Introduction

Reservoirs and hydropower plants in Brazil’s important São Francisco River Basin are representative of a particular development strategy in Brazil’s Northeast region, which has been pursued since the 1980s. Along with electricity provision, rural development has also been a prominent goal. However, for example, in the case of the Itaparica Reservoir, about 40,000 residents were resettled (Cernea 1991). Land management plans have primarily focused on irrigation schemes; first, by settling large commercial farmers often coming from other areas of the country and later on as an alternative for local farmers who lost their flood irrigation land alongside the river. Many irrigation schemes, however, have poor soils (shallow, with low organic matter content, very sandy in parts). Moreover, drainage water from irrigated agriculture and aquaculture systems, which care little about their feed and drug residues, is threatening the quality of local water bodies. The majority of local farmers are smallholders (according to the last census of 2006, for instance, 88% of all farms in the Itaparica micro-region had only up to 50 ha; IBGE 2009); they derive their grazing-based livelihoods from the non-irrigated Caatinga, a dry forest native to the semi-arid region. The population density in this semi-arid region is relatively high (15.5 persons/km2 in the Itaparica micro-region; IBGE 2017), adding pressure to the current scenario of vulnerable livelihoods and the degradation of natural resources.

Concerns about the unsustainable use of natural resources are widespread worldwide. Sustainable land management practices claim to minimize the negative impacts of land degradation, which otherwise results in the deprivation of human welfare (Mutoko et al. 2014). The ecosystem service concept can be useful in advocating for nature in human-environmental systems (Wamsler et al. 2014), eventually informing land use decisions (Förster et al. 2015). The extensive sustainability debate has elucidated the relevance of an integrated approach, with the negotiation of economic, ecological, and socio-cultural issues at its core. However, adopting such an approach remains tricky. Dynamic drivers come into play, such as economic and technological change, migration, and policies of different sectors (Nesheim et al. 2014). Governments have to oversee both local situations and macro development at the national or regional scale. Closing electricity gaps still triggers the planning of megaprojects such as large dams for hydropower, with huge implications on the ground (Zarfl et al. 2015). Enhancing irrigation schemes and water diversion projects also contributes to the ongoing conversion of ever less free-flowing rivers into dam- and reservoir-shaped hydrological systems: as documented by Zarfl et al. (2015), who assessed at least 3700 major dams (> 1 MW) being planned or under active construction, above all in emerging countries such as Brazil.

Water and land issues are intrinsically connected (Fig. 1). Multiple objectives are at stake and have to be addressed concurrently (Pittock et al. 2015). The nexus perspective is already complex in its own right and becomes even more so by virtue of the necessity to deal with external drivers such as climate change (Rasul and Sharma 2015) and the uncertainties such drivers create. Disciplinary research approaches repeatedly draw conclusions from single perspectives without sufficiently accounting for contradictions and trade-offs. In response, this paper synthesizes evidences from the comprehensive collaborative research project INNOVATE (“Interplay between multiple uses of water reservoirs via innovative coupling of substance cycles in aquatic and terrestrial ecosystems”). The paper draws on the results of interdisciplinary and transdisciplinary work, which addressed water and land management issues at a local to regional scale in Brazil. Communication is at the core of integration towards achieving greater sustainability (Cosens et al. 2014). Combining diverse knowledge is a cognitively demanding task, which depends on the willingness to share among peers (Salazar et al. 2012). Transdisciplinary research bridges the gap between understanding problems, finding out what works better and how this can be achieved in collaboration with various stakeholders (Hirsch Hadorn et al. 2006; Jahn et al. 2012; Siew et al. 2016).

Fig. 1
figure 1

A blessing and a curse: the nexus at hand. The figure illustrates how water, land, energy, and fish and agricultural produce are interlinked in the study region. The stocks of water and land allow the production of agricultural produce and fish, and the generation of electricity in the semi-arid environment near the reservoir. Depending on their scales and timing, uses are mutually exclusive, or can co-exist. The arrows indicate direction of flows

Full size image

The unifying research question of this paper, and similarly of the research project synthesized here, was which management options promote sustainable ecosystem services and economic viability under climate change conditions in the São Francisco River Basin? Moreover, we aim to offer insight into the outcomes of interdisciplinary and transdisciplinary practice as an example for collaborative knowledge production from local to landscape scale, which is necessary to address the depicted nexus situation.

The paper is results-oriented. The synthesized results are presented in four main sections: (i) availability and access to water at the basin scale, (ii) water quality in the semi-arid catchment, (iii) agroecosystems in the semi-arid environs of the reservoir, and (iv) the governance framework for implementing sustainable practices. Each section is broken down into issues which address major knowledge gaps in sustainable land and water management in the study region with interest to a broader audience. These are approaches to deal with increasing water scarcity in a scenario of growing water demand, while caring for water quality and the role soil and biodiversity in semi-arid environments play in achieving a balance between ecosystem health and securing livelihoods, while making the best use of the governance system.

