Smart Hydrology: Data, AI and the Future of Water Management

Hwirin Kim argues that AI is the most transformative technology for hydrology not because it replaces experts, but because it enhances their ability to make better decisions. She cautions against the misconception that AI can do everything without human expertise, stressing that hydrologists and domain experts remain essential. AI serves as a powerful support tool that augments human capacity rather than substituting it.

Kim explicitly stated that AI enhances availability to make better decisions and does not replace the expert or hydrologist . She warned that some people misunderstand AI as being able to do everything without people, which she considers totally untrue . Korea's national hydrological services use AI flood forecasting as complementary to traditional hydrological modelling, with AI first detecting where flooding is more likely and then human forecasters applying traditional methods for more detailed analysis .

Kim argues that the most robust forecasting outcomes come from combining traditional physics-based hydrological models with AI, rather than relying on either approach alone. Physics provides scientific consistency and grounding, while AI contributes speed, accuracy, and the ability to learn from large volumes of data. Together, they produce results superior to either method in isolation.

Kim stated her belief that the future lies in hybrid intelligence, combining physical-based hydrological models with AI, where physics provides scientific consistency . She noted that while AI enhances speed, accuracy, and the ability to learn from large data volumes, together they produce more robust forecasting than either approach alone . Korea's example was cited, where AI and traditional hydrological modelling are used together rather than one replacing the other .

Kim emphasises that AI models are only as effective as the data on which they are trained, making data quality and open sharing foundational requirements. Without high-quality, shared data, AI cannot produce accurate or reliable forecasting results. She also warns that multiple unverified information sources can cause confusion rather than clarity for decision-makers and the public.

Kim stated that AI with new technology without data cannot produce good or accurate results, and that everybody is asking to share data, which WMO has strong power in collecting and standardising . She warned that AI models are only as good as the data on which they are trained, and that too many unverified information sources at the same time causes confusion rather than trust . She also highlighted that success requires open data sharing, quality observation, common standards, and capacity development .

Kim argues that WMO fully supports open-source and interoperable systems so that member countries are not left dependent on external partners who hold proprietary source code. Past experiences where project partners retained source code after project completion left members unable to maintain or update tools without paying those partners again. Open-source, member-driven systems prevent this lock-in and ensure long-term sustainability.

Kim recounted past experiences where, once a project finished, the partner only shared the source code with themselves, making it difficult for members to maintain, update, or upgrade tools without paying the partner again . She stated that WMO now fully supports open-source, interoperable, member-driven systems to avoid repeating this situation . She also referenced CERN's open-source approach as something WMO truly supports .

Kim highlights a growing technological divide where some countries are advancing rapidly with AI and digital tools while least developed countries and small island states, which suffer the greatest casualties from water-related events, risk being excluded. She stresses that AI and digital water technologies must be equally shared and accessible to all nations, not just wealthy ones. WMO's mandate includes supporting the most vulnerable countries.

Kim noted that some countries develop very quickly while the rest of the world does not, and that least developing countries will be left behind if not supported . She emphasised that WMO is there to support not only rich developed countries but also least developed countries and small islands who face really big casualties every single water risk event . She called for AI to be equally shared as open source and open system .

Kim argues that effective flood response requires two-way communication, where affected communities share real-time information such as photos and ground-level observations, not just receiving warnings from agencies. This community-sourced data provides ground-truth information that AI systems cannot independently detect. Such real-time inputs are essential for validating warnings and identifying where people are in danger.

Kim described how communities share pictures and information in real time during floods, including whether warnings were accurate and where casualties are occurring . She noted that AI cannot detect everything and that community sharing is very important . She referenced discussions from a morning session about two-way communication, where communities share information because they are physically present at the event .

Kim argues that issuing a warning with numerical water level data is insufficient for effective decision-making; what is needed is dynamic, visualised information showing which specific infrastructure will be inundated and when. Digital twins and dynamic maps make it easy for decision-makers to determine evacuation priorities. This shift from abstract numbers to visual, actionable intelligence saves lives.

