Next generation forecast modelling for effective decision making

As climate change intensifies, water demand forecasting has never been more critical. Advanced modelling tools are transforming how utilities predict future needs, ensuring resilient infrastructure, efficient resource allocation and adaptive planning.

SA Water Metro Water Security Manager Steve Kotz and Stantec Digital Services Market Leader Alana Duncker are set to present at Ozwater’25 on implementing and operationalising next generation demand forecasting.

Ozwater’25 is set to hit Adelaide from 20-22 May and, with SA Water as a Principal Partner for the event, Source took the opportunity to catch up with Kotz and Duncker about why reliable and comprehensive demand forecasting is crucial for ensuring water security.

SA Water and Stantec collaborated to develop a new demand forecasting tool to address the challenges of managing water security in South Australia, which has diverse climates and rapid population growth.

The tool, replacing manual Excel-based processes, uses machine learning to predict water demand based on historical data, climate information and population demographics.

Kotz said moving away from manual processes towards more sophisticated modeling was essential for creating reliable forecasts.

"SA Water services more than 1.8 million people across South Australia, covering a range of climates and varying rates of population growth. Understanding what drives demand across these diverse regions is essential,” he said.

“Previously, our process involved manual spreadsheets, integrating any available data at the time. Maintaining these caused significant overhead. As we started experiencing more rapid growth than ever before, we realised there had to be a better way.

“We wanted a proper database that would allow us to focus on planning for our customers rather than managing spreadsheets. This led to our partnership with Stantec for design and delivery, and Tonkin Consulting for project management.

“There were many moving parts, and ensuring a successful integration of technical expertise, business needs and project management was critical."

Duncker said designing a systemised tool has ensured traceability and auditability.

“Forecasting is inherently uncertain, but a structured tool provides greater confidence in predictions and transparency in the assumptions that were used,” she said.

“Automating processes helps eliminate duplication and human error while creating a framework that produces reliable forecasts within a range, rather than a single prediction."

Investment insight

South Australia is the driest state in the driest inhabited continent in the world with highly variable climate conditions, Kotz said, creating a critical need to balance different water supplies while making large infrastructure investment. Demand forecasting is a critical foundation for these decisions.

“Accurate demand forecasts ensure long-lasting infrastructure is appropriately sized with a ‘no regrets’ approach,” he said.

“The most significant impact of this tool is enabling timely investment decisions. We don’t want to invest too early or too late. The tool supports our adaptive planning framework, ensuring water supply resilience under various population and climate scenarios.

"Beyond infrastructure planning, demand forecasting also informs energy portfolio management, chemical procurement for water treatment plants and long-term budgeting. Having high-confidence projections that are easily communicated to business partners and stakeholders is invaluable."

Kotz said SA Water is seeing unprecedented growth in areas like Adelaide’s northern suburbs, with reliable predictions required to guide allocation of resources.

"Forecasting is the foundation of all planning. By applying scientific methods to past trends and running multiple future scenarios, we provide the necessary data for the business to assess risks and investment priorities,” he said.

Detailed development

Stantec has had a long-standing relationship with SA Water, Duncker said, having already developed a short-term demand forecasting tool for operational decision-making in recent years.

“This project builds on that expertise to create a long-term demand forecasting system. The development process began with extensive stakeholder engagement. We consulted teams across SA Water to understand their forecasting needs, whether for master planning, supply-demand balancing or operational decision-making,” she said.

“This input shaped the tool’s requirements, including data integration, user interface and frequency of updates.

"The tool uses historical data, including past demand, weather conditions (temperature, rainfall, soil moisture), and demographic factors (population, customer mix, economic indicators).

“Using machine learning models, the tool is trained with historical data to predict future scenarios – considering different climate projections, population growth rates and economic conditions. The ability to run multiple scenarios enables us to assess a range of potential futures."

Duncker said a key challenge was ensuring spatial granularity.

“SA Water services 86 different operational regions, each with unique climate and population trends. The tool allows us to analyse demand at different scales – from state-wide projections to individual regions, ensuring highly specific and accurate forecasting,” she said.

Marked difference

Kotz said one of the key differences the new modelling approach has created is more time for planners to focus on planning, rather than processing or cleaning data.

“Having a centralised, transparent data source prevents reliance on isolated spreadsheets that can be lost or become outdated when staff changes occur,” he said.

From a modelling perspective, Duncker said there are a few critical differences that the new tool introduces, including a boost in transparency, flexibility and scalability for SA Water’s analytics.

"Regression-style models have been around for a long time, but the difference now is automation and accessibility. The tool isn’t replacing expert data scientists or modellers, it’s making their work faster, enabling them to run more scenarios and standardise data inputs for consistency across the business,” she said.

"Another key change is shifting to a cloud-based solution. This transition integrates SA Water’s data analytics environment, leveraging advanced tools while ensuring transparency and data security. The goal is to avoid a ‘black box’ system, so the tool remains flexible and accessible for ongoing enhancements.

"The tool isn’t static – it will continue to evolve. End users are always requesting new features, and because of the way it’s built, it can be enhanced over time. The flexibility of this system ensures it remains a valuable asset well into the future."

Interested in learning more about next generation forecast modelling and other critical developments unfolding across the water industry? Register for Ozwater’25 and join us at Australia's premier water conference.