Indicator Delta Scaling (IDS): A Consistent and Efficient Method for Bias-Correcting Climate Risk Indicators

Stakeholders across multiple economic sectors increasingly require ready-to-use and reliable climate information to support climate change adaptation and risk-informed decision-making across diverse sectors such as water resources, agriculture, energy, infrastructure, and health. For these applications, it is essential that climate estimates are as realistic and precise as possible, accurately characterizing both average conditions and climate extremes that underpin climate risk assessments.

Bias-correction methods represent a key processing step in the production of climate indicators derived from climate projections, aiming to reduce systematic model errors and enhance the usability of climate simulations. However, many studies have demonstrated that commonly used bias-correction approaches may introduce important inconsistencies. These include alteration of observed historical estimates, modification or even reversal of the climate change signal projected by climate models, changes in the model uncertainty spread, and strong sensitivity of method performance to the considered variable, climate indicator, region and observational reference dataset. These limitations highlight the risks of applying bias-correction techniques blindly, without careful examination of their implications for each specific case. This contrasts, however, with the strong need for a consistent and comprehensive provision of diverse climate indicators globally to support climate information needs across sectors and stakeholders.

 
Here, we propose a simple but consistent and accurate delta-based approach for computing adjusted climate indicators, the Indicator Delta Scaling (IDS). The method relies on two basic principles: historical estimates are derived exclusively from observational datasets, while future corrected indicators are obtained by simply updating the observational reference with the projected raw change signal. The method is evaluated globally using CMIP6 historical simulations against observations, which are used both as the historical reference and as a pseudo-future framework. A diverse set of simple, complex, and multivariate climate indicators is used to evaluate the performance of IDS in comparison with state-of-the-art bias-correction approaches, such as Quantile Delta Mapping and the ISIMIP3b method.

Results show that IDS outperforms existing bias-correction methods across multiple evaluation levels. In contrast to other methods, IDS ensures by construction a perfect representation of observed historical estimates, a strict preservation of the modelled delta change and a solid consistency across variables, indicators, and datasets. At the same time, it provides a similar but slightly more accurate estimate of most indicators for future periods. Moreover and importantly, by avoiding the bias correction of input variables' full data distribution, the approach delivers major computational efficiency gains when computing climate indicators.

In summary, the IDS provides a clear, consistent, accurate, and efficient framework for generating ready-to-use climate indicators, addressing key limitations of current bias-correction practices and supporting robust and comprehensive climate risk assessments. The method has been developed within a Copernicus Climate Change Service contract to streamline the global computation of indicators for assessing EU Taxonomy hazards, following the guidance of the European Investment Bank (EIB) for financial risk assessments.