The study region and the preparation of the synthesis

The study site, the Brazilian São Francisco watershed as a whole and its Itaparica Reservoir region in particular, was introduced in detail by Siegmund-Schultze et al. (2015a). In a nutshell, the nearly 3000 km long river studied lies entirely within Brazilian borders and is emblematic in the role it plays connecting diverse regions of the country. The Upper, and to a certain extent the Middle, sections of the watershed are where most of the river’s rainfall occurs. Two huge reservoirs have been installed in the semi-arid section, where evaporation is approximately four times the annual rainfall. Aside from generating hydropower for the integrated national grid and providing flood control, the reservoirs serve a number of irrigation projects. Irrigated commercial agriculture consists of export-oriented fruit production, while smallholder irrigation is widely a result of fierce negotiations between farmer representatives and the hydropower company regarding compensation for flooding their former plots and houses. The latter is the case for the irrigation schemes around the Itaparica Reservoir. Even though dam construction dates back to the late 1980s, livelihood options of farmers are still restricted due to soil types, outdated irrigation infrastructure, and unfavorable marketing organization, for instance. The booming production of tilapia in net-cages, in particular, is increasingly taking up space and impairing water quality along with the aquatic habitat. A prolonged drought started in 2012, which was still not fully over in early 2018 when revising this paper. It was so severe that it even affected domestic water supply, which is the priority use when water is scarce according to the Brazilian Water Act (Brasil 1997).

The written synthesis of research results resembles a literature review, which is, however, based on results of INNOVATE project partners, complemented by non-project work where considered necessary for the argumentation lines. In the list of references, the papers of project members have an asterisk in order to facilitate the tracing of information coming from either project members or other authors. Readers interested in specific methods are invited to consult the referenced papers. The development and synthesis of the research relied on a coordinated approach during the lifespan of the interdisciplinary and transdisciplinary project (compare Siegmund-Schultze et al. 2015a, Siegmund-Schultze et al., in preparation). The paper at hand draws on content from both scientific papers and interactions with scientists and stakeholders.

Availability and access to water at the basin scale

Responses to increasing variability of water availability are at stake

Global climate modeling and its regional projections of future rainfall have provided contradictory scenarios for the São Francisco River Basin. On average, however, the scenarios suggest a drier climate in the semi-arid region with long dry spells and rainfall falling during even shorter time periods. Even in the scenarios predicting an overall wetter climate, the number of consecutive dry years will increase (Hattermann et al. submitted). Based on 40 years of historic rainfall data from two tributary basins in the semi-arid portion of the São Francisco watershed, Assis et al. (2012) showed that annual rainfall is already decreasing, while the number of consecutive dry days is increasing. The trend has been confirmed for the whole Sub-Middle portion of the basin, pointing out a trend reversal of the rainy years which prevailed until the 1980s which were then followed by dry years, based on 50 years of historic rainfall data (Assis et al. 2015). Water scarcity will continue playing a major role and the multiple water requirements will be all the more competitive.

With ever more irrigation schemes planned in the Middle parts of the river system (Koch et al. 2015), water prices which do not reflect the real scarcity value of water (Alcoforado de Moraes et al. 2016), a conflictual diversion project in the Sub-Middle parts, and day-to-day challenges to managing ever more critical discharge budgets, good governance requires well-balanced “horizontal” and “vertical” decision-making processes. The “horizontal” challenge involves better interaction of more recent and regional (river basin committee) with established (often federal agencies) bodies (Siegmund-Schultze et al. 2015b). The “vertical” challenge of the basin’s land and water management system primarily consists of hydroelectricity-focused water management. This focus might not be maintained at the same level in the long run, as it has become increasingly adverse towards competing ecosystem services such as drinking water supply, agricultural usage, and habitat regulation. Initially, on the federal level, better integration of Brazilian hydroelectricity and water policies remains a challenge ahead. At the same time, the transmission grid availability in hydroelectricity-shaped river basins allows for a steady shift towards a more diverse renewable energy strategy. This has substantially progressed in the Northeast region of Brazil as regards wind energy (Brown 2011) and solar energy is becoming another player on the respective agenda (Sena et al. 2016). Rethinking the nexus of energy, water, and land presents new opportunities for the Northeast region given the availability of key elements: namely wind, an electrical grid, and solar radiation. Koch et al. (in press) conclude that adopting an integrated approach for hydropower and wind power generation can even serve a further objective, namely achieving a more environmentally sound flow regime. This involves higher discharges from reservoirs in the São Francisco River Basin during the rainy season and lower discharges during the dry season, representing a more natural flow regime while limiting hydroelectric power generation in the dry season. By decreasing the share of hydroelectric power generation, the flow regime in the Sub-Middle and Lower São Francisco River Basin might thus be modified to improve the ecological status of the riverine system. At the end of the day, the irrigation-based, but water scarcity vulnerable, agricultural land use paradigm might be substantially supplemented or even transformed into wind and solar energy landscapes: at best, even alleviating future conflicts over water supply in a semi-arid tropical setting.

