A Multi-step Approach for Downscaling Daily Precipitation Extremes from Historical Analogs

Future changes in extreme precipitation will have profound implications for agriculture, public infrastructure, and human health. According to the most recent assessment from the Intergovernmental Panel on Climate Change (IPCC), both the frequency and magnitude of extreme precipitation are expected to increase in midlatitude continental regions throughout the twenty-first century. Such changes will likely exacerbate the societal impacts of extreme precipitation in the future. In consideration of these issues, the Northeast Regional Climate Center (NRCC) has partnered with the New York State Energy Research and Development Authority (NYSERDA) to compare various methods of downscaling global climate model (GCM) output and create extreme precipitation projections for New York State. These projections will ultimately be incorporated into climate change adaptation planning.

One of the downscaling procedures used in this project involves a unique approach for downscaling daily precipitation extremes from historical analogs. Unlike previous analog downscaling methods, the new method utilizes a multi-step procedure in which the occurrence of extreme precipitation on a given future target day is first determined based on the probability of extreme precipitation on that day's most similar historical analog days. If extreme precipitation occurred on the selected analog day(s), daily precipitation observations recorded at stations on the analog day(s) are used to ascribe precipitation amounts on the corresponding future target day. The method was developed and tested for a historical trial period using daily precipitation data from 157 Cooperative Observer Program (COOP) stations and National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis data.

The performance of the analog method was tested by conducting a forecast verification and comparing analog-derived precipitation extremes with observed precipitation extremes. Overall, results were generally positive and indicate that the proposed analog method is suitable for downscaling daily precipitation extremes in New York State. While the analog method slightly underestimates the occurrence of extreme precipitation, it presents a significant improvement over climatology in terms of accurately predicting the occurrence of extreme precipitation on a given day. Return period precipitation amounts estimated from the analog method are similar to, but generally lower than those calculated from daily precipitation observations. Over the entire study domain, the median difference between downscaled and observed return period precipitation amounts is less than 10%. These precipitation biases compare favorably with those obtained from historical dynamically downscaled climate model simulations.