The semi-arid Jaguaribe River Basin (JRB) extends over an area of 75,000 km², comprising approximately 50% of the Federal State of Ceará, in Brazil. It forms part of the Drought Polygon (DP), a densely inhabited dry area located in the Northeast portion of the country (Figure 1). The Drought Polygon is characterized by extremely variable precipitation in space and time, which normally varies between 450 and 900 mm annually, and high potential evaporation ranging from 2,000 to 2,600 mm per year, while the runoff coefficient is usually between 5 and 10% (Gaiser et al., 2003; Pinheiro et al., 2016), falling below 1% in some areas (de Figueiredo et al., 2016).

Specifically, in the JRB, average annual rainfall ranges from 500 mm to 900 mm, of which almost 95% occurs from January to June and roughly 50% of the total rainfall, for the period 2004–2009, occurred on 10% of the rainiest days (Alves et al., 2012; Medeiros and De Araújo, 2014). The basin is set predominantly on a crystalline complex (85% of the area) with some sedimentary areas (15%), in which lithosols, luvisols and ferralsols are mainly found, with shallow soils of up to 1 m. Another important feature of the region is the high intermittency of all rivers, including the Jaguaribe, where annual runoff varies from 5 to 120 mm, depending on the geology of the site (Peter et al., 2014; De Araújo and Bronstert, 2016).

The current population living within the JRB is approximately 2.6 million with an average annual income that is 1.7 times the minimum wage (i.e., R$ 2,400), with an average per capita GDP of US$ 2,536 (exchange rate US$ 1.00 is R$5.57) (IBGE, 2021). The average Human Development Index of the municipalities that make up the JRB is 0.62, ranging from 0.56 to 0.84 (IBGE, 2010). Johnson and Kemper (2005) state that JRB follows the same economic structure as the state of Ceará, in which the service sector represents a major share of the composition of the GDP, whereas agriculture is of great social importance for the livelihood of the rural and most vulnerable population. Family farming in the region consists predominantly of short cycle crops such as corn and beans, as well as goats, sheep and cattle livestock (Mesquita et al., 2016).

Water scarcity, especially during the recurrent droughts, has frequently been pointed out as a key hindrance to regional development (Medeiros and Sivapalan, 2020). However, some disagree with the idea that water unavailability is the sole factor to explain why the region remains underdeveloped: for example, Salinas et al. (2019) showed that a water-infrastructure megaproject, built in a critically dry area, had very limited impact in terms of income and development for the local population. Still, a hydraulic solution focused on the implementation of surface reservoirs by damming the network of intermittent rivers that drain the region has been used as the major measure to cope with droughts since the late 19th Century (Medeiros and Sivapalan, 2020). After the great drought of 1878–1879, the Federal government ordered the construction of the Cedro dam, the first major water project in Brazil, located within the JRB. Further on, the Inspectorate of Works Against Droughts (IOCS) was founded in 1909 and the Ceará State Drought Secretariat was created in 1932. Following those early initiatives, a very dense network of reservoirs (approximately 13,000 dams with areas larger than 1 ha) was established over the course of the 20th Century, which we explore in the next sections.

Three main types of dams have been built in the JRB over time: (i) strategic dams, (ii) cooperation dams, and (iii) spontaneously built, small dams. Here we present their definition, in order to introduce how reservoir storage capacity was computed. Those definitions will later on be used to discuss relevant socio-hydrological feedbacks occurring at the JRB (Section 4).

The strategic dams are those serving relevant water demand centers such as cities, large irrigation areas and industrial zones, which were designed, built, and monitored by the public sector, therefore technical documentation is available. Cooperation dams, resulted from the collaboration between the public sector and landowners, and are also well documented. For those structures, storage capacity is known (Pinheiro, 2004). Small, non-strategic dams have also been built spontaneously by the rural population without proper documentation. Due to their large count (non-strategic reservoirs represent most of the existing structures), it was not feasible to characterize the entire reservoir network by field surveys. The storage capacity of these reservoirs was estimated using a remote sensing approach (Pereira et al., 2019), which consists of calculating the volumes of the reservoirs (Equation 1) based on geometric coefficients. The latter are estimated using lake geometry obtained by orbital remote sensing, namely the maximum flooded area and the maximum perimeter of the reservoir (Equations 2, 3).

