Does data lead to cooperation? Lessons from Water Accounting Plus in the Cauvery basin, India

ABSTRACT As the number of studies on remote sensing data for water management and governance increases, few articles reflect on their application in practice. This article shares learnings from the application of Water Accounting Plus (WA+) in a federal river basin in India. WA+ was applied to the Cauvery basin to contribute to solving transboundary water-sharing issues by providing a source of transparent data obtained through reproducible methods. By analysing how WA+ results and methodology were received, we show how data and models are also political and question the assumption that more data automatically lead to more equitable decision-making.


Introduction
There is an optimism around what remote sensing (RS) and RS-based models can offer to transboundary water management and governance.This relates to a hope that insights from space lead to more transparency and more engagement based on a common understanding, as well as the belief that technical developments will facilitate better data in the future and thus even more transparency and more engagement (Voss et al., 2013).This applies especially to data-scarce areas, for instance, when the political situation or available resources prevent in-situ data collection (Comair et al., 2014;Habib et al., 2012).However, there are calls for using such data sources with caution.Giordano et al. (2016) show that more precision and different definitions used with RS can counter efforts of practitioners when it comes to defining hydrological and political boundaries.
In a similar vein, Bretreger et al. (2021, pp. 7-8) stress that monitoring and compliance system based on RS need to take local cultures into account and be mindful of unwanted effects on rural communities if they are to contribute to water cooperation using the case of the federal Australian Murray-Darling river.Disconnecting a monitoring or modelling activity from the local context makes it impossible to understand the influence of the modelling activity, limits democratization in the decision-making process and limits CONTACT Rozemarijn ter Horst Rozemarijn.terhorst@wur.nlaccountability as well as the possibility to mitigate possible unwanted effects.Thus, especially in contexts where water allocation is a sensitive issue, understanding the interaction between the technology used and the (local) context is crucial, for instance, to ensure the tool indeed contributes to a level playing field.Milman and Gerlak (2020) show how river basin organizations consider carefully what science should be produced for facilitating cooperation over transboundary waters, and how it potentially influences different riparians and interactions between their members.However, there is potential gap between the development of science and practitioners, as the vast majority of articles on RS and models for transboundary water management and governance does not detail how the models contribute to water cooperation in practice, nor do they provide successful examples (for examples of the techno-optimistic approach, see Khoshnoodmotlagh et al., 2020;Mobariz & Kaplan, 2021;and Tian et al., 2015).Consequently, researchers have called for being more explicit about how models are developed, and how this development in turn influences water management and governance (Maeda et al., 2021;Melsen et al., 2018).In this article we seek to contribute to this research gap.
We focus the analysis on the application and uptake of the Water Accounting Plus (WA+) framework in an inter-state context, in particular on the Kaveri basin (from here on indicated by its anglicized name, Cauvery). 1 WA+ is a water-accounting approach which primarily makes use of RS and open-access datasets.One of the stated expectations of WA+ is that standardization in the presentation of water data and open data policies can contribute to cooperation at 'international basin level, also in conflict areas' (Karimi et al., 2013(Karimi et al., , p. 2463)).We explore this affirmation based on a structured assessment of the WA+ introduction, application and reception in the Cauvery context and contribute to knowledge on WA+ specifically, and to interactions between data-driven tools and water management in general.We are working together as a transdisciplinary team of engineers and a social scientist and have made the effort to write the article in such a way that it is accessible to a wide range of disciplines, avoiding jargon where possible and explaining it where pertinent.
In the following section we introduce the analytical and methodological framework, after which we explain the case study.We then present the analysis of the case study.We conclude the article with a critical discussion on how data and information are expected to contribute to transboundary water cooperation, relating it to the lessons drawn from the analysis of the introduction of WA+ to the Cauvery basin.

