 All right, it has gone one o'clock and we've got a few people in the Zoom session now. So I'll go ahead and start talking. My name is Dr. Julia Kazmaier and I'm the lead of the Computational Social Science Training Team at the UK Data Service. So we're really interested in getting social scientists and computational scientists to work together to combine their forces and achieve social science research projects with computational methods. You know, just try and hit some of those really big questions from a new angle. And today is the guest lecture portion of our agent-based modeling training series. I'm really pleased to introduce Dr. Kavan Narashaman. I think I got that wrong. She is a research fellow working at the intersection of computational science and social science and she has a very interesting real-world application of agent-based modeling. So please take it away. Thanks, Julia. Just checking everybody's able to hear me. Excellent. So hi, everyone. Thanks for joining us today. As Julia mentioned, I am Kavan. I am a research fellow at the Centre for Research in Social Simulation at the University of Surrey. I'm also an ESRC policy fellow, currently doing a second one role as data social scientist at BASE. In today's seminar, I would like to talk about the what, why, how and for whom aspects of our agent-based model called watering used to explore decentralized water management in sub-Saharan Africa and share our experiences, reflections and some resources from our research. I do hope to go through all of the different phases we went through to develop the agent-based model, but I'm sure there will be time at the end for questions and hopefully also to play with some of the sub-models that we have created from watering. So let's get started. So the one-percentage accessible fresh water that's available on our blue planet, which is almost 70 percentage of water, must meet the water demand of almost 7.8 billion people. Water management in simple term deals with how one plants and implements the rules and processes to store, divert and supply water to water users. Irrigation management specifically relates to water use principally for agriculture, but there may be other uses too, like domestic consumption and livestock consumption. In the 1950s and 60s, governments, international donors and development banks invested in large scale, centrally managed irrigation systems worldwide. Many of these irrigation systems underperformed or failed altogether due to improper planning, poor maintenance, as well as overlooking the local needs and traditions of farmers and communities as such. There are many examples in literature where such causes cost the failure of irrigation systems, like the Mesopotamian civilization, or more recently, the case of barley rice terraces where overlooking the locally coordinated cropping plants and rituals of rice farmers, organized in cooperatives called Subax, led to severe crop losses. The Green Revolution in barley was around 1971, so other failed cases of centralized irrigation management around the world in sort of a similar timeframe spurred the irrigation management reforms in the 1970s. And then further push for decentralized water management came in the 1990s with the integrated water resources management paradigm, which many countries signed up for, essentially suggesting that fresh water is a limited resource, so water management should use participatory approaches involving water users, planners and decision makers, and that water has an economic value in all competing users. Many developing economies have huge demand for fresh water for agriculture. In sub-Saharan Africa, where we conducted our case studies specifically, the demand for fresh water for agriculture is significantly higher than other sectors, as you can see in this plot, and there is huge potential for irrigation, estimated to be around 34 million hectares, but only around 10 percentage of cultivated land in sub-Saharan Africa is irrigated. So governments, donors and private sector investors are all keen to support irrigation development in the region, and decentralized water management is considered central to the irrigation expansion efforts. Decentralized water management broadly means water allocation and supply is handled within local communities with occasional or no technical and financial support from external actors once the project is live. There are further nuances to decentralized water management in sub-Saharan Africa. Firstly, there are differences based on how water is used for cultivation. There is traditional rain-fed farming where seasonal floods or rainwaters are utilized for growing crops, and then there is irrigation farming, and within that we have farmer-led initiatives or donor or state-sponsored schemes. Farmer-led irrigation schemes refer to small holder farmers organized in small groups and they use small affordable pumps to grow high-value produce such as vegetables. On the other hand, we have the sponsored schemes where state entities such as the Ministry of Food and Agriculture or an Irrigation Development Authority or other external donors would fund and implement large-scale irrigation systems and then often hand over operation maintenance and management responsibilities to community-level groups. The scheme-level groups responsible for routine water management are referred to as water user associations or OUAS in short. So, OUAS are semi-formal or informal institutions. Usually sponsored schemes have semi-formal OUAS and farmer-led schemes have informal OUAS. Semi-formal OUAS have institutional mandates which mainly focus on allocating water to users, the operation and maintenance of the irrigation system as well as collecting fees and fines from farmers and other members. Informal OUAS often work towards similar objectives even if they don't have a mandate as such. OUAS are generally small-scale