urban planning. The text goes on to examine the challenges associated with the integration of technology in urban contexts, including the critical aspects of governance and regulation of AI, water consumption, energy and carbon emissions from the use of this technology, as well as the regulation of water management in digital transformation scenarios. The study identifies the most representative patents that combat the problem, and in parallel proposes lines of research aimed at strengthening the water resilience and sustainability of cities. The strategic role of AI as a catalyst for innovation in the transition towards smarter, more integrated and adaptive water management systems is also highlighted.

1. Introduction

Water scarcity represents one of the most critical challenges in contemporary urban development, exacerbated by population growth, climate change, and aging infrastructure. Despite global efforts to improve water governance and sustainability, many cities continue to face severe limitations in supply, distribution, and resilience. In this context, artificial intelligence (AI) emerges as a promising tool to support advanced water resource management, from demand forecasting to catchment optimization and leak detection. However, there remains a significant gap in synthesizing how AI has been applied to address these challenges, particularly in urban settings. This article aims to fill that gap through a systematic literature review that explores the intersection between AI applications and strategies for enhancing urban water resilience.

Water scarcity in urban environments is one of the most pressing challenges of the 21st century. The exacerbation of the global water crisis has been attributed to several factors, including climate change, rapid urbanization, the obsolescence of water infrastructure and inadequate resource management. In this context, the search for sustainable, efficient and innovative solutions becomes imperative. In the domain of technology, artificial intelligence is regarded as one of the innovations with the most significant transformative potential.

Inappropriate management of water resources, sanitation systems, and catchment methods poses a significant threat to public health. The Ertop review [1] of the history of water is accurate and highlights how water harvesting, sanitation, and hygiene can help us understand the challenges and opportunities presented by these issues. This in turn allows us to draw lessons for sustainable development in the future. The evolution of water harvesting practices, from ancient infrastructure to data-driven systems, reflects the continuous human effort to secure water access. In contemporary times, the integration of AI technologies has precipitated a paradigm shift in the manner in which urban water challenges are addressed.

The utilization of AI has been identified as a strategic approach to address various challenges in the water sector. These include demand forecasting and leak detection, as well as catchment infrastructure optimization, water resilience modeling and adaptive urban planning. However, despite the exponential growth of research in this field, there remains a paucity of clarity regarding the actual scope of these applications, the challenges associated with their implementation, and the opportunities yet to be explored.

According to the World Economic Forum [2], global demand for freshwater is expected to exceed supply by 40% by 2030, and it is expected that approximately 1.6 billion people will not have access to safe managed drinking water. Currently, 4 billion people live in areas suffering from severe water scarcity, with 1 in 4 cities experiencing water insecurity. This is in part due to population growth, which means that more water is required to produce food and energy to run cities. Another key concern is water pollution, which endangers existing resources: it is estimated that 80% of wastewater from industry and municipalities is discharged untreated, contaminating water quality and other natural resources. The climate crisis is severely disrupting the water cycle that people and ecosystems depend on. Water resources are at the center of the crisis, as 90% of climate phenomena are directly related to water. Droughts, floods, and intensifying storms are becoming more prevalent, while groundwater resources are drying up, making access difficult. In addition, cities and farms face water shortages more frequently and glaciers are melting at an accelerating rate, contributing to rising sea levels.

In the ongoing quest for the sustainability of water resources, the development of alternative water sources has gained prominence and become a key focus of many countries. Ertop et al. [1] highlight rainwater harvesting as a means of storing significant volumes of water. The implementation of rainwater harvesting systems enables countries to conserve a substantial volume of domestic water, thus reducing the demand on the water network and generating economic benefits. In particular, the water obtained through rainwater harvesting is both free and of high quality. In addition, these systems can be deployed to manage emergency situations, including sudden droughts, extreme droughts, earthquakes, and fires. The stored water can be easily used to help conserve existing water resources and ensure the sustainability of rural areas. Beyond human consumption, water can be used for various purposes, including irrigation of green spaces, agriculture, personal hygiene, flushing of toilets, washing cars, and firefighting.

However, Yildirim [3] emphasizes the importance of selecting an effective and cost-effective catchment system. The decision-making process for selecting a catchment system consists of a cost analysis and the application of multi-criteria decision-making methods such as AHP (Analytic Hierarchy Process). The researchers emphasize that proper planning of water retention activities is an extremely important aspect in this context. They also consider rainwater harvesting (RWH) a viable method to address water-related challenges, although others have also questioned the feasibility of this method.

The impact of climate change is of great concern. Consequently, Botai et al. [4] point to drought indices, which define the duration and severity of drought in various ways, although they are generally based on the behavior of the index over time. In contrast, Habibi et al. [5] propose the drought severity index, defined as the accumulation of negative values of the index during the drought event. In the context outlined above, severity can be interpreted as the volume of the flow deficit, as defined by Șerban et al. [6] as the difference between the threshold level and the actual integrated flow during the drought period.

This systematic review addresses an increasingly pressing challenge in the context of urban environments: water scarcity and the need to strengthen water resilience through emerging technologies. In particular, the role of artificial intelligence (AI) as a strategic tool in efficient water harvesting, management and distribution is explored. Despite the existence of previous research in the fields of artificial intelligence and water infrastructure, there is a paucity of studies that offer a comprehensive perspective integrating the scientific literature with patent analysis, technology trends, and governance elements. This review addresses this knowledge gap by constructing a timeline that spans scientific and technological development, identifying patterns of innovation, and highlighting opportunities to implement effective solutions in urban settings. The findings provide valuable data for researchers, policy makers and decision makers seeking to integrate AI into urban planning with sustainability and resilience criteria.

In this sense, and given the interdisciplinary nature of water management, this study considers AI as a tool to be integrated into water governance schemes that are capable of balancing economic development, ecological conservation and social equity. Consequently, a systemic approach is proposed in which AI not only behaves as a technical efficiency mechanism, but also as part of a broader transformation towards resilient, inclusive and environmentally sustainable cities.

2. Materials and Methods

Documentary research work using the deductive method based on the SLR model (Systematic Review of the Literature) proposed by Garza-Reyes [7] has been adequate for this research. It should be noted that none of the phases were altered, which are: (1) formulation of research questions; (2) geographic location of the research development; (3) selection and evaluation of studies; (4) analysis and synthesis; and (5) reporting and use of results.

In order to mitigate the selection bias associated with open access, priority was given to searching in databases recognized for their global and multidisciplinary coverage (MDPI, Scopus, Web of Science and Springer Open), covering both indexed articles and publicly available and paid-for technical literature, since institutional licenses were available to access the papers. In the process of incorporating the inclusion criteria, priority was given to the representation of approaches developed in different regions of the world, including the Americas, Europe, Africa and Asia. In the course of the present study, a wide range of academic, governmental and private institution-based papers were considered, with particular emphasis on studies which demonstrated real or pilot applications in contexts characterized by high water vulnerability or low income levels. This approach enabled a more extensive coverage, which was not confined to the countries of the Global North, and consolidated the inclusion of diverse and culturally contextualized technological experiences.

The systematic review of the literature was carried out according to the steps described in Figure 1:

(1) The formulation of questions was carried out: What are the environmental, economic and social impacts of using AI in water management? How has AI been integrated with other emerging technologies (IoT, Big Data, smart sensors) to improve water harvesting? What are the current trends in the scientific literature regarding the use of AI in water management? What public policies could facilitate the adoption of AI in water management to mitigate water scarcity in urban settings? What are the current trends in the scientific literature regarding the use of AI in water management later on?

Subsequently, we are conducting a (2) global study using various scientific databases, including Scopus, Elsevier, and MDPI.

(3) In the selection and evaluation of studies, the following key words were searched: water scarcity, water management, water reuse, urban water systems, artificial intelligence, machine learning, neural networks, rainwater harvesting, rainwater harvesting structures, rainwater harvesting techniques, alternative water source, rainwater harvesting systems, hybrid systems, stormwater management, rainfall forecasting, real-time control, rainwater treatment, low impact development, sustainable urban drainage systems, water supply, rainwater reuse, wastewater treatment. After entering the relevant search terms, the next step was to narrow the search criteria by selecting only relevant publications from the last five years.

(4) The analysis and synthesis process was then carried out by selecting open-access articles, which were then downloaded for review of content, contributions, and conclusions. Following the analysis, it was decided whether the article should be included in this review.

(5) reporting and utilization of results, consisted of identifying the most significant contributions to combating water scarcity and identifying future proposals to solve the problem.

The following section details the taxonomy used to classify the articles identified in this literature review.

Furthermore, a thorough search was conducted on the global patent database Espacenet with the aim of identifying the most significant advances aimed at resolving the issue. Patents were identified for water collection devices, electrodes, aerosols, and chemical mixtures to stimulate artificial rain, and a taxonomy was constructed by year of registration in the patent book to visualize a timeline.

The search strategy used for the systematic literature review is detailed in

Table 1. Search Strategies.

PRISMA Diagram ID	Topic	DataBase	Keywords	Files Found	Files Selected
A	Optimization of Water Harvesting	MDPI and SCOPUS	(rainwater harvesting) AND (water quality) AND (water conservation) AND (rainwater tanks) AND (life cycle cost analysis) AND (multi criteria analysis) AND (urban flooding)	13	3
B	Technologies to improve water harvesting and storage efficiency	MDPI	(water sustainability indices) AND (systematic review) AND (SDG6) AND (water resource management) AND (water sustainability) AND (water resource sustainability)	7	7
C	Optimal location of catchment infrastructures	MDPI and SCOPUS	(Water infrastructures as complex systems) AND (resilience of multiplex water networks) AND (interdependent water networks)	9	5
D	Collection and storage efficiency	MDPI and SCOPUS	(Water and wastewater treatment) AND (Sustainable urban water management)	5	4
E	Artificial Intelligence (AI) Applications	SCOPUS	(Artificial intelligence) AND (Big data analytics) AND (Water resource management) AND (Water quality monitoring) AND (Water demand forecasting)	4	1
F	Machine Learning models for rainfall forecasting	SCOPUS	(System Modeling Algorithm) AND (Water-Driven Prediction Platform) AND (Applied Research) AND (Intelligent Platform)	8	5
G	Optimization Algorithms for Rainwater Harvesting System Installation	MDPI and SCOPUS	(Rainwater harvesting) AND (Sustainable urban water use) AND (rainwater) AND (Alternative water sources)	10	5
H	AI-based hydrology analysis	MDPI and SCOPUS	(aquatic ecosystems) AND (ecohydrology) AND (hydrological processes) AND (Monitoring/Modelling/ Prediction/Optimization)	24	10
I	Urban Resilience and Water Management	MDPI and SCOPUS	(water resilience) AND (urban floods) AND (storage of runoff water)	14	4
J	Reduction of losses in the system	MDPI	(resilience) AND (Wastewater treatment and reuse) AND (Water infrastructure management) AND (urban stormwater and wastewater)	8	6
K	Optimization of Water Distribution	MDPI	(resilience measures) AND (urban water systems) AND (assessment and diagnosis) AND (resilience)	10	1
L	Impact on Urban Resilience	MDPI and SCOPUS	(resilience of multiplex water networks) AND (decentralized and hybrid infrastructures) AND (strategic asset management) AND (adaptation)	9	3
M	Technology Integration	SCOPUS	(Smart irrigation) AND (IoT) AND (Server-Sent Events (SSE)) AND (Sensors) AND (Water management) AND (Embedded technology)	6	1
N	IoT sensors and Big Data for real-time monitoring	MDPI	(Internet of Things (IoT)) AND (monitoring) AND (notification) AND (efficiency) AND (water quality)	28	8
O	Hybrid IA + Renewable Energy Systems	MDPI and SCOPUS	(Smart water system) AND (Renewable energy resources) AND (Energy management) AND (Water monitoring) AND (Water pumping) AND (Water management)	16	4
P	Regulation of Water Management	MDPI	(socio-environmental regulation) AND (risk-based approach) AND (sustainable management) AND (watershed) AND (water resources) AND (socio-environmental risks)	5	1
Q	Water Governance	MDPI	(water & wastewater treatment engineering and ecotoxicity evaluation) AND (water-resources modeling) AND (digital elevation model)	17	10
R	Water Resilience	MDPI, SCOPUS and Springer	(water scarcity) AND (climate change) AND (blue-green infrastructure) AND (quality of life (QoL)) AND (livability) AND (climate resilience) AND (water in cities) AND (urban environments)	23	10
S	Governance and Regulation of Artificial Intelligence	MDPI	(ethics) AND (AI governance) AND (human rights) AND (AI regulation) AND (fundamental rights) AND (ethics washing) AND (GenAI regulations) AND (GenAI regulatory framework)	9	6
T	Cloud Seeding	MDPI	(cloud seeding) AND (evaluation) AND (change point) AND (rain gauge) AND (cloud seeding) AND (artificial precipitation) AND (effect evaluation)	15	6
U	Adsorption technology	MDPI	(adsorption) AND (mechanism modelling) AND (artificial intelligence) AND (emerging pollutant) AND (machine learning) AND (water treatment)	13	5
V	Environmental impact	Arxiv	(Machine Learning (cs.LG)) AND (Artificial Intelligence (cs.AI))	8	5
W	Systems Thinking	Spriger, Arxiv, Wiley and MDPI	(Viable system model) AND (Variety) AND (Systems science) AND (law of requisite variety) AND (Trust building) AND (Process-based trust Characteristic) AND (Network development) AND (Action research)	18	6
X	Patents	Espacenet	Water Harvesting	291,459	91
Total Resources				291,738	207

.

The following flow chart Figure 2 illustrates the selection process from the initial identification of records to the final inclusion of the analysed studies according to the PRISMA model [8]. It clearly details the exclusion criteria applied in a query comprising the period from 1 April to 18 May 2025.

3. Taxonomy

In this section, we propose a taxonomy of Artificial Intelligence (AI) applications, divided into six groups: 1. Water Harvesting Optimization, 2. Artificial Intelligence Applications, 3. Urban Resilience and Water Management, 4. Technology Integration: Use of sensors, IoT, Big Data and other advanced technologies in water management, 5. Regulation of Water Management: Works that consider the regulation of the use of artificial intelligence in order not to infringe regulations, 6 Others. The following section details the taxonomy used to classify the articles identified in this literature review, Figure 3.

3.1. Optimization of Water Harvesting

Rainwater harvesting is perhaps the oldest practice for meeting water supply needs. This technology has been used for many centuries. Since the 1970s, it has attracted renewed attention as a source of productive water, a means of water saving and conservation, and a tool for sustainable development. Historically, these technologies have played an important role in meeting the growing demand for water and in coping with climate change and variability. Rainwater harvesting is defined as a method of inducing, collecting, storing, and conserving local surface runoff for later use. The evolution of devices for harvesting rainwater from impervious surfaces such as rooftops, terraces, patios, and road surfaces or from the natural surface of the land and storing it in systems such as tanks, cisterns, and underground dams for indoor and outdoor use has been instrumental in this regard. As Rahman [9] notes, these practices improve access to water for domestic and agricultural applications in semi-arid regions, where catchment systems can offer a more robust and cost-effective approach to water security compared to complex public water supply systems.

Adham et al. [10] emphasize the need for a methodology to evaluate the performance of existing RWH (Rainwater Harvesting) structures. This methodology should integrate criteria related to structure design, engineering, as well as biophysical and dimensional factors, such as storage capacity, catchment area ratio, and socioeconomic impact. Furthermore, when the relationship between the catchment area and the cultivated area is not adequate, the performance of the RWH system can be improved by adapting the cultivated area to the effective storage area or by modifying the type of crop or planting density to match the amount of water stored.

Optimization should focus on the design of Domestic Rainwater Harvesting Systems (DRWHS) for communities. Sámano’s [11] approach involves identifying the relationship between local conditions, such as marginalization, expected service levels and rainfall, and the physical components of DRWHS. The effectiveness of these systems is calculated based on the catchment area and the size of the tanks required, with a focus on water demand, the runoff coefficient, and monthly rainfall. Determining water consumption is crucial to ensure a reliable and consistent water supply throughout the year.

In addition, it is imperative to perform a cost-benefit analysis to support the feasibility of the technical solution identified by design methods. Raimondi et al. [12] propose that financial analysis allows one to establish the price of water as the main variable, not only the actual price but also the expected price. The financial model should include the costs of maintenance, operation, and energy consumption. It is important to note that most approaches are simplified and do not holistically assess all potential benefits of water harvesting. The analysis should consider different climatic conditions, building types, rainwater use, number of users, and storage tank capacity. By comparing the results with those applied to traditional water supply systems, we can better understand the advantages of using harvesting technologies. Notable benefits include environmental protection due to reduced combined sewer overflows and groundwater recharge, which are difficult to quantify. In addition, indirect benefits have been identified, including a reduced need for improvements in the water infrastructure, energy savings, and reduced emissions due to the reduction in water collection, transportation, and treatment.

3.1.1. Technologies to Improve Water Harvesting and Storage Efficiency

Water scarcity has become the number one issue on countries’ agendas. In addition to the construction of dams and the reclamation of wastewater, cities must redouble their efforts to discourage wasteful consumption and unmetered use. In addition, cities are advised to explore ways to innovate in their water supply, for example, using microfiltration, acoustic nanotubes, ultrafiltration, and nanofiltration. Modern cities use a variety of technologies to maintain a stable water supply and avoid potential crises. Silva [13] says that prominent technologies include 60-story dams, centrifugal pumps, and ultrafiltration machines that convert wastewater to clean water for domestic and industrial use. Dams have become essential for both water supply and power generation. The 60-storey dam features a power plant where water passes through turbines at an impressive speed of 136.8 km per hour. The machine room houses a generator that converts the kinetic energy into electrical energy, sufficient to power a city and its surrounding towns. Investment in dams and other reservoirs enables cities to provide a consistent water supply throughout the year. Technological advances have enabled the development of sophisticated forecasting and tracking systems that enhance the accuracy and transparency of the water distribution. However, to ensure a stable and continuous supply in the future, cities will need to explore innovative solutions that go beyond today’s methods.

The most advanced technological approach is that it facilitates the incorporation of artificial intelligence, which is detailed in Section 3.2, the development of machine learning models, statistical analysis of seasonality, flow pattern detection, and quality control methods. They support the centralization of integrated water management and hydrological planning, especially in the context of reservoirs and groundwater sources. Jarzebski et al. [14] mention the need for informed decision making in the face of climate change and the need for effective strategies to address issues related to irrigation, rivers and water pollution. The use of cutting-edge technologies in conjunction with analytical techniques to improve water management practices can reduce uncertainty and improve the overall efficiency and accuracy of water-related processes. Recognize the interconnected nature of water resources and the need for comprehensive strategies that consider holistic management of surface and subsurface water systems. By encompassing the social, environmental and economic dimensions of sustainable development, Sustainable Development Goal (SDG) 6 reflects the interconnection of water with planetary and socioeconomic goals, including the interconnections with all other sustainable development goals.

In traditional paradigms, the water crisis is generally seen as a technological problem, but in emerging paradigms the focus shifts to recycling and reusing water, treating waste and storm water as resources, effective demand management, promoting green infrastructure, increasing community and stakeholder participation, effective governance, and a multidisciplinary approach to achieving water security. Citizen science is defined as “the participation of lay people in scientific research”. Based on Mishra et al. [15], it is also a way for the public to contribute to monitoring water pollution and progress in water management. For example, by participating in citizen science initiatives and reporting evidence of water pollution in their area. These initiatives have recently gained prominence due to advances in information and communication technologies (ICTs), such as smartphones with internet connection, GPS, and camera. The use of ICTs has made it easier for citizen volunteers to interact with professional scientists and provide useful information on water quality to decision makers.