Kim explained that traditional warnings give a station name and water level number, but dynamic maps showing which schools and roads will be inundated and when are much easier for decision-makers to act upon . She noted that Korea is using digital twins to forecast and predict flood inundation areas, providing visibility that makes evacuation decisions easier . She also highlighted that the challenge is no longer collecting data but extracting actionable intelligence from it .

Prof. Ehrlich describes VillageOS as an open-source orchestration platform that integrates earth observation, hydrology, ecology, and agriculture to inform human habitation and wildlife planning. The platform starts always with water, using permaculture design principles and indigenous wisdom to understand how water flows and behaves across landscapes. It aims to create self-sustainable communities by understanding what nature is capable of before designing interventions.

Ehrlich described VillageOS as an orchestration layer connecting earth observation, hydrology, ecology, and agriculture, most importantly informing human habitation and wildlife . He explained that the platform uses democratised, open-source, open-science approaches, integrating satellite imagery, LIDAR, radar, and observations to classify watersheds, drainage networks, vegetation, and soil sediment profiles . He noted that the platform has a generative AI design side starting with nature and water, and an edge AI implementation layer focused on sensing .

Ehrlich argues that VillageOS is designed biomimetically around mycelial networks — the fungal organisms under soil that act as elegant neural networks brokering minerals, carbon, and water. This approach informs a low-energy, decentralised, data-sovereign model for communities. The goal is to move away from siloed thinking across energy, water, food, waste, housing, and mobility, and instead apply whole-system design thinking as nature does.

Ehrlich explained that the VillageOS software is designed biomimetically around mycelial networks, which are large fungal organisms under the soil acting as neural networks that broker minerals, carbon, sugar, and water allotments . He described the goal as creating low-energy, eventually edge AI, low-power, decentralised, data-sovereign communities focused on AI as human planetary health . He stated the aim is to get away from silos of energy, water, food, waste, housing, and mobility, and look at whole-system design thinking as nature does .

Fiskå describes River Deep Mountain AI, an open-sourced initiative funded by public money that combines AI, geospatial data, and satellite imagery to gauge flow on rivers, brooks, and large water systems at scale. The tool has been tested on 90 catchments, peer-reviewed, and is now being considered for implementation in Denmark and the Netherlands. It demonstrates that publicly funded, open-source AI tools can produce scientifically validated, scalable water monitoring solutions.

Fiskå stated that River Deep Mountain AI has been implemented and tested on 90 catchments, is able to gauge flow on rivers, brooks, and large water systems at scale, and has been peer-reviewed . He noted that Denmark and the Netherlands are now looking at implementing it . He described it as an open-sourced, publicly funded initiative combining AI, geospatial data, and satellite imagery .

Fiskå argues that open-source systems built on freely available data sources such as Sentinel-2 satellite imagery and public governmental data can effectively address complex water challenges including pollution and flooding. The system developed for the UK used publicly available weather data, governmental sensor data, and satellite multispectral analysis to understand pollution sources and predict flooding. This approach is cost-effective and scalable to other regions.

Fiskå described spending 18 months building an open-source system using Sentinel-2, public governmental data, and weather data to address the UK's water pollution and flooding problems . He noted that Sentinel is free and the system is not costly to run, making it feasible to extend to other parts of the world . He also highlighted that the system is grounded with government sensor data on soil, water gauges, oxygen, and phosphorus .

Fiskå presents a case study of Shrewsbury, England, where flooding occurs during blue-sky, high-temperature days with no rain, puzzling local communities. By analysing soil carbon content, land use changes, and climate change interactions across the whole landscape, the system can now predict such flooding events 21 days in advance. He also proposes that AI can identify where small, systematic nature-based interventions — such as open beds and riverside strips — can slow water release and reduce pollution.

Fiskå described the Shrewsbury case where flooding of two to three metres occurs during blue-sky days because rising temperatures release water stored in soil carbon across miles of land, flooding the river . He stated that the system is now able to tell society 21 days in advance that this is going to happen . He proposed using AI to identify where small systematic interventions such as open beds and riverside strips could slow water release en masse across large areas, with the added benefit of nature absorbing pollution .