The controversial water diversion project needs monitoring and better communication

The water diversion project of the São Francisco River (transposição), which attempts to transfer water via two channel systems into neighboring watersheds, is a flagship project of the Brazilian Ministry of National Integration (Ministério de Integração Nacional). A minimum discharge of 26.4 m3/s is licensed for domestic supply and a maximum of 127 m3/s will be conceded under specific circumstances (Cirilo 2008). At first, these water quantities may sound small when compared to the river’s average annual flow at its mouth of 2850 m3/s (Torres et al. 2012). However, compared to the rough estimate of what can be withdrawn from the river without interrupting its functions, namely 360 m3/s (AGB Peixe Vivo and CBHSF 2011), the amounts of the diversion project become quite substantial. Still, the diversion project is only one of many users. Registration of all users is mandatory and payment is obligatory from 4 L/s onwards taken from the São Francisco River. While in the years 2000 to 2003, a water demand between 138 and 168 m3/s was observed (Maneta et al. 2009), Koch et al. (2015) simulated future demand being on average 267.3 and 474.3 m3/s, respectively, by applying different climate and socio-economic change scenarios. Simulations were based on existing water permits and irrigation schemes under construction, with one scenario including schemes which are still being in the planning process. Satellite images suggest that real withdrawals are currently about one third higher (irrigation schemes) or even double (private farms along the river) as deduced from the difference between declared areas and irrigated areas (Souza da Silva et al. 2014). Hence, registration is incomplete and the “real” amount of water which is withdrawn, and that which is available, is hardly known.

Eco-hydrologic modeling confirms a limited scope for attending to the needs of all users in the future, already accounting for climate and land use changes as outlined above (Koch et al. 2015). As the diversion channels were not yet functional, Rossiter et al. (2015) studied water quality parameters exemplarily in another (smaller) concrete transfer channel downstream, taking samples along the first 29 km of the particular channel. Temperature, pH, and conductivity showed increasing values with distance from the pumping station, while nutrients did not. All parameters, with the exception of one, complied with high Brazilian quality standards while phosphorous content proved to be elevated, forming a potential risk for eutrophication. However, the design of the studied channel is not fully comparable with the diversion channels under construction. Water quality responses cannot be directly extrapolated to the other bigger diversion project. The latter consists of a whole network of concrete channels, pumping stations, and a number of existing or to-be-established regional reservoirs. These reservoirs, even though artificially built, may provide relevant ecosystem services such as water purification. On the other hand, open water bodies in dry regions also have clear dis-services, i.e., reduction of water provision via evaporation. Pereira et al. (2009) estimated annual evaporation from the Sobradinho Reservoir to be between 1796 and 2149 mm (using different methods), which means a reduction of water flow of about 132 m3/s.

Results of high-resolution modeling do not suggest a major impact of the water withdrawal from the eastern channel on water levels at the very local bay scale, as the flow from the main reservoir stream will be high enough to level out the withdrawal (Matta 2018). Other impacts of the new outflow, however, are not yet fully understood. For instance, changes in velocities and residence times may alter water quality in the bay—either just in an area around the point of withdrawal or with a broader magnitude. Initial simulations by Matta showed that residence time of polluted water in the bay will be reduced substantially, namely about four times under standardized conditions of a low water level at 300 m.a.s.l. and a mean discharge of 2060 m3/s (mean regulated inflow to the Itaparica Reservoir). Hence, the selection of the withdrawal point could be considered beneficial for the water quality on a local bay scale, while the water quality taken up into the channel might also be better than anticipated (as long as one considers the recommendations for withdrawal discussed in the following section). In either case, water quality issues will require close monitoring efforts in the bay and inside the channel. Whether, overall, the withdrawal finally threatens or benefits the fishing and farming communities around the bay cannot be determined at this stage. What is clear, however, is that it will substantially reduce the water that is available for other users of the original watershed, while also increasing the overall losses to evaporation.

In 2016 (at the time of writing this paper), the construction of the diversion channels was well advanced. Fierce initial societal discussion for or against its construction shifted to a discussion about who will eventually manage the system and in which way. A federal decree from 2014 designated the rural development agency CODEVASF (which is linked to the Brazilian Ministry of National Integration) to be the national operator of the diversion system. The decree mentions 12 entities to compose the respective management council: primarily federal ministries, the four involved states, the São Francisco River Basin Committee, and the committees of the receiver basins. In early 2018, it remained still unclear how operation and maintenance would be done in practice, who would bear the infrastructure and management costs in the receiving watersheds, and which criteria would be applied for the issuance of water withdrawal permits, including its pricing system. This situation is, yet again, a clear example of prioritizing technical aspects while neglecting governance issues in conceiving a megaproject, thus creating uncertainty, resistance, and anger.

Water quality in the semi-arid reservoir

Water users: cause and distress of water pollution

Major initial concerns, as reported for tropical humid regions, about the river and reservoir water quality and possible greenhouse gas emissions from sediments, biomass decomposition, and the passage of hydropower turbines have not been generally confirmed. Sanitation plays currently the major role in sources and amplitudes of emissions in the entire watershed (Venohr and Fischer 2017). Major cities alongside the São Francisco River require substantial wastewater treatment, affecting water quality downstream. Many treatment plants exist, though treatment is basic, and thus the majority of the incoming nitrogen and phosphorus load to, for instance, the Itaparica Reservoir originates from upstream release.