However, Qian et al. [16] note that experiments have been conducted and analyzed to increase the accuracy of the assessment of water scarcity risk depending on the temporal scale, spatial scale, water supply and demand entities and assessment methods. To improve the overall accuracy of the quantification, Mekonnen and Hoekstra [17] used higher spatial resolution, assessed on a monthly time scale, and incorporated environmental flow requirements, resulting in a more accurate representation of water scarcity scenarios. To address the modifiable unit of area problem in spatial water resource assessments, which arises from significant differences in results due to variations in the choice of unit of analysis, they proposed a multi-area, multiscale approach to improve the robustness of the assessment.

Based on Gómez [18], recent technological advances in wastewater treatment and desalination, as well as the implementation of various programs (inspired by the European Green Pact and the new Circular Economy Action Plan), have led to uninterrupted growth in non-conventional water production. In addition, there is an expected increase in the use of reclaimed water, particularly desalinated water.

In a similar vein, Raimondi et al. [19] emphasize that the primary advancements and contemporary methodologies in rainwater harvesting and treatment are based on the following pivotal criteria: the technical and economic viability of rainwater harvesting systems, the efficacy of rainwater treatment through diverse solutions and techniques, and the impact of the first flush phenomenon on the degradation of rainwater quality. Financial support programs to promote the installation of rainwater harvesting systems and increase their economic viability. Real-time control of catchment systems in multipurpose management is also highlighted, with the importance of a real-time control system for rainwater detention basins being demonstrated through simulations showing their ability to meet non-potable use requirements and ensure hydraulic protection of the downstream system. In addition, the concept of “City Water Hubs” has been introduced, comprising decentralized treatment units to collect and treat rainwater from nearby rooftops using low-energy technologies. This approach demonstrates that sufficient rainwater can be provided for irrigation even in drought conditions.

Other harvesting technologies include Land surface harvesting is used to collect rainwater and runoff from surfaces such as paved areas, terraces, roads, patios, as well as unpaved areas such as soil, pastures, cultivated land, rocky areas, and natural slopes. The conveyance system is designed to transport water from the collection area to the storage area. The material used in this system depends on the collection system and can be made of different materials, such as galvanized iron or polyvinyl chloride. In roof systems, channels and ditches are used to transport water to the underground storage. The channels should have a minimum width of 12.7 cm and be designed to accommodate runoff during the highest rainfall. Microbasin methods are systems of a size generally less than 0.02 ha for farm systems and less than 0.1 ha for roof systems. The common flow is by an artificial slope or trough flow. Ertop et al. [1] propose calculating the potential for rainwater harvesting in greenhouses using roof systems, considering them as catchment areas.

Rainfall forecasting is a vital component in the management of storm-water and water harvesting systems. The technology applied to these infrastructures uses weather forecasts to calculate the expected volume of runoff from a rainfall event. Altobelli et al. [20] contribute to real-time control systems that manage the discharge of storage basins based on predictive models and the water level in the tank. The continuous monitoring and adaptive control methodology uses rainfall prediction to limit water outflow to the sewer while reducing potable water consumption. Real-time controls are integrated with weather forecasting and in-tank level sensors. Water is released to the downstream system via a pump if precipitation is forecast in the next 24 h and the available storage volume is insufficient. This proactive approach aims to optimize storage capacity in the storage basin before rainfall. The system’s capacity to maintain or restore the natural hydrologic balance of an area, thereby minimizing the negative impacts of urbanization on the water cycle, such as increased peak flow, is called hydrologic invariance. Detention tanks contribute to this by reducing and regulating outflow.

3.1.2. Optimal Location of Catchment Infrastructures

The optimal infrastructure for rainwater harvesting is dependent on the scale of domestic, agricultural and industrial application, and the intended use of the collected water. The roofs of buildings are typically used for the collection of rainwater. The quantity and quality of the captured water is influenced by the size and material of the roof. The runoff coefficients vary according to the roof material used, ranging from 0.6–0.8 for concrete and 0.8–0.9 for tiles. The conduction system, which is responsible for transporting the water from the catchment surface to storage, must also be considered. In the home model, 4-inch PVC gutters and pipes connected to the ridge and valley lines of the roof are mentioned to direct the water. Storage: Storage tanks are essential for storage of captured water for use. The capacity of these tanks should be calculated based on factors such as average rainfall, catchment area, and water demand. The option of using recharge wells to infiltrate collected water into the subsoil and recharge the aquifers is also being considered. Identification of sites prone to water accumulation (ponds) through analysis of the drainage network with ArcGIS can facilitate the determination of optimal locations for subway tanks. Filtration and Treatment To guarantee the quality of the collected water, it is essential to implement filtration systems. Mechanical pre-tank filters, made of stainless steel, are used to remove soil and debris before the water enters the storage tank. Ahmed et al. [21] define that filtration efficiency is measured in microns. Microscopic filtration involves the use of cartridges or sealed bags to capture very small particles. This process requires pressure to operate and is typically used with a pumped water supply. Finally, the integration of disinfection filters, employing methods such as chlorination, ozonation, ultraviolet light (UV) and membrane filtration, ensures the eradication of microorganisms. It should be noted that disinfection treatments can generate hazardous by-products. UV light sterilizes water by passing it over a glass tube and exposing it to UV radiation. Distillation is another purification method that involves boiling the water and collecting condensation. To consume the final product, a distribution system is required, which may include pumps to drive the water from the storage tank and piping to carry it to the points of use.

Large-scale, decentralized implementation: The widespread implementation of RWH systems, such as smart rain barrels, in individual properties may be a good alternative to the needed extensions of existing and underperforming urban water infrastructure. As with the application of Smart RWH Systems vs. Uncontrolled Systems smart RWH systems, equipped with remotely controllable discharge valves, can automatically release stored water prior to rainfall events to increase retention capacity. This improves performance in comparison to traditional RWH systems, whose retention efficiency is significantly influenced by user behavior and abstraction quantities. As Oberascher et al. [22] emphasize, it is essential to consider digital uncertainties, such as the accuracy of weather forecasts and the reliability of data communication, to ensure the optimal performance of these smart RWH systems. Deviations from perfect performance in these parameters can reduce or eliminate the performance improvements obtained. Therefore, when designing and implementing the optimal infrastructure, it is essential to consider these uncertainties. For example, overestimating precipitation can result in smart RWH systems not completely filling, reducing the available water for consumption. To assess optimal infrastructure, integrated resilience indices should be considered, combining classic urban water infrastructure metrics (e.g., sewer overflows) with metrics of RWH system performance and nonpotable water supply. The weighting given to these different metrics will influence the recommended extension strategy.

In their proposal, Kakoulas et al. [23] set a framework to assess the resilience of urban water infrastructure, encompassing the following interconnected elements: diversification of water sources: Decentralization of infrastructure can improve resilience by reducing vulnerability to failures or stress in a single water source. The implementation of alternative sources, such as RWH systems, can enhance resilience by providing a reserve in the event of shortages or interruptions to the main supply. It is essential to emphasize that resilience should be evaluated in the context of a region’s specific geographical, climatic, and social characteristics. The assessment should take into account the vulnerability of a region to extreme weather events, such as droughts and floods, and the infrastructure should have the capacity to function under these conditions. Furthermore, the economic factor should be sustainable in the long term; otherwise, the investment is less resilient.

The resilience of urban water infrastructure is assessed using a variety of tools and approaches that consider both demand management and supply, as well as environmental impacts. Teston et al. [24] observe that Water Balance Modeling is an effective method to determine the impacts that water management systems, such as stormwater harvesting systems, cause on the urban water cycle, providing insight into how urbanization can affect water quality, groundwater recharge timing, and surface runoff. Another highly significant tool for evaluation is the Life Cycle Assessment (LCA); this tool is used to evaluate the integral environmental impact of a product or service, from its manufacture to its final disposal. In the area of water infrastructure, LCA makes it possible to quantify the impacts generated by water supply systems, such as rainwater harvesting systems and integrated gray-water and rainwater reuse systems, throughout their life cycle. When comparing these alternative systems with conventional centralized systems, most LCA studies report better environmental performance of the alternatives.

The inadequacy of outdated drainage infrastructure in managing increased storm-water runoff, a consequence of rapid urbanization and population growth, poses a significant challenge, often resulting in flooding. Waseem et al. [25] cite the inadequate capacity of existing infrastructure to cope with changing urban conditions. The adoption of RWH can effectively mitigate the volume of runoff entering the drainage system, thus alleviating pressure and improving resilience to flooding. The exploration of sustainable alternatives to the current water distribution system is necessary due to aging infrastructure. Water harvesting systems represent a decentralized alternative with the potential to supplement or even reduce demand on the existing centralized system.

3.1.3. Collection and Storage Efficiency

The effectiveness of a rainwater harvesting system is influenced by various factors, including the catchment area and the runoff coefficient, which represents the proportion of rainfall that becomes runoff and can be collected. The National Association for Quality in Premises Installations (ANQIP) recommends values ranging from 0.7 to 0.9 for impervious surfaces, as well as the hydraulic efficiency of the filtration treatment to ensure effective management of collected water. Storage efficiency is directly related to the volume of the storage tank and its capacity to satisfy the non-potable water demand. Matos et al. [26] determined the tank size using the yield-before-spillage method, which calculates the maximum tank volume without increasing rainwater consumption. The efficiency of the system for all storage volumes used in the sizing process helps in the decision-making process. This efficiency is calculated by dividing the volume of non-potable water used by the volume of non-potable water required for the end uses in question.

The effective use of rainwater harvesting (RWH) is presented as a crucial strategy to address water scarcity. However, Judeh et al. [27] have identified significant challenges that must be overcome, for which various smart solutions have been proposed. Conventional RWH systems often lack effective control over the potability of harvested water, leading to the perception that it is not suitable for consumption. Contamination can come from the atmosphere, roofing materials, and organic wastes, causing physical (turbidity), chemical, and biological (fecal coliforms) problems. Uncontrollable factors, such as the scarcity and unpredictability of rainfall and the catchment area, limit the amount of water that can be collected. Conventional systems also frequently lack monitoring of tank filling and emptying and leakage, resulting in inefficient use of available water. An alternative to counteract water management may be intelligent water quality monitoring to ensure water potability for all users and intelligent leakage monitoring in the shared distribution network.

RWH Smart System Architecture: A six-layer architecture is proposed based on Judeh et al. [27]:

• Physical layer: It is equipped with components designed for the collection and primary treatment of water.
• Monitoring layer: Sensors are used to collect data on water quality and quantity.
• The data transfer layer facilitates seamless data transmission: Wireless technology is employed to facilitate the transmission of data to a server.
• The data processing layer: This layer is responsible for cleaning, storage, analysis, and visualization of the collected data.
• The control layer: It uses actuators (pumps, valves) to control water flow according to the data analyzed.
• The intelligent services layer: This layer offers functionalities such as early detection of contamination and leaks, water quality control, optimization of water resources and incident notification to users.

The proposed architecture diagram is shown in Figure 4 below.

The economic design of rainwater harvesting systems using tanks. A new numerical simulation model using the WASH_2D software version 2.0 is presented, which allows to simulate in detail the behavior of a RWH with a specific tank for supplemental irrigation activities. Nana et al. [28] note that the precision of the model was evaluated through two field experiments: one focused on optimizing the tank capacity for garlic cultivation and the other on evaluating the net revenue with different cultivated areas for beans, both using plastic sheeting as a catchment area. The results demonstrated reasonable agreement between the WASH_2D model simulations and observed field data, suggesting that this case study may be useful for guiding investments in small-scale irrigation and rainwater harvesting in semi-arid regions by allowing optimization of tank capacity and cultivated areas to maximize farmers’ net income.

The measurement of fog collection is typically expressed in terms of the water collection rate per unit area of the collector, typically in liters per square meter per hour (L m⁻² h⁻¹). Standard fog collectors can achieve average collection rates of approximately 1.26 L m⁻² h⁻¹. However, more advanced designs, such as porous liquid-infused slippery surfaces and their modifications, have demonstrated significantly higher collection rates, reaching up to 8.52 L m⁻² h⁻¹ and even 45.18 L m⁻² h⁻¹ with special three-dimensional structures. In laboratory tests, fog harps have been shown to be up to three times more efficient than traditional nets. Conversely, while dew uptake can be approximately twice that of fog, the absolute water collection rate is considerably lower, often an order of magnitude smaller. Hydrophilic surfaces favour spray nucleation, but low droplet mobility requires the use of super-hydrophobic surfaces for rapid removal and continuous condensation. Chen et al. [29] indicate that biphilic surfaces, which combine hydrophilic areas for nucleation and super-hydrophobic areas for droplet ejection, have been shown to significantly improve the collection rate, reaching up to 349% more, compared to the former. The primary metric for desalination is the ratio of freshwater production to energy consumption, along with the daily production rate (in m³/day). Reverse osmosis (RO) is considered the most energy-efficient technology to date. Multi-stage flash distillation and multi-effect distillation are capable of producing large volumes of fresh water (10,000–35,000 m³/day y 600–30,000 m³/day, respectively). Efficiency is also being enhanced through the integration of renewable energies, such as solar power. The sorption capacity of the materials and the rate of water production are key factors in evaluating water removal from the air. Organo-metallic frameworks have demonstrated excellent water sorption capacity. A water collector based on MOF-801 has demonstrated the ability to produce 100 g/day of water per kg of MOF. However, enhancing water productivity remains a significant challenge.

3.2. Artificial Intelligence (AI) Applications

Kamyab et al. [30] propose that artificial intelligence allows the monitoring and analysis of large volumes of real-time data on water resources, facilitating data-driven informed decision making. This computational technology allows the identification of trends, patterns and potential risks of rainwater harvesting, and machine learning allows the calculation of water allocation optimization and demand forecasting. Another significant scope of deep learning is the modeling of rainfall, river levels and topographic analysis to predict floods. Drought management can be facilitated by processing satellite imagery and climate data to allow continuous monitoring of the health of water treatment plants and soil moisture levels, facilitating early detection of pathologies and efficient water management through the application of models such as decision trees. linear regression, SVM (Support Vector Machine), KNN (k-nearest neighbors), Random Forest, XGBoost (eXtreme Gradient Boosting), ARIMA (Auto-regressive Integrated Moving Average), ANN (Artificial Neural Network) and LSTM (Long Short-Term Memory) to predict short-term water demand.

3.2.1. Machine Learning Models for Rainfall Forecasting

Machine learning methods for weather forecasting can be divided into two main categories: traditional machine learning methods and deep learning methods. These methods have the capacity to manage complex and high-dimensional data. 1. Traditional machine learning methods include techniques such as support vector machines (SVM), linear regression (LR) and random forests (RF), which have been successfully applied to simpler weather forecasting tasks. However, these methods present difficulties in capturing more complex relationships in weather data, especially when large-scale and high-dimensional input is involved. Classification approaches using Naive Bayes and the Chi-square algorithm to predict weather conditions have been presented, demonstrating their sufficiency for weather forecasting, although limited to predicting class labels (Good/Bad). Deep Learning Methods: Multi-layer neural networks are employed to automatically identify complex nonlinear patterns and spatio-temporal dynamics, reducing reliance on manual feature engineering. These networks have the ability to identify intricate patterns and achieve superior prediction accuracy and generalization in multi-scale and complex environments. The classification is shown in Figure 5. As Zhang et al. [31] have mentioned, transformer-based neural networks use attentional mechanisms to capture global dependencies and enable further parallelization. These networks are applied in immediate precipitation forecasting, global weather forecasting, and longer-term predictions.

The “System Modeling Algorithm” proposed by Hu [32] is a method focused on transforming real-world problems into mathematical representations. The algorithm’s ability to incorporate various sources and types of data allows it to build a model that captures the behavior of the system. This is in contrast to traditional prediction models, which are based on statistical models. The platform adopts a modeling approach based on analytical methods and physical models, allowing a better understanding of the physical characteristics of the system itself. This comprehensive analysis of the physical characteristics of the system allows the algorithm to identify the role of water in different states and to make predictions of the water production capacity with higher accuracy, stability, adaptability and lower Root Mean Squared Error (RMSE), MAE (Mean Absolute Error), MSE (Mean Squared Error).

DRL (Deep Reinforcement Learning) has been proposed by Hu et al. [33] as a promising alternative to non-linear and non-convex optimization problems. In a similar vein, traditional deterministic methods have been shown to simplify the complexities of WDS (Water Distribution Systems), though this may introduce biases and exclude potentially beneficial solutions. The paper proposes a real-time optimization model for water distribution systems, using an exploration-enhanced DRL approach that adds an entropy bonus to the policy objective of the proximal policy optimization algorithm. The approach shifts the computational load offline, ensuring sufficient exploration during the learning process, allowing the agent to discover suboptimal scheduling policies for various demand distributions. State normalization is also used to maximize initial entropy, allowing agile decision making in response to changes in water demand while effectively controlling energy costs and tank levels. Innovation in the design of the reward function, especially the tank level penalty, and improved scanning are crucial aspects of this proposal.

The K-star Ensemble machine learning model proposed by Tuysuzoglu et al. [34] applies meteorological data to predict the next day’s rainfall using an ensemble of K-star classifiers. This approach is based on the ensemble learning principle, in which the predictions of multiple models are combined to obtain a more accurate prediction. For the development of this model, meteorological data were obtained from observation stations, including the variables: temperature, rainfall, evaporation, sunshine, wind, humidity, pressure, and clouds. The data then undergo a preparation stage, which includes the elimination of missing data, data transformation, and selection of relevant characteristics. The training of the model begins by building an initial classifier on the data set, then for each instance of the training set, the classification probability is calculated, indicating the model’s confidence in assigning a “yes” or “no” category for the rainfall. Then a probability sampling approach is applied to create multiple training sets. The next stage of the process involves building an individual K-star model on each data set. Once these steps have been completed, a new data set is used to predict the rainfall. Each of the K-star models in the ensemble outputs a prediction. The final prediction of the EK-stars model is determined by a majority vote of the individual model predictions. A notable aspect of the model is the significance of the features revealed, with the Sunshine variable demonstrating the strongest correlation with the rainfall prediction, followed by the humidity and cloud-related variables. The application of feature selection further enhanced the accuracy of the EK-stars model.

Ensemble models have repeatedly demonstrated that ensemble-based algorithms, particularly CatBoost and XGBoost, outperform standard linear regression models and LGBM (Light Gradient-Boosting Machine Regressor) in daily rainfall prediction. This is due to their ability to process complex non-linear interactions between meteorological factors and rainfall patterns. These models have been shown to be more robust and reliable in predictions, as evidenced by the low percentage of prediction errors throughout the data set, as demonstrated by the scatter plots. The ability to combine predictions from multiple models allows for the reduction of variance and bias, leading to more stable and accurate predictions. Finally, as Kumar et al. [35] have demonstrated, these models have been shown to be resistant to overfitting, which is crucial to ensure that the model generalizes well to unseen data and provides accurate predictions in real scenarios. This results in more accurate, robust, and reliable predictions compared to traditional linear models.

3.2.2. Optimization Algorithms for Rainwater Harvesting System Installation

Optimizing the spatial design of rainwater harvesting systems for urban flood mitigation is achieved through a simulation-optimization model that considers the location, quantity, and capacity of the tanks. This approach, proposed by Huang et al. [36], allows the identification of the most effective arrangement to reduce flood losses and minimize installation costs. The fuzzy C-means (FCM) clustering algorithm is used to classify the study area into characteristic zonal sub-regions based on urban roof, land use and rainfall characteristics. Subsequently, regular spatial array specifications for tank quantity and capacity are designed using statistical quartile analysis for roof area and rainfall frequency analysis, considering different distribution patterns, such as region-wide dispersion, downstream concentration, upstream concentration, and concentration in easily flooded sub-basins. The stormwater runoff management model is then used to simulate floods and water levels in different spatial design scenarios during real rainfall events. Finally, a water level simulation model based on a backpropagation neural network (BPNN) is developed, allowing faster optimization and considering an interdisciplinary multi-objective model.