with only a few hundred members each. Previous research suggests that OUAS often struggle to fulfill their mandates. The main reasoning is that there is an expected feedback loop whereby improving irrigation performance through the provision of irrigation systems would boost agricultural productivity and income and thereby encourage users to comply with water regulations and pay water fees regularly. However, this expected feedback loop is disrupted by several factors, most notably the limited financial resources, the technical and management expertise of OUAS whose authorities are often elected members of the local communities. Furthermore, there are conflicts between competing water users and wider issues related to limited inputs and market opportunities for crop production, rigid water bureaucracies and so on. But despite all these challenges in implementing decentralized water management through OUAS, efforts are ongoing to expand irrigation and introduce OUAS or turn existing collectives such as farmer unions and village committees into OUAS. Against this backdrop, there is a recognized need to understand how OUAS operate and explore management options that can improve outcomes for OUAS, farmers and more broadly the local communities. But it is a significantly expensive and time-consuming task to do this research in real world, whereas computer simulations provide attractable and cheaper alternative to study the system. But to build meaningful computer simulations, we always need real world data. Due to practical constraints involved in collecting this data, we decided to focus on exploring specific aspects of OUAS. We particularly wish to understand how does participatory irrigation management work through OUAS. Participatory is the key word here as we were specifically interested in understanding and experimenting through our simulations how varying levels of participation, cooperation and conflicts among water users would affect water use, irrigation management and economic productivity within a work catchment area. Our second question was which work management option is better for performing the for improving the economic productivity of water users and the OUAS itself. And here the key words are management option which in simple words means the arrangements to store, divert and deliver water to end users. So we started our work with fieldwork to help us define the scope and the boundary of the system to model. This helped us develop a narrative of decentralized water management through OUAS and understand if there is data available that we could use to calibrate and validate our model or to see if it's feasible to collect such data. We conducted semi-structured interviews with different OUAS members, representatives and extension offices all mostly within the Upper East region of Ghana, particularly near Bolgatanga as is shown on the map. We also interviewed researchers based in Accra, Ghana and elsewhere in Sub-Saharan Africa and internationally who have all worked with OUAS in Sub-Saharan Africa. Some key insights emerged from our fieldwork. Firstly, we got to know that water use is not exclusive to farmers in most if not all irrigation schemes. Households who live within the catchment area of irrigation schemes and pastoralists also tend to use water. At the same time, it's fuzzy whether households and pastoralists can be OUAS members and pay water fees. Some OUAS say that they cannot be members and that should not use water and don't have to pay fees while others feel that households and pastoralists use water anyway. So they might as well be OUAS members and pay fees. Few words commented that water is for everyone. So pastoralists and households should not be charged even if they use water. Further complexity arises from some farmers who are not OUAS members using water from the scheme for irrigation but not paying for it. We have classified such farmers as independent irrigators. Their water use can be significant and it can cause severe water stress to the screen, especially during periods of water scarcity. The different types of water users and users within a scheme sometimes prompts conflicts among the water users and between water users and the OUAS. And finally, even within similar schemes in the same region, there is no standard management template for OUAS on how to allocate and supply water to communities. So the system we wish to explore through computer simulations look somewhat like this. Within each irrigation scheme is a OUAS. It is responsible for water management which involves allocating and supplying water according to demand, collecting fees from members, conducting meetings and resolving conflicts between water users. Since an irrigation system is primarily intended for agriculture, most OUAS members are farmers. They use water from the scheme, earn income by growing crops, pay fees to the OUAS and participate in meetings to learn about OUAS policies and how to comply with OUAS regulations. That said, almost always there are other water users within the catchment of an irrigation scheme. They are pastoralists, households and non-member farmers. Non-member farmers might be independent irrigators who use water from the scheme, but don't pay fees or non-irrigators who engage in traditional rain-fed farming. So they don't use irrigation and don't pay fees to OUAS. Pastoralists and households use water from the scheme for feeding livestock and for household consumption respectively, but they don't pay OUAS fees. Additionally, pastoralists might be forced to pay fines if their livestock destroy crops and households could have random demand for water for construction purposes. The technique we wanted to use for the simulations was becoming apparent to us by this point as we developed the narrative for the simulation. We understood that the system involves infrastructure, elements of human agency and the