Numerous scientific assessments have shown that human-induced climate changes are occurring, and more changes are expected as atmospheric composition is altered. This work focuses on how these changes affect extreme precipitation rates. In particular, design values are sought for extreme precipitation rates ranging from sub-daily to multi-day events. These values are typically quantified as extreme precipitation Intensity-Duration-Frequency values and, when plotted, are used by engineers and others for planning, design, and operations as Intensity-Duration-Frequency curves. The most comprehensive set of existing Intensity-Duration-Frequency curves developed over the past two decades is based on the assumption of a stationary climate. A key ultimate objective of this work was to transform Intensity-Frequency-Duration values into a new set that accounts for a non-stationary climate with varying degrees of climate change.
A prerequisite to developing non-stationary Intensity-Duration-Frequency values is to complete a comprehensive trend assessment across broad geographies. This includes both a representative set of Intensity-Duration-Frequency values and the weather and climate conditions known to affect these values across the U.S. and worldwide. Weather events of interest include those systems that generate upward vertical velocity over geographies with ample water vapor. Since it has been previously shown that particular weather types are needed to trigger extreme precipitation rates, an automated and objective method is needed to identify those weather types, not only in past data but in the enormous quantity of data generated from various global climate model (GCM) simulations, including their ensemble members of future climate.
Although GCMs are the backbone of understanding past and future climate, they are known to have deficiencies in simulating extreme precipitation rates of interest here. A number of methods have been developed to help overcome many of these deficiencies. Key objectives of this work include building on these existing methods to enable robust uncertainty estimates of future Intensity-Duration-Frequency values and, most importantly, provide an understanding of why changes in the Intensity-Duration-Frequency values are expected for specific locations and future times.
Two complementary, but different methods were used to provide best estimates for future Intensity-Duration-Frequency values. The differing methods provide a basis for assessing uncertainty of future changes. The sources of uncertainty include 1) differing rates of human-caused changes to atmospheric composition, 2) the sensitivity of the climate to changes in atmospheric composition, 3) parameter estimation uncertainty of model coefficients, 4) structural uncertainty related to the approach used, and 5) statistical sampling uncertainty. The first approach used to estimate future changes—the precipitation downscaled method—uses a statistical method known as generalized extreme value applied to a downscaled precipitation dataset, where changes in extreme precipitation rates, both historically and from global climate models, are calculated and used to project changes in Intensity-Duration-Frequency values by modifying existing Intensity-Duration-Frequency values based on a stationary climate. The second method, called the precipitation causes approach, makes use of changes in well-simulated meteorological factors shown to contribute to extreme precipitation rates. These factors include column-integrated water vapor (i.e., precipitable water [PW]) and the weather systems causing upward vertical velocity responsible for condensing the water vapor into precipitation. Changes in PW, weather fronts, and extratropical cyclones (ETCs) are calculated from GCM simulations of past and future climate, and they are used to transform Intensity-Duration-Frequency values from stationary to non-stationary estimates.
Machine learning is used to develop an objective and automated algorithm for identifying fronts both in GCM output and model-assimilated observed data. This includes warm, cold, stationary, and occluded fronts associated with extreme precipitation events. The frontal identification is validated through comparison with human expert weather map analyses regularly produced by the National Oceanic and Atmospheric Administration. Additionally, other objective methods were used to identify ETCs, both in observed data as well as past and future simulated climates. Changes in PW are also examined both historically and within GCM simulations of future climate and in relation to changes of extreme precipitation and upward vertical velocity.
Observations show that over the past several decades the interval between extreme precipitation events is decreasing as they become more frequent across a wide range of durations. This is not evident in all areas, but it is the predominant trend and it is strongly linked to increases in PW. Global climate models project widespread increases in PW as the climate warms with various scenarios of human-induced changes of atmospheric composition. As a result, as time evolves, the projected Intensity-Duration-Frequency values generally lead to greater frequency of, and shorter intervals between, extreme precipitation events for a wide range of thresholds and duration times. Additionally, the rarer events tend to increase more than less-rare events (e.g., 50-year return period versus one-year return period), regardless of duration. These projected changes are well beyond standard uncertainty intervals. The projected changes in other factors, such as fronts and extratropical storms, are not ubiquitous, and the magnitude of extreme precipitation is shown to be less sensitive to these changes compared to PW. This is because even when weather systems change in frequency, they still occur often enough to trigger copious precipitation when PW is high. The projected weather system changes do, however, combined with varying degrees of increases in water vapor, add to the spatial and temporal variability of projected Intensity-Duration-Frequency values.
Planners, designers, and operation managers can now get access to a set of extreme precipitation Intensity-Duration-Frequency values for thousands of locations to help assess future risks of extreme precipitation events. The Intensity-Duration-Frequency values incorporate varying degrees of climate change associated with human alterations of atmospheric composition. Additionally, it is now possible to unravel specific causes associated with projected changes to better understand the risks of extreme precipitation events.