In Equation 1, V is the estimated volume (m³), A is the area to which the volume is to be estimated (m²), A 0 and V 0 represent, respectively, the initial reference area (m²) and the volume (m³) stored at this level; K mod is the morphological coefficient (nondimensional); and α mod is the shape coefficient (nondimensional). The initial reference area was arbitrarily set at 5000 m² based on the spatial resolution of the satellite images (90-m resolution, with 5,000 m² representing 5 pixels) and V 0 was estimated at 2000 m³ according to average morphological conditions.

In the empirical Equations 2, 3, A max is the maximum flooded area (m²); P max is the perimeter of the maximum flooded area (m); D is the diameter of the reservoir (m), calculated as the ratio P max /π; and λ is A max / P max . The numerical values in these equations were calibrated by the authors using field data from 312 reservoirs, which resulted in a Mean Absolute Percentage Error—MAPE of 25% in their estimations.

We used the data available at the Global Surface Water Explorer (GSWE) database,¹ whose image-processing methods are described by Pekel et al. (2016). The GSWE product used in this methodology is the Maximum Water Extent, i.e., the maximum surface area of a given reservoir that was observed between 1984 and 2021. The equations were applied to the almost 13,000 reservoirs detected in the Jaguaribe River Basin, with flooded areas larger than 1 ha.

Water availability within the JRB was calculated as the water yield from the reservoirs with 90% annual reliability (Q90): this is the criterion used by the State Water Resources Management Company (COGERH) to simulate water management scenarios and to make decisions on water withdrawal from the strategic reservoirs. With the data provided by COGERH for 37 strategic reservoirs, excluding the three largest (Castanhão, Orós and Banabuiu) whose hydrological efficiency differs from the average, we established a regression between Q90 and reservoir capacity (Figure 2), and then used it to estimate water yield for the non-strategic reservoirs.

Based on the estimations of water availability obtained in the previous step, the per capita water availability was calculated based on the population size between 1900 and 2020. To assess the evolution of water availability at the JRB, we used the water stress indicator (Table 1) proposed by Falkenmark (1989), expressed as the ratio between the accumulated Q90 and the population of a given year.

Accessibility to water sources was assessed by calculating the distance from all points in the basin, displayed in a 1-km resolution raster, to the nearest reservoir, taking into account the strategic structures and those built in a public-private cooperation regime. Location and date of implementation of the strategic reservoirs were provided by COGERH, whereas the information of the cooperation reservoirs was derived from Pinheiro (2004). The distance from any point in the basin to the nearest water resource was calculated on a decadal basis between 1900 and 2020.

Met water demand is hereafter defined as the fraction of water demand that can be met by the system of reservoirs. Flow generation was estimated using the modified HYMOD model (Moore, 1985), a spatially concentrated, conceptual rainfall-runoff model that uses daily input data of precipitation and potential evapotranspiration, generating estimates of streamflow and actual evapotranspiration.

The streamflow is partitioned in the reservoir network using the Reservoir System Model (RSM), which divides the reservoirs into groups based on their volumes. Meira Neto et al. (2024) suggested the following classification of the reservoirs: Class 1 (0.0–0.1 million m³); Class 2 (0.1–0.5 million m³); Class 3 (0.5–1.0 million m³); Class 4 (1.0–20.0 million m³); Class 5 (20.0 million m³ - largest reservoir in the basin) and Class 6 (the largest reservoir within the basin). The model calculates the water balance for each reservoir class, considering a single representative, in which the local flows (evaporation and withdrawals) are estimated together with the state variables (local volume and water levels). The adoption of a representative reservoir per class is due to the large number of reservoirs in the network, over 10,000 structures, and lack of data for the small ones not monitored by the state water agency. This approach, proposed by Güntner et al. (2004), has been successfully tested in the study region for other catchments and at different scales (Krol et al., 2011; Mamede et al., 2018; Lima et al., 2023; Meira Neto et al., 2024). Figure 3 presents a schematic representation of the combined HYMOD-RSM model.

To estimate the water demand, the model takes into account rural ( D R ) , urban ( D U ) , irrigation ( D IP ) , and industrial required discharges ( D I ). The monthly unmet demand for each class of reservoir is used to assess the frequency of failures. This approach was applied to the Upper Jaguaribe River Basin as representative of the entire system (Figure 1).