Unpacking models and their interactions
We understand hydrological models as social constructs (Melsen, 2022;Melsen et al., 2019).This means that through conscious choices and unconscious actions they can have political effects (Morgan & Morrison, 1999;Pielke, 2003).To understand the (possible) effects of models, it is needed to both unpack the model and analyse its interaction with the context in which it is applied.To unpack WA+ we use the framework of MacKenzie and Wajcman (1999) who state that technology, and thus a model, consists of worldviews -which we specify into ontology and visions of the future, materiality and expertise (Figure 1).We briefly explain what these elements are and how they constitute a model.
Ontology is how (parts of) the world are represented in the model, including the assumed relationships between the different elements within the model, often most directly translated in the mental model that underlies a hydrological model that identifies what is to be included and excluded.Visions of the future is the part of the model that reflects ideas as to how the world should work in the future, including hopes of what can be achieved through the analysis provided (Konrad & Böhle, 2019).Expertise relates to what the developers of the model know, in terms of what they have learned in their studies and through experience, including specific norms and values on how to do their work.The expertise of the makers influences the design choices made in a model and thus what is included and not, as well as how (Addor & Melsen, 2019;Babel et al., 2019).Expertise and worldviews are thus closely related (Haas, 1992).Materiality relates to the physical properties of the model and its output, which will allow for only specific expertise and worldviews to be conveyed.The codes underlying a model as well as any outputs (such as numbers, graphs, maps) are all part of its materiality and allow for the model and its outputs to be shared independently of the makers (Knuuttila, 2005(Knuuttila, , pp. 1266(Knuuttila, -1267)).These three elements are closely interconnected and mutually influence each other, but also provide clear entry points to understand how a model interacts with, and influences, its environment.
To explore this interaction we use the theory of translation (Akrich et al., 2002a(Akrich et al., , 2002b;;Callon, 1984) as a methodology (Kanger, 2017).This theory focuses on how policies or technologies gain support, how they are rejected, or how they are changed and change others through interaction.Callon (1984) argues that the theory of translation and its broad application to a case is useful to gain a 'better understanding of the establishment and the evolution of power relationships ' (p. 201).He identifies four stages in the process of translation: • Problematization: in which initiators define a certain problem and convincing argument for the people, institutions and elements required to start working on solving the problem.• Interessement: Those involved in the problematization attempt to persuade others to buy into the defined problematization in various ways.• Enrolment: specific rolls are identified to solve the defined problems, and negotiations happen to get specific actors to play those roles.
• Mobilization: the rolls are activated and played out, are changed, or are rejected in the context of the specific actors.
These stages do not happen in a linear way, and it can be very likely that no successful mobilization is achieved.The structure provided by the theory of translation, in combination with information on the constitution of the model based on MacKenzie and Wajcman (1999), therefore provides useful insights as to how a model, or certain aspects thereof, gain support while others do not, as we will illustrate through the case study.
Data for this study were collected through interviews in the Netherlands over the course of six months in 2021 and three months of fieldwork in India in 2022, resulting in interviews with 38 people, which included representatives of the states and national institutions, as well as scientists and practitioners working on the Cauvery or WA+.The interviewees were selected through their connection to the development of WA+ and its implementation in the Cauvery.In addition to the interviews, we analysed the grey and scientific literature.Based on the theory of translation, and inspired by the research of Hasan et al. (2020), we have followed the WA+ modelling framework over time, from development in the Netherlands to presentation in the Cauvery.
It was difficult to gain access to government representatives, which is a well-known challenge in researching policy-related issues in India (Mollinga, 2005).This is due in part to highly politicized public debate in which there is little incentive for government representatives to be interviewed by journalists or scientists.The interviews were performed following a semi-structured format.We have chosen not to use quotations in this article as they can easily be attributed to certain people in the basin as few people work on the issue.This acknowledges both the sensitivities involved with analysing professional practices (Mosse, 2006), as well as the sensitivities related to water sharing of the Cauvery River.We have indicated which part of the article is based on an analysis of the interviews.For the sake of finding a balance between researching the WA+ modelling framework and the context in which it is applied, we have made the choice to prioritize relations between Karnataka and Tamil Nadu, instead of all four riparian states which also include Kerala and the Union Territory of Puducherry.As the WA+ analysis of the Cauvery basin has not been made public, we illustrate the article with figures made specifically with open-source data, and as example of a WA+ study, we use an openaccess report on WA+ applied to the three sub-basins of the Indian Krishna River (Salvadore et al., 2020).