ecology of the command area of an irrigation system, or in other words, we were trying to model a socio-ecological system. And there are different types of actors within the system, each having a need to use or manage an essential shared resource, water in this case, to meet the different objectives. In other words, we were dealing with a common poor resource problem with issues of excludability and subtractability. The system has multiple potential causes for failures, especially in terms of exploitation and scarcity of resources, but lacks even a partial causal explanation of how all of these different factors contribute to specific outcomes. But this knowledge is essential for improving decentralized water management through UWA's. Considering there is also limited individual and UWA level data on what demand and consumption, we needed a flexible modeling approach that was capable of working with reasonable theoretical assumptions where necessary. Finally, given that there isn't a standard UWA template, we needed a modeling approach that would allow generating meaningful counterfactuals to assess the impact of different management strategies based on the settings provided. All of these considerations made us to choose an agent-based modeling approach for simulating decentralized water management through UWA's. The agent-based model that we have developed is called Watering. It stands for Water User Associations at the Interface of Nexus Governance. We hope it can be useful as a tool to explore and plan community-level water management. More specifically, we hope the model will allow its users to understand and explain the combined influence of UWA policies and community participation on water use and income earned in an irrigation scheme. The agents in watering are farmers, pastoralists, households, and this one UWA agent. The environment is a simulated irrigation scheme comprising a physical asset made up of a reservoir, a primary canal, and secondary canals. Watering also has a social environment whereby neighboring water users influence each other. The watering model is implemented in NetLogo. It is an empirically-informed stylistic model, meaning that empirical data was used to inform model design, calibration, and validation. But the model itself is not a one-to-one representation of reality. We have made some simplifications in the model, namely the only model surface irrigation where water flows by gravity. We assume that water is pumped up to the reservoir and canals and flows from the canals and to the farm plots by gravity. We have not modeled groundwater irrigation or irrigation methods like drip or sprinkler systems. We made the simplification as many of the irrigation systems that we interviewed have surface irrigation. We also model only one oar within an irrigation scheme. Large irrigation schemes could have multiple oars, but we simplified because our focus was mainly to understand the dynamics of interactions between a oar and its member users. The next simplification we made is that each farmer cultivates one plot. In reality, some farmers cultivated multiple plots and grow different crops, but we simplified to one plot per farmer. But the size of the plots varies from less than a hectare up to a few hectares. We have also assumed that water users don't switch economic activity. In reality, if farmers or pastoralists incur any income losses, they might sell their farms or livestock and then take up another economic activity or migrate elsewhere. While farmers can still sell or rent out their plots in the model and pastoralists can sell off their livestock, we have not modeled water users switching their economic activity. Rental markets and migration are complex systems in their own right, but since we only focus on one irrigation scheme, we decided not to model these aspects in the current version of watering. We also made a simplified assumption that only one crop can be grown per plot per season, that is monocropping. While intercropping is possible, small holder farmers predominantly do monocropping. The model runs in monthly time steps. Each net logo tick represents one calendar month and each simulation is around 20 years. The image you see in this slide is the net logo worldview of watering. There's a reservoir in the top left primary canal just below the reservoir which runs across horizontally and the three parallel secondary canals which branch from the primary canal, that is the empty blue lines, that parallel lines that you see. Water users reside on the green patches. The varying shades of green indicate the different heights of patches. So darker the green, greater the height. Water reaches the patches with the highest elevation first assuming that water is pumped up and then flows to patches with lower elevations. Zooming into the image, the plant icons that you see here are the scheme farmers who are members. The flower icons are non-irrigators. The tree icons are independent irrigators. The house icons are households and the livestock icons are pastoralists. Also seen in this image is that scheme farmers are closer to the canals, the empty blue line, while independent irrigators, non-irrigators and households and pastoralists are further away. So this mimics the situation in the irrigation schemes that we studied. In terms of crop water demand, we model commonly grown dry season wet season and multi-season crops. For example, in the irrigation systems that we studied, tomato, maize and rice are commonly grown. While one crop can be grown per season up to two crops can be grown per year since the duration of each crop be modeled as anywhere between 120 to 180 days. The crop growth occurs in four stages, initial development mid-season and late season. The water requirement for crops varies by growth stage. The duration of each crop and crop stages are set via the interface tab in the NetLogo model. We