21 Mar Embracing Uncertainty Quantification: A Smarter Future for Water Management

Climate is increasingly uncertain. Australia has just dealt with the onslaught of tropical cyclone Alfred in Queensland on one side of the continent, causing floods and dam spills. In the meantime in South Australia, one of the worst droughts in decades is causing concerns about water security and water quality.

Such climate extremes immediately lead to discussions about insurance premiums, as they reflect the increased risk associated with these events. However, these discussions also highlight the need for better risk quantification in water management.

Mollee Weir, Namoi River (near Narrabri)

Understanding and quantifying risk is important for water security and water management as these are inherently uncertain. Water management relies on summarising the impact of highly variable climate events (such as Alfred) on complex catchment systems and predicting the downstream impacts. Many input variables related to water management (such as streamflow, rainfall and evaporation) are surprisingly difficult to measure exactly across large landscapes, such as in Australia.

The decline in global hydrological observation networks, resulting in fewer locations where on-the-ground data is collected, does not help. Finally, there is constant change in the landscape due to human activities and resulting ecological change.

As a result, hydrological predictions are also uncertain, and therefore risk is difficult to quantify, even if based on the best available science. This can result in sub-optimal decision-making and less equitable water sharing.

Quantified uncertainties can not only indicate risk; they also point to deficiencies in our data sets and even deficiencies in our theories. We can then use these to adjust our sampling and monitoring approaches, and we can adjust our theories. This is how science develops and improves, and we therefore should embrace uncertainty rather than be afraid of it.

Water scientists have for many decades grappled with this uncertainty, and together with mathematicians and statisticians applied many methods to understand uncertainty. However, uncertainty methods in hydrology tend to be computationally and theoretically complex, and not easily applied in operational systems.

As a result, standard application of uncertainty analysis is still lacking. In contrast, climate scientists and meteorologists have embraced uncertainty early on, for example, the forecasts by the Australian Bureau of Meteorology are all probabilistic.

Namoi River (near Boggabri) in early 2020 compared to early 2025 (Photo credit: Willem Vervoort)

The main challenge can be broken down into three areas:

• Complexity of formal uncertainty approaches
• Complexity of hydrological models and conceptual understanding
• Communication of uncertain outcomes

Of these challenges, the third is probably the most difficult, but also the most crucial. Without a clear understanding of the value and need for uncertainty quantification, progress in operational hydrology will continue to be difficult. I believe that we have made good progress in this area, but good examples of how uncertainty quantification has added value to decisions, or saved on operational costs, are still needed. At DARE we are working on two case studies that can contribute to this.

The first two challenges are more technical. Potentially with faster computing and toolboxes that are easier to use, we can solve this, but I feel that the greatest potential is in development of faster surrogate approaches using machine learning.

If these approaches include uncertainty and are explainable (in other words the algorithms can be interrogated and interpreted), then we can also address the third challenge more effectively. Close collaboration between experts in Data Science, Hydrology and the Water industry will ensure that this results in methods that are relevant and accessible.

To better understand the risks and opportunities to achieve the UN’s sustainable development goal SDG6 (Ensure availability and sustainable management of water and sanitation for all), we will need to embrace uncertainty quantification. This will ensure we can make more informed water management decisions, ensuring resilience in the face of climate change.

The most practical way to do this is through multi-disciplinary approaches, such as the DARE Centre, bringing together industry experts, Data Scientists and Hydrologists to build accessible, reproducible and explainable methods of uncertainty quantification for the water sector.