Geologic, Hydrologic, and Urban Hazards for Design in Desert Environments - Design with the Desert: Conservation and Sustainable Development - page 94

decreasing annual floods. Other rivers, such as the Santa Cruz, respond to different storm
types, and some of these types may well generate floods from a stationarity distribution
while others change with time. For example, using the partial-duration series, floods
caused by summer thunderstorms in the region demonstrably have time-invariant mean
and variance over the period of record, while floods caused by regional storms—either in
winter or dissipating tropical cyclones—are related to global-scale climate variation and
are nonstationary. 4 Because regional storms generally create the largest floods, the annual
flood series therefore is nonstationary as well.
The Santa Cruz River at Tucson is perhaps the best example to illustrate the impact of
nonstationarity in an annual flood series on the estimation of flood hazard. Starting in
1970, the 100 year flood is 20,870 ft 3 is based on the station statistics from 55 years of record
(Figure 5.3). Adding each new annual flood peak with each passing year and calculat-
ing the new 100 year flood result in an extremely interesting time series of flood hazard.
Because of large annual peak floods in 1977, 1983, and 1993, the 100 year flood rises from
20,870 ft 3 is to a maximum of 38,180 ft 3 is in 1993 (Figure 5.3). What is most interesting is
that in 2007, the 100 year flood is 31,610 ft 3 is and higher than the +95% confidence limit
estimated in 1970. In other words, the statistics of flood frequency calculated in 1970 and
2007 is significantly different, which would only occur if the time series is nonstationary.
In addition, the period of increased flood frequency is only for part of the record—from
about 1977 through 1995 (Figure 5.2B)—with what appears to be time-invariant flood fre-
quency for the gaging record before 1977 and after 1995.
In general, changes in surface-water flow illustrated by the gaging records of the San
Pedro and Santa Cruz Rivers represent some trends in southern Arizona. 23 Streamlow
in the San Pedro River at Charleston has declined by 50% in the period of record, 24 but
the annual flood series does not have an apparent trend (Figure 5.2A), illustrating the
decoupling of flood frequency with typical flows. Trends of increasing flood frequency
decrease northward in the region, and trends in streamflow decline are highly variable
in the region, although some general patterns remain. Large, regional floods occurred in
1862, 1884, 1891, 1963, 1977 through 1979, 1983, and 1993; similarly, streamflow was lower
during the mid-century drought than streamflow between 1960 and 1995. Differences are
also significant; although floods that scoured channels were common in southern Arizona

60,000

100 Ye ar flood
+95% Confidence interval
-95% Confidence interval

50,000

40,000

30,000

20,000

10,000
1970 1980 1990 2000 2010

FIGURE 5.3
Estimated 100 year flood as a function of time, 1970-2007, Santa Cruz River at Tucson, Arizona.