Water Accounting Plus (WA+)
We analyse WA+ based on the expertise, worldviews and materiality that constitute the approach.A summary of these elements is shown in Figure 2.
WA+ is an analysis that seeks to understand water availability, depletion and productivity in a basin in relation to land use (Karimi et al., 2013).The problems WA+ aim to address are manyfold, but boil down to the difficulty to make strategic decisions to secure water resources without information on the resource (Karimi et al., 2013).This position holds the assumption that information is a main obstacle to decision-making, and that data are a key to cooperation.Based on these ideas, WA+ aims to provide an 'unbiased reflection of the water resources and land use conditions' for an entire basin (UNESCO-IHE, 2014), which can also contribute to addressing conflict areas that cross political boundaries (Karimi et al., 2013).
WA+ has a strong focus on agricultural water consumption, facilitated by measurements and observations through satellites, specifically to derive evapotranspiration (ET).At the heart of WA+ is a division of the ET flux, the water vapour that ascends from the land surface to the atmosphere, into green and blue ET fluxes.Green ET corresponds to the rainfall stored in the soil and directly consumed through ET.Blue ET results from water that originated in a blue water source (e.g., lake, river or groundwater) before being consumed through ET (irrigation water would be counted in this flux; Bastiaanssen et al., 2012;Karimi et al., 2013).This allows for a distinction to be made between water that is naturally available and consumed in the landscape and water which is actively managed.
The assessment of water availability in WA+ is done based on the water-accounting approach of Molden (Karimi et al., 2013;Molden, 1998).Similar to the method of Molden (1998), it adopts the basin as main unit of analysis, and it separates water consumption into beneficial and non-beneficial consumption.Beneficially consumed water can provide benefits for agriculture, energy, leisure, the economy and the environment.Non-beneficial water is defined as 'the consumed water that is lost to the system for no benefits' (Salvadore et al., 2020, p. 38) or water that is unavailable for further use.This includes the water intercepted by wet leaves or impermeable surfaces, and evaporation from soil surface, weeds and degraded water.
The WA+ method specifically builds on public domain datasets, and especially those derived through RS, preferably available on a global scale in order for the method to be widely applicable.WA+ also relies on modelling needed for the split of blue and green ET, as well as to compute some important fluxes that cannot be measured through RS (e.g., discharge or water abstractions).While many types of models can be used to produce WA+ outputs, a simple distributed water balance-type model is often preferred as it keeps the spatially explicit nature of RS data and can be set up in data-scarce areas where parameterization of more complex models is subject to high uncertainties as few data points are available to compare the outputs of the model.This is important as WA+ especially targets basins in which in-situ data availability is limited.With support from institutions that aim to improve analysis and communication over water on a global scale, including IHE Delft, the International Water Management Institute (IWMI), the Food and Agriculture Organization of the United Nations (FAO), and the World Water Assessment Programme, WA+ is further developed to 'strive to achieve equitable and transparent water governance for all users and a sustainable water balance' and to improve the hydrological data democracy (IHE Delft, n.d.).The codes used to run WA + are therefore also publicly shared via the GitHub platform.
The outputs of the WA+ modelling exercises are depicted in 'sheets' that show how much water was thought to be available, and how it is used.The number of sheets produced differs per study, and for the Cauvery six sheets were produced, we add a brief description of these to further explain the WA+ framework.
Sheet 1 is an overview sheet that shows the general water balance in the basin and links consumption to general land-use types to give an overview of the manageable resource.The water balance includes the inflows (e.g., precipitation (P) and eventual inter-basin transfers), outflows (e.g., surface water leaving the basin and ET), and surface and groundwater storage change.Sheet 1 shows the outcomes of the analysis on green and blue water consumption and quantifies the available water for future developments, the water reserved for environmental or legal purposes and can be used to identify any overexploitation.Sheet 2 details the water consumption through ET for all land-use classes in the basin.ET is divided into evaporation, interception by the canopy and transpiration which is the portion of ET that directly contributes to plant growth.This separation allows for the quantification of beneficial and non-beneficial water depletion by sector.Sheet 3 focuses on the agricultural water consumption and crop production.This information is presented in the form of water productivity (amount of biomass produced per unit of water consumed) and land productivity (crop yield) for major crop groups (e.g., cereals and legumes).Sheet 4 provides information on estimated water withdrawals (blue water) from surface and groundwater sources per sector.The water withdrawn is then partitioned into consumed water (ET) and return flows to either surface or groundwater.Sheet 5 presents surface water flows and storages at the sub-basin and basin levels.Sheet 6 shows groundwater flows and storages in the basin.In addition to the sheets, maps and tables are developed based on the information derived from the model outputs.