Fiskå argues that combining AI with geospatial and satellite data within an open-source framework produces scalable, scientifically validated tools for water flow gauging. The River Deep Mountain AI system demonstrates this by being tested across 90 catchments and peer-reviewed, showing it genuinely works at scale. The open-source nature ensures the tool can be extended to new regions without prohibitive costs.

Fiskå described River Deep Mountain AI as combining AI, geospatial data, and satellite imagery, tested on 90 catchments, peer-reviewed, and confirmed to be really working . He noted it is being implemented in UK water companies and that Denmark and the Netherlands are looking at adopting it . He emphasised that Sentinel satellite data is free and the system is not costly to run, making it scalable globally .

Mathieu argues that AI dramatically lowers the barrier for non-specialists to access and use complex satellite datasets, enabling them to interrogate data and build reasoning scenarios without needing to know programming languages. By using AI as an intermediary that translates natural language into code and data queries, individuals who are not data scientists can plot satellite data, construct flood scenarios, and derive actionable insights. This represents a breakthrough in democratising access to earth observation data.

Mathieu described how, by talking to AI which then talks to Python which talks to Google Earth Engine, he can plot anything from a news headline without knowing Python himself . He demonstrated a level-zero application where AI constructs a flooding scenario per building from elevation model, LIDAR, and rain data in less than one minute and produces statistics . He cited the Valencia case where people did not act on early warnings because they could not visualise the risk, arguing that per-house flood scenario apps could change behaviour .

Mathieu explains that ESA's Copernicus programme, through its fleet of Sentinel satellites, provides routine, carpet-like monitoring of the planet with an open data policy, creating long time series of environmental data essential for hydrological modelling and disaster response. This data archive underpins scientific research and operational services globally. ESA's role is to create this database, document parameters, and steer the scientific community to extract value from it.

Mathieu described the Copernicus programme as a flagship European programme investing in a fleet of Sentinel satellites doing the equivalent of meteorology for the environment with an open data policy . He noted that the programme provides routine, almost carpet monitoring of the planet, creating long time series that generate significant scientific output . He stated that ESA's core business is to create this database, document parameters, and steer the scientific community to squeeze the juice of this data .

Mathieu highlights that the standard satellite data ordering process — involving tasking, competition with other users, and ground segment processing — can take 24 to 48 hours, which is far too slow for disaster response. By integrating satellite data providers directly into operational processing chains, delivery times can be reduced to three hours, representing a breakthrough for fast responders. This was demonstrated during the Storm Boris flooding in Poland.

Mathieu explained that the standard satellite data ordering process can take 24 to 48 hours due to tasking, competition, and ground segment processing . He described how, during Storm Boris in Poland, working closely with the disaster management organisation CKB and integrating satellite data directly into their processing chain reduced delivery time to three hours, which was a breakthrough . He noted that fast responders were then able to use maps directly in the field .

Mathieu points out that many disaster management organisations around the world are not using satellite data and some have never even heard of space technology, despite its enormous potential for disaster resilience. This represents a significant opportunity gap that requires targeted capacity building and education. ESA's approach is to work directly with these organisations, providing funding so that data providers can present their products in the organisations' own standards and jargon.

Mathieu stated that most disaster management organisations are not using ESA's data, and surprisingly some of them have never heard about space, representing a huge opportunity gap . He described ESA's approach of providing money so that data providers can mimic their products with the organisations' standards and jargon, enabling them to trade almost for free . He noted there is a capacity building and educational aspect to this work .

Tim Smith argues that quantum computing, while not a panacea, offers specific capabilities for highly complex, non-linear problems in water management that classical computing struggles with. Examples include water leak detection in distribution networks, where mapping pressure and flow changes across a massive graph is highly non-linear, and full water resource management modelling that integrates many system layers. Quantum techniques can also assist in molecular docking modelling to find better water treatment chemicals.