use the Blaney-Kirgel method to estimate crop water demand. It is a simple but commonly used approach to calculate crop water demand based on data like mean monthly temperature, monthly percentage of annual daylight hours and crop coefficient. We collected the necessary data for this from existing climate data sets and the FAO irrigation manual. Crop water demand can be met through irrigation rainfall or both depending on the season. And lastly, the number and types of crop each farmer grows depends on their affiliation to the ua. For example, ua members have access to irrigation so they can cultivate during both dry and rainy seasons, whereas non-irrigators only cultivate during the rainy season and they choose the crops accordingly. Next, in terms of what demand for livestock and households, pastoralists are agents in the model who have large numbers of cattle, goats and poultry. Some crop farmers also do livestock keeping but they have fewer animals. The water demand for livestock is calculated based on the number type and average water demand by type. The demographic characteristics of the household agents in our model is similar to the characteristics of the households in the irrigation schemes that we studied. The household water demand is calculated based on household size and the average water demand per person. Household agents also have random use of water for construction purposes. Normally in the irrigation schemes we studied ua member farmers are allowed to use water from the irrigation scheme for their livestock but pastoralists who own large numbers of livestock are not ua members so water demand for their livestock is not regulated by the ua. At times this results in conflicts between water users and even when ua members are willing to overlook the water consumption of livestock, conflicts arise when livestock destroy crops. Ua is expected to intervene in such situations to recover fines from pastoralists. We model these dynamics in watering. We also model counterfactuals where pastoralists and households can become ua members and pay fees for their water consumption. So ua's management policies are set via the interface tab in the NetLogo model and there are three options, membership, water allocation and strictness. The membership policy of ua can either be exclusive or inclusive. When set to exclusive only scheme farmers can be ua members. So even if pastoralists and households use water from the scheme, they cannot become members and pay for their water use. This is what we see in most irrigation schemes. The counterfactual is inclusive membership where scheme farmers, pastoralists and households can all become ua members and pay for their water use. In any case though, independent irrigators and non-irrigators don't become ua members. Non-irrigators don't have to as they don't use water from the scheme but independent irrigators do use water but they avoid membership and paying fees. The water allocation policy of ua can either be upstream downstream divide or equity. In the former upstream users get water first followed by downstream users whereas in the equity scheme the amount of water available for irrigation is divided equally across the scheme. So in case of an acute water shortage we can expect that farmers or more broadly water users wouldn't be disadvantaged just by their downstream position. In terms of the strictness policy we can set one of three stances for the ua which controls how strictly it enforces the water allocation decisions on water users limiting how much users can draw more than their quota. Enforced means full restriction on water use beyond quota. Incentivized means limited restriction on water use beyond quota or relaxed means no restriction on water use beyond quota. So in terms of the ua agents actions each year ua estimates annual water demand by month based on the cropping patterns of farmers who also includes estimates of water demand from pastoralists and households if ua membership is inclusive. Each month ua then allocates water to its members proportional to their demand based on the amount of water available in the scheme. Point to note here is that ua's estimates and allocations all based on the water demand of its members only. We assume that ua cannot know the demand arising from non-members because it won't be the same each month and ua cannot keep track of all the unregulated water use. This mimics the situation observed in the irrigation scheme. The ua agent also collects fees and fines and uses its income for operation and maintenance tasks in the irrigation scheme. ua loans money to its members affected by income loss to help them stay in business. So as we already saw in terms of water users there are three main types of water users farmers, pastoralists and households and there are three subcategories of farmers scheme farmers, independent irrigators and non-irrigators. ua membership can be exclusive to scheme farmers or inclusive of scheme farmers, pastoralists and households. Scheme farmers almost always form the majority of water users in the irrigation scheme. At the end of each year water users are in one of three economic states depending on their economic position. All water users are active by default. They become inactive when their costs exceed their income. In line with observed data we have modeled that water users who remain inactive for more than three years become obsolete. That is they will seize economic activity. However, within three years of becoming inactive users can go back to being active if they rent out their plots to other water users or borrow money from ua's to start this business again. So in terms of the water users behavior they have some reactive behavioral stances. So each month water users adopt a behavioral stance to either cooperate with a CP or not cooperate with the worst water allocations and CP. If they cooperate the