Citizen Science for Water Resources Management: Toward Polycentric Monitoring and Governance?

In many regions of the world, acute data scarcity still limits our understanding of hydrological processes and how humans interact with the water cycle. This in turn is a major bottleneck for advancing more efficient management of water resources systems. Indeed, official hydrometeorological monitoring networks are in decline globally, in particular in those places where socioeconomic development and environmental change are strongest (Fekete and Vörösmarty 2007; Hannah et al. 2011). This makes it paramount to explore new approaches to data collection and knowledge generation about hydrological processes and water resources systems. One promising approach may be the application of so-called citizen science.

Citizen science refers to the participation of the general public (i.e., nonscientists) in the generation of new scientific knowledge (Buytaert et al. 2014). Approaches are highly variable, ranging from grassroots and participatory data collection to soliciting contributions and carrying out scientific tasks with the help of large groups of people through the Internet (e.g., crowd-sourcing, Dickinson et al. 2012). Recent technological developments, in particular the advent of cheap and more reliable environmental sensors (e.g., Khamis et al. 2015), are opening new pathways for the application of citizen science and participatory approaches in a water resources context.

Compared to other environmental sciences and geography (Buytaert et al. 2014), the uptake of citizen science in water resources is slower. One of the reasons is the reliance of hydrological measurements on technologically more complex and expensive techniques. Additionally, because of the large spatial and temporal variability of the water cycle, data analysis ideally uses repeated measurements such as long time series of hydrological states and fluxes (e.g., precipitation, discharge) and requires these observations to be representative of a wider spatial area (e.g., for river flow this is typically the river basin-scale as this unit integrates all the input, storage, and transfer processes and so yields the useable water resource for people from a watershed). But new technologies increasingly enable cost-effective and robust measurements of various parts of the water cycle and their merging with other upcoming technologies such as remote sensing. Examples are automatic weather stations, water level measurements using acoustic sensors or photographic image analysis, electronic soil moisture measurements, and novel water quality sensors, among others (for a more extensive overview see Buytaert et al. 2014). Technology development in the context of the Internet of Things (Gubbi et al. 2013) holds further promise for improved data collection, transmission, and curation.

These new technologies support not only the application of citizen science in the strict sense (i.e., implemented by nonspecialists), but also enable a much wider range of specialist stakeholders within a water resources system to engage in, or increase their monitoring activities. For instance, farms may install weather stations as part of a drive toward precision agriculture (e.g., Mesas-Carrascosa 2015). Water supply and hydropower companies may install specific networks to support operational decisions and assess risks. Nongovernmental organizations (NGOs) install monitoring to quantify ecosystem services and degradation (e.g., Celleri et al. 2010). Such activities result in a growing monitoring community parallel to the established monitoring networks that are operated by national hydrometeorological offices.

Data generated by each of these activities will have different characteristics and potentially varying quality. Part of this is determined by the specific purpose of the actor. For instance, the initiative for hydrological monitoring of Andean ecosystems (iMHEA) participatory monitoring initiative in the tropical Andes described by Celleri et al. (2010) was set up to analyze the impact of land-use changes on the hydrological response of upland catchments. Using a network of paired catchments, each of which contains a single land-use type, the initiative is able to characterize the differences in hydrological response between land-use types robustly using high-resolution short-term monitoring (Buytaert et al. 2007). The high spatial and low temporal coverage of this network is highly complementary to the official network operated by national meteorological offices, which focuses on a low spatial density but long-term time series. Although the latter is less suited for evaluating land-use impacts, it allows for more robust analysis of climate trends or extreme events, which is a higher priority at the national policy level.

The evolution from a hydrometeorological monitoring landscape dominated by national networks to a more diverse community of multilevel and multipurpose monitoring bears strong resemblance to the emerging academic insight that water management in certain contexts may benefit from a polycentric governance model. The concept of polycentric governance accepts that many human systems are characterized by multiple centers of decision-making across different levels, thereby relying on a distribution of responsibilities, multiple sources of information, and cogeneration of knowledge (Andersson and Ostrom 2008; Cole 2015). Even if they are less streamlined than tightly integrated centralized systems, polycentric systems tend to “enhance innovation, learning, adaptation, trustworthiness, levels of cooperation of participants, and the achievement of more effective, equitable, and sustainable outcomes at multiple scales” (Ostrom 2010, p. 552).

In river basin management in particular, polycentric governance may be an alternative to the classic integrated water resources management (IWRM) paradigm, which uses a centralized, hierarchical command and control regulatory regime (Lankford and Hepworth 2010). Analyzing large ($15–\mathrm{150,000}{\mathrm{km}}^2$) watersheds in sub-Saharan Africa composed of disparate communities, institutions, and environments, Lankford and Hepworth (2010) found that water resources management in such basins is characterized by informal, localized decision-making based on ad-hoc, local knowledge and dialogue, and aimed at finding flexible, incremental solutions. They conclude that a decentralized, polycentric water resources management (PWRM) approach is more effective under such conditions.

