About the Project
This fully funded PhD project is part of the QUARTILES Doctoral Landscape Award, a BBSRC and NERC-funded research and training programme designed to equip PhD students with the skills, expertise, outlook, and real-world experience needed to become the next generation of scientific leaders capable of addressing pressing environmental grand challenges such as climate change, biodiversity loss, and sustainability.
Project in a sentence: build HydroFM, a physics-aware foundation model for hydrology that learns from global Earth observation and gauge data, adapts quickly to new basins, and delivers decision-ready metrics for droughts, floods, and water allocation.
Why this matters. Water systems are changing—climate, land use, regulation—and many models struggle when yesterday’s relationships no longer hold. In parallel, Earth sciences are moving to foundation models: broadly pretrained AI systems that fine-tune with little local data. This PhD applies that paradigm to hydrology, creating a reusable model that’s transferable, physically consistent, and transparent about its uncertainty.
What you’ll build. You will design and train HydroFM using well-established components for tractability: a UNet/ResNet-UNet spatial encoder for gridded EO/reanalysis, a time-series backbone (LSTM/TCN), and a lightweight graph layer for river-network routing. The model will predict key water-balance components (runoff Q, evapotranspiration ET, storage change ΔS), use probabilistic outputs for calibrated uncertainty, and enforce soft physics (water-balance penalties, sensible bounds). Pretraining will draw on harmonised global datasets (e.g., national hydrometric archives, Caravan/HYSETS), gridded forcings (MSWEP, ERA5-Land, GLEAM), satellite EO (SMAP/SMOS, MODIS), and hydrological reanalyses (GloFAS/EFAS). You’ll fine-tune with parameter-efficient methods (adapters/LoRA) so the model can specialise to new basins with minimal compute.
How it will be tested and used. You’ll run cross-region holdouts, temporal-shift tests, and forcing-product swaps against strong deep-learning and process-model baselines, reporting skill, water-balance closure, uncertainty calibration, and out-of-distribution detection. A capstone demonstrator will translate HydroFM into decisions—e.g., low-flow reliability and environmental-flow deficits for water planning, drought early-warning for irrigators, or reservoir rule-curve stress-testing—co-designed with stakeholders. You will also implement a future-scenarios pipeline (e.g., bias-adjusted CMIP6) to quantify risks under climate change.
What you’ll learn (training & support).
- First semester (front-loaded QUARTILES training): core AI/data skills, research methods, reproducible workflows, and professional development.
- Technical depth (project-specific): deep learning (UNet/LSTM/TCN/graph), self-supervised learning, parameter-efficient fine-tuning, uncertainty quantification, explainable AI, hydrological evaluation, and data governance/licensing.
- Tooling & compute: version control, containers, experiment tracking; access to HPC/GPU resources; curated data loaders for major hydrology/EO datasets.
- Supervision & community: weekly meetings, code reviews, a hydrology-AI reading group, and collaboration with partners (agencies, utilities).
- Placement: up to 3 months with an external partner aligned to your interests (e.g., regulator, utility, consultancy), focused on translating HydroFM outputs into practice.
Who should apply. We welcome candidates from hydrology, environmental engineering/science, computer science, geoinformatics, or applied maths. You don’t need to be an expert in everything on day one—curiosity, solid programming (Python/R or similar), and motivation to learn across AI + water is what matters. We are committed to an inclusive, supportive environment and will help you develop the skills you need.
Candidates should preferably have or be developing strong programming skills (Python, R); experience with time-series and/or geospatial/remote-sensing data; working knowledge of machine learning (train/validation, metrics) and readiness to learn deep learning; data handling; sound quantitative/statistical reasoning and clear written/oral communication; ability to collaborate with non-academic partners; high motivation to work across AI and hydrology, and to follow good data governance/ethics practices.
Where this takes you. Graduates will be competitive for roles in academia, research institutes, agencies, utilities, and data-driven consultancies—equipped to build trustworthy AI for environmental decision-making and to lead the next generation of digital-twin-ready water models.