• Rainwater harvesting systems, especially those equipped with Real Time Control (RTC), have the potential to increase the potable water supply by collecting rainwater and using it for various domestic purposes. Although these systems have historically been used in rural areas, they are now being adopted in urban settings due to the growing demand for water. These systems can supplement the existing potable water supply. By offering an alternative water source for non-potable uses such as irrigation, toilet flushing, and laundry, these systems can help reduce the overall demand for urban water supply networks. RTC technology improves storm water retention by using real-time rainfall forecasts to initiate water release before storm events. This technology generates additional storage capacity for runoff from the next storm and reduces the possibility of an uncontrolled system overflow. These systems have the potential to play a role in the controlled delivery of water to urban streams in a manner that mimics natural base flows, which are often diminished by urbanization. Xu et al. [37] state that the configuration of rainwater harvesting systems for base flow release, whether active or passive, does not significantly impact water supply performance, since the base-flow release volume is minimal relative to the water supply demand. Active RTC release systems generally perform best in delivering sustainable base-flow to receiving waters. In addition, they can be integrated into combined sewer systems and become a feasible complement to the existing centralized system, with great potential to reduce or even eliminate combined sewer overflow (CSO). This ability to provide decentralized stormwater control can reduce the need for improvements to stormwater infrastructure, offset by the cost of the active release system.
• According to Aghaloo et al. [38], the optimal sites for rainwater harvesting can be identified by proposing a new Decision Support System based on GIS (Geographic Information System). To generate a probabilistic rainfall analysis, a Monte Carlo simulation model was carried out, allowing consideration of the uncertainty associated with rainfall and selection of a level of probability of exceedance according to the needs of the population. Subsequently, biophysical criteria were selected, such as precipitation, slope, soil type, land use and/or land cover, and drainage density, as well as socioeconomic criteria, including distance to roads, rivers, cities and faults. Then, the weighting of the criteria was performed to determine the relative importance of each of them. The values were then transformed to a scale of possibilities from 0 to 1 using fuzzy logic. Then, a constraint map was generated to exclude areas not suitable for the implementation of structures due to technical, financial, and environmental constraints. Finally, based on suitability maps and comparing them with existing areas, a phased strategy for the transition from the current unsustainable infrastructure to a new paradigm based on rainwater harvesting was discussed.
• In contrast, the genetic algorithm proposed by Snir et al. [39] is designed to create policies that adapt to the initial conditions and characteristics of each rainfall event. The algorithm uses a drainage system model to predict its response to a given rainfall pattern by performing a benchmark simulation without controlled releases to establish a baseline and determine certain key parameters. From the baseline simulation, the constant flow rate (Qobjective) must be determined. This value is the total volume of overflows and the duration of the period in which these overflows occur. The purpose is to perform scheduled releases of water volume in a more uniform manner to reduce the maximum flow rate. It is important to note that the decision criteria are a function of the variables that the algorithm can adjust to control water release, including the percentage of opening of a valve at the bottom of each rainwater collection tank, at intervals 10%. These adjustments can be made at the beginning of each hour. Each “chromosome” in the algorithm represents a complete valve opening policy for each tank throughout the rain event. The “fitness” of each policy is evaluated using the objective function.
• The efficiency of a rainwater harvesting system is dependent on its capacity. However, as Jin et al. [40] point out, designing a system with the capacity to meet all expected demand may result in installation and operating costs that are too high. The capacity of a system is directly related to its reliability and the rate of reduction of runoff. Optimal capacity is achieved by maximizing revenue generation. A cost-benefit analysis is fundamental to establish the economic viability of catchment projects by comparing investment costs with expected benefits. The financial support programs implemented by local governments, such as subsidies for installation costs and relief on water utility bills, significantly increase the benefits and economic viability of water harvesting systems. Reliability, classified as temporal and volumetric reliability, is a crucial factor in determining the appropriate capacity. A comparative analysis of NPV (Net Present Value) and a cost-benefit analysis demonstrated that the capacity that maximized the value of the investment exhibited significantly low operational reliability, emphasizing the necessity of incorporating reliability considerations when optimizing the capacity.

3.2.3. AI-Based Hydrology Analysis

The importance of aquatic ecosystems and water resources, as well as the challenges to their management and conservation. Saha [41] highlights the impact of human activities, which exert increasing pressure on these systems through over-exploitation of water resources, alteration of flow regimes, channelization of rivers and discharges of untreated wastewater. It is therefore essential to understand the connections between hydrology, water quality and ecosystems. The collection of long-term hydrological and water quality data is considered essential to understand the inter-annual variability of flows and their impact on species life cycles. The development of an open source module to simulate the physical processes of the nitrogen cycle and predict nitrate leaching. In contrast, other authors explore the determination of water residence times in ponds, the use of tracers to delimit stream recharge areas and the impact of anthropogenic activities on lake eutrophication.

• The field of hydrological modeling has seen significant advancements in recent years thanks to the development of artificial intelligence. This technology has introduced novel methodologies for analyzing complex datasets, improving the accuracy of predictions, and expanding the capabilities of modeling across a range of areas. According to Biazar et al. [42] this transformation addresses the challenges of managing water and soil resources under increasingly dynamic and unpredictable environmental conditions. Advances in soil and land surface modeling technology have led to significant improvements in soil texture, water content, temperature, and erosion management. Together with access to faster, real-time knowledge, it has consistently outperformed traditional statistical methods by effectively capturing regional variations in soil properties. Significant advances have also been made in the prediction of groundwater levels, which have been hampered in the past by complex subsurface dynamics. Neural network models now predict groundwater fluctuations with greater accuracy, enabling better water allocation in agricultural and urban environments.
• A data-driven approach is employed with artificial neural networks, specifically auto-regressive neural networks, to predict drought in hyper-arid climates due to their cyclic nature. These networks combine the advantages of auto-regressive models, making them capable of capturing nonlinear behavior within an auto-regressive time series. The model proposed by Alsumaiei and Alrashidi [43] predicts the next value of the time series based on the past values, according to the target delay terms. The usefulness of drought forecasting lies in its ability to significantly improve the management of water resources and mitigate the severe consequences of water scarcity, especially in climates where water resources are limited and environments are hostile.
• Chang et al. [44] provide a comprehensive overview of key applications of artificial intelligence techniques in the field of hydrology. These applications have a wide range of significant implications, including the accurate modeling of complex, nonlinear, and dynamic hydrological processes. These processes can overcome the limitations of conventional approaches based on physics or statistics. This is crucial in an environment of increasing hydro-geo-meteorological uncertainty caused by climate change. Hydrological forecasting has great scope for short- and long-term hydrological prediction, such as precipitation forecasting to improve rainfall prediction accuracy and rainfall threshold estimation in unengaged watersheds, allowing better preparation for floods. Temperature forecasting involves the development of stochastic models that consider the daily occurrence of precipitation and its effect on maximum and minimum temperatures, which is vital for various studies and watershed management. Flow forecasting utilities probabilistic models with Long Short-Term Memory (LSTM) neural networks and Bayesian sampling (BLSTM) for multi-pass daily flow predictions, quantifying uncertainty and demonstrating superior performance compared to other models, especially for data with high variance and peaks.
• In contrast, Nunez et al. [45] highlight that the development of Explainable Artificial Intelligence (XAI) techniques has improved the understanding of hydrological processes in this mega-drought context in several ways. Firstly, the XAI techniques applied in the study allowed the identification of the importance of hydro-meteorological variables located outside the river basin boundaries. This suggests that information from neighboring areas can be crucial for understanding water flow patterns, especially in data-scarce conditions such as those found in arid and mountainous regions. By identifying these key variables, hydrologists can develop a more comprehensive understanding of the factors that influence flow, including those outside traditional watershed boundaries. Secondly, the partial dependence plots and cumulative local effects plots were consistent with pre-existing hydrologic knowledge. These plots revealed the existence of thresholds in the relationship between precipitation and snowbelt-driven flow, which is consistent with how hydrological relationships are typically parameterized in mountain areas.
• As Ma et al. [46] point out, the use of artificial intelligence has a significant impact on the efficiency and safety of hydraulic engineering. It has become an important research method for solving critical scientific problems and is widely involved in the optimal design, structural simulations, and safety assessment of water conservation projects due to its advantages in regression, classification, clustering, and dimensionality reduction. Peng et al. [47] employed a convolutional neural network model to achieve accurate segmentation of coarse-grained soil images. This approach outperformed traditional methods and discovered a new method for soil particle size analysis. The study identified significant potential for automation and precision in soil particle analysis.
• Extreme Machine Learning (ELM) was used by Forghanparast and Mohammadi [48] as a baseline to compare with deep learning models in intermittent flow prediction. Deep learning models, especially those based on short- and long-term memory networks, were found to outperform ELM in capturing hydrological extremes. However, Gonzales-Inca et al. [49] have highlighted the potential of GeoAI (Geospatial Artificial Intelligence) in analyzing vast quantities of spatial and non-spatial data to model hydrological and fluvial systems. The advantages of GeoAI include non-linear modeling, computational efficiency, integration of multiple data sources, and high predictive accuracy, which is crucial for understanding and predicting hydrological events. Ghobadi and Kang [50] employed deep reinforcement learning to automatically and consistently identify extreme events from anomalies in the water level.

3.3. Urban Resilience and Water Management

Xu et al. [51] define the concept of urban water resilience as the comprehensive capacity of a city to absorb, regulate, store, and use water in a sustainable manner, especially rainwater, with the aim of mitigating the negative impacts of extreme hydrological events such as floods and droughts, ensuring long-term water supply, and promoting a balance between urban development and the environment. To achieve urban water resilience, the following key aspects should be emphasized: The city infrastructure must be able to effectively regulate and purify rainwater during prolonged periods of precipitation. In the event of prolonged droughts, the stored rainwater must be released and utilized, which requires scientific planning, construction, and management of cities, taking full advantage of artificial urban ecosystems. The optimal utilization of rainwater is presented as a model of urban development that fits in with nature, protects it, and perfectly combines the city with the aquatic ecological environment, seeking a harmonious co-existence between humans and nature. This approach is regarded an effective means of mitigating the impact of urban flooding and water scarcity, while simultaneously improving water security.

Chitwatkulsiri and Miyamoto [52] stress the importance of establishing real-time urban flood forecasting systems to minimize risk and improve resilience of vulnerable urban communities. Urban flood risk management has benefited significantly from the development of hydroinformatics tools, such as rainfall forecasting and flood modeling. Harada [53] states that real-time flood forecasting, in combination with robust weather forecasts, is an effective method to deal with the uncertainties associated with the increased frequency of flood disasters in the context of current climate change. Furthermore, Bui et al. [54] posit that real-time forecasting of hydro-meteorological systems can enhance urban resilience in two key areas: real-time monitoring issues and urban flood warning systems.

3.3.1. Reduction of Losses in the System

The reduction of losses in the drinking water and wastewater infrastructure system is an important aspect of resilience. As Pamidimukkala et al. [55] note, the problem is exacerbated by the aging infrastructure, that is, pipes and other components are more prone to leakage and fractures, resulting in physical loss of treated water. Inadequate maintenance can further compound water losses and lead to inefficient system operation. Furthermore, population growth indicates that if demand exceeds system capacity and losses are significant, shortages and distribution problems can be exacerbated. Inadequate capacity can lead to combined sewer overflows, representing a loss of system and an environmental risk. The most effective strategy to combat these losses is the implementation of a geographic information system (GIS). A GIS allows the characterization of the condition and age of the infrastructure, facilitating the identification of vulnerabilities and improving the management of the system. The use of efficient pond sand filters is also recommended, as these filters can effectively reduce the need for reprocessing, thus minimizing losses. These measures must be implemented in conjunction with appropriate policies. This involves addressing inadequate maintenance and other operational issues that can lead to system failures.

Decentralization of the water and wastewater distribution network in the city can reduce water loss due to leakage in long-distance pipe networks. Asghari et al. [56] state that by distributing the network, the system becomes more flexible and leakage losses are minimized in long lines of pipes. System upgrades, involving improving and building the physical infrastructure of urban water infrastructure, such as replacing old pipes, can more effectively manage water loss caused by leakage and other waste, improving efficiency and resilience. In a decentralized system, these improvements could be more cost-effective than upgrading a centralized distribution system, as shorter pipe lengths would be required. The implementation of digital technology, such as smart grids, is also crucial to reducing losses. Smart grid modeling is a valuable tool that allows the detection of potential problems before they escalate, thereby optimizing system performance and enabling more informed decisions about infrastructure investments. The integration of IoT devices, wireless sensors and remote sensing applications with existing systems enables the development of intelligent systems that use forecasting models to monitor and manage water quality, detect leaks and blockages in near real-time, providing operators with the information they need to take appropriate action. Incorporating intelligent sensors and monitoring systems can help detect leaks and failures, resulting in a more agile and intelligent regime for system repairs and maintenance.

In the context of Wang et al. [57], the term ‘loss of fresh groundwater lenses’ refers to the decrease in the amount of exploitable freshwater due to overexploitation, which in turn results in saltwater intrusion. In contrast to Lu et al. [58], the loss is the reduction of water area or volume due to evaporation or decreased water flow from tributary rivers. Determining the minimum ecological water demand aims to prevent these losses, which endanger the survival of the lake. For urban water systems, losses are defined as leaks in drinking water distribution networks, as well as inefficiencies in wastewater treatment or suboptimal reuse of reclaimed water, as outlined by Zuo et al. [59]. According to Ju et al. [60], the impact on river ecosystems can be summarized as follows: 1. Degradation of aquatic habitats, 2. Decrease in biodiversity, 3. Reduction in ecological flow, which is necessary to maintain the health of the ecosystem

3.3.2. Optimization of Water Distribution

The process of evaluating the performance of a water distribution network by incorporating contextual and future uncertainty through the use of scenario building. The methodology begins with an analysis of the system context in order to identify the key factors that impact the performance of the distribution network. Crucially, planning encompasses the establishment of objectives, criteria, and metrics. Optimization requires a clear understanding of what is to be achieved and how success will be measured. According to Carneiro et al. [61], exploratory scenarios are developed to analyze how different political, economic, sociocultural, technological, environmental, and internal management factors could affect the system in the future. By modeling the behavior in these different scenarios, possible variations in hydraulic variables and performance indicators can be identified. In addition, the weight of temporal uncertainty, similar to a discount rate, must be considered to reduce the relevance of performance results in the distant future. This approach acknowledges the inherently higher uncertainty in future predictions and helps prioritize optimization decisions in the short and medium term, while also considering the potential long-term consequences identified in the different scenarios. When seeking optimal solutions, it is possible to consider not only the cost or technical performance but also the robustness of the solution to different future scenarios.

3.3.3. Impact on Urban Resilience

This is based on Feng et al. [62] under ecological constraints. A high index indicates less encroachment on ecological infrastructure, thus strengthening urban adaptability. However, unplanned growth can diminish this resilience by reducing ecological spaces and increasing environmental sensitivity. In contrast, density resilience, measured by ecological footprint, evaluates the balance between human demand and ecosystem capacity. A high ecological deficit reduces this resilience, limiting sustainable development. Similarly, morphological resilience is a function of the spatial balance between “source” and “sink” landscapes. An optimal spatial distribution strengthens resilience, while an unbalanced allocation weakens ecological permeability and robustness of the system. Finally, functional resilience is linked to the diversity of ecosystem functions. Greater diversity and interaction between functions reduce the risk of urban collapse. The loss of ecosystem services has a detrimental effect on this resilience. Taken together, these factors emphasize that sustainable urban planning must strike a balance between growth and ecological protection, avoiding overexploitation of resources and maintaining functional and spatial diversity.

Currently, a comprehensive understanding of the inherent resilience of the entire urban water system is being developed, rather than focusing only on individual subsystems. This is achieved through the use of simulation models that stress-test and assess resilience under long-term uncertainties. These models consider water resource management models, hydraulic distribution models, and water demand generation models, using a stochastic assessment methodology to model disturbances. As Sitzenfrei et al. [63] point out, green infrastructure is gaining ground as it seeks to improve the sustainability and resilience of urban drainage systems, bringing multiple benefits such as reducing flood risk and improving water quality. In conjunction with smart rainwater harvesting, it is also considered a strategy to reduce potable water consumption and improve urban drainage performance. A decentralized approach to drainage networks and smaller wastewater treatment plants is suggested to increase flexibility and adaptability, increasing structural resilience, and possible cost reduction.

The different scales of urban water systems have a significant impact on their overall resilience to various challenges. Resilience is defined as the ability to anticipate variability, absorb disturbances, adapt to changing conditions, maintain functionality, and recover from disruptions, ensuring present and future services while protecting nature. As Arnaud et al. [64] point out, this concept of resilience can be described through three capacities: absorption, adaptation, and restoration. Centralized Water Systems (SUACs): These systems consist of large treatment plants, extensive distribution networks, and collection systems that transport water over long distances. These systems typically feature a linear design that encompasses the entire process from freshwater intake to the discharge of wastewater and stormwater into water bodies. In contrast, Decentralized Water Management Units (DWMUs) have gained significant prominence in the context of Integrated Urban Water Management and Water-Sensitive Urban Design. These focus on the use of local water sources, such as rainwater harvesting, gray-water reuse, and recycled water, promoting closed loops for nonpotable reuse and using green infrastructure for stormwater management. Finally, the Hybrid Urban Water Systems (SUAHs) are worthy of note: These systems seamlessly integrate UGADs into the existing centralized network, providing a robust backup in case of failures. They combine decentralized treatment and reuse at the site level and mid-scale with centralized facilities.

3.3.4. Comparison Between Decentralized and Centralized Systems

The implementation of SUAHs is a decentralised and flexible alternative that has gained relevance in urban water management, particularly in areas with scarce infrastructure or in contexts of high climate variability. However, a comparison with traditional centralized systems reveals multiple advantages and limitations that should be considered.

With regard to capital expenditure, HEDs generally necessitate reduced initial outlay when implemented on a small scale or in domestic settings. However, large-scale replication has the potential to result in a substantial escalation in the costs associated with coordination, monitoring, and standardized maintenance [65]. In contradistinction to decentralised systems, substantial upstream investments are required by centralised systems. However, it is important to note that these systems benefit from economies of scale, resulting in greater efficiency in metropolitan contexts.

In terms of scalability, Gude [66] points out centralized systems offer a robust solution for large urban territories due to their unified infrastructure and hierarchical planning. Conversely, decentralized systems necessitate modular and adaptive planning, which can render expansion challenging in the absence of adequate local governance.

Rodrigues et al. [67] mentions in the context of long-term maintenance, centralized systems offer distinct advantages in terms of operability and quality control, attributable to their homogeneous infrastructure. Conversely, SUAHs present considerable challenges, as they depend on local technical capacities to ensure operational efficiency and water quality, which can result in disparities between communities.

In summary, both approaches exhibit complementary strengths. A hybrid strategy that integrates decentralised solutions with the support of a centralized infrastructure can represent a resilient and sustainable path to contemporary water challenges.

3.3.5. Assessment of the Resilience of SUAHs to Extreme Weather Events

Recovery Time: The aforementioned interval corresponds to the time required for the system to restore normal operation after a disruptive event, such as flooding or drought.

Operational redundancy: The implementation of alternative routes or units is crucial to ensure the functionality of the system in the event of failure of one or more components.

Hydrological adaptability: The system has the ability to make dynamic adjustments to its operation through a RTC system, allowing it to adapt to changes in precipitation patterns efficiently and in a timely manner.

Energy sustainability: A key focus of this study will be on assessing the energy balance of the system, with particular attention given to its integration with renewable energies and low-consumption IoT sensors.