chances are higher that they will comply with the worst water use quotas. If they do not cooperate the chances are higher that they will withdraw more than their quota. We assume that ua members cooperate if their water allocation exceeds their demand their baseline demand. Non-members simply have a 50% chance of cooperating with ua's water allocations. Once water users have their behavioral stance they will then calculate a utility to cooperate with ua's water allocations based on their own behavioral stance the dominant behavioral stance of their neighbors and the strictness position adopted by the ua. So in mathematical terms the strictness of the ua is a factor added to another factor which focuses on the neighborhood influence to determine each water user's utility to cooperate. Based on the utility to cooperate each water user will then calculate the actual amount by which they would reduce their water demand. Finally any reduction in water demand as a result of this process has a proportionate effect on water users. For example, less water results in reduced crop yields and lower income for farmers and less water for livestock results in poor health or death of livestock. So based on our assumptions about the behavior actions and decisions of the water users and the ua which I have presented so far we set up and ran net logo simulations using the integrated behavior space tool. We have already discussed the first three model settings that you see in the slide but additionally we also included a setting to model damages caused by livestock. There is some damage to crop or infrastructure caused by livestock in several irrigation schemes. So true means livestock costs damages and false means livestock don't cost damages but still there's a random chance of infrastructure damage. Example maybe it's due to age. Finally we have some settings controlling the duration of the frequency and the intensity of irrigation events. The combination of the parameters and the parameter settings in this slide resulted in 24 unique settings which we ran 10 times each resulting in 240 model runs in total. So before going on to look at the results I will now show a quick demo of the watering model implemented in net logo and then we'll go on to the results. I'm going to stop sharing for a second and share the model just a second. It's just taking a second to load. Okay, hopefully we should be able to see it now. There we go. Okay, hopefully you are seeing the net logo model. Can I get a quick thumbs up? Excellent, thanks Julia. So this is the watering net logo model. So as you can see here on the left we have most of the input settings and controls that a user of the model can change to kind of explore the different options that they would like to investigate using this as a tool. In the middle we see the net logo world and on the right we have major output plots. We also have some controls further down below. So the model ships with these input controls people can change it if they want to or just leave it as it is and just explore the effects of the main management controls here on the top left. So just quickly if I set up so I'm just setting the total irrigation allocation for the scheme meaning 60% of water available at source is allocated for irrigation. The rest could be used for other purposes. The scheme irrigation efficiency essentially is one factor that combines the effect of multiple factors like how effective is the infrastructure to retain the water? Like what about water losses and things like that? The discharge rate controls at what rate water goes from the main outlets into the canals. The irrigation hours per day is the duration of each irrigation event and the irrigation days per month is the frequency of irrigation events each month. So if I'm happy with these settings now so if I just press go we see that water flows through the canals. There was some orange line in between that indicated there was damage to the canal which the world would then have had to fix. So the green and the yellow changing plant icons show here that whether or not a particular plot is actively cultivating. The blue color of the patch indicates what a flow to that patch and on the right we have various outputs resulting from the model. Some of the results I will discuss now in a second. So hopefully that's just a quick demo of how the watering model as such looks in its entirety. So now we'll go on to discuss specific results. So I will pause this, stop share and go back to the slides. Hoping you're back on the slides again. Excellent, thanks, Celia. So over the next few slides we will look at some results from our watering simulations. So the results seen in this slide shows that the average annual income over a 20 year period which is the duration of each model run of different categories of farmers under different water management policies. So the results here show that scheme farmers earn more than independent irrigators and non-irrigators which is expected considering that they have better access to irrigation. Non-irrigators earn the least which is also expected as they only cultivate during the rainy season. Farmers earn more with an upstream, downstream water sharing arrangement. Again, it is expected as most independent irrigators who engage in unregulated water use are located downstream. But surprisingly though, we found that scheme farmers earn more when water is relaxed and conversely independent irrigators earn more if water is strict. This happens because when OOIR enforces its decision of limiting water supply and scheme farmers are the ones who readily comply with OOIR's decisions, they suffer most water shortage which works to the advantage of independent irrigators who would abstract water anyway from the scheme. So in the event OOIR is not able to control unregulated water use within a scheme it might be better if OOIR is less strict in