Modeling exercises visualize which regions show a low or high risk for nutrient loss (taking a land perspective) and which portion of the river and its tributaries might be at risk of nutrient overload (the water perspective). A land risk map is sustained by P sorption measurements and subsequent calculations of risks of P losses from available measurement datasets (Fischer et al. in press; compare soil section below) and can help to allocate limited funds for monitoring to the major areas at risk.

In the Itaparica Reservoir itself, aquaculture, primarily with net-cages, is the most important anthropogenic factor of trophic upsurge and water contamination, followed by drainage water from irrigated agriculture and wastewater inflow (Selge et al. 2016). The authors estimate a lower overall risk of eutrophication for the reservoir today as compared to the loads shortly after damming, almost 30 years ago. Hence, the upsurge of nutrients happened in the early stages of reservoir formation.

Eutrophication processes, along with the development of harmful algae and noxious waterweeds, trigger acute problems in shallow bays with reduced water exchange with the river’s main flow (Matta et al. 2016). In such isolated bays, the establishment of aquaculture is not recommended. The identification of similar vulnerable bays (not every bay is isolated and affected in the same way; prevailing wind directions play a role, among others) can be used in planning and monitoring. From an ecological perspective, permits for lake-based aquaculture should not be issued for such bays. It remains to be justified whether installations up to a certain size might be acceptable from a developmental perspective. Limiting the expansion of lake-based aquaculture in the Itaparica Reservoir as a whole and promoting land-based aquaculture instead are the main conclusions from a limnologic perspective, which still need to be analyzed for their broader ecological impacts, socio-economic preconditions, institutional implications, and acceptance by fish farmers.

In particular, there was a clear nutrient gradient within the studied bay during the rainy period, which starts at the drainage points of the irrigation scheme of Icó-Mandantes; hence, wash out of nutrients from irrigated agriculture is occurring (Gunkel et al., accepted with revisions). The observed gradient was highest shortly after heavy rainfall, leading to temporary eutrophication hotspots, including toxic cyanobacteria. In general, one must assume that eutrophication is not a continuous phenomenon in the semi-arid reservoir, but is always driven by such sudden and strong nutrient pulses, with evaporation contributing to nutrient concentration to some extent.

Water withdrawal (pumps) should avoid such locations of higher nutrient and contamination loads. For example, residents in the irrigation schemes near the Mandantes Creek should refrain from withdrawing water shortly after heavy rainfalls, since the ephemeral tributary creek will wash out potential pollution and the polluted water might be pumped as indicated by modeling the effects of flow and transport coming from the tributary using the TELEMAC modeling system (Matta et al. 2016). Hence, adaptations of the timing of withdrawal and the positioning of facilities are at stake. From a limnologic perspective, the annual water level amplitudes of the reservoir should be kept as limited as possible, as should the daily water level amplitudes from a local user perspective. This strategy challenges the predominant aforementioned hydroelectricity-driven water level regime.

Water level fluctuations: friend and foe

While the tested water purification scheme successfully uses the water plant Egeria densa (“water pest”), limiting its growth is a challenge for water quality management in the reservoir. E. densa acts as an effective phosphorous sink; it can absorb 10 times more P than the surface sediment (Keitel et al. 2016). The drying of both E. densa and sediment reduces, however, P sorption. The authors observed a 30% reduction of P uptake capacity of dried sediment. Due to the changes in water level, whole stands of E. densa dry out and release GHG and nutrients, which accumulate in the reservoir water. This is particularly critical when water bodies are shallow and sediments are vulnerable to (intense) drying at low water levels. At the same time, the sediments are crucial in their role in retaining external P flushes after runoff events. Hence, the more water levels fluctuate, the higher the concentrations of soluble reactive phosphorus in the water will be due to leaching of desiccated sediments. Abrupt P releases outweigh retention in these dry-wet cycles and eutrophication hotspot pulses occur. Water level fluctuations, i.e., reservoir management, will eventually control the quality of E. densa and sediments as either a source or a sink of nutrient loads (Keitel et al. 2016).

Daily water level fluctuations reached up to 24 cm in the study period from January 2010 to September 2015, while increases occurred more often at low flow conditions than high flow conditions, hence fluctuations were rather uncoupled from water availability (Selge 2016). Reservoir management was conducted according to national needs (the Itaparica hydropower plant is connected to the national grid) and even showed a difference between weekdays with more demand and weekends with less. In times of low flow rates, some additional high flow rates occasionally occur upon demand from ships needing a temporarily higher flow for mobility. A stricter coupling of hydropower production to the natural water supply could reduce eutrophication impacts from E. densa and sediments (Keitel et al. 2016). Under average rain conditions, seasonal fluctuations can be as large as 5 m. These seasonal differences are still low when compared to naturally occurring water masses from rainfall with clear seasonal differences and subsequent filling of the river.

If water levels change gradually along clear seasonal patterns, then harvesting the desiccated E. densa could be a predictable activity, while less steady fluctuation would also mean less overall development of E. densa (Lima and Gunkel 2015). Removing E. densa is a means to control the P load of the reservoir’s water as described above. Removal also addresses other negative outcomes of mass development of the plant, such as the reduction of oxygen content in deeper water layers, the disturbance of turbines and boats, and forming habitat for the snail which is the host of Schistosomiasis (Gunkel et al. 2015). Underwater harvesting of the submerged plant, which consists of about 90% water (Lima and Gunkel 2015), is tedious; a technical solution has yet to be conceived. Harvesting underwater is also not advisable due to additional ecological impacts, such as sediment resuspension and harvesting of other organisms living in the bed. Moreover, the many submerged trees in the shallow areas of the reservoir would make the undertaking very difficult.

