The livelihoods of several billion people in developing countries are heavily dependent on agriculture, making them highly vulnerable to water shortages and flood events. In addition, in many of these countries hydropower is a major contribution to energy production. Accurate forecasts for extreme rainfall and droughts are therefore required to mitigate against the risks associated with hydrological extreme events, with climate change further increasing the probabilities for extreme events and thus the need for their prediction.
However, precipitation forecasts can still have substantial errors, in particular in the tropics and for extreme precipitation. These forecasts are primarily based on global weather forecasting models. Often these models are used to drive high-resolution regional weather forecasting models to improve the forecast quality and spatial resolution, but this approach is computationally costly and cannot be applied to all members of forecast ensembles from global models. An efficient and promising alternative for improving precipitation forecasts from global models is statistical postprocessing. The potential for this approach over India has been demonstrated for instance within the HEPPI project (‘HEavy Precipitation forecast Postprocessing over India’ as part of the ‘Weather and Climate Science for Services Partnership (WCSSP) – India’ program (Angus et al. 2021, Angus et al. 2022). It was found that a statistical postprocessing method for ensemble forecasts (‘Ensemble Model Output Statistics’, EMOS, Schefzik et al. 2017) substantially improves over most parts of India the forecasts for daily precipitation including for extremes.
The aim of this PhD project is to implement and evaluate the EMOS postprocessing method for other developing countries and to investigate its benefits for forecasting of hydrological extremes including surface floods and soil moisture droughts. It is expected to lead to improved operational forecasts through the following objectives:
- assess the value added by EMOS-postprocessing for precipitation forecasts over India, South Africa and Brazil with a focus on extreme precipitation events;
- determine the spatial scale on which raw and postprocessed forecasts are most informative to guide the optimal communication of forecasts to end users;
- compare the skill of hydrological simulations using raw and postprocessed precipitation as input.
Applicants should have a background in a related field such as climatology, meteorology, hydrology, geosciences, engineering, physics or mathematics. Interest in statistical analysis and programming are essential. Some working experience with UNIX and weather/climate or hydrological models would be beneficial.
Methodology:
Weather forecasts (e.g. from ECMWF, UKMO, NCEP, NCMRWF) will be obtained from the TIGGE archive (https://confluence.ecmwf.int/display/TIGGE) and Indian partners. Gridded daily rainfall observations for fitting the statistical postprocessing and model evaluation will be obtained from publicly available sources (e.g. APHRODITE) and regional partners. An EMOS implementation is available from the HEPPI project.
River flow will be simulated on the BlueBEAR supercomputer at UoB with the spatially distributed VIC model (Liang et al. 1994; Hamman et al. 2018), using raw and postprocessed precipitation forecasts as input, for three catchments: a sub-catchment of the Paraguay River in Brazil, the Ganges in India, and the Orange River in South Africa. The different catchment characteristics allow for a comprehensive assessment of the added value of precipitation forecast postprocessing from an end user perspective.
The evaluation will focus on heavy precipitation, extremely high and low river flows, and soil moisture droughts using verification measures such as Brier scores, Reliability Diagrams and Receiver Operating Characteristics curves.
Training and skills:
Students will be awarded CENTA2 Training Credits (CTCs) for participation in CENTA2-provided and ‘free choice’ external training. One CTC equates to 1⁄2 day session and students must accrue 100 CTCs across the three years of their PhD.
The student will gain skills in working on High Performance Computers and in using state-of-the-art dynamical and statistical models. Training in analysing meteorological and hydrological data, and in using the BlueBEAR supercomputer will be provided at the University of Birmingham. There is also the opportunity to attend lectures on meteorological and hydrological processes, statistics and data analysis in the MSc programs ‘Applied Meteorology and Climatology’ and ‘River and Environmental Management’.
The supervisors have many well-established national and international links, which will help the student to build a research network.
Partners and collaboration (including CASE):
The School of Geography, Earth and Environmental Sciences at UoB includes strong groups in meteorology, climate science and hydrology, which makes it ideal to host this interdisciplinary project. The student will benefit from links to past and current international projects on regional climate change (EU COST Action VALUE, CORDEX), hydrological processes (UNESCO-FRIEND, IAHS Panta Rhei), hydroclimatic changes in India (India UK Water Centre, WCSSP-India project HEPPI (led by M. Widmann)), and weather and climate research in South Africa and Brazil (WCSSP-South Africa, CSSP-Brazil). There are also well-established links with the UK Met Office and the Indian National Centre for Medium Range Weather Forecasting (NCMRWF).
COVID-19 Resilience of the Project:
The project is fully resilient against impacts of Covid-19. There is no fieldwork involved. The BlueBEAR supercomputer on which the modelling and analysis will be done can be accessed remotely. Supervision meetings can be hold on Zoom if needed. The team of four supervisors ensures resilience of the supervision in case supervisors get ill.
Possible timeline:
Year 1: Pre-processing of weather forecasts and observations for India, South Africa and Brazil. Implementation and evaluation of EMOS postprocessing for the three study regions. Analysis of dependence of forecast skill on spatial scale.
Year 2: Implementation of the VIC hydrological model for all three study regions. Assessment of the VIC model performance using observed meteorological drivers, and optimisation of model parameters.
Year 3: VIC simulations with raw and postprocessed meteorological forecasts. Assessment of the added value of postprocessing of the precipitation forecasts in the context of simulating extremely low and high river flows, and soil moisture drought.
Please email potential supervisor Dr Widmann ([Email Address Removed]) for more information.
Continue with Facebook
