Downscaling Extremes of Rainfall: Sensitivity to Gridded Observations and Downscaling Technique

Statistical downscaling has become an important tool to assess the impact of a changing climate on local rainfall and the hydrology of local regions. This includes assessing changes to extreme events which drive flooding and the risk to life and property. However, statistical downscaling uses two inputs - global climate models and gridded observations - to derive projections of local change to rainfall. Prior studies have shown that global climate models are one major source of uncertainty for precipitation projections generally. However, little attention has been given to the effect of downscaling technique and gridded observations on precipitation projections, and particularly projections of precipitation extremes. This presentation focuses on the sensitivity of downscaled projections of precipitation extremes to the statistical downscaling technique and gridded observations using the south-central United States as an example. The analyses make use of multiple downscaling techniques and gridded observations applied to downscale precipitation from three GCMs forced with representative concentration pathway (RCP) 8.5. This study also focuses on how different downscaling techniques captures the projected changes in the precipitation distribution and the effects to derived variables representing rainfall extremes such as annual one-day maximum precipitation. Results from this study indicate that the choices used in the creation of gridded observation can be translated through downscaling to affect the change signal for projections of wet day frequency and the intensity of precipitation extremes. The change signal for precipitation projections is also effected by the downscaling technique, reversing the change signal in some cases. Gridded observations and downscaling technique are sources of uncertainty in downscaled precipitation projections which results from decisions made to correct measurement errors and capture GCM change signals. These two sources of uncertainty are potentially reducible with respect to projections of regional change. However, they also have the potential to influence impact assessments that rely upon daily outputs of precipitation from downscaling, including assessments connected with projections of precipitation extremes.