Cycling nutrients between productive sectors is feasible

The desiccated biomass of Egeria densa can be used more easily than the wet biomass. Options include as a fertilizer, an animal feed, or for biogas production, although currently, these possibilities appear to be restricted by high removal costs (Lima and Gunkel 2015). As a soil amendment, E. densa biomass could provide organic matter and nutrients. According to Batista et al. (2004), E. densa hay has potential as a fodder for ruminants, provided that it is supplied in a low proportion in the feed or for a short period of time, while the fresh material might be rejected due to possibly high fish odor. Using its biomass in a biogas digester, for instance, alongside, sludge (Zhen et al. 2015) is another option to be considered in the long run, coupled with a comprehensive business plan.

The land-water nexus has further been addressed with the so-called green liver approach, demonstrated with flows of nutrients and critical substances from land-based hatcheries. The green liver approach is essentially a system of connected ponds filled with macrophytes (i.e., water plants), meant to purify water from toxins (Pflugmacher et al. 2015). Aquacultural wastewater and drainage water from irrigated agriculture are application possibilities. Marques et al. (accepted with revisions) tested the nutrient retention capacity of different macrophytes, while Vilvert et al. (2017) focused on the purification potential of the system regarding antibiotics. Diclofenac is one drug which potentially threatens fish populations in the reservoir by causing estrogenic activity in tilapia (Gröner et al. 2017). While an up-scaling of the prototyped and preliminarily tested purification system could not yet be undertaken, further monitoring and subsequent approval might allow for substantial implementation efforts, contributing to a more sustainable land-water-interface.

The biomass of the macrophytes utilized in the water purification schemes also needs a clear strategy for disposal. Its turnover period is about every 2 to 4 weeks under warm conditions. The difference of this source of biomass is its potential accumulation of toxins and drug residuals due to the water’s passage through aquaculture ponds. This biomass might have strong restrictions for use in agricultural production and ultimately human consumption. Composting prior to application could be one means to reduce biotic contamination but studies on efficacy and feasibility are required.

Another promising approach for recycling nutrients is using the sludge from land-based fingerling production as a soil amendment (Silva et al. 2017). The sludge from the fish ponds must already be removed after each production cycle. Until now, it has still been common practice to dispose of it on land areas near the ponds, which amounts to an environmental threat to the reservoir when washed away with rainfall. The authors tested different substrate mixtures. Three quarters sludge and one of local soil turned out to work best for lettuce production. Still, the substrate’s components need to be tested continuously since the sludge may contain traces of antibiotics or other drugs commonly used in commercial fish farming. Minimizing the use of pharmaceuticals altogether in the nexus would be an important step towards greater sustainability.

Agroecosystems in the semi-arid environs of the reservoir

Poor soil quality and runoff risks—measures for improvement

Sandy soils, featuring low nutrient availability and low water holding capacity, prevail in many parts of the lake’s environs (Araújo Filho et al. 2013). Research in INNOVATE focused on identifying locally available and economically feasible substrates for soil amelioration, such as clayey sediment and biochar (Beusch and Kaupenjohann 2015) to thus improve the productivity of the agroecosystem with endemic multipurpose trees (Spondias tuberosa, umbuzeiro; Mertens et al. 2015). Umbuzeiro had little to no response at all to the tested soil amendments; these might be mainly suited for non-native crops, which are not inherently adapted to water and nutrient stress. Positive results regarding the soil water and nutrient content due to clay addition (available from waterhole sediments, for example) were not the only positive results to be identified. The use of clay and biochar (e.g., available after charring of invasive plants, or parts of these; in our case Prosopis juliflora, the algaroba tree, which in another study municipality already comprises 6% of land cover, Barbosa Neto et al. 2015) can support food and income security in areas with sandy and nutrient-poor soils in the semi-arid tropics (such as arenosols in the study region). Furthermore, the application of biochar in irrigation schemes as well might minimize the leaching of nitrate and P to prevent eutrophication. The implementation of innovative agricultural practices requires, however, further consideration and guidance from involved stakeholders, for instance, EMBRAPA, the Brazilian Agricultural Research Corporation.

In another section of the environs, soils were characterized as shallow, stony to rocky, with a relief from flat to gently undulated and poor drainage. Barbosa Neto et al. (2015) found places with severe erosion in the Itacuruba municipality. While Caatinga still covers the major part of the municipality, the vegetation is becoming thinner and sparser over time (81% in 1986 and already 95% in 2010; Barbosa Neto et al. 2015). Measuring the iron oxide content in the sub-soil has been suggested as a low-cost method for determining the spatial variability of environmental degradation (Sousa et al. 2015). The lack or removal of vegetation facilitates erosion; deforested set-aside areas (fallow, wasteland) are at major risk. The conservation of the native vegetation would conserve plant and animal species, safeguard the soil carbon stock, and favor better coverage of the soils.

