Comparison of uncertainties in projected flood frequency of the Zhujiang River, South China

摘要

This study investigated uncertainties in the modeling of hydrological impacts of climate change on projected flood frequencies of the Zhujiang River, South China. The hydrological model HBV-D was applied to simulate and project future stream flow based on a multi-model ensemble. As this implies high uncertainties, the magnitude of three uncertainty sources, i.e. emission scenarios, GCM structure, and downscaling techniques, were determined in relation to the observed and projected natural variability. The relative change in each uncertainty source and the overall dominance among the three sources were further analyzed. The changes in flood frequency are projected for five return periods (2, 5, 10, 20, and 50 years) and three future time periods (2020s, 2050s, and 2080s). The results suggest that in comparison to the natural variability of the multi-model ensemble, the uncertainty sources show much stronger variations. The range of their relative change and their dominance vary with the lead-time and return period. In most of the return periods, the dominant uncertainty can primarily be attributed to downscaling techniques and emission scenarios, while GCMs structure is minor in the 2020s. However, downscaling technique is the second dominant source behind GCM structure, while emission scenarios represent the lowest uncertainty ranges of the three sources for the projected flood frequency in the 2050s and 2080s. The uncertainty and projected impact of climate change differs also between the four applied GCMs, as compared to the natural variability MK3_5 shows higher ranges than CCSM3, MK3_5 and ECHAM5. The upper bounds (95% percentile) in uncertainty mostly show an increasing tendency with increasing return period, and partially with increasing lead-time. Hence, the more extreme the return period (higher flood frequency) the higher is the uncertainty of the model projections. It is therefore essential that climate change impact assessments consider a wide range of climate scenarios derived from different GCMs under multiple emission scenarios and including several downscaling techniques. The uncertainty due to natural variability should also be considered more intensely. The projection of flood frequency and the identification and quantification of the uncertainties in the modeling is important for the implementation of adaptation policies into water resource planning in the Zhujiang River basin, South China. This study will enrich the scientific research on the uncertainty from different sources of modeling results in river basin parameters, and help to obtain conclusive results on the importance of different sources of uncertainty.