rationing its water supply to its members. The next slide seen here shows the average annual income of OOIR over a 20-year period of each model run under different OOIR management settings. Here we can see that the income trend is slightly reversed based on the water allocation stance and that OOIR actually earns more under an equity arrangement. This is because more users make a marginal profit under an equity arrangement and are thus able to pay more fees compared to fewer users making greater profit under an upstream downstream divide scenario. But in terms of strictness, it is also in OOIR's interest to be more relaxed with its water allocations and avoid strict rationing to improve its income. With strict rationing, however, adopting an upstream downstream water sharing policy better favors OOIR income. The next slide seen here shows the average annual income of OOIR over a 20-year period but under two different OOIR management settings. First is whether livestock frequently damaged infrastructure or not and second is OOIR membership being exclusive to scheme farmers or inclusive of scheme farmers, pastoralists and households. We find that OOIR has more income if livestock damages infrastructure but this counterintuitive result stems from the model assumption that OOIR successfully recovers fines from pastoralists if they are livestock damage infrastructure. If this is combined with OOIR membership being exclusive to scheme farmers then OOIR has more income due to having to make fewer and lower payouts to support members who face income losses. But if livestock don't damage channels though then OOIR has more income if membership is inclusive of all water users. The results we see in this slide is a radar plot showing the performance of the three categories of farmers against different measures. We find that even if their income is low non-irrigators are more economically active than scheme farmers and independent irrigators as they have no water use costs as such. We also find that scheme farmers are more water stressed than independent irrigators due to complying more often with water rationing decisions. But at the same time we find that scheme farmers do have the highest yield productivity that is crops harvested per unit of water consumed compared to independent irrigators and non-irrigators. So based on all of these results going back to our research questions focusing on the role of participatory irrigation management through OOIRs and which management options work better to boost incomes our simulation results suggest that OOIR being relaxed about water users exceeding their quarters might improve the income of scheme farmers and potentially encourage them to continue participating and cooperating with OOIRs. This is true especially when unregulated water use is an issue that OOIR is aware of but is not able to control effectively as reported in some of the irrigation schemes that we studied. Next is that OOIR income is higher under the equity water sharing option because more members make just enough profits to pay OOIR fees compared to fewer members. Predominantly the scheme farmers located upstream making higher profits in the upstream downstream water sharing scenario. Another result is that OOIR membership being exclusive to scheme farmers is better when damages caused by livestock to crops are infrastructure is a prevalent issue and cannot be controlled or managed effectively by OOIRs through other means. Alternatively, if damages caused by livestock is not an issue to an irrigation scheme then inclusive membership works to the advantage of both OOIR and its water users. So some expected and some surprising results here from our simulations which helped unpack the conditions that lead to different outcomes. Of course at this point we had questions about the validity of the results and also bigger questions about what fair and sustainable water management is. So we needed a way to assess if watering has captured the key processes and dynamics if it's doing the right things and if and how it could be a home with if and how it can find a home with stakeholders involved in decentralized water management, individual or OOIR level data on water consumption and other aspects of participatory water management is hard to combine like I mentioned at the start. And yet validation is a key step in the process of seeing agent-based models to completion but the data required for validation may not be available at all, be difficult to access or not be in the format that we need. And in these circumstances as modelers I think we need to think carefully about how we think about the real world usefulness of our model. One way we did this is by organizing evaluation and knowledge exchange workshops with stakeholders and citizens in Ghana. We ran full day workshops using different methods to allow stakeholders and citizens to assess the plausibility of the simulation results but not just that we also conducted sessions through which we got participants to look under the hood of our model as it were getting their feedback on our assumptions and the way we implemented the model too. So here's an overview of how we did that. So we ran a participatory systems mapping session to develop a system level understanding of decentralized water management through OUS. The objective of the exercise was to encourage and help participants to think widely about factors and conditions affecting water management and success of OUS. I will illustrate the main steps in the participatory systems mapping activity here with a simple example. So we start off by identifying the key factors. So the success of OUS was our focal factor that you see in the red box there and the yellow boxes where the other factors which directly or indirectly influence the focal factor. We then moved on to identifying key connections. So each factor has incoming and outgoing links