DEPARTMENT FOR SCIENCE, INNOVATION AND TECHNOLOGY

Weather and Climate Science for Service Partnership (WCSSP) South Africa - Calls- tender-UKCEH

Project disclaimer

Disclaimer: The data for this page has been produced from IATI data published by DEPARTMENT FOR SCIENCE, INNOVATION AND TECHNOLOGY. Please contact them (Show Email Address) if you have any questions about their data.

Description

WARD SA: Water Availability duRing Drought in South Africa South Africa is a semi-arid to arid country, with low rainfall (~500 mm annual average) (Abiodun, Naik , Mogebisa, & Makhanya, 2022) and high evaporation rates. Rainfall over South Africa is highly variable, both spatially and temporally, and the country often experiences meteorological drought. The intensity of meteorological drought is projected to increase under climate change (Engelbrecht F. A., Steinkopf, Padavatan, & Midgley, 2024), and is a significant risk to sustainable development into the future. Because of South Africa’s high levels of terrestrial, freshwater and marine biodiversity and endemism, South Africa’s progress towards environmental sustainability is critically dependent on being able to predict and manage hydro-climate risks into the future (Mutanga, et al., 2024). While South Africa has a long tradition of research-based approaches to the management of its water resources (Morant & Quinn, 1999) and there has been significant research on future climate over Southern Africa (Engelbrecht F. A., Steinkopf, Padavatan , & Midgley , 2024; Engelbrecht & Monteiro, 2021), there have been relatively few studies on the impacts of climate change on hydrology and water resources (Kusangaya, Warburton, van Garderen, & Jewitt, 2014). This information is key to future planning of water resource management, to assist decision making and to inform potential adaptation and mitigation strategies. Until relatively recently, land surface models, such as the Joint UK Land Environment Simulator (JULES), have tended to ignore water resource management activities, instead representing a more nearly natural system. More recently land surface models have started to recognise the importance of better representing human management of the natural world, often using approaches pioneered in earlier global hydrology models such as H08, WaterGAP and GWAVA (Hanasaki N. , et al., 2008; Döll, Kaspar, & Lehner, 2003; Meigh, McKenzie, & Sene, 1999). However, to date JULES includes very little description of water resource management – just a simple scheme to acquire water for irrigation from soil and then rivers. Land surface models are used across a range of resolutions (i.e. the model gridbox size) from of the order of 50 km for global earth system applications, to of order 1 km in so-called “K-scale” applications. An area of active research is the extent to which model formulation and parameterisations need to be adapted for K-scale, and water resource management is no exception here. In WARD SA we will address these knowledge gaps by building on existing Water Resource Management (WRM) functionality in JULES to create a modelling tool which can be applied at coarse and fine spatial scales to assess the impacts of climate change on available water resources in Southern Africa. Driving data from the Inter-Sectoral Impacts Model Inter-comparison Project (ISIMIP) at 0.5⁰ spatial resolution, and km-scale data from convection-permitting models, will be used to produce simulations that can inform effective climate adaptation and mitigation action. Working closely with South African partners, different land and water resource management options will be investigated through a set of exploratory runs, demonstrating the potential for this tool to inform critical policy decisions. Alongside the development of JULES, WARD SA will explore the potential of Machine Learning (ML) methods to aid water resource forecasting in reservoirs. Machine Learning methods are being increasingly applied to hydrological modelling, particularly Long Short-Term Memory (LSTM) models (Kratzert, Gauch, Klotz, & Nearing, 2024). There are far fewer studies applying ML methods to reservoir storage or outflow (Dai, et al., 2022; García-Feal, González-Cao, Fernández-Nóvoa, Dopazo, & Gómez-Gesteira, 2022), perhaps because this data is far less available than streamflow data, but the restricted nature of reservoir data is what makes it a good candidate for ML methods. Operating rules for managed reservoirs are rarely openly available, so process-based models rely on generic equations with calibrated or approximated parametrisations, but ML methods can “learn” the operating rules provided there is sufficient training data. These ML models can then be run independently with appropriate driving data, or used as part of a hybrid modelling approach, to improve reservoir storage predictions and downstream flow simulations.

Objectives

This work intends to meet the need for assessments of the impact of climate change on water resources across Southern Africa on multi-decadal timescales by developing and applying of various models for water resources under future scenarios. Specifically, we will address the following scientific objectives: Objective 1: Develop a model of large-scale water resources over southern Africa and use this to develop an analysis of regional water availability under a range of multidecadal projections of climatic and socioeconomic scenarios. Objective 2: Refine the model for higher-resolution (km-scale) applications and develop case studies for selected river catchments. Objective 3: Explore the value of Machine Learning approaches, potentially as part of a hybrid modelling approach. These objectives will be delivered through a programme of activities described below, in continuous collaboration with South African partners.

---

Location

The country, countries or regions that benefit from this Programme.

South Africa

Disclaimer: Country borders do not necessarily reflect the UK Government's official position.