In a broader approach covering the whole watershed, Fischer et al. (in press; in press) and Fischer and Venohr (2017) suggest (i) never fertilizing soils in a way where artificial fertilizer is left on top of the soil (still a common practice in Brazil) because of the resulting high risk of P losses with runoff, which means P losses and subsequent P transport to in unintended places such as water bodies, and similarly (ii) avoiding erosion (through landscape engineering techniques) to reduce particular P losses with runoff (e.g., buffer zones and intercropping), and (iii) promoting a unified evaluation of Mehlich 1 P (M1P) data for the whole São Francisco catchment area, which can be used to estimate the degree of P saturation of soils. This monitoring protocol would allow for better identification of areas at high risk of P losses and consequently help to direct scarce resources and special attention to these areas. The studies indicate that the used method of estimating P saturation (Pöthig et al. 2010) can be applied to tropical soils with highly differing properties. The use of M1P data, as the only soil P data available in high spatial resolution, provides a unique means to consistently assess the risk of dissolved P losses by surface runoff (Fischer et al. in press).

The beneficial interplay of biodiversity and irrigated crop production

Farmers should avoid herbicide use and favor organic fertilizer, particularly the available goat manure, when cultivating perennial crops (coco, banana, fruit trees; Cierjacks et al. 2016). The reduced application of herbicides and pesticides promotes the preservation of crop pest predators’ habitats; Cierjacks et al. (2016) showed that yields are still similar, making this cost-effective and ecologically-friendly practice economically relevant to farmers. The majority of the relevant crop pest predators, and in particular the amphibians studied, are known to the farmers (Oliveira and Freire 2015). Weed diversity and crop yield were positively correlated—a certain layer of weeds has a positive impact on yields (Cierjacks et al. 2016). When the weed layer becomes too massive (weed biomass and crop yield were negatively correlated), then farmers can shift to integrated management options such as understory grazing by small ruminants in order to keep biomass of weeds low, while also benefiting from animal production.

Whitefly damage in coconut stands was a common problem in the irrigation schemes and had a significantly negative effect on yields (Marr 2016). This can be counteracted by using coconut varieties with high phenolic content in their leaves, which show a significantly decreased infestation rate. Another positive management measure analyzed in the same study was micro-spray irrigation which showed a significantly positive effect on yield. Many farmers still have old irrigation infrastructures from the scheme’s establishment phase. Renewing the systems with the most adequate technology also eventually solves part of the water usage problem outlined above. The responsibilities regarding who pays for infrastructure improvements is however unsolved: some farmers believed (mentioned during interviews and workshops) that it should be the hydropower company (respectively the development agency who took over supervision) since initial supply has been less complete than agreed upon.

High plant diversity contributes to the diversity of agricultural land on a landscape level, since each crop species has a unique accompanying flora. Moreover, diversification of production might be a good way to withstand hardship and establish a more constant flow of income. However, due to small plot sizes, diversification is limited and marketing shortcomings may also play a role. Income from fruit production in smallholder irrigated agriculture is currently rather low and farmers are economically vulnerable (Hagel et al. 2014). The same authors found clear differences among gross margins of different fruits; however, current income strongly depends on the availability of farm workers who accept low wages. Figuring out systematically what works well under different conditions might be a useful way to adapt production systems, as would a shift to permanent crops in irrigation schemes, avoiding herbicide use in coconut, banana, and fruit plantations, and preferring goat manure as fertilizer.

Caatinga: appreciation of species and land use diversity makes sense

The major question for the non-irrigated Caatinga ecosystem and its smallholders focusing on goat and sheep grazing has been to which degree grazing intensity could be adjusted to a still feasible state while safeguarding development of the ecosystem’s biodiversity, biomass, and carbon storage capacity. Could a process of endemic restoration of the Caatinga dry forests be safeguarded along with viable socio-economic outcomes for the smallholders in the region? Adaptations of the grazing system and intensity have been proposed to serve as ecosystem services oriented practices, to tackle desertification processes and ongoing losses of habitat quality. In particular, heavy grazing (measured by number of animal droppings detected) was found to significantly reduce the organic carbon content of the upper 5 cm of the soil (Schulz et al. 2016) as well as the richness of plant species (Tavares et al. 2016). In order to avoid substantial removal of vegetation, farmers should comply with the commonly suggested annual carrying capacity of < 1 goat/ha in the natural Caatinga and apply basic practices such as pasture rotation. Interestingly, the Caatinga, although classified as dry forest, clearly features a predominant share of herbaceous species (Tavares et al. 2016) including a number of palatable grass species. Many woody species also serve as animal feed, and some cacti and bromeliads as well, either by directly browsing on the plants, feeding on litter, or through cut-and-carry systems.