As for water security, a water pollution index (I_wp) was estimated for the Upper Jaguaribe Basin, defined as the ratio of the annual volume of water potentially polluted to the volume of runoff. The index is a combination of two fractions of pollution - sewage and agricultural runoff - and was estimated for the period 1920–2020. Sewage production was calculated as the product of the population by the average per capita water consumption (assumed as 100 L.cap⁻¹.day⁻¹) and by a fraction (0.8) for sewage generation from clean water supply. Population data was obtained from official census reports and interpolated for the years when a census had not been carried out. The yearly volume of sewage was then divided into three fractions: (i) collected and treated; (ii) collected and untreated; (iii) uncollected. The fractions of collected and treated sewage were obtained from the Brazilian National Sanitation Information System (SNIS) for the municipalities located in the UJB.

Only the shares of the collected and untreated (V_col-unt) and the uncollected (V_unc) sewage were admitted as harmful, with the potential for polluting the runoff and the water accumulated in the reservoir system. The volume of sewage that is collected but not treated was fully considered as potentially polluted. The portion of sewage that is not collected was admitted as being released directly into the environment without any treatment. However, some assimilation by the environment is possible, e.g., the sewage disposed of in septic tanks, therefore a fraction of 0.5 of the sewage pollution not assimilated by the environment and potentially polluting (f_sew,pot-pol), was admitted. The volume of sewage with the potential to pollute the water resources (W_ps, m³) was calculated as the sum of the collected and untreated sewage and the uncollected share (Equation 4).

Water potentially polluted by agriculture was computed as the sum of the runoff volumes in irrigation projects and in rainfed agricultural fields. Information on the extension of public irrigated perimeters (A_ip, m²) was retrieved from official records, whereas the diffuse rainfed agricultural areas (A_rf, m²) were estimated as the product of population by a per capita cultivated area of 0.32 ha per capita. Runoff (Q_gen, m³.year⁻¹) was simulated for the UJB using the HYMOD+RSM model (Meira Neto et al., 2024), and the volume of water potentially polluted by agriculture (W_pa, m³) was then calculated using Equation 5, where A_UJB is the area of the Upper Jaguaribe Basin (m²).

The pollution index is calculated yearly by Equation 6.

The focus of historical data used in this research was on the occurrence of droughts, the creation of concentration camps to assemble together the droughts’ migrants, and the establishment of the dam policy. The term “concentration camps” was firstly used in 1916 by the state governor in a message to the Legislative Assembly:

However, in order to avoid any theoretical confusion with the term used decades later for concentration camps used in Europe, we will hereafter adopt the term “drought camps” when referring to the event in the state of Ceará.

Historical data presented in newspapers and official documents were surveyed in the Brazilian Digital Library.² In the search we used keywords related to the occurrence of droughts and their impact on society, as well as information obtained from technical reports on the drought camps and the construction of dams, including the number of dams built per year, their respective storage capacities, the value of the awards provided for those that built dams in their own lands and other technical data.

Books were also an important source of information in this stage of the research, regarding: drought occurrence (Alves, 2003; SUDENE, 1981; Guerra, 1981); drought camps (Liberato Barroso, 1916; Rios, 2014; Alves, 2023); and construction of dams in cooperation (Pinheiro, 2004).

The panel shown in Figure 4 summarizes some key variables over the course of the last century for the JRB. Population size in the JRB in the early 1900’s was in the order of 300,000 and grew at increasing rates throughout the 20th Century (Figure 4a). By 1970 the population of the JRB had reached approximately 1,500,000, including an intensification of rural–urban migration. The 1991 demographic census indicated a population of 1,850,000, and currently (circa 2024) over 2.5 million people live in the area. Figure 4b shows how the Falkenmark water stress indicator varied over the study period, changing from a period of absolute scarcity that was present before the 1900’s and lasted until the 1960’s, followed by a scarcity period from the 1960’s to the 2000’s, which then gave rise to a period that was mostly in the water stress range. The transition between water stress levels is associated with the expansion of surface water storage capacity, where the large changes can be attributed to the construction of large dams such as the Orós (1,940 million m³ storage capacity) in 1960, Banabuiú (1,600 million m³) in 1966 and Castanhão (6,700 million m³) in 2002.

Although the construction of large dams explains much of the basin scale changes in water stress, many more dams were built throughout the study period, and the mode in which those dams were built and managed are an important aspect of the evolution of the JRB. A first look into this matter is presented in Figure 4c, showing the rate of construction of strategic and cooperation dams (See 3.1 for their definition). It is possible to see how cooperation dams were built mainly between 1930 and 1960, while strategic dams were built throughout the whole study period, with a somewhat constant rate between 1920 and 1990, a sharp increase between 1990 and 2000 and decreasing rates since then. A natural consequence of the construction of reservoirs can be captured by the change in distance from any point within the JRB to the nearest reservoir, shown in Figure 4d, where the distances to both strategic and cooperative dams are depicted.

