Geologic, Hydrologic, and Urban Hazards for Design in Desert Environments - Design with the Desert: Conservation and Sustainable Development - page 93

year is defined, purely for convenience, as the federal government fiscal year (October 1
through September 30). Another useful subset of flood data is the partial duration series,
which contains all floods above an arbitrary and fixed base discharge; some years have
many floods while some years have only one or no floods that exceed the base discharge.
Flood frequency can be estimated from the annual flood series using standardized tech-
niques that are applied throughout the United States.*
Not all streams and rivers that can cause flood damage have streamflow gaging records;
in fact, the number of streamflow gaging stations in the United States has declined sig-
nificantly since its peak in the 1970s and 1980s. 15 Flood frequency at ungaged sites com-
monly is estimated using regional-regression techniques that estimate flood frequency
as a function of drainage area, climate, elevation, and geographic area. 16 Online geo-
graphic information system tools are in development to allow a user to click on a map
showing rivers and streams in the United States and access flood frequency information
for that ungaged site. †
Flood-frequency analysis, whether performed directly on gaging data or indirectly using
regional-regression techniques, requires some assumptions that have been severely criti-
cized as unrealistic, particularly given the specter of future climate changes. As currently
applied worldwide, flood-frequency analysis requires some data assumptions that may
not be reasonable, given past and potential future trends in climate and flood-producing
storms (see Chapter 3). The critical assumption is that the annual flood series is station-
ary, a term defined as a series with time-invariant mean and variance over the period of
record. Some have argued that historical gaging records are nonstationary 4 ; others claim
that stationarity is no longer a viable assumption in hydrologic time-series analyses, 18,19
particularly in flood-frequency analysis. Proponents of nonstationarity in surface-water
hydrology either offer different techniques for analyzing flood records or advocate models
that rely on future climate predictions.
The San Pedro and Santa Cruz Rivers well illustrate the response of middle-elevation
watercourses to climate variation (Figure 5.2). Both records illustrate the concept of
nonstationarity in flood frequency, 4,5 although for different reasons. The statistical
concept of stationarity holds that the moments of a time series are temporally invariant;
hydrologic nonstationarity occurs when the mean and (or) the variance of flow or flood
magnitude changes with time (e.g., has a trend or fluctuation). The San Pedro River had
a large flood in 1926 that was considerably larger than the second largest flood; but,
otherwise, the annual flood series does not show significant trend in magnitude through
the twentieth century (Figure 5.2a) and therefore is considered to be a stationary time
series. However, the season during which the annual flood occurred has changed, with
a shift from predominantly summer floods in the middle of the twentieth century to a
mixture of mostly fall and winter floods since the mid-1960s. In contrast, the annual flood
series of the Santa Cruz River (Figure 5.2b) shows changes in seasonality of flooding as
well as flood magnitude.
The issue of stationarity, as applied to flood-frequency analysis, is as complicated as
the surface-water hydrology of the southwestern United States. Large rivers, such as the
Colorado and Rio Grande, are subject to decreasing snowpack 20-22 and its impact on spring
flood flows; these rivers therefore would be expected to have a nonstationary series of
* Techniques for estimating flood frequency in the United States are contained in U.S. Water Resources Council. 14
Software for estimating flood frequency is available from the U.S. Geological Survey at http://water.usgs.gov/
software/PeakFQ/ (accessed June 24, 2009).
† An example is the State of Utah, one of the first states that has implemented the StreamStats program within
U.S. Geological Survey; see http://water.usgs.gov/osw/streamstats/ (accessed March 16, 2009). 17