FFFai Fung

Fai Fung

*This is an algorithmically generated author record based on publicly available information.

Fai Fung has a total of 62 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Climate Service Recipes: automatic multi-hazard climate information workflow generation using agentic Large Language Models (LLMs) and knowledge graphs

    by Anrijs Abele, Hailun Xie, Arjun Biswas, Hang Dong, Fai Fung, Hywel Williams

    Climate Service Recipes (HACID-CSR) is an agentic system designed to assist providers of climate services in developing their advice for a wide range of clients. HACID-CSR guides providers by navigating the large and ever-increasing corpus of knowledge as well as an area without established standards and with limited access to scientific experts. It automatically generates detailed workflows (or “recipes”) by leveraging both a large language model’s internal reasoning and contextual knowledge from a domain knowledge graph for climate services (CS-DKG). The CS-DKG is an expert-curated ontology of climate service concepts with mapped relationships between climate variables, emission scenarios, indices, hazards, sectors, and key datasets (CORDEX, CMIP5, UKCP18), built as part of the Horizon Europe-funded HACID project (Hybrid Human Artificial Collective Intelligence in Open-Ended Decision Making).The HACID-CSR architecture consists of a memory-enabled supervisor agent orchestrating multiple specialised agents. A planning agent first proposes an initial workflow outline, and a preliminary recipe agent uses only the LLM’s knowledge to draft answers to key workflow steps. The system then engages a knowledge graph retrieval sequence: a class selection agent identifies relevant classes in the CS-DKG, an instance selection agent finds specific instances (entries) highly relevant to the query within those classes following a two-stage selection process, i.e. semantic similarity based pre-selection and LLM-enabled refined selection, and a subgraph extraction agent retrieves the corresponding subgraph of related knowledge entities. Next, a recipe generation agent creates each step of the workflow by combining the LLM’s reasoning with the retrieved graph context using graph retrieval-augmented generation (GraphRAG). Finally, a recipe refinement agent compares the preliminary LLM-only solution with the knowledge-enhanced solution and refines the output, yielding a diverse and context-aware workflow.By using this multi-agent approach, HACID-CSR increases the diversity of solutions and fills the knowledge gap between climate information and domain specific applications, helping experts to identify suitable methodologies and datasets. The resulting workflows are more traceable and transparent, improving user trust compared to answers from a general-purpose chatbot. We have also developed a bespoke automatic evaluation method to complement human expert validation of the generated recipes. We highlight the potential of the HACID-CSR approach for multi-hazard climate service design, and discuss remaining challenges and opportunities for further refinement of this agentic LLM-based system.

  • Open Access

    Exploring climate service purveyor engagement and use of national climate projections

    by Marta Bruno Soares, Peter Walton, Jillian Schacher, Neha Mittal, Fai Fung

    The increased focus on climate change impacts over recent years and the recognised need for adaptation across society has led to a rise in the use of climate information such as the Climate Projections (UKCP18) in the UK. However, while public sector engagement with the Projections has been explored, there is very limited knowledge on how the private sector engages with and uses climate information to provide services to their end-users. The REVEAL project aimed to understand how these climate service purveyors use UKCP18 and other climate data, their relationships with both producers and users, and emerging climate information needs. For the purposes of this study, purveyors were defined as “Organisations beyond government bodies that use, enhance or add value to UKCP18 climate information to address specific users' needs.” Building upon initial Met Office research conducted in 2023 the project initially carried out a landscape assessment of purveyors in the UK who would likely use climate data and information. A total of 31 semi-structured interviews were then conducted with purveyor organisations to explore in more in-depth themes around: organisational and client profile; current and future use of UKCP18; barriers to using UKCP18; suggested developments for the next UK Projections; and wider recommendations for climate services in the UK. This study highlighted key findings around the expanding and heterogeneous landscape of purveyors in the UK, their perceptions and current use of UKCP18, the type of activities pursued and their clients, requirements for improving the next phase of projections as well as recommendations for linking up with other derived products such as the recently launched Local Authority Climate Service and Climate Risk Indicators. The study also offers insights into the expanding role of these intermediaries in the climate services value chain over the last few years as well as wider considerations on the governance of climate information and services in the context of climate adaptation efforts in the UK and Europe.


  • Open Access

    Risk of extreme climate impacts on European Norway spruce forest: drought and frost in the climate emergency

    by Duncan Ray, Fai Fung, Jennifer Pirret, Fiona Plenderleith, Felix Trotter, Maurizio Marchi