In their analysis, Lankford and Hepworth (2010) associate PWRM with data-scarce basins. Yet, the distributed nature and informality of citizen science-based monitoring would seem highly compatible with a polycentric governance model, given that any actor may take initiative to generate evidence to pursue a particular issue. As such, citizen science contributes to the scientific evidence base of localized and multileveled decision-making.

An example of polycentric monitoring is the Piura basin in North Peru, which is part of the iMHEA network (Celleri et al. 2010). The basin has an extreme altitudinal gradient in precipitation between the coastal desert and the highlands (Fig. 1). The network of the Peruvian National Hydrometeorological Service (SENAMHI) is biased toward the lowlands, because of the cost of installing and maintaining stations in the upper mountain areas. Water resources estimations are based on a linear interpolation of precipitation with altitude (Fig. 1). The iMHEA data show an exponential rather than a linear trend, obliging official publications on water resources to be revised.

We identify several challenges to the integration of hydrological monitoring in a model of polycentric river basin management and governance, which relate to the generation of data themselves, the way they are processes and interpreted, and how they influence governance processes.

The most common criticism of citizen science is that the generated data are likely of lower quality for reasons such as the lack of standards, limited technical capacity, and lower-quality equipment (Cohn et al. 2008). Although there is very little scientific evidence on the quality of citizen science data, a lower quality may not inhibit the generation of useful scientific evidence to support decision-making in a particular policy context. It is paramount, however, that uncertainties are properly accounted for in the data analysis to avoid misinterpretation and potential overestimation of the generated evidence. From a scientific perspective, the integration of heterogeneous datasets with different statistical and epistemic properties also poses specific challenges (e.g., Beven 2009). At the same time, increasingly powerful methods exist for uncertainty analysis and data assimilation.

Polycentric monitoring may also challenge the monopoly on data held by hydrometeorological departments in many countries, and may therefore encounter resistance. The example of Piura (Fig. 1) shows that different datasets generated in a polycentric monitoring landscape can in effect be highly complementary. Yet in practice, polycentric approaches to monitoring and governance inevitably lead to some degree of redundancy, in terms of measurements but also in terms of responsibilities. However, such redundancy is potentially of great value to make networked systems more robust: if some nodes in the network fail, other nodes can take over. At the same time redundancy is at odds with efficiency optimizations where any overlap in tasks is meant to disappear. The ideal level of redundancy will be context-dependent. For instance, a higher level of resilience may be desirable in monitoring networks focused on operational disaster risk reduction, compared to networks focused on scientific data gathering.

Nevertheless, polycentrism will inevitably require a higher level of coordination between actors to avoid unnecessary duplication and waste of resources. In that respect, the value of polycentric monitoring depends on the connections between the different centers of monitoring. Because polycentrism avoids tight integration, polycentric monitoring will therefore need to rely on connections between the nodes. Such connections need to be established at a technical level, e.g., database connections for data merging and visualization, but also at between information sources and relevant users (e.g., Vitolo et al. 2015). Within a polycentric environment, data discovery and optimal use of available data will inherently be more challenging.

Lastly, the polycentric approach to decision-making on natural resources itself is challenging (Gelcich 2014; Pahl-Wostl and Knieper 2014). Greater clarity is needed of what polycentricity means in practice, in particular what capacity exists for actors at different scales and levels within a watershed to take genuinely autonomous decisions. Recent work on polycentrism favors a pragmatic approach to understanding autonomy in decision making (Norström et al. 2014; Pahl-Wostl et al. 2012), rather than focusing on formal legal delegation of decision-making powers. Implementing polycentric monitoring for water resources management would enable more granular exploration of the degree of autonomy afforded local actors in their everyday management practice. Secondly, it is unclear what the consequences of polycentrism might be for coherence of decision making within social-ecological systems. Again, implementation and examination of polycentric monitoring would afford greater understanding of whether and how systemic coherence over decision-making emerges over time. Overall therefore, we argue that polycentric monitoring provides valuable opportunities to advance current understandings of polycentricity both theoretically and empirically, enabling closer approximation of this concept to real-world conditions and expectations.

Recent developments in sensing technology provide strong support for a more diverse and heterogeneous community of environmental monitoring including citizen science. Hydrology is still a very data-scarce science and it is therefore likely that this trend will give new impetus to hydrological data availability despite technological challenges, thus benefiting water resources management. The nature of citizen science bears strong resemblance to that of polycentric governance, which is captured in the concept of polycentric monitoring. Because of the increasing relevance of polycentric river basin management as a potential alternative to the centralized, hierarchical IWRM approach, we identify a potentially significant role for polycentric monitoring in river basin management. However, at this stage, there is still considerable uncertainty on the concept of polycentric governance itself, as well as the type of basins it applies to, and how polycentric monitoring needs to be implemented to ensure an optimal support of decision-making in such contexts. But the increasing availability of technology, and the increasing emergence of citizen science initiatives themselves, provides a unique opportunity to analyze and understand the potential impact and relevance of polycentric monitoring for hydrological science and water resources planning and management.

We thank Boris Ochoa of Imperial College London for elaborating Fig. 1. The ideas expressed in this paper are developed as part of U.K. Natural Environment Research Council grant number NE-K010239-1.