Smith described a use case of water leak detection in Mexico's water distribution networks, where sensing pressure and flow changes to identify leaks is a highly non-linear problem in a massive graph, and quantum techniques are being applied to assist identification and predict better sensor placement . He also mentioned the American University in Beirut exploring full water resource management modelling using quantum solvers to handle the complexity of many different models and data points . He noted quantum computing could assist in molecular docking modelling to find replacements for chlorination and extract pollutants more effectively .

Smith argues that one of the keys to data sharing is placing data in a single, shared, open location where people can opportunistically find it and experiment with combining datasets that were never designed to work together. CERN's Zenodo platform is presented as a model for this approach, enabling cross-fertilisation across scientific disciplines. This open access allows creative interoperability and discovery of new insights from combined datasets.

Smith described Zenodo as a service used by every science around the world, which is one of the keys to data sharing by putting things in the same place where people can opportunistically find them and get cross-fertilisation . He noted that people can openly access and try playing and interoperating with different datasets that were never designed to be together, enabling creative discovery . He also referenced CERN's 30-year history of building open data services and advocating for open science .

Smith argues that running quantum hackathons and challenge-based investigations in countries around the world helps local students engage with the next technology wave before it arrives, enabling local universities to build relevant curricula. This approach also connects local industry with trained students, helping retain talent locally rather than having it absorbed by remote actors. The goal is to make local communities participants in shaping how the next technology wave addresses their specific challenges.

Smith described running hackathons and quantum hackathons in countries around the world to get local students interested in the next technology wave before it arrives, so that local universities building curricula can see challenges relevant to them . He noted the aim is for local industry to be interested in taking those students locally rather than remotely . He also described challenge-based investigations requiring use cases to be in a country with a national institute and a local actor that wants to implement what they find .

Dr. Zhao expresses his interest as an ecology student in the convergence of AI and environmental protection, particularly for water and climate solutions. He has come to the session specifically to learn about cutting-edge technologies at this intersection. His presence represents the growing interest among the next generation of researchers in applying AI to ecological and environmental challenges.

Zhao introduced himself as a student of Zhejiang University studying ecology, expressing strong interest in the combination of environmental protection and AI . He stated he came to the session specifically to learn about cutting-edge technology at this intersection .

Nayab Sayed argues that upskilling individuals on technology and AI is a prerequisite for meaningful engagement with digital tools, including those used for water management. His organisation HIP Digital has achieved significant scale in this work, demonstrating that large-scale digital upskilling with strong employability outcomes is achievable. This foundation of digital literacy is essential for communities to benefit from AI-driven water solutions.

Sayed described HIP Digital's achievement of 115,000 candidate training days in 71 countries over five years, with 85% employability success . He noted that a big chunk of the programme is tied into employability as well as upskilling on technology and AI .

Sayed highlights that his NGO IEEE International has established over 3,000 water points across 22 countries, including hand pump rehabilitation, solar solutions, and municipal water distribution networks. This practical work reveals the enormous infrastructure gap in the Global South that digital technologies must ultimately address. The scale and diversity of this work underscores the urgency of making digital water tools accessible and relevant to underserved communities.

Sayed described IEEE International's work across seven regions, primarily in the African subcontinent, with a core focus on provision of potable water, small-scale rural electrification, and building infrastructure for villages and communities . He noted that the organisation is now up to just over 3,000 water points across 22 countries, including rehabilitation of hand pumps, solar solutions, and municipal water distribution networks .

Prof. Abbasi argues that AI, combined with satellite data and smart irrigation, can support resilient agriculture and food systems in the Global South, particularly when implemented with a focus on women's empowerment at the grassroots level. She is running seven projects with final-year students at universities in the Global South, demonstrating a practical model for embedding AI for SDGs into higher education. This approach links AI capability building with gender equity and food security outcomes.

Abbasi described herself as a professor for Ethical AI and AI for SDGs, particularly interested in working with universities in the Global South on resilient agriculture and food systems . She stated she currently has seven projects with final-year students looking at smart irrigation and how satellite data can be used practically to build more resilient systems, focusing on women's empowerment at the grassroots .