Functional decentralization index: Assessing the degree of autonomy of decentralized subsystems within the SUAH is a very useful measure for evaluating their operational independence in the event of possible failures in the general network.

3.4. Technology Integration

As stated by Morchid et al. [68], the objective of this technological integration in the management of water systems is to achieve robust and more efficient management for communities, as well as real-time monitoring of climate conditions and population needs. One example of this is embedded systems, which are responsible for collecting data from specialized sensors, such as the DHT22, which take temperature and humidity readings, soil moisture sensors, and water level sensors. These sensors communicate with a web server and have the ability to connect via Wi-Fi to various peripheral interfaces. In contrast, there is a constant search for minimizing the excessive use of water and ensuring its more equitable distribution in urban areas. The integration of sensors that capture data allows the development of analytical models to facilitate informed decision-making. This facilitates the automation of leak detection to avoid wasting valuable liters of water and the intelligent closing of valves through the Internet of Things (IoT). The IoT transmits data to field devices, which are received by the management platform through Server-Sent Events Protocols. This enables automatic and continuous updates to real-time data without the need for constant user requests. This is crucial for real-time sensor monitoring. Promoting the operational efficiency of water systems and reducing the costs associated with manual management. At the same time, we are committed to promoting sustainable practices and building resilience to the challenges of climate change.

3.4.1. IoT Sensors and Big Data for Real-Time Monitoring

The use of advanced remote sensing technologies and solutions based on the Internet of Things (IoT) to solve water problems is essential. The design of smart solutions that provide access to real-time data using state-of-the-art technology is a key. Solutions must be adaptable, modular, scalable, and easy to install for the end user. The use of mobile applications to analyze water coloration and monitor red tides through image processing is proposed. The use of unmanned aerial vehicles for water quality sampling in rivers, taking into account flight limits, power consumption, load, and river size, in combination with low-power, wide-range communication technology. Zulkifli et al. [69] have proposed the development of adaptable systems that integrate a variety of sensors without requiring complex interconnection techniques between the sensor and the main board of the data acquisition system. The text also highlights the importance of incorporating power and reverse polarity protection circuitry in the design of low-cost microcontroller boards to improve the protection and reliability of the data acquisition system. It is recommended to develop a software agent-based model for the monitoring of the health and consumption of the subway lines. This model would utilize intelligent agents to alert water engineers to any problems, enabling them to take a rapid control action and restore supplies.

Technological trends, with a particular focus on artificial intelligence, indicate that in the near future the integration of enhanced sensors capable of detecting a wider range of critical parameters in real time will be a reality. These sensors will be able to detect specific pollutants, such as E. coli, chlorine, and harmful algal blooms. The trend towards smaller, more accurate sensors with lower power consumption will also continue. The application of technologies such as NB-IoT (Narrowband Internet of Things) for low-power, long-range data transmission is a key trend, enabling the deployment of monitoring systems in more remote areas with less developed communication infrastructure. In addition, the cloud-based architecture for data storage, processing, and visualization is essential for the management and analysis of large volumes of data generated by distributed monitoring systems, as Wiryasaputra et al. [70] have also highlighted. However, there is an increasing move towards systems that allow more comprehensive real-time environmental monitoring, i.e., involving the integration of data from various sources such as meteorological stations and satellite data, to obtain a more holistic understanding of the state of water resources.

The Internet of Things (IoT) is transforming traditional water quality monitoring methods in several significant ways. Introduces a new dimension to field research, allowing researchers to access their data and information in real time from anywhere. Manual data collection is a more laborious and resource-intensive process. In addition, it facilitates the deployment of sensor networks within a designated region, allowing these networks to communicate with each other to provide comprehensive real-time insights. Raspberry Pi devices are capable of measuring various parameters, including temperature, oxygen, dissolved solids, turbidity, conductivity, redox potential, and pH, enabling continuous real-time monitoring of water quality. Miller et al. [71] highlight the components that comprise a figure of the Internet of Things system Figure 6. The information is collected and analyzed in programming languages such as Python to obtain information to monitor the health of lakes or rivers, detecting changes that could indicate problems such as algae blooms or changes in the ecosystem, and make informed water management decisions. The benefits of such systems include cost reduction, improved safety and reliability of drinking water supplies, significantly reduced water waste and lower energy consumption, and access to clean, safe, and sustainable water sources.

Flores et al. [72] developed a case study applied to the aquaculture industry, highlighting its role as a key activity that contributes significantly to global food security and as one of the main suppliers of fish inputs. A worldwide growth in consumption is accentuated, with an impact of 87.5 million tons of aquatic animals valued at USD 264.8 billion. The growing adoption of more efficient technological alternatives, mainly the use of sensors based on the Internet of Things, is also a key factor. The integration of these technologies has optimized processes and improved resource use through the collection of real-time data on pH, temperature and dissolved oxygen in conjunction with remote monitoring of water quality. The development of models based on artificial intelligence allows the prediction and timely detection of atypical levels of total ammonia nitrogen, which is crucial for the health of aquatic organisms. This is in contrast to conventional methods of sampling and laboratory analysis, which are very complicated and expensive and do not allow for the immediate detection of critical changes. The bibliometric analysis conducted indicates a significant 74.79% increase in water quality monitoring in aquaculture between 2020 and 2024.

The hydrochemical monitoring proposed by Wurl et al. [73] for seawater intrusion and hydrothermal activity uses various methodologies to define the impact of seawater intrusion, such as the “seawater fraction”, the “chromium-alkali indicators”, the “hydrochemical causes evolution diagram” and the “saltwater mixing index”. Copetti [74] proposes electrical conductivity as a fundamental indicator, emphasizing the necessity of incorporating hydrothermal activities when interpreting hydro-chemical data to formulate effective management strategies. Boryczko et al. [75] emphasize the use of hydraulic modeling and simulation for the effective management of water supply systems, with the objective of preventing failures. The simulation is carried out using the EPANET 2.0 software to estimate the consequences of failures and create risk maps of the water supply. As Tchórzewska-Cieślak et al. [76] emphasize, risk reduction measures include system modernization, alternative water sources, emergency capacity in water tanks, correction of water treatment technology, redesign of the water supply network, alternative energy sources, reservation of strategic network facilities, and introduction of remote monitoring and control of the system.

3.4.2. Hybrid IA + Renewable Energy Systems

Energy management strategies are crucial to optimize the use of renewable resources such as water. The overarching objective in this context is to reduce energy costs and emissions. As Zakariazadeh et al. [77] point out, an effective strategy is to reschedule the water demand to coincide with periods of lower electricity prices or when renewable energy generation is higher. This can be achieved by shifting the filling of storage tanks to a lower energy cost schedule or off-peak hours of the power grid, or by analyzing energy consumption to determine power factor correction, the use of high efficiency transformers and the integration of variable frequency drives with the pumps. Another alternative is through the application of mixed integer linear programming heuristics to minimize electrical energy costs in the operation of water distribution, determining when and how many pumps should be used to collect and distribute water.

• As stated in Wołosz et al. [78], some energy optimization modeling tools have been developed to decarbonize wastewater treatment plants. Another example is based on the circular economy, using used vegetable oil for cogeneration in wastewater treatment plants. This approach represents a way to optimize the use of resources and potentially reduce the demand for external energy in water treatment. It is also important to note that the analysis of water resources and energy consumption is essential for sustainable urban planning.
• Sahin et al. [79] explore innovative data-driven energy services and business models in the domestic building sector, proposing the use of pay-for-performance approaches to improve energy efficiency and create new markets for energy service providers. In contrast, Rangoni Gargano et al. [80] analyze the interrelationships between the water, energy and hydrogen production sectors, identifying major opportunities for hydrogen production from water and wastewater treatment processes, representing a technological innovation for sustainable resource management.

3.4.3. The Energy-Water Nexus in the Deployment of AIoT Technologies

The implementation of smart technologies, such as IoT sensors for water monitoring and AI algorithms for predictive analytics, has proven effective in optimizing water use efficiency, leak detection and real-time catchment optimization [81]. However, it is important to note that these solutions are associated with an increase in energy consumption, which poses a significant challenge in terms of environmental sustainability [82].

This phenomenon, termed the energy-water nexus, posits that enhancements in water efficiency through technological means may incur an additional energy expense. In order to address this tension, a number of strategies have been proposed, including the use of solar- or hybrid-powered systems, the design of low-power sensors, and the implementation of perimeter computing architectures, which reduce the need for constant data transfer to the cloud, which in turn reduces energy consumption [83].

In this regard, the selection of computationally efficient algorithms, such as decision trees or compressed LSTM-type lightweight networks, can reduce energy impact without compromising predictive capacity. These actions enable the alignment of the digitalisation of the water sector with the sustainability goals established by the Sustainable Development Goals (SDG 6 and SDG 7), thereby ensuring a balance between technological innovation and environmental responsibility [84].

3.5. Regulation of Water Management

Loukas and Garrote [85] state that the most significant global challenges in the management, policy and governance of water resources are as follows: strong competition for limited resources, regional disparities in water supply, wealth, growth of global water demand, depletion, pollution of surface and groundwater, and linked water stress induced by climate change. It is imperative that novel methods and approaches are developed for the integration of water resource management and protection. This requires the establishment of suitable policies and the delineation of viable governance structures in the face of increasing water demand. However, there are challenges that must be acknowledged and overcome to ensure a reliable and adequate supply of high-quality water for future generations. Currently, there is a basin-scale and potentially globally applicable Water Rights Analysis Package (WRAP) which is a modeling system used and updated for water allocation and planning. Provides a detailed simulation of water rights systems, international treaties, agreements, and operations of storage and conveyance facilities. On a global scale, the ANEMI3 model is an integrated global change assessment model that emphasizes the role of water resources. It is used to assess water supplies needed for population growth and the global economy, as well as to analyze future water stress and the effects of water quality on surface supply. In order to comprehend the intricacies of water governance across various levels—local, state, national, and even global—a Water Governance Complexity Framework has been put forward. This framework identifies the modes of governance at the local level that result from institutional interactions on different scales. This framework is regarded as a universally applicable pragmatic approach for the formulation of novel policies and legal reforms, or transitions to poly-centric governance models. At the scale of trans-boundary river basins, cooperation for drought management is of great relevance, as demonstrated by the definition of common drought and water scarcity indicators with the aim of developing a joint international drought management plan.

3.5.1. Water Governance

Water governance refers to the political, social, economic, and administrative systems established to develop and manage water resources and the provision of water-related services at different levels of society. It is important to note that the effective management of shared resources is contingent on collective decision-making and the implementation of effective environmental governance models. Currently, the governance of water resources is facing several challenges that have a detrimental effect on its effectiveness. Although developed nations have established regulatory frameworks for the management of water, emerging economies often lack coherent policies for the sustainable governance of water resources. Sáez-Ardura et al. [86] emphasize that regulatory efforts, both formal and informal, often do not converge into cohesive governance frameworks, creating significant challenges for effective water management. Additionally, the presence of multiple public and private stakeholders frequently complicates decision-making and inter-agency coordination efforts. In order to establish effective governance, greater community participation and integrated management approaches that address local contexts are required. Standardized technical norms often prove inadequate, as governance necessitates context-sensitive strategies to address specific socio-environmental challenges. The challenge is further compounded by the inherent complexities of socio-environmental regulation in the water domain. A key challenge lies in the interconnectedness of actors.

Lee et al. [87] propose the exploration of quantifying the impact of human activities such as antibiotic pollution, land use changes, and mining on water resources in order to formulate strategies to mitigate negative impacts. In contrast, Bonilla et al. [88] seek collaborative governance for the management of water resources, water conservation facilities, and socioeconomic systems within a river basin, advocating an integrated watershed management approach for sustainable development, highlighting the need for sustainable water management practices. Li and Wu [89] developed a linear programming model to optimally allocate water supplies from upstream reservoirs to meet water needs, which considers different storage capacities and hydrological conditions.

The following key implications should be taken into account in the context of environmental policy making and public health. Montuori et al. [90] emphasize the need for a continuous and comprehensive monitoring of pollution in various environmental matrices, given the carcinogenic risk. De Rosa et al. [90] agree with Montuori on the importance of regular monitoring of pollution to assess ecological risks over time. They confirm that environmental policies should prioritize and fund long-term monitoring programs to identify trends, evaluate the effectiveness of implemented measures, and detect new risks.

Wang et al. [91] propose a novel ecological risk assessment method. This is based on toxicity testing and the probability of ecological risk to the efficiency of water treatment plants. It introduces weighted cumulative indices that better reflect the impact of each water parameter on water quality and human health. Bărbulescu and Barbeș [92] observe that environmental agencies should take these methodological advances into consideration to improve the precision and relevance of risk assessments that inform policy.

The absence of an effective governance model at the three levels of government (federal, state, and municipal) is identified as a primary cause of water pollution and scarcity in many countries. This is due to the inability of the model to facilitate effective coordination in the adoption and implementation of coherent policies for water and wastewater management. This, coupled with the lack of a holistic perspective, hinders the adoption of sustainable approaches to water management. Although regulations are in place, de Anda and Shear [93] emphasize that compliance and proper enforcement alone are not sufficient to prevent water body pollution. Furthermore, the lack of transparency in the management of budget allocations for infrastructure and sanitation at the municipal level has led to distrust and complicates accountability. Changes in government can also lead to loss of capacity and interruption of long-term projects in the water and sanitation sector.

Smart Water Metering is an advanced water consumption measurement system that goes beyond traditional water meters, examining how water policies have influenced its adoption. Msamadya et al. [94] observe that this system measures the volume of water supplied during a specified period for billing purposes, thus providing water consumption data with high accuracy, resolution and frequency. This enables the application to enhance water demand management capabilities in response to various sociodemographic, contextual, external, and internal factors. It enables national governments to overcome supply-side management constraints. Performing operational improvements in stable water supply by reducing per capita consumption, leakage, and operational and maintenance costs. Despite the clear benefits, the expansion of this initiative has been slow due to a lack of regulation.

3.5.2. Water Resilience

According to Barth et al. [95], water resilience refers to the process of improving the financial and operational stability of water and wastewater utilities. This enables them to proactively plan for, manage, and recover from external hazards. Such threats are strongly related to climate-related risks, such as water stress and flooding, as well as aging infrastructure. To this end, it is recommended that long-term resilience be prioritized at the state level, with the objective of establishing dedicated resilience funds and providing the necessary support to direct more funds to long-term resilience projects. Defining goals to effectively combat resilience with possible mandates and incentives to help reduce pressure on utilities is also recommended. In addition to seeking funding to integrate new technologies into utilities, such as advanced metering infrastructure and predictive maintenance, which can reduce the risk of new technologies, it is essential to consider the financial implications of these changes.

• In this context, rainwater harvesting systems present a promising solution to increase water resilience in regions facing water scarcity. Based on research by Feloni and Nastos [96], various climate change scenarios have been considered, including factors such as catchment area, storage tank volume, and household size. By diversifying water sources, these systems can significantly reduce dependence on potentially vulnerable systems, such as water transport from the mainland or desalination plants. This decentralization makes the water supply more resilient to large-scale failures that can affect centralized infrastructures. They also reduce pressure on aquifers and surface water sources, can mitigate storm-water runoff and generate substantial savings in household water costs.
• According to Ptak-Wojciechowska et al. [97], the aspects of water have a fundamental influence on urban quality of life and are crucial to achieving high quality of life. However, these aspects are often underestimated or omitted in many quality of life assessment tools. The manner in which cities manage their water resources can determine their livability, safety and competitiveness. Disruptions or deficiencies in water services can have a detrimental effect on the health, well-being and daily lives of urban residents. The presence and quality of blue-green infrastructure, including parks, wetlands and sustainable drainage systems, are crucial for effective water and climate management in urban areas. This infrastructure plays a crucial role in rainwater regulation, mitigating the risk of flooding, and enhancing the quality of life for citizens by providing green and blue spaces. The importance of proximity and access to water bodies for various aspects of urban life, including water supply and sanitation, urban culture, health, well-being and disease prevention, should not be underestimated. Consequently, water pollution can lead to health problems, while access to blue spaces can have benefits for mental and physical well-being.
• In their research, Richards et al. [98] explored strategies to improve water resilience, with a particular emphasis on rainwater harvesting and cost-effective water treatment methods. These approaches are presented as crucial for adapting to water scarcity and future demand, with the goal of reducing pressure on existing water resources. A rooftop rainwater harvesting system is evaluated as part of a decentralized wastewater treatment system designed for a circular economy and a more reliable water supply. Rainwater harvesting is acknowledged to have the potential to complement existing water supplies and alleviate pressure on groundwater sources. It is also acknowledged that the use of rainwater that would otherwise be wasted can play a crucial role in ensuring water security and potentially mitigating flood risks. Rainwater harvesting has a long history and is encouraged for the recharge of wells, wellheads, and groundwater aquifers.
• Adaptive management, as defined by Bohlas-Haddad et al. [99], is a key strategy to increase water resilience and mitigate the adverse effects of climate change. As Baghban et al. [100] emphasized, it is essential to adopt adaptive strategies to address the significant risk of water shortage. In the context of climate change, it is essential that water infrastructure be able to adapt. However, the capacity to do so may be limited due to high costs and restricted opportunities. However, optimizing the operation of existing water systems and maintaining and improving infrastructure for regional and local transmission are vital to improve efficiency and thus contribute to resilience. Bozorg-Haddad et al. [101] state that water management policies must adapt to climate change in order to strengthen the region’s capacity to manage the water supply more effectively. This includes the management of inter-basin diversions, water recycling and optimizing water use. In disadvantaged communities, self-organization programs to strengthen water delivery contribute to people’s resilience to water stresses.
• Krzymowski [102] proposes that water diplomacy has considerable strategic importance in several aspects crucial to global development and international security. Together with Lu and Wang [103], he also emphasizes the crucial role of water diplomacy in the architecture of global security and conflict prevention. It has the potential to become an effective platform for international cooperation in the face of numerous current and future global water challenges. Tan et al. [104] propose a comprehensive approach that integrates preventive and reactive measures, along with mediation and the implementation of solutions. This approach is essential to ensure regional and global security. The absence of collaboration in the domain of water resource sharing has resulted in the emergence of hydro-hegemony in numerous regions worldwide. Water diplomacy is a tool that can be used to address these challenges and provide a comprehensive approach to international security. It is also necessary to support trans-boundary water management.

3.6. Others

Dumouchel [105] asserts that these misconceptions stem from the erroneous interpretation of AI as a singular, ’intelligent’ entity akin to humans. Instead, it should be regarded as a diverse array of technologies with distinct characteristics. This misperception hinders our ability to foresee the implications of its advancement and societal propagation, as well as to safeguard against the political and social risks it poses. The heterogeneity of technical innovations and research efforts in AI makes it complex to identify common characteristics to establish effective regulations. Attempting to control “intelligent” machines through ethics, either by warning about the ethical consequences of interaction or by promoting artificial moral agents (AMAs), presents serious difficulties. Furthermore, it is important to note that ethics does not equal regulation. It is crucial to acknowledge that regulations invariably carry a political dimension, since the advantages and disadvantages of technologies are not uniformly distributed across diverse social groups, thus reflecting existing power relations. It is therefore vital that regulation should focus on how AI technologies are introduced into the social world, at what price and with what consequences, seeking to protect the public from any potential dangers and optimizing the advantages. The fundamental differences in cognitive domains between humans and AI systems must be taken into account when formulating regulatory frameworks. In contrast, the adoption of AI-driven automation can result in a concentration of power among large companies that develop and implement these technologies. This shift in power relations between different actors and social groups is a central consequence of the growing importance of AI; the fundamental problem here is political, not ethical, or metaphysical.

