IMPLICATIONS OF MULTI-DECADAL CLIMATE VARIABILITY FOR STREAM FLOW FREQUENCY ESTIMATION AND FLOOD MITIGATION IN EASTERN PENNSYLVANIA

The assumption that floods are random events, identically and independently distributed in time, is the statistical foundation on which dams, levees, and stormwater systems are designed, and floodplains managed. In recent years, as it has been recognized that climate changes on all time scales, there have been increasing doubts about static models of flood probabilities. I report on step-like changes in the amount and intensity of precipitation in eastern Pennsylvania over the last century. The steps define two “dry” epochs spanning a total of 39 years in which precipitation averaged 11% lower than the 1904-2000 average, and two “wet” epochs spanning 58 years in which precipitation was 7% higher. Precipitation events were larger and more intense during wet periods. For example, at Palmerton, PA the 5-yr rainfall event within wet epochs was 5.4 inches compared to 4.2 inches within dry epochs; the 5-year single-day intensity was 3.6 inches (wet) compared to 2.8 inches (dry).

Flow frequency estimates based on 17 stream gages monitoring watersheds from ~10 sq. mi. to ~10,000 sq. mi. show that the magnitudes of 100-yr floods based on data from wet epochs exceed those based on dry epochs by over 50% (median). For example, the 100-yr flow at the Chadds Ford gage, Brandywine Creek, is 30700 cfs based on wet epochs but 20000 cfs based on dry epochs. In many cases, the 50-year flow in wet epochs is nearly equivalent to the 100-year flow in dry epochs. There are also differences in seasonality: dry epoch floods are relatively more likely in early spring and less likely in late summer and fall.

Data from stream gages with records extending back more than 30 years are mixtures of measurements from wet and dry epochs. Since the span on which data are available differs from gage to gage, current estimates of flow frequencies are not reliably consistent from gage to gage. Furthermore, these estimates have an unknown meaning for the future if precipitation continues to change in a step-like fashion. These results show that there is a need to take multi-decadal variations in precipitation and stream flow into account throughout the U.S. as has been recognized in the west (e.g., California’s American River). Developing the ability to predict multi-decadal change, possibly through oceanic indices, will make our planning efforts more effective.