Leveraging normalized data to improve point-scale estimates of precipitation–temperature scaling rates

Abstract

Sub-daily to daily extreme precipitation intensities are expected to increase in a warming climate, consistent with the Clausius–Clapeyron (C–C) relationship, which predicts a ∼ 7 % increase in atmospheric moisture-holding capacity per °C of warming. Many studies have benchmarked observed extreme precipitation–temperature (P–T) scaling rates against this theoretical value, finding that globally averaged scaling rates of daily extreme precipitation are largely consistent with C–C, while hourly extremes have been observed to scale at rates greater than C–C. Significant challenges remain, however, in accurately estimating and interpreting P–T scaling rates, particularly at point scales. In this study, we use observational station-based data from the Upper Colorado River Basin to illustrate these challenges and propose methodological improvements. Specifically, we compare multiple approaches, including those using raw (non-normalized) and normalized data, to estimate P–T scaling for hourly and daily extreme precipitation. Model performance is assessed using a cross-validation framework. Our results demonstrate that when estimating P–T scaling rates using data pooled from multiple stations and/or months, it is essential to account for spatial and temporal climatological variability. We find that using normalized data allows us to more effectively leverage pooled data, and thus improve our estimates of P–T scaling rates.