Speaker 5 describes the lived reality of communities in Kenya facing extreme and rapidly shifting weather patterns, including a recent dry season that has reduced food availability and an imminent flood season predicted for the coming weeks. This highlights the urgent, practical need for accessible predictive tools that can help communities prepare for and respond to such events. Her dual role as a government worker and founder of a grassroots NGO supporting rural women underscores the human stakes involved.

Speaker 5 described Kenya as experiencing extreme weather seasons and patterns, including a recent dry season that has left less food for the following year . She noted that predictions show imminent floods in the coming weeks, creating a situation of back-to-back climate extremes . She described her interest in the session as being about how to help her communities navigate these challenges .

Nakul Prasad argues that water is a cross-cutting resource that underpins energy, food, health, transportation, and other domains, making interdisciplinary domain expertise essential for effective system-wide water management. He emphasises that getting away from siloed thinking is key to addressing water challenges comprehensively. This aligns with WMO's move towards Earth System Modelling, which integrates the atmosphere, hydrosphere, and biosphere.

Prasad stated that water is everywhere, undercutting all domains including energy, food, health, and transportation, and that interdisciplinary domain expertise is key when looking at system-wide modelling or system design thinking . He noted that WMO is now looking at Earth System Modelling, taking in all different elements - the atmosphere, hydrosphere, and biosphere - and how they interact with each other . He affirmed that this aligns with VillageOS's approach of getting away from silos .

Feoma Ozochukwu, representing the Nigerian Communications Commission as the telecom regulator for Nigeria, expresses interest in understanding how digital and AI technologies can be applied to water and climate challenges. Her presence as a participant from a regulatory background highlights the importance of regulatory frameworks in enabling or constraining the deployment of digital water solutions. This signals potential regulatory engagement with the intersection of ICTs and water management in Nigeria.

Ozochukwu introduced herself as representing the Nigerian Communications Commission, the telecom regulator for Nigeria, and stated she is interested in hearing about the work being discussed .

Carrie Chow raises the importance of digital inclusion and addressing AI bias and data equity as essential considerations when deploying AI-driven solutions, including those for water management. She brings this perspective from her work with the Kithwell AI Alliance, which focuses on digital inclusion in AI. Her interest in the water session is framed through this lens of ensuring AI solutions are fair, inclusive, and free from bias.

Chow stated she had just had a session about digital inclusion in terms of AI, covering bias, data, and related issues, and expressed exceptional interest in learning from the water session through this lens . She described representing the Kithwell AI Alliance, an initiative focused on digital inclusion in AI .

Anna, an intern at WESIS, participates in the session as part of WMO's focus on youth engagement and empowerment. Her presence reflects the moderator's explicit encouragement of youth participation in discussions about water, climate, and digital technologies. This signals the importance of including early-career professionals and young people in shaping responses to water-related challenges.

Anna introduced herself as an intern working for WESIS, participating in the session . The moderator explicitly encouraged her participation, noting that youth engagement is a focus area and that young people should not be shy about contributing .

Speaker 1 briefly identifies themselves as a colleague of Nayab Sayed, connecting them to the work of HIP Digital and IEEE International in digital upskilling and water infrastructure provision. This collegial relationship suggests shared engagement with the intersection of technology capacity building and practical water access challenges. Their presence at the session reflects the value of bringing together practitioners from organisations working on both digital skills and water access.

Speaker 1 identified themselves as a colleague of Nayab Sayed , who leads HIP Digital's upskilling work across 71 countries and IEEE International's water infrastructure work across 22 countries .

Alina, the technical moderator working in IT, plays a supporting role in ensuring the session runs smoothly and that all participants, including remote attendees, can engage effectively. Her role as technical moderator reflects the importance of reliable ICT infrastructure and technical facilitation in enabling inclusive international discussions. Without such technical support, the multistakeholder dialogue on water and digital technologies could not function effectively.

Alina introduced herself as the technical moderator working for IT, responsible for the technical facilitation of the session .