3.6.1. Governance and Regulation of Artificial Intelligence

According to Lemke et al. [106] the implementation of AI based technologies in regions experiencing water scarcity gives rise to a number of significant ethical and governance challenges. Firstly, the collection, analysis and use of vast datasets require strong privacy and security safeguards, particularly in contexts where data protection legislation is absent or weak. Secondly, asymmetries in digital infrastructure and technical capabilities have the potential to accentuate existing disparities between urban and rural areas, or between developed and developing regions. This, in turn, can limit equitable access to advanced technological solutions. Similarly, Rehman et al. [106] point out the utilization of AI in critical decision-making processes, such as water resource allocation, has the potential to result in opaque outcomes if transparent and participatory governance mechanisms are not implemented. These challenges necessitate a regulatory framework that is ethical, inclusive, and adapted to local circumstances. This is to ensure that technological solutions do not reproduce or exacerbate pre-existing inequalities.

• According to Hogan and Lasek-Markey [107], an ethical framework based on human rights is the most effective approach to address the ethical challenges associated with AI and its regulation. As human rights frameworks can be considered the common denominator between law and ethics, they play a crucial role in the ethical-based legal governance of AI. The proposal has emerged in response to concerns regarding the practice of ‘ethics washing’, which refers to the superficial adoption of moral values by the technology industry, often through the implementation of voluntary codes of conduct.Currently, the EU AI Law 2024/1689 is in effect. Although the law is strongly committed to protecting fundamental rights, as expressed in the Charter of Fundamental Rights of the EU legal system, the question of implementation remains unresolved. However, by contextualizing them within a values-based framework, specifically the “European values” enshrined in Article 2 of the Treaty on European Union (TEU), such as respect for human dignity, freedom, democracy, equality, the rule of law, and respect for human rights, it may be possible to avoid the issue of “ethics washing”. This could provide an interpretative framework to support effective regulation of AI and avoid ethics washing.
• Bodini [108] has stated that the regulatory complexity of Generative Artificial Intelligence (GenAI) stems from its numerous applications and the variety of risks it introduces. Current regulatory efforts have primarily concentrated on issues such as data privacy, security, and transparency. The European Union’s AI Regulation classifies AI systems according to the potential damage they could cause, from minimal to unacceptable risk, imposing greater regulatory oversight on high-risk systems. In the United States (US), it is described as legally non-binding and evolving, with a primary focus on consumer and/or data privacy, prioritizing encouragement of innovation while considering security. The UK has outlined principles for AI regulation, encompassing safety, transparency, fairness, accountability, governance, con-testability and redress. Its approach is considered legally non-binding, but with oversight, the main priority is to balance innovation with safety. The People’s Republic of China (PRC) has introduced legally binding and sector-specific regulations related to recommendation algorithms (2022), deep synthesis (2023), and GenAI (2023). The primary focus in China is on state control and supervision.
• As Monasterio et al. [109] point out, the ethical governance of artificial intelligence is a topic of growing debate in a globalized scenario, where power relations and inequalities between countries and regions need to be addressed. While developed countries are leading the establishment of an ethical governance framework, the countries of the Global South face a situation of vulnerability and reliance that leads them to import digital technology, capital, and organizational models from the Global North. This situation, without ethical reflection, can have a significantly negative impact on their already excluded, oppressed, and discriminated populations. The geopolitics of AI demonstrates that technology and, particularly, AI, is being used as a weapon. The new world order is no longer based on geographic control, but on the control of data flows and connections to technology. A significant yet often unacknowledged aspect of this is the hidden workforce, comprising activities such as data tagging, moderating harmful content, and training machine learning models. This training is often contributed by end consumers who unwittingly act as “ghost workers” by training recognition algorithms free of charge. The philosophy of data-ism, which underlies the development of AI, considers that all data can be exploited and mined, which leads to people becoming data points. In light of this, there is an increasing call for the establishment of international guidelines for the ethical governance of AI, with a view to ensuring that human rights are respected. It is considered fundamental that AI systems are designed in a manner that aligns with these rights, which are intrinsic to the human condition.
• Quinn et al. [110] state that decision support systems are science-based tools that are considered necessary to manage and mitigate the harmful effects of water pollution under climate change. It is clear that decision support tools are very useful in addressing various types of problem related to water quality management. To illustrate this point, concrete examples of these problems are presented, showing the application of contemporary science and technology to overcome the associated challenges. Bornstein et al. [111] emphasize the pivotal role of stakeholders in supporting program and implementation initiatives, and the crucial need for their participation to guarantee the success of water quality management programs.

3.6.2. Regulatory Frameworks and Ethical Principles for the Adoption of AI in Water Governance

The implementation of artificial intelligence in the field of water management gives rise to a number of ethical and regulatory questions that must be addressed with the utmost urgency. From a governance perspective, it is imperative to establish principles that ensure equity, transparency and accountability in urban environments affected by water scarcity.

The OECD Recommendation on IA [112], which promotes responsible and human-centered use, is among the recommended regulatory frameworks. In this regard, UNESCO has proposed principles of inclusion, sustainability and non-discrimination, which are particularly relevant for regions with unequal water infrastructure.

From a more robust regulatory perspective, the European Union’s AI Act is a significant development in this field, as it classifies AI systems according to risk levels, providing a useful approach for evaluating technologies applied in the water sector [113]. This approach can be adapted to classify predictive consumption models, intelligent monitoring platforms or AI-based maintenance systems.

Furthermore, the utilization of a systems thinking approach is advocated as an ethical framework for intervention, in which the design and evaluation of technologies should encompass not only technical efficiency, but also their impact on equity of access, citizen empowerment and community resilience [114].

In summary, the regulation of AI for water governance must combine global principles with contextual adaptations that integrate local stakeholders, algorithmic transparency, data protection and socio-environmental auditing mechanisms.

3.6.3. Cloud Seeding

Cloud seeding is a form of weather modification that aims to increase rainfall by intervening in the micro-physical processes of specific cloud types, either from the air or on the ground. Al Hosari et al. [115] argue that the introduction of large artificial aerosol particles into clouds increases the absorption of liquid water available in the cloud with the objective of larger seeding particles triggering a “competition effect” that favors the production of large droplets that can trigger the collision-coalescence process and enhance rainfall generation. The implementation of this technology is more common in various countries in their arid and water-scarce regions.

• Negative ion-based cloud seeding technology works by enhancing the micro-physical processes of precipitation formation in multiple stages, which ultimately leads to increased precipitation on the ground downwind of the emitter. As described by Zheng et al. [116], a high direct current (DC) voltage is applied to a device with tip electrodes, creating aerosols with negative ions on the particles near the electrode. The plume of charged particles then moves with the upward current toward the cloud layer, which can form precipitation. This transport is facilitated by orographic lifting and dispersion by mechanical mixing, the atmospheric electric field, and possibly other turbulent processes, filling a relatively large volume of cloudy air. The charged particles act as condensation nuclei as a result of electrostatic forces, accelerating the nucleation of cloud droplets and ice crystals. These cloud droplets and ice crystals then grow until they are large enough to fall to the ground.
• The target area for successful cloud seeding is determined by calculating the dispersion and transport of the materials. Wang et al. [117] note that there are uncertainties in the input data for the diffusion model, such as the temperature and wind speed of the seeding layer derived from radiosondes, that may present limitations in accuracy or spatial and temporal representation. The process begins by entering the release rate in conjunction with the horizontal wind data into a diffusion model. With this information, it determines the diffusion of planting materials by time intervals using the advection transport algorithm. It is worth mentioning that for this calculation, a diffusion model of moving point sources based on the Lagrange methodology is previously applied. Finally, the displacement of the terminal fall of the raindrops is considered. During this process, there may be some deviation between the actual distribution of surface precipitation.
• As Wan et al. [118] have demonstrated, this technique can be employed from aircraft. They used mesoscale numerical modeling coupled with a catalytic process to simulate seven cloud-seeding operations, providing a scientific method to evaluate and forecast the potential of these operations. As Al Homoud et al. [119] also note, commonly used agents include silver iodide and sodium chloride to stimulate ice crystals to increase precipitation. The model proposed by Wan found that when water vapor conditions were suitable, the aerial dispersion of silver iodide significantly increased the large particle content of high-altitude ice crystals and snow, resulting in an increase in low-level rainwater content and, consequently, an increase in precipitation at the surface. In contrast, insufficient water vapor conditions hindered the triggering of effective precipitation by silver iodide dispersion.
• Water management considering climate modification, as outlined by Ma et al. [120], introduces a significant degree of complexity due to a variety of interconnected factors. The effectiveness of the cloud-seeding technique is not uniform and depends intrinsically on the specific characteristics of each watershed and reservoir. Furthermore, the efficiency of this technique is a function of the volume of water that can be captured by the increase in artificial rainfall, considering the direction and wind speed, which affect the spatial and temporal distribution of rainfall. Similarly, this increase leads to an increase in runoff, which encompasses variable complexity and is impacted by soil characteristics, land use, and vegetation, including evapotranspiration rates.

3.6.4. Environmental Risks of Rainfall Stimulation Technologies and the Role of IA

In this review, we have identified a number of patents that propose the use of aerosols and hygroscopic or nucleating chemical compounds to stimulate rainfall. While these techniques may offer short-term solutions to water scarcity, there are also environmental risks involved that must be given full consideration. The main risks associated with this include the disruption of local ecological cycles, the accumulation of chemical residues in soils and water bodies, and the possible runoff of polluting particles. These risks could have negative impacts on biodiversity and human health [121].

3.6.5. Adsorption Technology Is a Key Solution for the Treatment of Water and Wastewater

Adsorption technology is a promising solution for treating various types of pollutants in water and wastewater. Tran [122] notes that this technology has gained interest due to its cost-effectiveness and rapid removal capacity, generating minimal secondary pollutants. It is highly efficient in removing pollutants, even at low initial concentrations. Based on Elayadi et al. [123] advantages of this technology include its simple design, low cost, ecological benefits, high capacity, and the absence of secondary pollution generation. In contrast to processes such as coagulation and flocculation, adsorption techniques do not generate by-products such as sewage sludge, the discharge of which represents a significant waste management concern.

Tran [124] underlines the fact that adsorption is used for the removal of a wide variety of contaminants present in water, including emerging pollutants, radionuclides, potentially toxic metals and dyes. Paredes-Laverde et al. [125] state that machine learning processes and statistical physics models are applied to estimate the adsorption capacity or removal efficiency. De Oliveira et al. [126] discuss the kinetics, isotherms, thermodynamics and mechanisms of adsorption with a view to improving the technique.

3.6.6. Environmental Impact: Energy and Water Consumption

Luccioni and Yacine [127] conducted a thorough examination of the cost of machine learning system inference, identifying the primary factors that significantly impact energy consumption and carbon emissions. The energy consumption and emissions of a model depend on the type of task it performs. Classification tasks involving text or images tend to consume the least energy, with an average of 0.002 kWh per 1000 inferences for text classification and 0.007 kWh for image classification. Generative tasks consume more energy; for example, text generation consumes 0.05 kWh per 1000 inferences. Multi-modal tasks such as image captioning and image generation represent the highest energy consumption, at 2.9 kWh per 1000 inferences.

Ren [128] draws attention to the fact that the consumption of water by AI can be divided into two main stages: 1. On-site water consumption refers to the utilization of water in data centers for the purpose of cooling servers. This often involves cooling towers that rely on water evaporation or the use of outside air with evaporation assistance and humidity control. The on-chip liquid cooling system also facilitates the transfer of heat to the data center cooling system. This heat is then rejected using water-consuming methods such as cooling towers or outside air. 2. With regard to external water consumption, please see the following information: The electricity that powers the data centers is generated using water. In addition to operational consumption, there is the water consumption embedded in AI supply chains, with approximately 2200 gallons of ultrapure water needed to produce one microchip.

Li et al. [129] highlight that the proliferation of AI products and services is a pivotal factor contributing to the accelerated rise in data center water consumption. AI represents the fastest growing workloads in data centers. If not adequately addressed, the water footprint of AI has the potential to become a significant obstacle to sustainability and generate social conflict, given the extremely limited and unevenly distributed freshwater resources suitable for human use. Training a large language model such as GPT-3 can consume a total of 5.4 million liters of water, including 700,000 L of direct consumption for training purposes alone. Each conversation with the GPT-3 model requires approximately 500 milliliters of water for every 10 to 50 responses of average length, depending on the deployment parameters. In 2023, the data center level of large technology companies saw one of their own data centers withdraw 29 billion liters and consume more than 23 billion liters of fresh water for cooling. Of this, almost 80% was potable water. This level of annual water consumption is comparable to that of a major beverage company. Projections indicate that the total water consumption of US data centers in 2028 could reach approximately 150 to 280 billion liters, further exacerbating pressure on water infrastructures.

O’Brien et al. [130] have highlighted that training AI models can have a substantial environmental impact, and the cost of semiconductors can be significant, leading to a notable increase in water consumption. In addition, data centers must use water-pumping systems to maintain optimal server temperature, particularly during periods of high heat. Microsoft reported a 34% increase in its global water consumption from 2021 to 2022, reaching approximately 1.7 billion gallons, while Google also reported a 20% growth in water use over the same period. In July 2022, OpenAI used approximately 11.5 million gallons of water to cool its GPT-4 training facility in Iowa. This constituted approximately 6% of all water used in the West Des Moines Water Works district, which also supplies drinking water to residents. Microsoft has announced that it is investing in research to measure the energy and carbon footprint of AI, while working on ways to make large systems more efficient in both training and application.

Saenko [131] notes that there are strategies to mitigate the environmental impact of using artificial intelligence, one of them being 1. Use of renewable energy: By introducing computing capabilities in areas where green energy is more prevalent or by scheduling computing operations to coincide with times of day when renewable energy is more readily available, emissions can be substantially reduced by 30% to 40% compared to a conventional grid that is heavily reliant on fossil fuels. Transparency and social pressure: It may be beneficial to encourage companies and research laboratories to disclose the carbon footprints of their AI models. In the future, consumers may use this information to select a ‘greener’ chatbot. 3. Geographical location: One method of reducing an organisation’s environmental impact is to perform computing in locations where green energy is more abundant.

3.6.7. Reducing the Water Footprint of AI in Contexts of Water Scarcity

Use efficient, lightweight models: Rather than relying on extensive architectures, it is advisable to employ compact models that have been optimised through techniques such as pruning, distillation, and transfer learning. This measure significantly reduces energy and water consumption without compromising accuracy [132].

Deployment in green data centres: It is vital to prioritise the training and deployment of models in infrastructures powered by renewable energy and located in areas with abundant water resources, in order to minimise the direct environmental impact on vulnerable regions [133].

Federated AI and edge computing: These technologies enable decentralised processing, reducing the need for centralised infrastructure and, consequently, the associated water footprint [134].

Sustainability assessment before deployment: In order to ensure efficiency and accountability in the management of water-related projects, it is imperative to incorporate sustainability metrics, including water use, as a fundamental criterion for the selection and application of models [135].

3.6.8. Technological Applications for Forced Water Retention

According to Suits et al. [136], the forced retention of water in urban environments is carried out by active systems that dynamically regulate the flow, retaining or releasing water depending on hydrological and climatic conditions. Slaboch et al. [137] highlight solutions such as pipes with integrated hydraulic control and forced retention stations equipped with sensors and valves controlled by smart grids. Firstly, Chen et al. [138] evaluated three configurations of forced retention and were able to model their efficiency by using neural networks, identifying key elements such as the flow reduction coefficient and the level of storage allowed. These technologies have been demonstrated to increase the resilience of urban areas by enabling proactive management of extreme rainfall events, optimizing water storage, and releasing flows in a controlled manner, thereby leading to a reduction in flood risk. The integration of AI and the IoT facilitates the implementation of adaptive retention, a process in which decisions regarding the retention of data are made in real time.This development has led to a substantial enhancement in the effective utilization of space, achieved by decreasing the necessary storage areas. This approach is of vital importance in densely populated urban environments with impervious infrastructure, where forced retention can be a key strategy for urban water resilience.

3.6.9. Channel Retention and Its Integration with Intelligent Technologies

As stated by Teschemacher et al. [139] channel retention is a vital technique in the design of sustainable urban drainage systems (SUDS). Its primary function is to reduce the velocity of water flow in open channels or artificial structures in order to promote infiltration, sedimentation and volume control. According to Kwon et al. [140] this technique is especially relevant in the context of heavy rainfall or sudden runoff events, contributing to flood mitigation and reducing pressure on urban sewerage networks. Boluwade [141] presents an account of recent developments incorporating water level and water quality sensors in retention channels. These intelligent systems, based on artificial intelligence (AI) algorithms, enable regulation of sluice gates, prediction of overflows, and transmission of real-time information to urban authorities on critical conditions. These systems represent a convergence between traditional hydraulic infrastructure and information technologies.

3.6.10. Leasing of Retention Potential Through Real-Time Control (RTC)

As Sun et al. [142] have highlighted in recent literature, the technique of water retention enhancement through real-time control (RTC) has been identified as a highly effective solution. Liang et al. [143] report that, for domestic rainwater harvesting tanks, a reduction in peak flow rates is possible in the range of 35% to 85%, depending on design and climatology. In contrast, Liang et al. [144] state that research applied to urban storage networks has revealed that the Target Flow Control Systems (TFCS) approach, an RTC method without basin calibration, improves infrastructure capacity and durability, based solely on real-time measured levels. As demonstrated, the integration of artificial intelligence with sensors has the potential to optimize available holding capacity, eliminating the need for new construction. In contrast, Brasil et al. [145] observe that hybrid green infrastructure systems, such as green roofs and bioretention, have demonstrated that retrofitting using RTC can enhance their effectiveness in runoff management and reduce hydraulic peaks. Jiali et al. [146] findings indicate that RTC is a significant emerging technique for optimizing retention capacity in various water infrastructures, particularly in conjunction with AI, IoT, and predictive modelling.

3.6.11. Systems Thinking

According to Zhichang [147], complexity science has led to the recognition that predictability in the natural and human domains is severely limited, if not completely impossible. This is due to key notions such as self-organization, emergence, non-linearity, and evolutionary attractor. Complexity arises from the interaction of a large number of agents at the local level, producing coherent patterns at the global level without an overall plan. The fundamental question in complex adaptive systems (CAS) is how these non-linear systems function to produce ordered patterns of behavior, especially novel patterns, in the absence of an overall blueprint. In contrast to systems thinking, which criticizes dualism and spatial metaphor, the process perspective emphasizes the temporal, social, and dialectical nature of human consciousness. In this way of thinking, interactions give rise to further interactions, relationships evolve from previous relationships, and patterns emerge from patterns. These processes are continuous and self-referential, involving gesture-response and generalization-particularization.

