In the sprawling landscapes of hydrological modeling, a new study is challenging the status quo, with significant implications for the energy sector. Led by W. J. M. Knoben from the University of Calgary’s Schulich School of Engineering, the research, published in Hydrology and Earth System Sciences, delves into the heart of model selection for predicting water availability and threats across vast geographical domains.
The energy industry, with its thirst for water in power generation and cooling processes, is particularly invested in accurate hydrological predictions. However, the current practice of selecting models based on aggregated efficiency scores, such as the Nash–Sutcliffe efficiency and Kling–Gupta efficiency (KGE), may not be as reliable as previously thought. Knoben’s study reveals that these scores are subject to considerable sampling uncertainty, meaning the exact time steps used to calculate them can significantly impact the results.
“This uncertainty makes it very difficult to unambiguously select one model over another,” Knoben explains. “In fact, if models were selected based on their validation KGE scores alone, almost every model would be chosen as the best in at least some basins.”
The study, which analyzed 36 conceptual models across 559 basins, found that model equifinality—where very different models produce similarly accurate simulations—is a significant challenge. When accounting for sampling uncertainty, the number of models needed to cover 95% of investigated basins dropped to just four. To cover all basins, only ten models were needed.
This finding suggests that using a wide variety of models may not necessarily lead to appreciable differences in simulation accuracy compared to using a smaller number of carefully chosen models. For the energy sector, this could mean more efficient and cost-effective modeling practices, with less time and resources spent on model selection.
But what does this mean for the future of hydrological modeling? Knoben’s research indicates that we may need to rethink our approach to model selection. Rather than relying on aggregated efficiency scores, we might need to consider other factors, such as model complexity, computational cost, and the specific needs of the energy sector.
Moreover, this study highlights the importance of accounting for sampling uncertainty in model selection. As Knoben puts it, “Understanding and accounting for this uncertainty is crucial for making informed decisions about model selection and for improving the reliability of hydrological predictions.”
As the energy sector continues to grapple with water scarcity and climate change, accurate hydrological predictions will become increasingly important. Knoben’s research, published in the journal ‘Hydrology and Earth System Sciences’ (which translates to ‘Hydrology and Earth System Sciences’ in English), offers a timely reminder that our modeling practices must evolve to meet these challenges. By embracing a more nuanced approach to model selection, we can ensure that our hydrological predictions are as reliable and useful as possible, helping the energy sector to navigate an uncertain future.