Brief overview of the origin of the conflict
The Cauvery is located in the South of India, connecting four riparian states including Karnataka, Tamil Nadu, Kerala and the Union Territory of Puducherry (Figure 3).The century-long conflict over its shared waters is created and sustained by colonial legacies, cultural differences and political rivalries (Settar, 2018).Historically, the Cauvery delta in Tamil Nadu is known for paddy cultivation.Till the end of the 19th century, irrigation in the entire basin was based on run-of-the-river diversion structures without much storage capacity.During British rule, the delta, in the then Madras province, was favoured for irrigated agriculture as it is fertile and easy to irrigate.Upstream Mysore princely state (now Karnataka), which has more hills and highlands, did not use Cauvery River waters much.However, in 1870, downstream Madras complained about the impact of newly developed irrigation systems in Mysore on water availability.The two states negotiated from 1890 to 1892, leading to the Madras-Mysore agreement regarding new irrigation works, requiring the approval of downstream Madras for any upstream developments.After a disagreement over an upstream dam, a new agreement was signed in 1924.This agreement detailed the allocation of water of the Cauvery, favouring use in the delta as agricultural production was easiest there.The agreement solidified the already large power imbalances, and cultural and political differences became more pronounced (Anand, 2004).Before the reviewing period of the agreement, 50 years after signing, India had become independent in 1947.The new states of upstream Karnataka and downstream Tamil Nadu interpreted the validity of 1924 agreement differently; Tamil Nadu expected the states to uphold the agreement, whereas Karnataka saw it as a chance to develop its irrigation schemes more independently.The building of four new dams and impounding of reservoirs after 1974 entailed a diminution of the flow towards Tamil Nadu and further sparked the conflict between the two riparian states (Cauvery Water Disputes Tribunal, 2007).Still, the idea that water is more abundant downstream is strongly embedded in contestations between Karnataka and Tamil Nadu.Our interviews with farmers show that these ideas are persistent.It led initiators of the Cauvery Family, a grassroots initiative to connect people from the two states and foster water cooperation, to dedicate time to having farmers visiting each other's fields across state borders to understand each other's water realities (Iyer, 2007;Janakarajan, 2016).
Water sharing between the states remains highly politicized to this day.For example, after a verdict on water allocation of the Cauvery in 2007, a leader of an influential political party called for a day-long statewide fast to protest the outcome (News 18, 2007).The politicization also affects researchers, with researchers from either party state having difficulties in obtaining data from the other state.In the public debate, anyone who speaks up must take into account that reactions can be harsh and personal.This makes basin-wide research challenging, and as a result there are no agreed-upon numbers beyond the measurements carried out by the Central Water Commission under the Government of India on the amount of water that flows across the border between Karnataka and Tamil Nadu at the village of Biligundulu, and of water levels in the large dams.The fear thus remains that political conflict will flare up.The Cauvery dispute is known to escalate in dry years, when Tamil Nadu requests for water to be released and protests erupt against these requests in Karnataka, pointed towards people from Tamil Nadu who live in Karnataka or business owners who travel between states (Agnihotri, 2016;Joy et al., 2008).The latest violent protests took place in 2016 (Pokharel, 2016).

Division of authority over water
The management of water at the state level, or the role of WA+, cannot be well understood without insights into the constitutional and legal framework related to water in India.Water is still the legal responsibility of the states in India, unless the parliament enacts any legislation regarding regulation and development of inter-state river waters (Cullet & Gupta, 2009).In case of a conflict, states can ask the central government for mediation, and if that fails an independent tribunal can be set up to adjudicate over the issue (Chokkakula, 2014(Chokkakula, , 2015)).This also happened in the Cauvery dispute.In 1972, a report was published with findings of the Cauvery Fact Finding Committee; and in 1976, an understanding between the riparian states was reached over sharing of surplus waters between the states.However, the government that was voted in after in Tamil Nadu did not ratify the same.In 1986, Tamil Nadu made a request to the central government to constitute a tribunal, as mediation was felt to have failed.The tribunal was set up in 1990, and in 2007 a final verdict was reached, dividing the water amongst the four riparian states (Iyer, 2007).None of the parties in the conflict agreed to the verdict.A final allocation was decided upon in February 2018 by the Supreme Court of India and this verdict has not been officially challenged.It stipulates that additional water is allocated for the growing city of Bangalore, at the loss of the volume allocated to Tamil Nadu (Supreme Court of India, 2018).The Supreme Court did not define how to divide resources in drought years and allocated this task to a basin institution that was to be set up.

Basin institutions
To facilitate the implementation of the final allocation, the central government has constituted the Cauvery Water Management Authority (CWMA) and the Cauvery Water Regulatory Committee (CWRC) on 1 June 2018 (Ministry of Water Resources, River Development and Ganga Rejuvenation, 2018, Art. 10 §3).The tasks of the authority are, amongst others: to determine the total residual storage in the specified reservoirs; to take stock of the actual yield in the basin; to ensure implementation of the final allocation including the carryover storage during good year and the water releases for environmental purposes; identify situations of distress -caused by diminishing water flows in the river basin; to provide guidance for the integrated operation of important reservoirs; to maintain an account of cropping patterns, cropped area and irrigated area for each party state; to maintain an account of domestic and industrial water usage by each party state; and lastly, to consider proportional and temporary reduction in the allocations.Moreover, the authority is tasked to 'set up a well-designed communication network in the Cauvery basin for transmission of data and a computer-based control room for data processing to determine the hydrological conditions including distress' (Art 10 §3(vi)).The tasks of the CWMA closely relate to what the WA+ can offer in terms of information.It is therefore interesting to analyse how the WA+ connected and disconnected with existing practices when introduced to India and the Cauvery.