    Shifts in weather patterns causing more severe and frequent climatic extreme events, particularly periods of hot drought and late-spring frost, cause stress and damage to many tree species. Here we use 12 members of the UKCP Regional (12km) ensemble to assess the changing risk of tree stress posed by hot droughts and late-spring frosts. As an exemplar tree species, we use Norway spruce (Picea abies) to demonstrate a method that links the European distribution of the species to extremes in the projected climate data ensemble at the continental scale. Monthly climate projections were used to calculate the standardised precipitation and evapotranspiration index (SPEI) and bias corrected daily temperature projections were used to calculate a frost buffer of the number of days between last date of spring frost and the date of leafing of Norway spruce. The UKCP Regional climate model ensemble shows substantial changes in the distribution of SPEI likelihood and severity between the pseudo-global warming threshold temperatures of 10C and 20C indicating concern in relation to drought severity and frequency in Alpine, Atlantic, and Mediterranean biogeographic regions. In addition, late spring frosts particularly in the Atlantic and Continental biogeographic regions will continue to affect sensitive young trees such as Norway spruce. The approach removes the time series estimation of climate extremes providing a single method of estimating effects of the climate emergency across shared socio-economic pathways. Finally, the approach could be useful to direct the deployment of suitable selected forest reproductive material from tree-breeding programmes in Europe.


  • Open Access

    Impacting on our Lives: Using British sports and culture to explain uncertainty in climate projections

    by Lottie Woods, James Pope, Fai Fung

    Whether its memories of a cold, frosty Christmas or an August bank holiday beach trip interrupted by rain, many cultural, sporting, and social events in the United Kingdom have strong associations with particular weather conditions. As the average global temperature increases, the impacts of a changing climate are likely to be felt across many aspects of British life, including in the public’s experiences of these popular events. Several recent works conducted by the UK Met Office have sought to make the local day-to-day impacts of climate change more understandable for the public by exploring likely climatic conditions of popular events by the 2050s. These works have received strong engagement from the public, demonstrating the demand for relevant and understandable climate information.We serve this demand by using the 2018 UK Climate Projections (UKCP18) and HadUK-Grid observations data to evaluate how climate change will affect the climatology of a diverse range of British social, cultural, and sporting events. To explore and communicate the uncertainties in UKCP18 due to inherent model biases, several bias correction methods are applied to the data and the resulting data is analysed together to give an improved uncertainty range. The research will focus on assessing changes to temperature variables at global warming levels of 1.5°C and 3.0°C to illustrate these two future scenarios and the uncertainty within each scenario.We will show that some events are likely to have a significantly altered climatology which is likely to substantially change the nature of these events or force them to change when they occur during the year to give the best chance of having favourable climatic conditions. By assessing the impact of climate change on popular British events such as the London Marathon and Glastonbury Festival the findings of this research will prove useful in communicating the impacts of climate change in a way which will resonate with the British public.


  • Open Access

    Vulnerability of estuary water quality to climate change

    by Peter Robins, Charlotte Lyddon, Gemma Coxon, Thom Clough, Aaron Furnish, Mirko Barada, Laura Devitt, Tom Coulthard, Davey Jones, Andy Barkwith, et al.

    Estuaries are crucial for freshwater and nutrient cycling throughout shelf seas that drives the biodiversity and ecology of coastal and marine wildlife, and provide ecosystem services that sustain the livelihoods and wellbeing of coastal communities. These ecosystems are, however, potential pollution corridors and sinks carrying sewage and other loads containing harmful pathogens and contaminants – a serious health issue that is worsening with littoralisation and population growth. Being at the interface between oceanographic and fluvial processes, estuaries are the most dynamic coastal system, where water quality processes and habitat dynamics are shaped by complex geo-physical, chemical, and biological interactions that change over small spatio-temporal scales and are unique to each estuary. It is essential that these systems maintain safe water quality standards and that we are prepared for future changes in water quality that will affect their ecological status and public health risk. This research aims to characterise variability and potential change in indicators of estuary health across the UK, using a robust analysis and modelling strategy, that can be built upon to evaluate a range of water quality degradation processes and used to inform future management strategies. We will present the first analysis of both riverine and marine climate projections for the 21st Century (UKCP18 RCP8.5 perturbed parameter ensemble), downscaled to hourly- and sub-meso-scales, and applied to all estuaries in England. In particular, characterising projected changes in hydrology, temperature, salinity, sea level, and coincident conditions. Additionally, we have developed fine-scale estuary hydrodynamic models (Delft3D) of all estuaries and present potential changes in simulated estuary residence times as a result of projected sea-level rise and changing hydrology. The analyses and simulations highlight estuaries and estuary types that are vulnerable to changes in the physical stressors of coastal water quality – where coastal management efforts and hazard response should be focused the coming decades.