Skills and Experience: Candidates should preferably have or be developing strong programming skills (Python, R); experience with time-series and/or geospatial/remote-sensing data; working knowledge of machine learning (train/validation, metrics) and readiness to learn deep learning; data handling; sound quantitative/statistical reasoning and clear written/oral communication; ability to collaborate with non-academic partners; high motivation to work across AI and hydrology, and to follow good data governance/ethics practices.
Informal enquiries are encouraged! For further project information please contact the lead project supervisor by selecting the first listed name at the top of this advert and sending your enquiry.
———————————
ELIGIBILITY:
Promoting equality, diversity and inclusion is core to the QUARTILES Doctoral Landscape Award. We actively encourage applications from diverse career paths and backgrounds and across all sections of the community, regardless of age, disability, ethnicity, gender, gender expression, sexual orientation and transgender status, amongst other protected characteristics.
We also invite applications from those returning from a career break, industry or other roles. We typically require a minimum 2:1 in your first degree (or equivalent), but exceptions can be made where applicants can demonstrate excellence in alternative ways, including, but not limited to, performance in masters courses, professional placements, internships or employment – this will be considered on a case-by-case basis, and is dependent upon approval from the relevant host institution. We offer flexible study arrangements such as part-time study (minimum 50%), however this does depend on the nature of the project/research so will be considered on a case-by-case basis.
If you have any questions about your eligibility, please email us at quartiles-admissions@abdn.ac.uk
———————————
APPLICATION PROCEDURE:
- Please visit this page for full application information: How to Apply – QUARTILES DLA
- Please send your completed QUARTILES application form, along with academic transcripts and certificates to quartiles-admissions@abdn.ac.uk
- Please provide two academic references (we are unable to directly request references from your referees. If you would like to include references to support your application, please ensure they are provided directly to us. Some project supervisors may choose to contact your referees – please also include their contact details on your CV.
- Please ensure you submit all the required information and documentation.
- If you require any additional assistance in submitting your application or have any queries about the application process, please don’t hesitate to contact us at quartiles-admissions@abdn.ac.uk
Funding Notes
This 45 Month (NERC) opportunity is open to UK and International students (The proportion of international students appointed to the QUARTILES DLA is capped at 30% by UKRI).
QUARTILES studentships include a tax-free UKRI doctoral stipend (£19,795 for the 2025/26 academic year, the 2026/27 rate has yet to be released), plus a training grant of £9,000 to support data collection activities throughout the PhD.
QUARTILES does not provide funding to cover visa and associated healthcare surcharges for international students.
References
Shen, C. (2018). A transdisciplinary review of deep learning research and its relevance for water resources scientists. Water Resources Research, 54, 8558–8593. https://doi.org/10.1029/2018WR022643
Kratzert, F., Klotz, D., Shalev, G., Klambauer, G., Hochreiter, S., & Nearing, G. (2019). Towards learning universal, regional, and local hydrological behaviors via machine learning applied to large-sample datasets. Hydrology and Earth System Sciences, 23(12), 5089-5110. https://doi.org/10.5194/hess-23-5089-2019
Tripathy, K. P., & Mishra, A. K. (2024). Deep learning in hydrology and water resources disciplines: concepts, methods, applications, and research directions. Journal of Hydrology, 628, 130458. https://doi.org/10.1016/j.jhydrol.2023.130458
Frame, J. M., Kratzert, F., Klotz, D., Gauch, M., Shalev, G., Gilon, O., … & Nearing, G. S. (2022). Deep learning rainfall–runoff predictions of extreme events. Hydrology and Earth System Sciences, 26(13), 3377-3392. https://doi.org/10.5194/hess-26-3377-2022
Szwarcman, D., Roy, S., Fraccaro, P., Gíslason, Þ. E., Blumenstiel, B., Ghosal, R., … & Moreno, J. B. (2024). Prithvi-eo-2.0: A versatile multi-temporal foundation model for earth observation applications. arXiv preprint arXiv:2412.02732. https://doi.org/10.48550/arXiv.2412.02732