The Cauvery and WA+
The following analysis is based on the interviews with 38 people who are either connected to WA+ or the Cauvery.The interviews show that the phase of problematization does not start in the Cauvery, but rather in the Netherlands where WA+ is developed since 2012.The lead developer of WA+ at IHE Delft has functioned as its most prominent policy entrepreneur by widely promoting WA+, supported by a representative of the Ministry of Foreign Affairs in the Netherlands.It is not only the current technology, but also especially the future potential of WA+ that was conveyed to potential users.This is based on high expectations of developments in RS (Water Accounting+, 2016).WA+ is first brought to India through the IWMI, which has an office in India, and the National Institute of Hydrology (NIH) in Roorkee in North India.It connects well to a campaign of the World Bank that promotes water accounting for India (2030Water Resources Group, 2017).Based on good existing contacts, WA+ as specific method for water accounting is subsequently presented to the then-Secretary of the Ministry of Jal Shakti, Department of Water Resources, River Development and Ganga Rejuvenation, by the policy entrepreneurs connected to IHE Delft, the IWMI and the NIH.
The worldviews that WA+ promotes, which includes, first, the idea that with transparent data more strategic decisions can be taken, and second, the idea that more transparent data promote cooperation, connects well with the National Water Policy of India.It especially connects with the aim of India's National Water Mission to 'optimize water use by increasing water use efficiency by 20%' (Ministry of Water Resources, River Development and Ganga Rejuvenation, n.d.), as well as to long-term efforts to collect homogeneous data from the different states through the National Hydrology Project and the Water Resources Information System (WRIS).The problems are thus understood in a similar way, which enables the translation of WA+ from the Netherlands to India.WA+ is thought to contribute well to the National Water Mission and ongoing programmes, with WA+ being accepted as a framework to provide evidence to help to 'manage excessive and conflicting water demands and negotiate trade-offs especially in adeficit year' (Central Water Commission, 2019, p. 72).Two basins are selected to function as case studies, which are the Tapi and the Cauvery, the latter seen as the most difficult case and opportunity for WA+ to prove itself in a basin with little datasharing between riparian states.The idea is to roll out the project all over India in case of success.
In the phase of interessement, support is gained from four national institutions including the Central Water Commission, the NIH, the Central Groundwater Board and the National Remote Sensing Centre, who are asked to second personnel to develop WA+ for the Cauvery and Tapi basins.A total of 12 officers are seconded (enrolment), and are trained in using the WA+ method for three months in the Netherlands and two months in India in 2017 and 2018 (mobilization).During this training, the WA+ framework is applied to the Cauvery.It is in this period that the final verdict on allocation of the Cauvery's water is given by the Supreme Court, and the establishment of the CWMA and CWRC was notified on 1 June 2018, with directions for the CWMA and CWRC that connect directly to the aims of WA+, including taking stock of water flows in the basin, making accounts of cropping patterns, and setting up a communication network (see the section on basin institutions).
Translation, from problematization to mobilization, worked well at the national level.However, the riparian states do not wish to buy into the problematization, or the solution that is presented through WA+.In September 2018, the results are shared with the four riparian states in a workshop (Central Water Commission, 2014).The states all accept the invitation, though this does not necessarily imply an openness to the problematization or to the roles they are asked to play in engaging with the information and in any foreseen cooperation based on the WA+ outputs.In this workshop, time is first taken to show that the developers understand the political challenges, after which the modelling approach of WA+ is explained, and the results of the WA+ study are presented.Ultimately, the state representatives indicate no interest in WA+, and it is not used by the CWMA and CWRC.The process of translation is halted at the problematization phase.The justification for the problematization and suggestions made to solve this problem through WA+, that are so well-accepted at the national level are not easily accepted at the state level.In the next section we will focus on possible reasons why the process stopped at the problematization phase, how this prevented an adoption of WA+ to happen, and how answers can be found in the relation between the worldviews, expertise and materiality of WA+ and the context of the Cauvery.

WA+ disconnects in the Cauvery
We highlight here four elements that constitute a disconnect between the approach of WA+ and management in the basin, derived based on an analysis of the interviews.It allows us to show how WA+ engages with the context it is applied in and shows how worldviews -including ontologies and visions of the future, expertise and materiality all play a role.Although these four are closely intertwined, we selected examples that speak for each.