The system’s evolution differed within its various types of reservoirs with respect to their capacity to provide water with high reliability. Figure 4e shows the evolution of the fraction of water demand being met by the strategic reservoirs (usually the larger ones) as compared to the non-strategic reservoirs (including the cooperation ones). It is possible to see how the average met water demand at small dams are lower over time, with the strategic dams achieving higher levels of met water demand earlier in time (predominantly after the 1960’s).

Our estimate of how water quality has evolved over time at the JRB is shown in Figure 4f. It is possible to see a steady increase in the polluted water ratio, with marked variability as sudden increases, associated with very dry years, as the index is inversely proportional to average streamflow. In 1930, it is estimated that the sewage generation was of the order of 3.9 million m³.year⁻¹, reaching 13.0 million m³.year⁻¹ in 1970 and 16.7 million m³.year⁻¹ in 2020. In this same period, the volume of treated sewage increased from 5% to almost 85% of the collected fraction. On the other hand, agriculture (both irrigated and rainfed family farming) has continuously increased during the study period, following the growth of population and water availability. Consequently, the polluted water index (i.e., the ratio of potentially polluted water produced annually to the annual runoff volume) has risen from roughly 0.01 in 1920 to values fluctuating around 0.1 in the 2010’s.

Lastly, the time series of the number of wells drilled by the state government over time shows how well construction started in the 1980’s, at a rate that is somehow kept constant until around 2010’s, where a sharp increase is noted: more than 50% of the wells by the Ceará Water Resources Secretariat were drilled between 2015 and 2020, in conformity with the Drought Coexistence Plan, presented by the State government in February 2015 (SOHIDRA, 2021). This change in the rate of well construction is likely a response to the long-lasting drought from 2012 to 2017, which reduced the storage in surface reservoirs to nearly 6% of its capacity in the JRB.

In this section we present a classification of different periods within the 100 years of dam expansion within the JRB by analyzing the changes in the coupled human-water system. The result of the classification of human-water coevolution at the JRB is presented in Table 2, where four distinct periods are summarized.

The choice for defining each period was based on how four different factors behaved distinctly within the time-period studied, namely (i) the impacts of droughts on society, i.e., the different ways droughts’ negative outcomes have been observed; (ii) the main ways humans responded to droughts by constructing different types of dams collectively or individually; (iii) the feedbacks of the system and their respective unintended consequences; and (iv) the main limitations for water security to be met over time.

To define the periods and identify the main factors of human-water interactions, we explored the hydrology-related variables shown in Section 3.1 through a socioeconomic and political lens, resulting from the assessment of: (i) official documents, news articles and literature describing the impacts of droughts; (ii) technical reports and implementation of policy regarding dam construction; (iii) scientific documentation and interpretation about system feedback; (iv) data on water quantity and quality that could be provided by the reservoir network over time. Table 2 guides the reading of the next sub-sections, where we explore in more detail each of the four periods and their characteristics.

As a consequence of the adaptation of society to water scarcity and the construction of numerous dams, by demanding strategic ones from the governments and constructing own small structures, a very dense network of surface reservoirs emerged. Using the Global Surface Water Explorer (Pekel et al., 2016), Pereira et al. (2019) identified more than 13 thousand dams with flooded areas of at least one hectare in the Jaguaribe Basin, totaling 15,000 million m³ of storage capacity, of which 74 are strategic reservoirs monitored by the Water Resources Management Company of Ceará and 190 are cooperation reservoirs (Pinheiro, 2004). Figure 9 (left) illustrates dam construction over time, from which it is possible to note that reservoirs in cooperation were built mostly between 1930 and 1970, whereas construction of strategic dams was intensified after the 1970’s. Figure 9 (right) shows the distance from any point in the basin to the nearest water source, taking into account the strategic dams, those in cooperation and those built spontaneously by society (with an area of over 1 ha). The map illustrates the high density of the network, with most areas of the basin presenting a distance of <3 km to a water source.

