Climate change impact assessments form the basis for the development of suitable climate change adaptation strategies. For this purpose, ensembles consisting of stepwise coupled models are generally used [emission scenario → global circulation model → downscaling approach (DA) → bias correction → impact model (hydrological model)], in which every item is affected by considerable uncertainty. Large uncertainties inherent in future climate projections may, however, reduce the willingness of regional stakeholder to develop and implement suitable adaptation strategies to climate change.
The presentation provides an overview of different possibilities to consider uncertainties in climate change impact assessments by means of (1) an ensemble based modelling approach and (2) the incorporation of measured and simulated meteorological trends.
The ensemble based modelling approach consists of the meteorological output of four different climate downscaling approaches (DAs) (two Regional Climate Models (RCMs) and two statistical downscaling approaches (113 realisations in total)), which drive different model configurations of two conceptually different hydrological models (HBV-light and WaSiM). As a study area serve three near natural subcatchments of the Spree and Schwarze Elster river catchments (Germany). The objective of incorporating measured meteorological trends into the analysis was twofold: measured trends can (i) serve as a mean to validate the results of the DAs and (ii) be regarded as harbinger for the future direction of change. Moreover, regional stakeholders seem to have more trust in measurements than in modelling results. In order to evaluate the nature of the trends, both gradual (Mann-Kendall test) and step changes (Pettitt test) are considered as well as both temporal and spatial correlations in the data.
The results of the ensemble based modelling chain show that depending on the type of DA (RCM or statistical) used, opposing trends in precipitation, actual evapotranspiration and discharge are simulated in the scenario period (2031-2060). While the statistical DAs simulate a strong decrease in future long term annual precipitation, the dynamical DAs simulate a tendency towards increasing precipitation. The trend analysis suggests that precipitation has not changed significantly during the period 1961-2006. Therefore, the decrease simulated by the statistical DAs should be interpreted as a rather dry future projection. Concerning air temperature, measured and simulated trends agree on a positive trend. Also the uncertainty related to the hydrological model is comparably within the climate change modelling chain low when long-term averages are considered but increases significantly during extreme events.
This proposed framework of combining an ensemble based modelling approach with measured trend analysis is a promising approach for regional stakeholders to gain more confidence into the final results of climate change impact assessments. However, climate change impact assessments will remain highly uncertain. Thus, flexible adaptation strategies need to be developed which should not only consider climate but also other aspects of global change.