Ontology: a basin versus a reservoir approach
WA+ provides an analysis of the water balance at the river basin scale, mainly built on data collected through RS translated into pixels, and inspired by the idea that water should be managed in an integrated way (Karimi et al., 2013(Karimi et al., , pp. 2461(Karimi et al., -2462;;Molden, 1998).WA+'s strength is that it computes pixel-level water balances throughout a basin and provides spatially distributed information on water availability and consumption.This includes all green water and all blue water, such as reservoirs, but also smaller tributaries.While calculating the amount of water in specific reservoirs, but not all, is possible from RS, it requires additional information on bathymetry in addition to the WA+ method.This information is not always available, and it is not an analysis that is incorporated in WA+ as the focus is on the water balance based on flows of water at the pixel and basin scale, not at stocks in specific water bodies.A result is that WA+ implicitly and explicitly suggests to the riparian states to adopt a basin view while using WA+, which entails taking all water in the basin into account, instead of the current focus the states have on the water in the reservoirs.
This basin approach, facilitated by the worldview of an integrated approach to water management as well the materiality of the RS data, is one factor in creating a disconnect.Historically, the conflict over water allocation in the Cauvery only concerns water in the main reservoirs created by dams on the river.This thus excludes discussions on how surface water in smaller tributaries or lakes, as well as groundwater is used and shared, creating a resistance to engaging with WA+.

Visions of the future: water productivity, beneficial uses and implied objectives
The main causes of contention between farmers in Karnataka and Tamil Nadu are the ideas that farmers in the delta have historically had access to more water and that farmers in the upstream project command areas have started irrigating water-intensive crops that diminish downstream flows.In times of resource scarcity, this results in a limited willingness to share.WA+ provides information about the gross biomass water productivity, calculated in kilograms of biomass produced per m 3 of water consumed.This is shown in sheet 3 of the WA+ analysis, but can also be shown in a map (Figure 4).Often a conclusion is drawn that high water productivity is good, while low productivity is bad, based on dealing with situations of water scarcity and ideas of what beneficial and nonbeneficial water consumption is.
To illustrate this, in the case of the Cauvery we can show with open-access data that a quick conclusion could be that water can be saved in the paddy cultivation regions of the basin (reddish colours in Figure 4 indicate low water productivity in the delta region and in the command areas of Krishnarajasagar and Mettur projects).This links to historical causes of contention between farmers in Karnataka and Tamil Nadu described in the brief overview of the origin of the conflict, and can easily be politicized without a further context.The information can be interpreted in bad faith as careless management of a precious resource, and used as leverage to demand the release of water or to ignore claims for more water.Especially judgements based on biomass alone are misleading.Context is needed.For instance, detailed regional and local parametrization is required to be able to determine what amount of the biomass produced can be used for either consumption by humans or animals.Also, the biomass figure does not provide information about what agricultural alternatives could be available that also work in the specific socio-economic contexts that are analysed through RS.Thus, the WA+ model, presented as neutral, can seemingly provide politically very sensitive results for the actors in Cauvery to engage with.

Materiality: pixels for management versus discrete monitoring
WA+ analyses are largely based on RS data.The resolution of this data directly influences what is observed, why and how.There are great differences between the information provided by WA+, which is intended to manage farming practices, and the data used by the CWRC and CWRA are used for monitoring.
Data collected by the CWRC and CWRA are supplied by the states to monitor the water usages in the basin to potentially inform decisions on water allocation in times of shortage.They include the accounts of cropping pattern, area cropped and area irrigated, and domestic and industrial water usage which can be collected at potentially very spatially explicit level and shared at any administrative unit level.If and how such data are verified is not clear, yet the minutes of meetings available in public domain show that there is no discussion about the data itself, although discussions are being held on what data are admissible within the framework of the CWRC and CWRA (Cauvery Water Regulatory Committee (CWRC) Secretariat, 2018aSecretariat, , 2018b)).In contrast to these data, the WA+ data can be used to develop outputs that can zoom in to pixel level, which can draw the attention to management at panchayat (village) level.However, the pixels do not necessarily tell the whole story.To demonstrate what can be seen and not seen in WA+, we discuss the spatial resolution relating to P, ET and land use land cover (LULC).P and ET are measured and estimated at spatial resolutions of approximately 5 and 1 km 2 , sometimes at different timescales (Salvadore et al., 2020, p. 72).LULC maps can in theory be done at a very high resolution (currently 10 m), but making a detailed and accurate LULC map requires extensive field data and managing high volumes of data.For the Cauvery, a LULC map with a 270 m resolution was chosen, which was also the resolution at which the modelling was carried out.The mismatch between the resolution of P and ET that are measured at the spatial resolutions of 5 and 1 km 2 for the LULC pixels of 270 m requires a generalization of P and ET for every LULC pixel.Very locally, a situation may be different with some of the pixels receiving water, and others not.We also want to draw attention for what cannot be seen within the pixels, but what would be potentially influenced through management decisions based on WA+.The LULC, but to an even larger extent the ET and P, pixels can represent a large variety of farmers and farming practices, as well as different socio-political practices that influence if and what actions may have more or less beneficial effects on water use.Zooming-in to WA+ results at the pixel level should therefore be used as an indication of where to conduct further research, but not as a sole source of information on which to base policies.Figure 5 is an illustration of the diversity of areas contained within one ET pixel.