• Critical systems thinking, as outlined by Jackson [148], has facilitated the development of systems thinking as a trans-disciplinary field. This approach has established methodologies for integrating diverse system approaches to promote effective intervention in complex social problem scenarios. In essence, the contributions of critical systems thinking demonstrate that success is achieved by integrating social sciences’ theories and practices through systems methodologies, thereby significantly enhancing the trans-disciplinary nature of systems thinking. This integration of approaches offers valuable insights that can be applied in various applied disciplines. Utilizing tools such as “system of systems methodologies”, metaphor analysis, and paradigm analysis, it can highlight the theoretical limitations of these disciplines and how these hinder the achievement of successful practical outcomes. This approach has been successfully implemented in fields such as evaluation research, logistics, information systems, knowledge management, and operations research/management science.
• In order to effectively address the water scarcity problem, the authors Gausdal and Hildrum [149] define trust-building processes in water technology business networks in several ways, facilitating both feature-based trust and process-based trust at different stages of network development. The first step is to facilitate exploratory meetings using dialogue conference methods to inspire, discuss common challenges, and foster interest in creating a network. These meetings allow companies to discover the challenges and possibilities in network cooperation. Subsequent network meetings are to be organized and facilitated using the Network IGP (Individual, Group, and Plenary) method and temporary groups. These dialogue processes are instrumental in the promotion of relationships and the establishment of initial trust. Following these meetings, the focus shifts to practical collaboration through teamwork and workshops. These interventions have been instrumental in turning companies from passive observers to active collaborators in high-risk projects, highlighting the pivotal role of action researchers in the promotion of trust within business networks.
• According to Dongping [149] the complex holism is a multilevel and multidimensional co-evolutionary perspective to understand organizational evolution and management. The “complex holism” is characterized by analyzing the existence of a network of causation between the levels of a complex system, overcoming the attempt of causal reductionism to eliminate downward causation, and it also recognizes the limitations of theoretical reduction and adds the autonomy of theories and the “up-duction” of theories as a supplement to reductive explanation, which allows analyzing the causality between society, organization and individuals to demonstrate that individualism is incomplete as a methodology to explain complex social organizational phenomena, which makes holism necessary.
• According to Wulun [150] systems thinking emerged as a transition from modernism to postmodernism, with Bertalanfy’s General Systems Theory as a key connection. It focuses on holism, seeking to understand the system as a whole. It defines a system as an aggregation of interconnected elements or an organic body with a specific function composed of interacting and independent parts, in turn being part of a larger system. It tends to view the system as an assembly composed of various parts and to understand it by reducing it to its parts, believing that the combination and analysis of these parts are necessary to understand the system. In contrast, complexity thinking marks the movement towards postmodernism and emerges with the development of complexity sciences, such as chaos theory and dissipative structure theory.
• Vahidi et al. [151] state that the Viable System Model is an operational model introduced by Stafford Beer in the field of cybernetics of management. It was presented as a multidisciplinary concept that combines elements of systems theory, control sciences, and management sciences. The model is based on the idea that organizations can exist independently and share a common structure inspired by natural systems, particularly the human nervous system. For an organization to thrive in a dynamic and ever-changing environment, it must emulate the key characteristics of living organisms that have demonstrated resilience and adaptability. The model is based on the existence of five interacting subsystems that are essential to the identity and survival of any viable system. The first subsystem is the operating units, the second is the autonomic nervous system, the third optimizes collective operations and seeks synergies between operating units, the fourth is analogous to the conscious nervous system, observing the environment, gathering information, and making predictions for optimal adaptation, and the fifth defines the identity of the system, its overall vision or reason for being, and establishes the policies and guidelines to be followed.

3.6.12. Applicability of the Viable System Model (VSM) in Urban Contexts with Limited Capacities

In order for VSM to be implemented in these contexts, the following strategies are suggested.

Modularization of the model: In order to facilitate effective monitoring and decision-making in a decentralized manner, even in situations with limited resources, it is imperative to divide the urban water system into manageable subsystems, such as catchment, distribution and governance [152].

Gradual implementation: VSM should be introduced in phases, starting with the monitoring function (System 3) with available data and accessible technologies (low-cost sensors, community surveys).

Use of artificial intelligence as a catalyst: AI has the potential to complement the scarcity of structured data through predictive inference, satellite image analysis and scenario modelling. This makes feasible approaches possible, even in the absence of consolidated databases [153].

Citizen participation and distributed governance: Collaboration in the creation of indicators and protocols with local stakeholders is essential for the collection of qualitative information and the generation of legitimacy in weak governance contexts [154].

In other words, VSM has the capacity to adapt to complex urban contexts if implemented with institutional sensitivity, appropriate technology, and a progressive approach. Its value as a tool for diagnosis, organizational learning, and resilient design makes it a powerful framework for addressing urban water scarcity, even in disadvantaged environments.

Then, analyzing the most relevant sources of the last 5 years, Figure 7 shows the bibliometric visualization of the literature where the 36,746 existing articles in Scopus are shown, highlighting the central node that links to the main ideas of the systematic review of the literature; if we follow the 4 quadrants we can observe that in quadrant I water harvesting is strongly correlated with climate change and emerging systems to combat drought, followed by quadrant II, Then in quadrant III we observe research into water treatment through catalysis, photodegradation, to name but a few, and finally in quadrant IV we observe the correlation with biochemistry such as metabolism, biomass, and genetics.

In addition, the yellow color of the nodes indicates the most recent articles. In other words, water remains an entity that needs to be continuously studied in all areas.

The purpose and scope of the bibliometric map (Figure 7) is to provide an enriched overview of the field through a keyword concurrence analysis. The construction of this bibliometric map is detailed in the flowchart, which can be found in Figure 8.

This analysis was performed for exploratory and visual purposes only and did not influence the final selection of studies included in the systematic review. The syntax used in Scopus was as follows: TITLE-ABS-KEY (“water” AND “harvesting”) AND PUBYEAR > 2020 AND PUBYEAR < 2025.

It is imperative that the authors’ research can be verified through experiments carried out by third parties, as well as allowing the development of devices or technologies that address the problem through innovation processes. To this end, a second taxonomy was performed, identifying the most significant and innovative patents.

Lee and Lee [155] define a patent as a legal right granted in exchange for the disclosure of an invention, which allows the holder to prevent others from making, using, or selling the invention for a specified period of time. Inventors often obtain patents to protect their innovations; companies use them not only to protect their intellectual property, but also as key intellectual assets to ensure market dominance, i.e., by acquiring patents, companies can achieve various strategic competitive advantages; likewise, patents serve primarily as a barrier against unauthorized copying by competitors or outside entities, making them critical tools to help companies develop, manage, and market their technological advances.

Intellectual property plays a fundamental role in meeting the challenges of innovation and economic growth by stimulating research and development (R&D). Beltran-Urvina et al. [156] mention that there is a broad consensus among researchers on the need for optimal management of intellectual property rights, that minimum levels of protection are essential to sustain innovation and its long-term growth, while excessive levels could hinder the development of innovation, and that it is important to strike a balance in the use and management of these rights.

For organizations, sustainable development means a form of management that simultaneously and equally takes into account the economic, environmental, and social aspects related to their operations. Deren and Skonieczny [157] introduce the concept of “green” intellectual property, which refers to the protection of innovations in the field of green technology. It is a concept in which innovations that are useful to the environment in one way or another are legally protected and serve as a strategic resource for an organization working toward sustainable economic development. It allows the design of a strategy based on the definition of objectives and actions to ensure homeostasis between economic, social, and environmental development. They also define the concept of “green intellectual entrepreneurship”, which is defined as a continuous process of creation and transformation of an organization’s green intellectual potential and capital.

One of the challenges identified by Hamza and Pradana [158] is that overcoming intellectual property barriers to the use of large amounts of data is a significant obstacle, particularly in applications involving multiple databases. Despite existing legal agreements related to security policies and intellectual property rights in this context, there are still complex issues that require further protection. Therefore, the emergence of big data requires the development of new conceptual frameworks, security standards, and laws that address intellectual property challenges. It should be noted that intellectual property rights are becoming a crucial tool for companies to control their information assets, exploit R&D results, facilitate technology licensing and promote better or new products and services based on cross-platform service innovation.

Mechanisms to prevent monopolisation

• With regard to fundamental technologies related to drinking water, treatment and monitoring, the use of open licences is recommended.
• In the area of international agreements specifically, the inclusion of mandatory clauses assumes even greater importance in cases where water emergencies are a possibility.
• There is a need to implement multisectoral financial resources aimed at promoting technology transfer between South-South and North-South regions under equitable schemes.
• Public bodies and multidisciplinary entities are encouraged to implement Creative Commons licences and other models of collaboration in technological development.

It is important to note that the various existing alternatives to combating water scarcity are not a recent phenomenon. There have been inventors who have tried to develop technologies for water harvesting, cloud seeding and the use of chemical mixtures, such as aerosols, to stimulate artificial rain. Other technologies include atmospheric generators that capture moisture from the air.

For this reason, the time-frame was broadened to cover the period from 1963 to 2024. This allows us to trace a technological timeline showing the evolution of knowledge and the different solutions proposed from a technical and innovative perspective. This enables us to identify the main development trends, technological transitions, and opportunities to integrate emerging approaches, such as artificial intelligence, in

Table 2. Table of patents filed to combat water scarcity.