Figure 10 illustrates the network configuration in terms of accumulated frequency of reservoir count, storage capacity and water yield with 90% reliability, sorted by size: 97% of the reservoirs have individual capacities of <1 million m³, which altogether can store and yield only roughly 5% of the total of the system. On the other hand, the three largest strategic reservoirs (Castanhão, Orós and Banabuiú, with capacities of 6,700, 1,900, and 1,600 million m³, respectively), can store and yield over 65% of the water of the system. Such configuration of the network depicts the representativeness and role of the small reservoirs in distributing water in the Droughts Polygon, as discussed by Nascimento et al. (2019) and Meira Neto et al. (2024), whereas the large ones promote water yield with high reliability.

The main features preventing society from fully achieving water security in the Jaguaribe River Basin changed over time, as the system evolved. We argue that the four distinct periods, or eras, as identified in Table 2, can be conveniently structured into the Availability, Accessibility, Acceptability and Quality—or AAAQ—framework. The following paragraphs expand on the definition of each component of the AAAQ framework and its relationship to the water security evolution at the JRB.

In the AAAQ framework, Availability is quantified for a given geographical location and assessed from a supply perspective measured by quantitative data (Jensen et al., 2014). Era 1 (before 1930) in the JRB was characterized by an imbalance between water supply and demand, when population suffered due to severe drought events. During that period, the main constraint to water security was the very low water availability with high reliability, since the region lacked structures to store the runoff that was concentrated in a few months per year. This was when the policy of damming rivers was initiated, whose inception was the construction of the Cedro dam in 1906. In this stage of the adaptation, society pursued continuous water supply in sufficient quantities.

Accessibility, as defined by Jensen et al. (2014), refers to the level and who has access to water, i.e., it may be available in abundance, but there are factors that influence the right of access to this resource. The authors list three aspects to be considered in the access to water: physical access; economic access (affordability); accessibility to information (participation in decision-making). In Era 2 (1930–1970), the feature restricting water security was accessibility, with some areas in the JRB distant over 200 km from the nearest water source at the onset of this period. During this era, the Dam Policy focused on expanding the reservoir network through the public-private cooperative regime. However, these dams were initially unable to meet their objective of facilitating access to water, since the owners of the land on which these infrastructures were built denied access to the population. Srinivasan et al. (2012) studied more than 20 areas in four continents and concluded that water scarcity in JRB has been primarily caused by political reasons, in a process that the authors called “water capture by elite,” which agrees with the findings of other authors. For example, Boelens et al. (2019) argue that political and economic elites are able to capitalize on dam projects and marginalize other social groups. Bakker (2010) shows that the culture of political exclusion of indigenous farmers is among the reasons for the lack of water supply in Cochabamba, Bolivia.

Era 3 initiated in the 1970’s, after the expansion of the network of cooperation dams and the implementation of some important strategic reservoirs such as Orós and Banabuiú in previous decades, and lasted until the early 2000’s. During this era, water security was constrained by the acceptability of marginalized social groups that relied on water sources with low reliability, mostly small reservoirs. In the JRB, cultural issues of these groups such as family farming and dietary habits, as well as other cultural perceptions such as the memories of drought events and their negative effects, have a strong influence on the acceptability of a water system that had been put in practice in the previous eras. Without access to the reservoirs built in cooperation and not being supplied by the strategic ones, these groups intensified the construction of their own dams spontaneously.

With the expansion of the network of reservoirs, by the end of Era 3 in the early 2000’s, the water security aspects of availability, accessibility and acceptability had been partially met in the JRB, and a decrease in the rate of dam construction was observed. The system reached a hydrological limitation to increase water availability, and the high storage capacity (almost two-fold the average annual runoff volume) promoted high sediment and nutrient trapping, giving rise to the water quality effect (Medeiros and Sivapalan, 2020).

In Era 4 (2000-present), the major restriction on water security has been water bad quality, whose adequacy should be defined according to international standards and considering parameters that might pose risks to human health, as proposed by Jensen et al. (2014) in the AAAQ framework. The pursuit of water security by society in the Jaguaribe River Basin through the Dam Policy, followed by the decline of this policy in Era 4 is illustrated in Figure 11.

Socio-hydrological studies have demonstrated that construction of dams might be followed by unintended consequences due to the bi-directional interactions between society and hydrology. In the Jaguaribe River Basin, a major negative feedback of storage capacity expansion was the water quality effect, i.e., water quality degradation due to nutrient retention in the reservoir network. The monitoring program conducted by COGERH since 2008 indicates a general trend to high eutrophication in Era 4 of the Dam Policy, with the three largest reservoirs remaining in eutrophic or hypereutrophic state over more than 50% of the time. The study by Silva et al. (2025) suggests eutrophication to be accelerated by the practice of damming rivers in the region even under preserved conditions: the authors analyzed the Boqueirão reservoir, inserted in a preservation area, and observed high phosphorus loads from the catchment and concentration in the reservoir.