Expertise: complexity and specificity
Expertise of the makers influences the design choices made in a model and thus what is included and not, as well as how (Addor & Melsen, 2019;Babel et al., 2019).The development of WA+ has attracted people with a specific expertise that is required to work with RS and hydrological data, requiring people who want to implement WA+ to be able to code in the Python language, or at least have a basic understanding of how the data are produced.In addition to coding, expertise is required to interpret the RS data.For instance, knowledge of the field level is required to interpret an LULC and to identify whether the classes suggested relate to what is happening on the ground.Related to materiality and expertise, the WA+ analysis is based on different data sources with RS data being an important source, as well as calculations based on generalized assumptions, for instance, the percolation rate.It takes considerable time to learn to understand and assess the WA+ analysis and results.
Although efforts are being made to make these steps transparent and explain how and why, and based on what data and theories calculations are made (Salvadore et al., 2020), the high level of complexity still renders the process opaque to many.The 12 engineers seconded to the project had time to learn about the worldviews, expertise and materiality of WA+, although most focused on aspects related to their professional backgrounds.One interviewee remarks how this resulted in discussions that were difficult to follow for outsiders, due to jargon specific to WA+ that became familiar to the engineers.The difficulty to understand all the technicalities of WA+, however, does not necessarily create an immediate disconnect.Three interviewees active at the international and national levels indicated they did not have the expertise to understand WA+ fully, but they had created specific and positive expectations of what the analysis could deliver, which contributed to the way WA+ was accepted at the national level in India.This difference in expectations created different imagined versions of WA+, disconnected from the technology itself, for which people had higher expectations than could be delivered.

Discussion
In this article we aimed to understand how WA+ connected with water management in the Cauvery basin in India, and to draw lessons on if and how models and RS data can be used in the context of transboundary water conflict.The case study shows that the act of collecting and analysing data, even if intended as a neutral contribution, is in essence political.This confirms studies that emphasize the social construction of models (Addor & Melsen, 2019;Krueger et al., 2012;Melsen et al., 2019).At the same time this research adds to studies that remain within 'model-land' (Thompson & Smith, 2019), as well as studies that disregard how a model is made when analysing its impact.The model and the way it gains influence are better understood through the interactions of different elements of the model with the context it is applied.
The model travels and changes in interaction with the people and institutions that use it, and becomes political in different ways for different people (Bijker et al., 1987;Latour & Woolgar, 1986).Seeing a model as an interplay of worldviews, expertise and materiality has proven a useful entry point to analyse interactions between the model and context (MacKenzie & Wajcman, 1999).In combination with the theory of translation it provides insights as to why a model or data-driven tool is accepted or rejected.It provides an indication for both modellers and model-users as to what could be changed in a model and modelling process to improve the interactions between model and the context it is applied in.This is challenging as it requires modellers and model-users to think and work outside of traditional disciplinary boundaries.
What our analysis has not covered in-depth is the (potential) influence the use of WA+ has on water management and governance and thus water users at different levels, such as done by Sanz et al. (2019) for a groundwater model.We have not discussed questions of morality and justice explicitly, which could be useful avenues to explore impact.The WA+ analysis is not intended as being a sole input for decision-making, but makers cannot prevent that it may be used for setting standards and monitoring them.The use of global datasets enables decisionmaking from afar, disconnected from a local reality (for instance, the work of, e.g., Haraway, 1991;Litfin, 1997; and specifically for WA+, see Zwarteveen et al., 2018).This can also impair democratic decision-making processes.Moreover, in a country in which conflict over water is both highly politicized and localized, and where caste and class play a large role in water allocation and management at the micro-level, bringing in the local context is required to understand water within a pixel and between pixels (e.g., Shah, 2008).
Lastly, the theory of translation draws our attention towards those who support a certain technology, but also those who aim to change or reject it (Latour, 1984, p. 267), to which we emphasize the importance of knowing who disagrees or avoids interaction.The latter group is often forgotten in (participatory) modelling efforts, but crucial when it comes to using data and data-driven tools in conflict settings.These tools are not automatically a useful contribution, nor accepted, based on a seemingly neutral scientific contribution.Often, the conflict is not about data, or data can be politicized, as we have seen in the Cauvery case.
In this discussion we make space to share the following lessons from the WA+ process we, as authors, were part of.We invite others to do the same, whenever possible, to facilitate learning on how data and tools can be used in more conflict-sensitive ways for transboundary water governance and management.
• Involve actors from the beginning, to define the problem and possible solution together and situate problems and solutions to develop a model (if needed) that is fit for purpose: WA+ was designed as a one-size-fits-all model -with one goal of standardization being to allow for comparison between basins.Unfortunately, the specific needs of the CWMA and the riparian states were not fully addressed by this approach.
• Take time for conflict sensitive modelling, with space to adjust a modelling process: There are high expectations of funders/commissioners of what data and models can achieve.This can influence how a modelling process is designed, with a model presented as a solution instead of an input for debate or decision-making.Be clear about the limitations of the model and the time required for a conflict-sensitive modelling process, be aware that modelling may not be a solution, and create safe spaces for political discussions.
• Diversify the team: knowledge diversification is highly recommended, for instance, to limit biases in the modelling process through challenging assumptions, or to be able to design the process based on local as well as modelling expertise.
• Recognize that the capacity to use the model depends on many factors, including technical, institutional and political: An understanding of institutional capacities and expertise is required to ensure the proposed tool can be critically assessed, appraised and applied by those who use it.For WA+, institutional specificities, including a design of local institutions based on disciplines, as well as the position of the CWRC and CWMA as young institutions in a highly political environment, influenced the possibilities to critically assess, appraise and apply WA+.• Be aware of the realities of those who are modelled, and potentially affected by modelling: In the WA+ project, local realities were not taken into account in the problematization phase, but they were partly included in the WA+ modelling activity and the expected action.This brings up critical questions on the data democracy that is aspired by WA+, and how to relate this global aim to local needs.