  • Open Access

    Blueprinting for collaborative climate service design

    by Lottie Woods, James Pope, Fai Fung, Neha Mittal

    Since the launch of the UK Climate Projections (UKCP) in 2018, the UK Met Office has continually updated the underlying climate model datasets and operated a supported climate service on behalf of the UK government. In addition to providing technical support for its datasets, the UKCP Climate Service monitors the evolving user need, supports activities to enable knowledge sharing, improve understanding and co-design additional services. This work has resulted in the release of new datasets, reports and user guidance as well as a community-contributed monthly webinar series led by the co-chairs of its user group, the Development and Knowledge Sharing network. In 2023, a user consultation process was conducted and has revealed data needs that are not satisfied by the current UKCP18 corpus of information. As the UK Met Office assesses how to meet these data needs through new scientific capabilities and services, understanding their impact on climate action and their priority is of great importance.New climate service design often focuses on the delivery system, leaving much of the user requirements defined in terms of climate variables, uncertainty, scenarios, data formats and visualisations. This can lead to misinterpretation by producers of the need and users of the most appropriate scientific tools available. We present the application of service blueprinting, which allows the explicit co-design between producers and users of a climate service which is outcome focused, i.e. it outlines the steps on how new information provision can potentially affect user actions. We will show draft blueprints co-designed with industry partners through interviews and focus group sessions, addressing user needs for information on extreme events, ocean chemistry, uncertainty and hydrological variables. The blueprints provide a holistic view of the whole climate service production process, starting at the potential outcome and working through the steps to information delivery and formats, data processing and interpretation as well as climate model simulation requirements. The blueprints offer a shared understanding for all stakeholders of the climate service.

  • Open Access

    Investigating the resilience of vulnerable children to future groundwater scarcity

    by Fai Fung, Mary Zhang, Mostaquimur Rahman

    Climate change and water scarcity are intertwined natural and humanitarian crises, disproportionately affecting vulnerable populations. By 2040, one in four children globally will face severe water scarcity, exposing them to water-borne diseases. Many, particularly girls and young women in resource-limited regions, are forced to abandon education and career opportunities to fetch water due to traditional gender roles. As climate change reduces surface water availability, groundwater demand is projected to rise significantly over the next 30 years.We will present our research framework to investigate how slow-onset climate events, such as prolonged droughts, impact child poverty - characterised as malnutrition and educational deprivation - through groundwater availability by integrating theories, data, and methods in child development, water security, climate change, and sustainable development . The aim is to develop solutions to improve resilience, adaptation strategies and public services for vulnerable children by answering three important questions:Where are groundwater-dependent regions globally, continentally, and regionally, and what climate factors drive uncertainty in groundwater availability? How does groundwater scarcity affect child poverty in semi-arid and arid developing regions? Why do regions with comparable climate hazards and groundwater availability exhibit different levels of child poverty, and what lessons can be shared?

  • Open Access

    Investigation of Future Climate Change Over the British Isles using Weather Patterns

    by James Pope, Kate Brown, Fai Fung, Helen Hanlon, Robert Neal, Erika Palin, Anne Reid

    For those involved in planning for regional and local scale changes in future climate, there is a requirement for climate information to be available in a context more usually associated with meteorological timescales. Here we combine a tool used in numerical weather prediction, the 30 weather patterns produced by the Met Office, which are already applied operationally to numerical weather prediction models, to assess changes in the UK Climate Projections (UKCP) Global ensemble. Through assessing projected changes in the frequency of the weather patterns at the end of the 21st Century, we determine that future changes in large-scale circulation tend towards an increase in winter of weather patterns associated with cyclonic and westerly wind conditions at the expense of more anticyclonic, settled/blocked weather patterns. In summer, the results indicate a shift towards an increase in dry settled weather types with a corresponding reduction in the wet and windy weather types. Climatologically this suggests a shift towards warmer, wetter winters and warmer, drier summers; which is consistent with the headline findings from the UK Climate Projections 2018. This paper represents the first evaluation of weather patterns analysis within UKCP Global. It provides a detailed assessment of the changes in these weather patterns through the 21st Century and how uncertainty in emissions, structural and perturbed parameters affects these results. We show that the use of these weather patterns in tandem with the UKCP projections is useful for future work investigating changes in a range of weather-related climate features such as extreme precipitation, or impacts on the energy sector.

  • Open Access

    Does AI mean we need to do climate services differently?

    by Fai Fung, Neha Mittal

    The knowledge base to support climate change adaptation is rapidly growing with the latest Intergovernmental Panel on Climate Change’s 2022 report on impacts, adaptation and vulnerability with over 34,000 citations; nearly three times compared to its predecessor published in 2014. This information is complex, requiring specialist knowledge to interpret the science and further expertise to understand how to act upon the information. The small pool of experts across the world presents a challenge with the need of an ever-growing set of decision-makers who need tailored information.To date, most attention has focused on the use of machine learning for climate science rather than the application of AI technologies to enhance climate services provision. With the advent of public access to large language models and more generally, generative AI technologies, we consider here the opportunities that they bring to solve the problems around delivering climate services but also the implications on the pillars of saliency, credibility and legitimacy.We present the findings from the “Towards AI for Climate Services” workshop held on 22 February 2024 in the UK which aimed to understand if and how AI technologies could help to address existing challenges in the sector. Participants represented academic, public and private providers of climate and AI services. Discussions focused on three aspects: challenges in climate services that AI technologies are addressing, emerging risks with employing AI technologies and pathways to mitigating these risks.We will show that while many of the issues raised are not exclusive to using AI technologies, their rapid development and accessibility suggest that the climate services sector need to urgently engage with the debate