The water quality effect might also trigger other undesired consequences, such as inequalities on water security already observed during the system co-evolution: the Drought Industry in Era 2 and low reliability of water supply from small reservoirs in Era 3. Treatment of the poor-quality water from the reservoirs implies high technological and financial costs, impacting its tariff and threatening the supply of the most vulnerable social groups with low payment capacity. Furthermore, throughout the evolution of the Jaguaribe system, signs of other socio-hydrological phenomena have been identified (Medeiros and Sivapalan, 2020): the supply effect (Di Baldassarre et al., 2018)—increase in water demand due to higher water availability promoted by reservoirs, offsetting the initial benefits to the supply–demand balance—and the reservoir effect (Di Baldassarre et al., 2018)—vulnerability of society due to overreliance on reservoirs as secure water sources.

Arguably, enhancement of water availability with high reliability enabled the continuous growth of population and the economic development in the JRB and the Metropolitan Region of Fortaleza, connected to this system. With the flow regularization by the reservoirs, the high water-consuming agricultural sector could establish and, at a later stage, also the service and industrial sectors. The loop of increasing water availability by reservoir implementation, followed by demand increase, which required building new dams, is illustrated in Figure 4b: a thorough raise in the water scarcity index is observed after a dam being built, with a continuous reduction in the following years due to demand growth until the construction of the next dam.

High dependence of society on the reservoirs for water supply is another consequence of the focus put into the Dam Policy throughout the 20th Century. Currently, over 90% of the water demand of the Ceará State is met by dams (Peter et al., 2014). Most surprisingly, the two largest reservoirs alone (Castanhão and Orós) account for 57% of the total storage capacity of the system, meeting not only the local demands of the JRB but also being responsible for the water security of the Metropolitan Region of Fortaleza, outside of the basin. By the end of the dry period of 2012–2017 (Zhang et al., 2021), those reservoirs reached nearly 2 and 4% of their capacities, respectively, affecting the water availability (Pereira et al., 2023). On that occasion, only the irrigation sector suffered rationing, which started in September 2013 and lasted until the end of the drought. This example demonstrates that relying on a few surface reservoirs in times of extreme drought can lead to the collapse of the system if management actions are not taken in advance and encompassing all users.

The evolution of the reservoir network in the JRB compares with other examples, such as the case of the ancient Maya society. In the Mayan region, the rainy season extended for 7 months, so there was a need to harvest rain to provide drinking water during the dry season. Manmade reservoirs provided drinking water for millions of Mayans in the Classic Period (c. 250–900 AD), but just like in the JRB, streams and the smaller reservoirs dried up as the dry season progressed (Lucero, 2023). Lucero (2006) states that the annual wet and dry seasons affected the socio-political systems of the Maya, so that urban forces dominated during the annual drought, i.e., there was appropriation of water by the elite, and rural forces dominated during the rainy season, when many Mayans lived dispersed across the landscape to cultivate the land.

Kuil et al. (2016) applied a socio-hydrological model to the case of the ancient Maya, whose population peaked during the Classic Period (600–830 C. E.) and then declined during the 9th Century. The authors’ hypothesis is that periods of drought played an important role in the collapse of society: from simulating the possible feedback between water and the people, it was demonstrated that a modest reduction in rainfall could lead to an 80% population decrease due to food scarcity. The simulations performed by Kuil et al. (2016) indicate that the construction of reservoirs resulted in less frequent drought impacts, but if the reservoirs dried up, the impact of the drought could be more severe due to the reservoir effect described by Di Baldassarre et al. (2018). At least eight droughts lasting between 3 and 18 years hit the region between 800 and 930 C. E (Medina-Elizalde et al., 2010) and triggered problems such as overuse of resources, crop failures, decreased water supply and quality (Lucero, 2023).

Beyond the system feedback already noticed in the Jaguaribe Basin, climate change may play a major role on water resources in the future. Rodrigues et al. (2024) conducted climate simulations with regional climate models and showed different trends in future evaporation from the network of reservoirs in Ceará. The authors concluded that the simulated future scenarios of evaporation can impact water availability in the range of −7 to +9%. In a region where there is considerable reliance on reservoirs, changes in evaporation and/or precipitation patterns can lead the system to collapse (Kuil et al., 2016; Di Baldassarre et al., 2018).