Small ruminant production in mixed crop-livestock systems, and partly as solely a livestock system, plays a significant role for poor and more affluent farmers alike. There are basically five smallholder goat farming systems in the reservoir-affected region: resettled farmers in public irrigation schemes, resettled farmers with their own irrigation supplies, resettled farmers without any irrigation facilities, and non-resettled farmers with or without irrigation infrastructure (Stock de Oliveira Souza et al. 2014). During the severe drought of 2012, livestock keepers within the irrigation schemes, who are the less numerous, had the least animal losses since feeds could be derived from more sources. Studies highlighted farmer’s innovation and adaptation strategies to changing markets and environments: Farmers experimented on-farm with drought-tolerant livestock breeds and forages, forage conservation, pasture rotation, alternative worm control, and the recycling of nutrients from manure in crop irrigation water (da Costa von Gehlen et al. 2015). It appears that people with a higher educational level are more inclined towards implementing innovations in their production systems. Although many farmers claim a lack of extension services in their regions, communication and exchange among peers appears to be at least of similar significance. Nevertheless, the number of pioneers is still low so that exposure to innovations is not common for many of the farmers, especially in remote areas. Attitude and willingness are crucial for the adoption of sustainable land management practices. Generational turnover and better education may trigger more change in the future, though many of the more highly educated people may even prefer to cease being farmers and look for opportunities elsewhere.

The conservation status of the Caatinga is patchy, nevertheless, there was a general tendency of increasing forest stands in the period from 2001 to 2011, which indicates lower pressure, which is partly explained by decreasing grazing intensity (Schulz et al. 2017). In the long run, this will increase carbon stocks in the soil and the biomass. Despite this positive trend, some actions towards conservation should be taken now. A recurrent problem in the study region (and probably beyond) is the missing follow-up of promising actions. For instance, the “Serra da Canoa” conservation area, a so-called environmental station located in the project’s study region, has been designated, but measures to implement effective conservation strategies are sporadic. The project supported a key actor, a local NGO, by providing species lists, which are necessary in justifying and maintaining the conservation status.

Monitoring and control of existing regulations and laws often lag behind the generally very comprehensive documents (Rodorff et al. 2013b). Buffer zones along water bodies are mandatory as described by the revised Forest Act (Brasil 2012); the widths of such buffer zones are defined by biome. Monitoring the establishment and maintenance of buffer zones, called areas of permanent preservation (APP), would be an important step forward to reduce erosion since the APP are expected to impede surface runoff, along with its nutrient loads such as high P concentrations, from directly entering water bodies. Moreover, the monitoring of farm conservation areas (reservas legais), required by the same Act, would contribute to a more consistent system of biodiversity and soil conservation. Protected sites with a unique Caatinga biodiversity should serve as backbone habitats, supplemented by stepping stone sites to support species migration. The implementation of such a differentiated conservation approach depends on the willingness and power of a range of stakeholders at the local (land users) to state or national scale (agencies, ministries).

Slowly growing, ultimately promising: the native umbuzeiro tree

Umbuzeiros are emblematic trees of the Caatinga with a very characteristic shape, especially in older trees, and unique water storage capacities. The project’s vegetation assessments detected, however, very few young trees. Apparently, the regeneration of this tree has been interrupted. Seed damage has been ruled out as a reason for regeneration failure (Oliveira et al. 2014). Genetic diversity of umbuzeiro is largest in the unused areas, and is smallest in the humanly influenced areas near the river. However, natural regeneration currently takes place exclusively in the agricultural areas. Consequently, the genetic diversity is presently deteriorating unless varieties of the more undisturbed areas are artificially propagated as well. Being endemic, farmers are not allowed to cut umbuzeiro trees. The tree is multipurpose, providing shade, food for people, feed for animals, and its existence is appreciated by all groups of rural dwellers, while indigenous peoples also attach a religious-cultural connotation to the tree. Despite its many benefits, the tree is still undervalued scientifically (Mertens et al. 2015).

In order to support the (re-)establishment of umbuzeiro trees, a field experiment investigated the response of umbuzeiro rooting systems to the addition of soil amendments, namely goat manure, biochar, and mineral fertilizer (Mertens et al. 2016). A Bayesian network approach has further been applied in order to detect major levers for future adoption of the studied technology, starting from the assumptions that umbuzeiro tree planting benefits both the environment and the farmer’s revenue (Rodorff et al., accepted with revisions). Factors for successful adoption were determined in collaboration with potential adopters. These were the use of faster-growing grafted seedlings, the access to financial support, and organization of marketing and sales within cooperative structures. Keeping livestock away from tree seedlings is also crucial for the establishment phase of the trees. The illegal cutting of Caatinga vegetation or even entire areas offsets, however, the conservation and livelihood objectives. Much will depend on whether such initiatives will actually be backed by, for instance, the Brazilian rural development agency for the São Francisco valley (CODEVASF), in cooperation with further authorities and civil society.

The governance framework for implementing sustainable practices

Finally, major land management and governance phenomena have been identified as driving forces or relevant barriers for sustainable land and water management (Rodorff et al. 2013a). This involves multi-level governance challenges, the manifold actors involved at different scales (with different objectives, a more active or passive voice, and partly overlapping responsibilities), and persistent drought crises as a driver for change in the region (e.g., a controversial discourse on payments for irrigation water). Rodorff et al. (2013a) also found missing exchanges between institutions, for example, dealing with adjacent producer groups in the same municipalities. At the end of the day, different concepts regarding “development” remain: while up-stream (as in Minas Gerais and Bahia state) commercial irrigation farming on larger scales prevails, smallholders still struggle on relatively small plots in the Itaparica region. The towns and the irrigation scheme “islands” in a semi-arid and poor environment face immigration dynamics and vulnerability. Strategic and participatory land use planning (involving institutions from different sectors at different scales), including environmental and social impact assessments, remain a missing link thus far in Brazil.