Title	Inventors	Abstract	Publication Number
Systems And Methods For Plant-Localized Atmospheric Water Harvesting [159]	Cavote Collin et al.	A water harvesting system captures atmospheric moisture using a plant-connected structure with air inlet and outlet vents linked by an internal air conduit.	WO2025059225A1: https://worldwide.espacenet.com/patent/search/family/095022529/publication/WO2025059225A1?q=pn%3DWO2025059225A1 (accessed on 11 September 2024)
Scalable And Passive Mof Assisted Atmospheric Water Harvester [160]	Yaghi Omar and Song Woochul	A passive, scalable water harvester uses sorbents and natural sunlight to extract drinkable water from air without external energy input.	WO2024167697A1: https://worldwide.espacenet.com/patent/search/family/092263333/publication/WO2024167697A1?q=pn%3DWO2024167697A1 (accessed on 26 January 2024)
Automatic harvesting depth adjusting device of water chestnut harvesting machine [161]	Ge Shuo et al.	An automatic device adjusts the depth of a water chestnut harvester using a sliding vertical rod, sensors, and an electromagnetic valve for precise control.	CN221203341U: https://worldwide.espacenet.com/patent/search/family/091543274/publication/CN221203341U?q=pn%3DCN221203341U (accessed on 30 November 2023)
Water Harvesting And Collecting Multifunctional All-In-One Machine [162]	Luo Huan et al.	The invention is a multifunctional overwater cleaning and harvesting machine featuring a slanted front pressing plate, a U-shaped fixing rod, and a lifting seat with a rotating head and clamping piece.	CN117099563A: https://worldwide.espacenet.com/patent/search/family/088800076/publication/CN117099563A?q=pn%3DCN117099563A (accessed on 13 September 2023)
Atmospheric Water Harvesting Using Fog Harvesting Fabric [163]	Waite Tyler et al.	The system uses atmospheric data to identify optimal conditions for harvesting water from the air, enabling efficient and responsive water collection.	US2025032977A1: https://worldwide.espacenet.com/patent/search/family/094373502/publication/US2025032977A1?q=pn%3DUS2025032977A1 (accessed on 25 July 2023)
Methods And Apparatuses For Harvesting Water From Air [164]	Richard Boudreault	The system uses porous walls where air flows cause water to condense on one side, then seep through the pores to a collecting surface for water collection.	US12186702B2: https://worldwide.espacenet.com/patent/search/family/066438719/publication/US12186702B2?q=pn%3DUS12186702B2 (accessed on 1 April 2023)
Atmospheric Water Harvester Having Subcooler Heat Exchanger [165]	Kapustin Levgen et al.	A water harvester uses a secondary heat exchanger to improve efficiency by absorbing ambient heat during desorption and releasing excess heat back to the environment.	US2024246026A1: https://worldwide.espacenet.com/patent/search/family/091951861/publication/US2024246026A1?q=pn%3DUS2024246026A1 (accessed on 19 January 2023)
Atmospheric Water Harvesting Generator [166]	Do Yoonseo	An atmospheric water generator uses a nanoporous adsorbent and amphiphilic substrate to efficiently extract water in dry climates, operating with minimal power and simple airflow control.	US11745117B1: https://worldwide.espacenet.com/patent/search/family/085985215/publication/US11745117B1?q=pn%3DUS11745117B1 (accessed on 22 December 2022)
Heat Pump-Based Water Harvesting Systems, And Methods Of Using There [167]	Kuo David and Kapustin Ievgen	WHS are designed to capture moisture from air while minimizing energy use and enhancing the efficiency of the harvesting cycle.	US12151199B2: https://worldwide.espacenet.com/patent/search/family/083848631/publication/US12151199B2?q=pn%3DUS12151199B2 (accessed on 8 December 2022)
Heat Pump-Based Water Harvesting Systems [168]	Kapustin Ievgen and Kuo David	The system employs sorbent materials, such as metal-organic frameworks, to extract water from the air. The captured water is desorbed as vapor, condensed into liquid, and then utilized to dehumidify or humidify the air within the system.	US2023063572A1: https://worldwide.espacenet.com/patent/search/family/085230590/publication/US2023063572A1?q=pn%3DUS2023063572A1 (accessed on 5 August 2022)
Atmospheric Water Harvester With Climate-Adjustable Adsorbant Properties [169]	Kapustin Ievgen et al.	This disclosure presents atmospheric water harvesting systems designed with an optimal adsorption threshold, determined by energy cost and water availability factors.	US11536010B2: https://worldwide.espacenet.com/patent/search/family/082549029/publication/US11536010B2?q=pn%3DUS11536010B2 (accessed on 18 July 2022)
Atmospheric Water Generation Systems And Methods [170]	Stuckenberg David et al.	The system utilizes a fluid-desiccant flow that is cooled to absorb water vapor from atmospheric air. The desiccant is subsequently heated to release the vapor into a controlled air stream circulating within the system.	US2023010376A1: https://worldwide.espacenet.com/patent/search/family/082742750/publication/US2023010376A1?q=pn%3DUS2023010376A1 (accessed on 6 July 2022)
Atmospheric Water Harvesting Apparatus [171]	Wood Ryan	The catcher consists of a cylindrical wall surrounding a central post, with an air passageway that has an inlet and outlet. The inner surface of the wall is coated with a triboelectric material to facilitate specific functional interactions	US12060698B2: https://worldwide.espacenet.com/patent/search/family/083116010/publication/US12060698B2?q=pn%3DUS12060698B2 (accessed on 2 March 2022)
Atmospheric Water Harvesting System With Cross-Flow Configuration [172]	Yaghi Omar et al.	Atmospheric WHS use a sorbent cartridge with permeable trays and spacers arranged to allow cross-flow, enabling efficient adsorption and desorption of water through airflow pathways.	US12000122B2: https://worldwide.espacenet.com/patent/search/family/074660754/publication/US12000122B2?q=pn%3DUS12000122B2 (accessed on 16 February 2022)
Water Caltrop Cultivation System Provided With Water Caltrop Harvesting Equipment [173]	Shi Lin et al.	The system includes a pool filled with water, with a water caltrop cultivation area on the surface. A fish culture area is created by water between the cultivation area and the pool bottom.	CN216415143U: https://worldwide.espacenet.com/patent/search/family/081339980/publication/CN216415143U?q=pn%3DCN216415143U (accessed on 23 November 2021)
Atmospheric Water Harvesting System [174]	Yu Guihua et al.	Improved soils for agriculture are disclosed, incorporating atmospheric WHS networks that efficiently hydrate the soil, reducing the need for external water during germination and growth stages.	US2023365865A1: https://worldwide.espacenet.com/patent/search/family/080950963/publication/US2023365865A1?q=pn%3DUS2023365865A1 (accessed on 4 October 2021)
Positive And Negative Electrode Type Artificial Rainfall Device [175]	Peng Haiming	The invention relates to artificial rainfall devices and discloses a positive and negative electrode type artificial rainfall device.	US2023365865A1: https://worldwide.espacenet.com/patent/search/family/080950963/publication/US2023365865A1?q=pn%3DUS2023365865A1 (accessed on 1 September 2021)
Atmospheric Water And Carbon Dioxide Harvesting For Farming [176]	Marchon Bruno and Kapustin Eugene	The system uses materials, such as metal-organic frameworks, to capture water and carbon dioxide from the air. The captured water is desorbed as vapor, condensed into liquid water, while the carbon dioxide is also desorbed.	S2023302394A1: https://worldwide.espacenet.com/patent/search/family/080353830/publication/US2023302394A1?q=pn%3DUS2023302394A1 (accessed on 25 August 2021)
System For The Condensation Of Atmospheric Water Vapour [177]	Rivera Soria Cayetano	The system captures atmospheric water vapor with minimal energy by using a straight duct with an air impulsion system, a convergent nozzle, and a nozzle neck with a convergence zone for condensation.	EP4356997A1: https://worldwide.espacenet.com/patent/search/family/077910821/publication/EP4356997A1?q=pn%3DEP4356997A1 (accessed on 15 June 2021)
Atmospheric Water Harvesting Device And Method [178]	Perkin Richard	The chamber features a first and second plenum space separated by a partition. A rotating vessel with a base and sidewalls extends from the base to an opening, positioned within the chamber.	US2023228066A1: https://worldwide.espacenet.com/patent/search/family/070278225/publication/US2023228066A1?q=pn%3DUS2023228066A1 (accessed on 10 March 2021)
Atmospheric Water Harvester With High Efficiency, And Methods Of Using There of [179]	Smith Taber Hardesty et al.	The system captures water from the air using specific materials, desorbs it as vapor, condenses it into liquid water, and collects it for use as drinking water.	US2023338891A1: https://worldwide.espacenet.com/patent/search/family/077292637/publication/US2023338891A1?q=pn%3DUS2023338891A1 (accessed on 2 February 2021)
Apparatus For Electro-Spray Cloud Seeding [180]	Fluhrer Helmut et al.	An apparatus for mobile electro-spray cloud seeding includes a storage tank, at least one nozzle connected to the tank by a pipe or hose, and a high-voltage direct current source to charge droplets ejected from the nozzle.	EP3994976A1: https://worldwide.espacenet.com/patent/search/family/073288368/publication/EP3994976A1?q=pn%3DEP3994976A1 (accessed on 6 November 2020)
Calculation Method Of Total Artificial Precipitation In Seeding Area Compared To Non-Seeding Area [181]	Ro Yonghun et al.	A method is provided to calculate the total amount of artificial precipitation in seeded vs. non-seeded areas, improving the reliability of seeding effect data, observational data, and numerical model predictions from artificial precipitation experiments, ultimately enhancing water resource management through weather modification.	US2022113450A1: https://worldwide.espacenet.com/patent/search/family/078716786/publication/US2022113450A1?q=pn%3DUS2022113450A1 (accessed on 14 September 2020)
Rocket For Artificial Rainfall Using Ejection Hygroscopic Flare [182]	Park Ji Man et al.	The rocket for artificial rainfall uses an ejection hygroscopic flare and includes a rocket body with a parachute for descent, a hygroscopic flare discharge outlet, and a communication module to send and receive launch and ejection commands from a ground station.	US2022065599A1: https://worldwide.espacenet.com/patent/search/family/080353582/publication/US2022065599A1?q=pn%3DUS2022065599A1 (accessed on 28 August 2020)
Artificial Rainmaking By High Power Laser Initiation Endothermic Reactions Through Drone Aircraft Remote Control System [183]	Chopkar Shivshankar and Kanhuji	The invention presents a system and method for artificial rainmaking using a drone or unmanned aerial vehicle (UAV) controlled remotely, based on the principle of endothermic reaction.	AU2020101897A4: https://worldwide.espacenet.com/patent/search/family/072513331/publication/AU2020101897A4?q=pn%3DAU2020101897A4 (accessed on 19 August 2020)
A Method, Apparatus And System For Collecting Atmospheric Water [184]	Ward Jarrod and Reiter Gerard	The water collection system includes a reaction chamber with desiccant, heating means surrounding or integrated within the chamber, and an air inlet for efficient water collection.	AU2020204336B1: https://worldwide.espacenet.com/patent/search/family/072750413/publication/AU2020204336B1?q=pn%3DAU2020204336B1 (accessed on 29 June 2020)
Atmospheric Water Generation Systems And Methods [185]	Stuckenberg David	The desiccant flow is heated to promote the evaporation of water vapor, which is then directed into a controlled air stream circulating within the system.	US11000799B2: https://worldwide.espacenet.com/patent/search/family/060991620/publication/US11000799B2?q=pn%3DUS11000799B2 (accessed on 5 February 2020)
A Method Of Cloud Seeding With Use Of Ice-Nucleating Agents [186]	Anastasopoulos Ilias Gavriil	The invention relates to a cloud seeding method for controlling atmospheric precipitation (rain, snow, hail, fog) using ice-nucleating agents of natural, plant, or mineral origin.	GR20200100042A: https://worldwide.espacenet.com/patent/search/family/075639921/publication/GR20200100042A?q=pn%3DGR20200100042A (accessed on 29 January 2020)
Water Harvesting Systems, And Methods Of Using There [187]	Marchon Bruno et al.	The system desorbs water as vapor, which is then condensed into liquid and collected, making it suitable for use as drinking water.	US12054402B2: https://worldwide.espacenet.com/patent/search/family/071735523/publication/US12054402B2?q=pn%3DUS12054402B2 (accessed on 22 January 2020)
Method And System For Determining Cloud Seeding Potential [188]	Xue Lulin et al.	A method and system for determining cloud seeding potential involves receiving temperature and liquid water content (LWC) data.	US11256000B2: https://worldwide.espacenet.com/patent/search/family/071608878/publication/US11256000B2?q=pn%3DUS11256000B2 (accessed on 17 January 2019)
3d Reduced Graphene Oxide/Sio 2 Composite For Ice Nucleation [189]	Zou Linda and Liang Haoran	The invention introduces an ice-nucleating particle for cloud seeding, which initiates ice nucleation at −8 °C. The particle number increases rapidly as the temperature decreases.	US2022002159A1: https://worldwide.espacenet.com/patent/search/family/071613727/publication/US2022002159A1?q=pn%3DUS2022002159A1 (accessed on 14 January 2019)
Method And System For Autonomous Cloud Seeding [190]	Zaltzman Rafi	The disclosed system and method involve using an autonomous vehicle to seed clouds through a computer-implemented process.	WO2020121301A1: https://worldwide.espacenet.com/patent/search/family/065910817/publication/WO2020121301A1?q=pn%3DWO2020121301A1 (accessed on 13 December 2018)
Solar Water Harvesting Device [191]	Pokharna Sohil et al.	The device includes a thermoelectric module powered by electricity, a cold plate to condense water from humid air, and a heat sink for dissipating heat from the enclosure.	US2019368168A1: https://worldwide.espacenet.com/patent/search/family/068693226/publication/US2019368168A1?q=pn%3DUS2019368168A1 (accessed on 4 June 2018)
Method And System For Expressing Airborne Cloud Seeding Line Considering Cloud Water [192]	Chae Sanghee et al.	The method and apparatus disclosed are for defining a seeding line to artificially enhance rain in airborne experiments, taking cloud water into account.	EP3574746A1: https://worldwide.espacenet.com/patent/search/family/064426654/publication/EP3574746A1?q=pn%3DEP3574746A1 (accessed on 29 May 2018)
Portable Rainwater Harvesting System [193]	Armisen Bobo Pedro et al.	The system offers an efficient and economical solution, especially for regions with limited resources. It includes a rainwater collection device made of affordable, easily available, and maintainable components, making it a more accessible option than other known systems.	ES1215364U: https://worldwide.espacenet.com/patent/search/family/062791847/publication/ES1215364U?q=pn%3DES1215364U (accessed on 20 April 2018)
Cloud Seeding System Through The Use Of Hoses (Machine-Translation By Google Translate, Not Legally Binding) [194,195]	Munoz Saiz Manuel	The cloud seeding system uses vertically arranged hoses to inject ice particles or water vapor into the cloud or rain zone. These particles act on supercooled liquid water, causing it to melt or condense, ultimately inducing rainfall.	ES1217974U: https://worldwide.espacenet.com/patent/search/family/063580076/publication/ES1217974U?q=pn%3DES1217974U (accessed on 9 March 2018)
Atmospheric Water Harvesting System [196]	Yu Guihua et al.	The disclosed water harvesting networks enable moisture extraction and collection from the atmosphere without the need for electrical energy.	US11326327B2: https://worldwide.espacenet.com/patent/search/family/063107073/publication/US11326327B2?q=pn%3DUS11326327B2 (accessed on 9 February 2018)
Device For Seeding A Cloud Cell [197]	Cardi Philippe	A cloud seeding device that uses a pyrotechnic torch filled with an active substance to initiate precipitation.	US2020196539A1: https://worldwide.espacenet.com/patent/search/family/060202081/publication/US2020196539A1?q=pn%3DUS2020196539A1 (accessed on 6 June 2017)
Intelligent Systems For Weather Modification Programs [198,199]	Defelice Thomas Peter	The system gathers data on potential cloud locations and directs a vehicle to approach one or more of these clouds for seeding.	US10888051B2: https://worldwide.espacenet.com/patent/search/family/063710945/publication/US10888051B2?q=pn%3DUS10888051B2 (accessed on 11 April 2017)
Artificial Rainmaking By High Power Laser Initiation Endothermic Reactions Through Drone Aircraft Remote Control System [200]	Chopkar Shivshankar and Kanhuji	A system for artificial rainmaking using remotely controlled drones or UAVs, based on the principle of endothermic reaction.	WO2018167797A1: https://worldwide.espacenet.com/patent/search/family/063521878/publication/WO2018167797A1?q=pn%3DWO2018167797A1 (accessed on 15 March 2017)
Device For Seeding A Cloud Cell [201]	Cardi Philippe	A cloud seeding device equipped with systems to convey and deliver an active substance, using aerostatic components to meet safety and regulatory standards.	US2020178481A1: https://worldwide.espacenet.com/patent/search/family/056263955/publication/US2020178481A1?q=pn%3DUS2020178481A1 (accessed on 10 May 2016)
Rainwater Harvesting Roof For Water Storage Tank [202]	Delost Jeremy	The RWH roof features a water-impermeable cover with a central peak supported by a central structure. Near its outer edge, a permeable band allows water collection into the underlying storage tank.	US2016258137A1: https://worldwide.espacenet.com/patent/search/family/056850426/publication/US2016258137A1?q=pn%3DUS2016258137A1 (accessed on 4 March 2016)
Cloud-Seeding Operation Effect Radar Detection Time Sequence Contrast Analysis Method And System [203,204]	Yao Zhanyu	A method and device for analyzing the effects of cloud seeding using radar time sequence data. It compares seeded cloud units with similar natural cloud units by analyzing radar data after the seeding operation.	CN105353378A: https://worldwide.espacenet.com/patent/search/family/055329379/publication/CN105353378A?q=pn%3DCN105353378A (accessed on 15 October 2015)
Materials For Moisture Removal And Water Harvesting From Air [205]	Yeung King Lun and Ferdousi Shammi Akter	A moisture-removal material composed of a hydrophilic, microporous structure that holds a low water activity substance to efficiently harvest water from air.	US10486102B2: https://worldwide.espacenet.com/patent/search/family/055746136/publication/US10486102B2?q=pn%3DUS10486102B2 (accessed on 14 October 2015)
Method And System For Manually Influencing Weather [206]	Xu Wanchen et al.	A manual weather modification system composed of a floated pipeline, a floating platform, and a remote control platform for coordinated operation.	CN105075758A: https://worldwide.espacenet.com/patent/search/family/054557834/publication/CN105075758A?q=pn%3DCN105075758A
Apparatus And System For Smart Seeding Within Cloud Formations [207,208]	Martínez De La Escalera Lorenzo	A smart cloud seeding system using devices and materials to disperse micro- and nanoparticles of sodium chloride or similar compounds with high precision to induce rainfall.	US2016299254A1: https://worldwide.espacenet.com/patent/search/family/057112163/publication/US2016299254A1?q=pn%3DUS2016299254A1 (accessed on 13 April 2015)
Warm Cloud Catalyst, Preparation Method Thereof And Application Thereof [209,210]	Yan Yougui	A warm cloud catalyst, a preparation method thereof and an application thereof.	CN104322334A: https://worldwide.espacenet.com/patent/search/family/052397264/publication/CN104322334A?q=pn%3DCN104322334A (accessed on 30 September 2014)
Full-Automatic Rocket System Integrated With Radar Detection And Integrated Seeding [211]	Yu Ziping et al.	A fully automated rocket system with integrated radar detection and seeding capabilities, filling the gap in prior systems where cloud-seeding opportunities were missed due to failed communication between the aircraft and the ground station.	CN104412877A: https://worldwide.espacenet.com/patent/search/family/051009672/publication/CN104412877A?q=pn%3DCN104412877A (accessed on 10 September 2013)
Novel Artificial Influenced Weather System Designed By Ordnance Science Institute Of China [212]	Liu Jianping et al.	An artificial weather modification system using a pilotless aircraft to gather meteorological data and distribute materials for cloud seeding and rain induction, based on the analysis of the data to determine optimal rainfall conditions.	CN104412879A: https://worldwide.espacenet.com/patent/search/family/051505718/publication/CN104412879A?q=pn%3DCN104412879A (accessed on 10 September 2013)
Unmanned Aircraft System For Artificial Influence Type Weather Detection [213]	Yu Ziping et al.	Aircraft system for artificial weather modification, designed to address the high manpower demands of manned cloud seeding aircraft used for cloud seeding and rain induction.	CN104412878A: https://worldwide.espacenet.com/patent/search/family/051505717/publication/CN104412878A?q=pn%3DCN104412878A (accessed on 10 September 2013)
Hygroscopic Flare Composition For Weather Modification, And Cold Cloud Dissipation Method Using The Same [214]	Bang Ki Seok et al.	A combustion composition for weather modification, specifically designed to dissipate cold clouds (below 0 °C) using a silver iodide (AgI) seeding material and a tailored combustion composition.	KR101348115B1: https://worldwide.espacenet.com/patent/search/family/050144659/publication/KR101348115B1?q=pn%3DKR101348115B1 (accessed on 15 February 2013)
Method For Stopping Severe Weather Disasters [215,216]	Liao Yimin	A method for stopping severe weather disasters.	CN103766180A: https://worldwide.espacenet.com/patent/search/family/050559314/publication/CN103766180A?q=pn%3DCN103766180A (accessed on 19 October 2012)
Method And System For Accelerating Dissipation Of A Landfalling Tropical Cyclone [217]	Mardhekar Dhananjay	A method and system for accelerating the dissipation of tropical cyclons or hurricanes as they make landfall, reducing their strength and limiting their impact on further geography in an irreversible way.	US2014048613A1: https://worldwide.espacenet.com/patent/search/family/050099378/publication/US2014048613A1?q=pn%3DUS2014048613A1 (accessed on 9 August 2012)
Harvesting System Of Water Hyacinth Harvesting Ship [218]	Wei Guo et al.	The utility model discloses a harvesting system that allows for segmented harvesting of the roots and stem leaves of water hyacinth, facilitating their recycling.	CN202310590U: https://worldwide.espacenet.com/patent/search/family/046422607/publication/CN202310590U?q=pn%3DCN202310590U (accessed on 11 November 2011)
Water Harvesting Device [219]	Allan Scott	The WHS device consists of multiple longitudinal walls, with at least one wall adjacent to an existing structure. The walls define a water storage area with sufficient height to enhance water pressure, and it engages with at least one water source.	US2010096390A1: https://worldwide.espacenet.com/patent/search/family/042107824/publication/US2010096390A1?q=pn%3DUS2010096390A1 (accessed on 22 April 2010)
A Rainwater Harvesting System [220]	Brittain Graham	A RWH that includes an inlet cover tile, a concealed roof gulley with an outlet, pipe connectors, an overflow pipe, a bleed valve, a filter with a releasable access porthole, and a storage tank with a bull-nose shaped end.	GB2475924A: https://worldwide.espacenet.com/patent/search/family/041642059/publication/GB2475924A?q=pn%3DGB2475924A (accessed on 7 December 2009)
A Rainwater Harvesting System Comprising An Auxiliary Water Storage Tank And Float Valve [221]	Brittain Graham	A RWH, comprising one or more supplementary water storage tanks 1 to store and automatically release harvested rainwater.	GB2475923A: https://worldwide.espacenet.com/patent/search/family/041642057/publication/GB2475923A?q=pn%3DGB2475923A (accessed on 7 December 2009)
Atmospheric Water Harvesters [222]	Hill James et al.	The refrigeration system, using a gas vapor-based circuit, may feature a variable-speed compressor and a variable-speed fan or impeller to regulate air movement through the system.	US2009241580A1: https://worldwide.espacenet.com/patent/search/family/041115085/publication/US2009241580A1?q=pn%3DUS2009241580A1 (accessed on 1 October 2009)
Method And System For Modelling Water Treatment And Harvesting [223]	Mccarthy David Thomas et al.	The method comprising: receiving modelling data, the modelling data comprising a plurality of parameters relating to intended application of the water harvesting and treatment system.	US2011166848A1: https://worldwide.espacenet.com/patent/search/family/041443902/publication/US2011166848A1?q=pn%3DUS2011166848A1 (accessed on 24 June 2009)
Seeding And Verification Method For Targetted Cloud Seeding [224,225]	Jung Jae Won et al.	A method for targeted cloud seeding, which involves moving to the windward side of a target area, dispersing cloud seeds, and verifying the seeding effect using drop size distribution measurements and aviation radar, based on weather conditions.	WO2010071305A2: https://worldwide.espacenet.com/patent/search/family/042269201/publication/WO2010071305A2?q=pn%3DWO2010071305A2 (accessed on 19 December 2008)
High Altitude Atmospheric Alteration System And Method [226,227,228]	Chan Alistair et al.	The system and method include providing a high altitude conduit 700 and expelling a material through the conduit into the atmosphere at high altitude.	GB2448591A: https://worldwide.espacenet.com/patent/search/family/039433686/publication/GB2448591A?q=pn%3DGB2448591A (accessed on 18 April 2007)
Hurricane Mitigation By Combined Seeding With Condensation And Freezing Nuclei [229]	Rosenfeld Daniel et al.	A method for treating a tropical cyclone, a tropical storm or a tropical depression.	US2010170958A1: https://worldwide.espacenet.com/patent/search/family/039296065/publication/US2010170958A1?q=pn%3DUS2010170958A1 (accessed on 12 December 2006)
Charged Seed Cloud As A Method For Increasing Particle Collisions And For Scavenging Airborne Biological Agents And Other Contaminants [230,231]	Diehl Steve Ray	A system and method is provided for increasing interaction between seed particles in a seed cloud and target particles to be neutralized, detected or knocked down to the ground.	US2012123181A1: https://worldwide.espacenet.com/patent/search/family/039083128/publication/US2012123181A1?q=pn%3DUS2012123181A1 (accessed on 18 August 2006)
Weather Modification By Royal Rainmaking Technology [232]	His Majesty King Bhumibol	The “Royal Rainmaking Technology” process for weather modification involves four steps: “Triggering” to activate cloud formation, “Fattening” to promote cloud growth, “Moving” to direct clouds to a target area, and “Attacking” to initiate rainfall.	US2005056705A1: https://worldwide.espacenet.com/patent/search/family/035160534/publication/US2005056705A1?q=pn%3DUS2005056705A1 (accessed on 15 September 2003)
Method And Apparatus For Controlling Atmospheric Conditions [233,234]	Khain Alexander et al.	A method for controlling atmospheric conditions by encouraging controlled collisions between water droplets in the atmosphere, leading to their coalescence for weather modification.	WO03061370A1: https://worldwide.espacenet.com/patent/search/family/027590147/publication/WO03061370A1?q=pn%3DWO03061370A1 (accessed on 25 December 2001)
Method Of Modifying Weather [235]	Cordani Peter	The polymer is dispersed into the cloud, where the storm’s wind agitates the mixture, causing the polymer to absorb rain and form a gelatinous substance that precipitates, reducing the cloud’s ability to rain.	US6315213B1: https://worldwide.espacenet.com/patent/search/family/024396424/publication/US6315213B1?q=pn%3DUS6315213B1 (accessed on 21 June 2000)
Method And Apparatus For Modifying Supercooled Clouds [236]	Fukuta Norihiko	The invention critically examines factors affecting cloud seeding and introduces a new horizontal penetration seeding method. This method uses liquid homogeneous ice nucleants at lower levels of supercooled clouds to enhance microphysics-dynamics interaction and optimize phase change energy for improved cloud seeding.	CA2306651C: https://worldwide.espacenet.com/patent/search/family/025681758/publication/CA2306651C?q=pn%3DCA2306651C (accessed on 26 April 2000)
Device For Change Of Space Charge In Atmosphere [237]	Uskov	The system is designed for creating and destroying space charge in the atmosphere, with applications in meteorology and agriculture for cloud seeding, fog dispersion, and protection against drought, precipitation, and hail.	RU2124820C1: https://worldwide.espacenet.com/patent/search/family/020189734/publication/RU2124820C1?q=pn%3DRU2124820C1 (accessed on 7 February 1997)
Method And Apparatus For Modifying Supercooled Clouds [238]	Fukuta Norihiko	A horizontal penetration seeding method using liquid homogeneous ice nucleants at lower levels of supercooled clouds to improve cloud seeding by enhancing microphysics-dynamics interaction and optimizing phase change energy.	US6056203A: https://worldwide.espacenet.com/patent/search/family/021870011/publication/US6056203A?q=pn%3DUS6056203A (accessed on 15 December 1996)
Method Of Cloud Seeding [239]	Mather Graeme	Releasing hygroscopic particles from a seeding flare, created by burning a pyrotechnic composition containing potassium chlorate or potassium perchlorate as an oxidizing agent.	US5357865A: https://worldwide.espacenet.com/patent/search/family/025580552/publication/US5357865A?q=pn%3DUS5357865A (accessed on 22 February 1991)
Increasing Nucleation Activity With Lichens And Fungi [240]	Kieft Thomas	The invention enhances ice nucleation in liquids and gases using lichens, fungi, and ice nuclei derived from them, making it particularly useful for snowmaking, cloud seeding, and industrial freezing and cooling processes.	US5169783A: https://worldwide.espacenet.com/patent/search/family/022988426/publication/US5169783A?q=pn%3DUS5169783A (accessed on 20 October 1988)
Liquid Propane Generator For Cloud Seeding Apparatus [241]	Fukuta Norihiko and Milosevic Dragan	A liquid propane-filled bag in a chamber that applies continuous pressure as the propane exits for cloud seeding. A valve mechanism is triggered by a fuse ignition to release a lever action.	US4600147A: https://worldwide.espacenet.com/patent/search/family/024047148/publication/US4600147A?q=pn%3DUS4600147A (accessed on 18 July 1983)
Seeding Clouds To Make Rain, Esp. Using A Rocket—Using Silver Iodide And A Urea Compsn. [242,243]	Lacroix Tous Artifices	The process is used for forming rain, increasing the amt. of rain. preventing hail and for dispersing cloud.	FR2394979A1: https://worldwide.espacenet.com/patent/search/family/009192343/publication/FR2394979A1?q=pn%3DFR2394979A1 (accessed on 21 June 1977)
Pyrotechnic Cloud Seeding Composition [244]	Slusher Thomas	The pyrotechnic cloud seeding composition contains 60-85% silver iodate, 10-40% fuel (aluminium or magnesium), 5-15% binder, and 0.1-10% halogenated organic compound with a melting point of at least 50 °C.	US4096005A: https://worldwide.espacenet.com/patent/search/family/025192692/publication/US4096005A?q=pn%3DUS4096005A (accessed on 13 June 1977)
Combustion Monitoring System [245]	Solheim Fredrick and Erb Lee	The system uses an indicator light to show the combustion state, activated by a comparator. The comparator is connected to a thermocouple in the combustion chamber to monitor the combustion state.	US4062006A: https://worldwide.espacenet.com/patent/search/family/024729332/publication/US4062006A?q=pn%3DUS4062006A (accessed on 26 April 1976)
Cloud Seeding System [246]	Womack	A solid propellant rocket motor vehicle and launching means to transport seeding material to clouds for increasing precipitation and/or suppressing hail growth.	US3785557A: https://worldwide.espacenet.com/patent/search/family/023232454/publication/US3785557A?q=pn%3DUS3785557A (accessed on 21 December 1972)
Uniform Size Particle Generator [247]	Amand P. and Koff I.	A method and apparatus for generating fine particles, like silver iodide, using a nebulizer. The nebulizer maintains a constant level of a volatile solution, such as acetone with silver iodide, and uses an ultrasonic generator to produce micron-sized droplets for cloud seeding.	US3788543A: https://worldwide.espacenet.com/patent/search/family/023110485/publication/US3788543A?q=pn%3DUS3788543A (accessed on 14 September 1972)
Weather Modification Utilizing Microencapsulated Material [248]	Nelson Loren and Silverman Bernard	The hygroscopic chemical agent for cloud or fog seeding is contained in a liquid-permeable capsule shell, optimizing particle size for improved seeding effectiveness.	US3659785A: https://worldwide.espacenet.com/patent/search/family/022255789/publication/US3659785A?q=pn%3DUS3659785A (accessed on 8 December 1970)
A Method For Aerosolization With The Purpose Of Fog Clearing, Cloud Seeding And Weather Modification [249]	Consiglio Nazionale Ricerche	The composition and additive are located in containers which may be of alluminium or magnesium or alloys thereof, and the charge is provided with a fuze.	GB1110768A: https://worldwide.espacenet.com/patent/search/family/025673490/publication/GB1110768A?q=pn%3DGB1110768A (accessed on 9 April 1964)
Use Of Asphalt Coatings For Weather Modification [250]	Black James Francis	The materials used may be asphalt, a residual petroleum product, powdered gypsum, limestone or calcined lime and the coating may be discontinuous, part of the area being covered with a highly radiating material and part with a relatively low radiating material.	GB988109A: https://worldwide.espacenet.com/patent/search/family/009991301/publication/GB988109A?q=pn%3DGB988109A (accessed on 25 March 1963)

shows the most important inventions are highlighted in the table, clearly indicating the patent registration title, the inventor(s), a brief description of the invention, the publication number, access to the publication details, and the registration date. Where appropriate, access to official documentation filed with the patent office and diagrams are also provided to facilitate a better understanding of the invention.