Conclusions
This article contributes to a critical and constructive discussion on the application and adoption of RS-based models, specifically by looking at the implementation of WA+ in a setting of water conflict across state borders, with a clear intent to contribute to water cooperation.It calls into question the objectivity of satellite data-driven models and the idea that satellite data could create a level playing field and spark cooperation in a conflict-ridden basin.It shows the influence of ideas on how the world works and should work, strengthened and influenced through expertise and materiality, that are embedded in modelling practices, as well as how models travel and interact with the context they are embedded in, sometimes in different ways than the developers intended.The study clarifies how the worldviews embedded in WA+ which aligned with national-level objectives related to ensuring water-use efficiency were not accepted at the state level where more resistance to management by other parties and the introduction of new potentially contentious data existed.
The study draws our attention to how introducing data and data-driven tools into a conflict needs to be carefully adjusted to the local context.In the case of WA+ and the Cauvery, this was especially challenging due to disconnects and difference in needs between the national and state levels.The approach to create a data-democracy by sharing the same data to all parties involved may sometimes overlook power differences and the political aspects of data and models.By discussing how WA+ was introduced, we therefore raise questions on how data and information, that hold in themselves a clear idea about how water should be managed, can contribute to water cooperation.It shows that working towards transparency and a data democracy should take into account the ideas and ideals that are embedded in data and datadriven tools.Not being aware of these in a situation of conflict over water can inadvertently contribute to more conflict.Bringing these ideas to the fore from the start could ensure that the data and models do not interrupt a potentially very sensitive and long-term process of building trust between parties in a conflict.

Figure 1 .
Figure 1.Unpacking a model as technology, through ontologies, visions of the future, expertise and materiality.Source: Based on Mackenzie and Wajcman (1999).

Figure 2 .
Figure 2. Worldviews, expertise and materiality of WA+, as intended by the designers of WA+.Source: Figure based on analysis by authors.

Figure 3 .
Figure 3.The Cauvery basin.Sources: Outline digitized by the authors based on Central Water Commission (2014, map 2(b), p. 4); main rivers and reservoirs obtained from Lehner et al. (2022); and national and state boundaries obtained from https://www.diva-gis.org/gdata

Figure 4 .
Figure 4. Gross biomass water productivity of agricultural areas for the period June 2006-April 2007.Sources: Authors using the following data sources: biomass from MOD17 (Running & Zhao, 2015), ET from Senay et al. (2013) and land use from Zanaga et al. (2022).