The recent paradigm of public participation is being integrated gradually into governance procedures. The Water Act (Brasil 1997) is one of the new laws which prescribes it. However, a thorough revision of existing decision-making procedures has not yet taken place and results in contradictions of current participation practices (Siegmund-Schultze et al. 2015b). For instance, public participation is essentially restricted to meetings of the watershed committee, since other existing institutions widely practice business as usual without visibly accommodating the committee’s recommendations.

Consistent, relevant, up-to-date, and easy-to-use data is an important basis for evidence-based decision-making. Many organizations provide access to data online, though these datasets are often inconsistent, contradictory, or use different or unrevealed measurement protocols. Overall, the accessibility of datasets is not yet sufficiently secured. As demonstrated above, a unified evaluation of P sorption would be constructive. Similarly, the different Brazilian states should adopt a unified basis for water management and planning in order to facilitate the integration of data (for instance, applying the same flow percentiles across states). While online data management and communication are developing quickly in Brazil as a whole, limitations still exist. Inclusive citizenship is still struggling at the local scale as evidenced by a study on public participation GIS (Gottwald 2015). The difficulty to read maps and the availability of consistent cartographic material were current limitations in the studied municipality.

Ultimately, processes of change need a functioning governance system and good governance is multi-faceted. Rodorff et al. (submitted) suggest a systematic approach to account for the various attributes, namely (i) the functional institutional support, policy environment and jurisprudence, (ii) the legitimacy and accountability of actors, (iii) the fairness and transparency of decision-making processes, and (iv) the means to assure quality control and adaptiveness. They prototyped a template of good governance criteria, which offers systematization for planning and steering the implementation of sustainable land management practices.

Conclusions and outlook

Critical cornerstones of sustainable management in the reservoir context (and beyond) are as follows: minimize the resource users’ footprint by, above all, emitting and consuming less, e.g., reducing the use of contaminators, such as fish pharmaceuticals, mineral fertilizers, and pesticides, using water-saving irrigation systems and crops with higher water use efficiency. Secondly, mitigate the environmental impact: effluent purification, wastewater treatment; and finally, cycle the residual nutrients and substances: sludge and manure as soil amendments, water from aquaculture to grow crops, animals fed on weeds and crop by-products, and E. densa reuse. Stock management is likewise crucial, addressing temporal dynamics and multi-use infrastructure (feed conservation for the dry season, smart planning of reservoir flow regarding times of scarcity, using power lines not only for electricity from hydropower but also wind turbines). Strategic and participatory land use planning, including environmental and social impact assessments, has to integrate the demands of the sectors today and in the future, seeking win-win situations where possible. Unconventional solutions should also be considered, such as shifting to electricity generating landscapes instead of focusing on agricultural production alone, or rethinking options for rural tourism. A fair resolution and mediation of conflicts are mandatory in democratic systems. While the government needs to clarify competencies and improve continuity, citizens should take a responsible stance towards advancing sustainable management.

Land becomes gradually more important under water scarcity. The increasingly erratic availability of water resources should encourage a transparent discussion about what to do with land resources. Aside from concentrating water use in some areas to foster water use efficient crop production, other land areas could be allocated to electricity generation by wind and solar power, possibly integrated with livestock production. Diversity of uses appears to be the key to sustainable livelihoods and environmental well-being, along with clear definitions of their objectives and domains. Communication requires room and training, and has to be based on well-sustained and accessible knowledge. In the past, the large dam region of the São Francisco River suffered from a monodisciplinary view on resource management: water for electricity first. Flood protection was indeed a major reason for dam creation, while irrigation came second in flow management. Local residents and the environment did not play major roles at early planning stages. In implementing participatory governance, introduced by the constitution in 1988 with the rising democracy, land and water management started to face much more complex expectations. The nexus unfolded.

Over the course of the scientific cooperation, the project members derived major recommendations for practice from their research results. An overall guidance manual aims at making research results more accessible over time and safeguarding knowledge beyond in-person interaction with current stakeholders (Siegmund-Schultze 2017). Some research results became even clearer when deeply thinking about their ultimate usefulness. Systems thinking (e.g., Bosch et al. 2007), the core of the nexus approach, needs to be mainstreamed within society as a whole in order to appreciate and grasp the complexities of sustainable land and water management. Not only the selling of new technologies, but also the elaboration of their externalities and governance implications should be on the agenda of researchers and actors. Large dams, or any other megaprojects, should not be demonized per se nor sold as silver bullets. In collaborative learning, people must be able to withstand the discrepancy between the importance of one’s own partial convictions or specific results and the significance of the big picture. Open-mindedness and embracing continued learning are prerequisites and skills needed in this process. Tolerance, endurance, and thinking outside the box are further keys to coordinated collaboration in both science and public decision-making towards sustainable land management.