depending on whether it influences or is influenced by another factor. Green arrows are positive connections meaning if one factor increases the connected factor also increases. Red arrow means negative connections meaning if one factor increases the connected factor decreases. Gray arrows indicate complex or uncertain connections between factors. We then moved on to clarify the nature of the connections between factors as weak, medium, or strong. And finally, in a consolidation phase we just got participants to agree on the map produced and the knowledge generated. We used the resulting systems maps to refine and extend the narrative of the watering model. I must note here that participatory systems mapping is a complex systems approach probably worth a separate session of its own. But I have included some links in the slides here which might be useful as further reference. And it is my understanding that these slides will be made available after today. The second type of stakeholder engagement session will be ran used flow charts each showing a sequence of actions performed by different agents in the watering model such as farmers, pastoralists, and households. So stakeholders worked in small groups to review the flow charts commenting if the sequence of actions are experienced or expected from the different actors. The objective of the flow chart session was to allow stakeholders to evaluate the logic and the behavior of the model or in other words, the way in which we implemented watering. Finally, we ran another stakeholder engagement event called watering stories through which stakeholders evaluated the assumptions and the outputs of watering. But rather than presenting the assumptions and results as statistical charts we presented them as narratives and asked stakeholders to identify if the stories were far removed from reality corresponded to reality or sat somewhere in between. So our short talent discuss approach looks something like this where you can see a written description of the assumptions and results alongside plots and the pictorial depiction of the results. The story session was useful for collecting stakeholders' opinions and reflections on the simulated results and their plausibility. I must note here that our stakeholders involved different types of actors like extension officers who are members, who are representatives and even irrigation authorities. So this slide here shows an overview of how we brought together the stakeholder feedback gathered through the different sessions and that this is only part of the feedback. We found that watering assumptions and results aligned with some of the experiences and expectations of stakeholders shown in the green box. For example, they agreed with certain assumptions and results about actions, membership, water users' cooperation with the water and unregulated water use being a prevalent issue. On the other hand, some of our assumptions and results either aligned partly or diverged from stakeholder feedback and that's shown in the red box. For example, they commented that mid-season water shortages in the scheme are less likely as the area cultivated in the scheme would be decided based on the amount of water available at the start of the season. So water shortages would still impact farming, irrigation and income generation, but instead of impacting directly how much water is available to users, it would impact the area of the land cultivated. Another point of disagreement was about strictness. Stakeholders expressed that worse strictness should impact cutting off water supplies entirely rather than rationing quotas and they expect that this would improve water users' cooperation with water as well as boosting their income. At the same time, stakeholders also agreed that implementing the effect of worse strictness is not as straightforward as they know from experience. We also got suggestions for new dimensions to consider in watering. For example, members influencing and being influenced by one another through our membership. So based on the feedback we collected from stakeholders, we are in the final phase of refining watering, but in the meantime, as a researcher and agent-based modeler, I'm keen to follow open research practices to promote the openness, transparency and reusability of agent-based models. I mean, let's consider all the great work done by the open-source software community which allows us to use libraries, packages, extensions and the like which make our work easy or even doable in some circumstances. In that spirit, we felt that the irrigation and crop growth sub-models of watering implemented in NetLogo could become components or building blocks useful in a range of socio-ecological systems modeling applications and have thus made them standalone components. As they stand now, these models only simulate irrigation or irrigation and crop growth, but the idea is you can use them as building blocks and add new categories of agents and new procedures to model a socio-ecological system of interest to you. You can download the NetLogo files and run them from the link shared in this slide. We will also pop these links into the chat if you'd like to try soon after now when I finish the presentation. So a quick recap of the things I wish to achieve in this session before tanks and closing. So I wanted to share the experience and reflections from developing an agent-based model called watering to explore decentralized water management through us based on real-world data for potential real-world application which was co-developed with stakeholders and developed in an iterative fashion. I also touched on the importance of open research practices and the usefulness of it in agent-based modeling. With that, I would like to thank all my future dams colleagues here in the UK and Ghana for their contributions to watering. Thanks very much. Happy to take questions.