Uni- and multivariate bias adjustment methods in Nordic catchments: Complexity and performance in a changing climate

Abstract

For climate-change impact studies at the catchment scale, meteorological variables are typically extracted from ensemble simulations provided by global and regional climate models, which are then downscaled and bias-adjusted for each study site. For bias adjustment, different statistical methods that re-scale climate model outputs have been suggested in the scientific literature. They range from simple univariate methods that adjust each meteorological variable individually, to more complex and more demanding multivariate methods that take existing relationships between meteorological variables into consideration. Over the past decade, several attempts have been made to evaluate such methods in various regions. There is, however, still no guidance for choosing appropriate bias adjustment methods for a study at hand. In particular, the question whether the benefits of potentially improved adjustments outweigh the cost of increased complexity, remains unanswered.

This paper presents a comprehensive evaluation of the performance of two commonly used univariate and two multivariate bias adjustment methods in reproducing numerous univariate, multivariate and temporal features of precipitation and temperature series in different catchments in Sweden. The paper culminates in a discussion on trade-offs between the potential benefits (i.e., skills and added value) and disadvantages (complexity and computational demand) of each method to offer plausible, defensible and actionable insights from the standpoint of climate-change impact studies in high latitudes.

We concluded that all selected bias adjustment methods generally improved the raw climate model simulations, but that not a single method consistently outperformed the other methods. There were, however, differences in the methods' performance for particular statistical features, indicating that other practical aspects such as computational time and heavy theoretical requirements should also be taken into consideration when choosing an appropriate bias adjustment method.