The objective of this study is to analyze the patents consulted in Espacenet, a database that is managed by the European Patent Office (EPO). In order to achieve this objective, the reliability that is ensured by the standardized processes of registration, technical classification and access to the original documentation has been given due consideration.

The selection was based on a targeted review strategy that considered four primary criteria: first, thematic relevance to the study’s focus (IA, water optimization, urban resilience); second, the degree of technological innovation and practical applicability; third, the level of citation in technical and/or academic literature; and finally, evidence of implementation or subsequent development. This methodological approach facilitated the integration of historical patents with foundational influence and recent documents with transformative potential, thereby ensuring the maintenance of a coherent and analytically useful narrative for comparative purposes.

This study acknowledges that patent analysis does not ensure the effective implementation of the described technologies. For regulatory, economic or technical feasibility reasons, many inventions may remain at the conceptual stage.

Similarly, articles published before 2020 are only included if they make a fundamental contribution to our understanding of the studied problem or technology. This enriches the review from a historical and contextual perspective without compromising the integrity of the systematic approach applied to the main corpus.

4. Discussion

The first patent for cloud seeding was filed in 1963, and from this milestone various solutions to combat droughts have been presented. This patent development has spread across Europe, reaching Italy in 1964, and then the United States with a proposal in 1970. Although patents could still be obtained on an annual basis, there were no significant inventions to speak of. The patents identified above allow us to visualize the different solutions proposed by inventors based on their environment and resources. Some considered methods to stimulate artificial rain using aerosols or electrodes, while others suggested installing tanks to collect rainwater on residential roofs. In a similar vein, solutions such as air humidity capture represent waves of innovation. The application of artificial intelligence has the potential to enhance the functionality of existing rainwater harvesting systems. The Figure 9 illustrates the relationship between the list of systems and the research taxonomy.

The vectors in the Figure 9 resemble a continuous flow in scientific and technological development as addressed in the taxonomy. The use of artificial intelligence in combination with machine learning has been instrumental in refining existing techniques and enhancing the planning and design of robust infrastructure. This approach has been shown to improve the management of the complexity of different environmental agents that can be encountered in the water supply. The second vector, which addresses the issue of water scarcity, has been found to be a fruitful area of research. It is evident that each country faces unique challenges; for instance, Afghanistan is dealing with severe droughts, while Japan is dealing with frequent floods and water overflows. The review has highlighted the importance of examining these challenges from a cross-country perspective, with the aim of identifying and implementing solutions that can be applied in regions experiencing both situations. The third vector addresses sustainable solutions, and the governance of water and artificial intelligence must be set within a framework that must be respected so that we do not consume resources that belong to a nation and that the data with which artificial intelligence is trained is transparent and does not compromise people’s privacy. These initiatives are a significant step towards the creation of resilient and intelligent cities, aligning with the 2030 agenda established by the United Nations.

As is well known in the Figure 9 presents a convergence between science, scarcity and sustainability as the foundations for solutions guided by AI. It is imperative to recognize that technological innovation cannot be considered an independent solution. Indeed, techno-centric solutions run the risk of obscuring water management practices deeply rooted in local and indigenous knowledge. Examples include watershed management based on environmental observation, agricultural terrace systems, and the traditional use.

It is therefore crucial to move towards convergence between artificial intelligence and traditional forms of knowledge, fostering an epistemic dialogue that allows for the development of hybrid models of governance. This involves incorporating community perspectives into the design of algorithms, ensuring free, prior and informed consent, and validating data using cultural criteria, not just scientific ones. AI should be regarded not as a substitute for local knowledge, but as a tool to support community decision-making and adaptive capacity in the face of climate change.

As Michel Foucault’s perspective, drawing on Kimble and Bourdon [251], shows, discourse is key to understanding how power is exercised and how social control is achieved in the context of sustainable water. This enables an understanding of how regulatory discourses on transparency and corruption prevention influence systems of control. Furthermore, it emphasizes the importance of incorporating a range of perspectives during the establishment of power relations through discourse, thereby ensuring the consideration of diverse viewpoints.

Martinez [252] presents a proposal for knowledge management in the context of corporate water responsibility. This addresses both codified and uncodified knowledge related to the extraction, use, and disposal of water by companies and stakeholders. The proposed model aims to establish dialogues with a wider community of water-related stakeholders to respond to their needs and move toward a more holistic and inclusive knowledge management.

In contrast, Russ [253] presents the quantum organizational decision-making model, highlighting how misalignment of the time horizons of different actors can hinder the effective use of knowledge in relation to the sustainability of water systems. Simulation-based learning and role-playing facilitate consistency of time horizons and time frames of events, thereby improving the decision-making process.

There has been a notable surge in interest in the implementation of artificial intelligence in the field of water management. However, there is still a tension between its trans-formative capacity and its environmental impact. Bolón et al. [254] mentioned that Large-scale language models have proven to be highly energy and water intensive during their training phase, which raises questions about their viability and sustainability in contexts where water is precisely the scarcest resource. In this sense, it is a priority to promote the use of lighter, decentralized and specialized algorithms, capable of operating efficiently in limited infrastructures without compromising the accuracy and value of the results. This study addresses the problems faced by researchers, developers and policy makers seeking to implement AI technologies as a structural solution to water scarcity. These actors must navigate the critical balance between the operational efficiency offered by these intelligent systems and the environmental sustainability demanded by the current context.

Water is an indispensable resource for humankind, which is why we need to adopt a holistic approach to give it the importance it deserves. We need to explore new alternatives to combat scarcity and droughts, as proposed by Checkland. Based on Zexian [255] the Checkland approach is based on systems thinking, starting from the “two umbrellas” of emergence and hierarchy, as well as communication and control. This new approach extends systems thinking to complex systems thinking by introducing a “third umbrella”: self-organization and evolution. The integration of this third umbrella reflects advances in systems science and complexity studies.

While LSTM and SVM models demonstrate exceptional accuracy in predicting water demand, their opacity has prompted the development of tools such as SHAP and LIME, which facilitate comprehensible visualization of the influence of variables such as temperature, historical consumption and urban growth. This approach is essential to increase the confidence of decision-makers and stakeholders in the application of AI models in water management.

As illustrated in the Figure 10, the integration of systems thinking with patents is central to the proposal. The device used captures rainwater and air humidity, with considerations for evaporation losses and potential overflows. The result is the calculation of the available water, which is in agreement with the predictions made by artificial intelligence. This available net water is then stored in a tank and made available to the population.

The model was developed using Vensim software [256] in the student version to facilitate simulations that will determine the functionality of the proposed model. This facilitates the representation of phenomena such as feedback, causal loops, and accumulations associated with water collection, distribution, and scarcity. However, its ability to predict long-term behaviour under conditions of climate uncertainty has significant limitations.

Firstly, it is important to note that the model is sensitive to input values and initial assumptions. It is important to note that climate scenarios such as prolonged droughts, extreme events, or abrupt changes in precipitation patterns may not be adequately captured if they are not explicitly incorporated. Secondly, it should be noted that the deterministic nature of the model in question limits the incorporation of stochastic or non-linear variability, which is of vital importance in the context of the climate crisis we are facing.

To mitigate these limitations, future phases will involve multiple simulations, sensitivity analyses, and the integration of hybrid models that include agent-based or machine learning approaches. It is also recommended that adaptive scenarios with iterative updates be considered as new climate or policy information becomes available.

4.1. Limitations and Challenges in the Application of Emerging Technologies in Stormwater Management

The increasing interest in advanced technologies such as artificial intelligence, the Internet of Things, sensors and RTC to optimize urban water management is accompanied by significant economic barriers to their practical adoption. In accordance with the findings of Camargo et al. [257] concerning the selection of low-cost IoT sensors for water quality monitoring, it is reported that the cost of devices can range from USD 6 to 169. However, in certain specialized cases, the cost can exceed USD 500 per device, without taking into account additional expenses associated with calibration, integration and replacement. In a similar vein, the research conducted by Lit et al. [258] concentrates on densely populated cities. The findings of this research indicate that the establishment of a robust RTC network, which provides approximately two-thirds of the coverage, necessitates investments close to 1 million USD. This results in an improvement of only 37% in flood resilience over a 30 year cycle.

However, it should be noted that there are a number of hidden costs associated with the integration of IoT technology. These include the development of data management platforms, ongoing maintenance, cybersecurity and technical training, which typically represent 20% to 30% of initial expenditures in the first year of the project alone. The technology demonstrate that emerging technologies offer operational advantages. However, their implementation necessitates a thorough evaluation of the return on investment and the total budget, inclusive of maintenance and training costs. In municipalities with limited financial resources, these economic burdens can become significant constraints to their adoption.

4.2. Answers to the Research Questions

What are the environmental, economic and social impacts of using AI in water management?

In the domain of water harvesting, artificial intelligence (AI) emerges as a pivotal instrument, facilitating the optimization of diverse processes. In this regard, applications that are oriented towards the accurate prediction of rainfall, through the utilization of time series models such as LSTM or hybrid ARIMA, are particularly noteworthy. These models facilitate the estimation of rainfall patterns, thereby enhancing the efficacy of the design and operation of rainwater harvesting systems. Furthermore, the integration of evolutionary algorithms and neural networks is presented as a novel strategy for the optimization of these systems. Similarly, the geospatial identification of optimal areas for infrastructure installation is enhanced by the analysis of satellite images and machine learning algorithms. This approach enables the precise determination of suitable areas for implementing hydraulic systems. Consequently, this reduces structural failures or losses and optimizes the efficiency of water use, thereby maximizing the use of alternative water resources.

How has AI been integrated with other emerging technologies (IoT, Big Data, smart sensors) to improve water harvesting?

The implementation and adaptation of artificial intelligence in the urban environment has proven to be an effective strategy to mitigate water scarcity. This is achieved by optimizing water resource management through the following mechanisms: early leak detection, water demand forecasting and dynamic supply redistribution. This redistribution is carried out in accordance with consumption patterns and weather conditions in real time. Moreover, it has been demonstrated that this integration facilitates the incorporation of decentralized catchment systems with smart grids, thereby enabling a responsive approach to water stress events. However, the effectiveness of these applications is contingent on three key factors: the quality and availability of the data, the technical capacity of urban operators, and the digital governance and ethical framework that regulate the responsible use of algorithms.

What are the current trends in the scientific literature regarding the use of AI in water management?

Among the most relevant emerging technologies are the integration of IoT sensors for continuous monitoring of hydrometeorological parameters, the use of high-resolution satellite images combined with computer vision algorithms, and the use of edge computing techniques that enable local data processing in low-energy consumption devices. These solutions are complemented by low-weight machine learning models (e.g., decision trees or reduced neural networks) and the use of open platforms that allow system interoperability. Furthermore, there is an emerging trend towards hybrid models that integrate passive technologies, such as dew or fog collectors, with intelligent tools that optimize their performance in real time.

What public policies could facilitate the adoption of AI in water management to mitigate water scarcity in urban settings?

The integration of artificial intelligence (AI) in urban water management is encumbered by structural and operational challenges. The structural challenges that must be surmounted are manifold. These include the limited availability of reliable and representative data, especially in resource-poor areas; institutional and cultural resistance to technological change; high initial acquisition and training costs; and ethical concerns related to algorithmic transparency, data privacy, and the environmental impact of AI systems, particularly in relation to their carbon footprint and energy consumption. Operational challenges include the need to improve the quality and reliability of available data, as well as the creation of mechanisms to facilitate access to information by citizens. Furthermore, it is imperative to deliberate the repercussions of AI on environmental sustainability, encompassing its influence on the carbon footprint and energy consumption of systems. In order to surmount these challenges, it is imperative to implement inclusive public policies, establish clear regulatory frameworks and foster cooperation mechanisms between the public, private and community sectors of science.

What are the current trends in the scientific literature regarding the use of AI in water management later on?

Despite the increase in the number of publications, the existing literature still presents significant deficiencies. In the field of research, there is a predominance of theoretical or small-scale experimental validation studies, with limited evidence on the long-term implementation of artificial intelligence systems in real urban contexts. Likewise, there is limited exploration of the social and economic impact of these technologies in vulnerable communities, as well as an absence of analysis regarding technology transfer and the adaptability of the models to contexts with limited infrastructure. In addition, there is a need for further research on governance mechanisms, interoperability between platforms and sustainable financing mechanisms for their implementation.

5. Conclusions

The proposed taxonomic structure will organize existing knowledge and reveal research gaps. The convergence of smart technologies, regulation, urban integration and water resilience makes it clear that the solution to the scarcity problem requires a multidimensional approach. These findings establish a foundation for future research that aims to enhance adaptive and inclusive water governance.

A review of the scientific literature reveals that the development of water optimization strategies has begun to incorporate regional variability through data-driven approaches, especially through models trained with local historical precipitation records. In the domain of technology and innovation, an escalation in the utilization of distributed sensors, commonly designated as IoT, has been observed. These sensors have been incorporated into a variety of systems, enabling the dynamic adaptation of collection and storage strategies. This technological development has contributed significantly to the increased applicability of systems in variable climatic environments, such as those found in Central America and Southeast Asia.

Water scarcity is an issue that is becoming more prevalent due to various factors, including population growth and climate change. In response, governments and scientists must collaborate to identify sustainable solutions to address this challenge. Since ancient Roman times, advanced aqueducts have been used to manage the water supply. However, it is necessary to modernize techniques and infrastructure by leveraging advanced technologies such as artificial intelligence and machine learning to improve the resilience of urban water.

The integration of Information and Communication Technologies (ICTs) with established water harvesting techniques facilitates the storage of rainwater, thus mitigating the risk of catastrophes such as floods and landslides triggered by the force of water. Although these systems demonstrate high functionality during rainy seasons, advances in technology have led to the incorporation of warning systems that provide advance notification of impending rainfall, a measure that contributes to the prevention of overflows or substantial water losses. These systems are not limited to domestic use, but are also scalable to agricultural and industrial use. The architecture of intelligent catchment systems is another key support, as it allows the establishment of a complete system from the installation of hardware in the field that sends signals that are converted into data. This in turn facilitates the development of models to support decision-making.

Rapid advancements in science and technology, in conjunction with the digital era, have paved the way for the integration of sophisticated techniques such as artificial intelligence. This, in turn, has enabled the analysis of vast amounts of data in real time. Similarly, the development of sophisticated mathematical models, such as SVM (Support Vector Machine), Nearest Neighbor Analysis (K-nearest neighbors), Random Forest, XGBoost (eXtreme Gradient Boosting), Artificial Neural Networks (ANN) and LSTM (Long Short-Term Memory), has enabled the prediction of short-term water demand. These advances in mathematics facilitate informed decision making supported by data.

However, sustainable solutions are not only based on software or digital models. It is also vital to incorporate a tangible element that complements the system and generates results. Deficiencies in infrastructure were identified, such as abandonment of facilities and a lack of preventive and corrective maintenance. Water decentralization is essential to ensure that communities receive water equally. Therefore, it is vital to ensure that the capacity of the water supply systems is robust to prevent leaks or overflows, as well as the ability to direct wastewater to treatment plants for use other than human consumption. This type of infrastructure is known as hybrid water systems (SUAH).

The integration of the Internet of Things (IoT) technology, e.g., sensors detecting leaks or deficiencies in the system that emit a signal and trigger the opening or closing of valves in an automated manner, together with real-time monitoring, facilitates the analysis of large data sets to identify patterns and/or pathologies in the systems and enable effective, rapid reactions. In addition, there is a substantial reduction in associated costs.

Technological advances promote innovation and development (I + D), encouraging healthy competition that drives researchers to seek intellectual property protection for their inventions, utility models, processes, or algorithms. Identifying the 80 most significant patents from 1963 to 2024, we can visualize the progress and adoption of technologies worldwide over time. Given the integration of technologies, water is a human right. The unequal distribution of this resource between countries complicates water regulation and management. In terms of politics and governance, competition for limited resources is promoted, and the proposal of the ANEMI3 model, together with the Water Rights Analysis Package, promotes water allocation and planning.

Similarly, the governance and regulation of artificial intelligence is sought through the proposal of an ethical framework based on human rights. However, the existing AI Law of the European Union (EU2024/1689) is firmly committed to protecting fundamental rights.

In contexts where sectors such as health and education compete for the same funds, AI should be understood not as a technological luxury, but as an enabling tool that strengthens urban resilience, protects the human right to water and, ultimately, contributes to multisectoral sustainability. Therefore, its strategic incorporation must be accompanied by governance frameworks that prioritise social impact, equity and long-term returns.

The elevated water consumption, in conjunction with the substantial energy consumption and carbon emissions that are inherent to artificial intelligence technology, as it engenders water competition, must be given due consideration. Furthermore, there is a need for rigorous research into the optimization of resource consumption, including green energy, and alternative methods for regulating data center temperature.

Given that water is an essential resource for life and ecosystems, its nature and management present a high degree of complexity, thus rendering the integration of complexity sciences with systems thinking essential for the future of our field. The work of Stafford Beer’s Viable System Model exemplifies this integration, as it demonstrates the capacity for evolvement and adaptation to its environment.

In addition to technological interventions, it is essential to implement non-technological strategies in parallel to achieve true water resilience. It is essential to reform water governance to ensure transparent, equitable management based on principles of water justice in order to strengthen decision-making. Furthermore, the active involvement of local communities in the design, implementation and monitoring of solutions promotes the social adoption of technologies and their adaptation to specific contexts.

It is vital to strengthen technical and institutional capacities if AI-based solutions are to remain operational, scalable, and adaptable to changing conditions. Investing in education, technical training and inter-institutional collaboration networks is essential to enhancing organizational resilience, a fundamental aspect of responding effectively to water crises.

When combined with AI tools and other innovations, these non-technological strategies form a holistic and inclusive approach to water management in diverse urban contexts.