Logo: to the web site of Uppsala University

Climate change poses undeniable impacts on hydroclimatic processes due to simultaneous effects of rising temperature and changing precipitation patterns. To quantify these impacts, simulations of climate variables are typically retrieved from climate models, which are then downscaled and bias-adjusted for a particular study site.

The literature holds various methods for bias adjustment, ranging from simple univariate methods that only adjust one variable at a time, to more advanced multivariate methods that additionally consider the dependence between variables. 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 thesis primarily sought to provide an answer to this question by offering practical guidelines for the application of uni- and multivariate bias-adjustment methods in hydrological climate-change impact studies. To this end, the thesis includes a practice-oriented overview of copulas, one of the most widely used multivariate methods in climate-change studies. Furthermore, it presents an evaluation of two commonly used parsimonious univariate and two advanced multivariate methods. The assessment focused on their ability to reproduce numerous statistical properties of precipitation and temperature series, and on the cascading effects on simulated hydrologic signatures. The thesis culminates in a practical application of one bias adjustment method as part of a modeling chain to quantify future droughts. 

The results elucidate that all bias adjustment methods generally improved the raw climate model simulations, but not a single method consistently outperformed all other methods. Univariate methods generally adjusted the simulations reasonably well, while multivariate methods were favorable only for particular flow regimes. Thus, other practical aspects such as computational time and theoretical requirements should also be taken into consideration when choosing an appropriate bias adjustment method.