Estimating Climate Impacts to Precipitation-Frequency: Non-Stationarity Detection and Qualitative Analysis of Climate Projections

The Probabilistic Flood Hazard Assessment (PFHA) framework is sensitive to changes in precipitation-frequency. Precipitation-frequency relationships are mathematical expressions of the precipitation magnitude for a series of return periods, from the common events up to extremely rare events with an annual exceedance probability (AEP) of one in a million or less, useful for assessing the safety of dams. For two dams owned by the Tennessee Valley Authority, sensitivity analyses showed that the primary driver of changes to hydrologic loading is potential changes to the precipitation-frequency relationship in the watersheds induced by climate change.

This analysis focuses on qualitative and quantitative methods to evaluate potential changes in precipitation-frequency in the two watersheds using both historical data and future climate projections.

Two types of historical observed precipitation data were tested at ten stations within the vicinity of the two dams: annual maxima series (AMS) for the risk-driving mid-latitude cyclone (MLC) storm type and total annual precipitation. Both time-series were found to be stationary with no observable or statistical trends or break points in the data. Climate covariates related to precipitation, such as cool season temperature when MLCs are most common, temperature anomalies, and sea surface temperatures, were tested against the historical observed AMS and the total annual precipitation.

The projections from six downscaled general circulation models (GCMs) were used to compare a historical period (1990-2010) and future period (2060-2080) at the ten station locations. The projections exhibited an upward trend in AMS and in total annual precipitation though the entire period. Using an L-moments regional statistical analysis, precipitation-frequency analyses were conducted on the historical and future GCM AMS separately and compared to one another. From a regional perspective, the future regional point precipitation-frequency curves exceed the historical regional point precipitation-frequency curve at all AEPs. The rate of increase differed with AEP, such that the difference increased with smaller (rarer) AEPs; for example, the differences between the two periods increased from 24.7% at 10-2 to 28% at 10-4 AEP.