Application of Streamflow Reconstruction to Water Resources Management - Dendroclimatology - page 241

climatic episode in a temporal perspective (Biondi et al. 2005 ) . In this method, a
stochastic model is used to describe the joint distribution of run sum and run length
(severity and duration).
Streamflow series have been observed to have long-term persistence, which
increases the overall variability of the series and compounds the problem of predic-
tion from a short record (Dunne and Leopold 1978 ) . Long-term persistence cannot
be effectively summarized by simple low-order autoregressive models, as described
above, and requires different analysis approaches. The Hurst coefficient, or rescaled
adjusted range, has long been used by hydrologists to describe long-term persis-
tence (Hosking 1985 ) . A tree-ring reconstruction of the Gila River, Arizona, was
analyzed to place the Hurst coefficient of the gauged record in a long-term context
(Meko and Graybill 1995 ) . The empirical distribution of sample Hurst coefficients
for 254 overlapping 70-year segments of the reconstruction indicated that the most
recent 70-year segment provides a slightly biased (high) estimate of the long-term
persistence of annual flow.
The continued development of networks of tree-ring chronologies over the past
several decades has expanded the possibilities for studying multi-basin aspects
of streamflow and runoff. In semiarid regions, such studies are especially rele-
vant to water resources planning because of inter-basin transfers of water. The
starting data for such studies might be existing streamflow reconstructions. For
example, co-occurrence of drought episodes in the Sierra Nevada of California
and the Rocky Mountains of Colorado was summarized in an analysis of recon-
structions for the Sacramento River, California, and Blue River, Colorado (Meko
and Woodhouse 2005 ) . An evolutive cross-spectral analysis was used in that study
to point out subperiods of enhanced coherence in runoff in the two regions. The
extended records provided by the reconstructions strengthened the statistical evi-
dence for a greater-than-chance joint occurrence of extreme low flows. Joint drought
was also the subject of the previously mentioned study for the Salt River Project
(SRP) (Hirschboeck et al. 2005 ) . The focus in that study was the potential for
drastically reduced runoff simultaneously in the upper Colorado River basin and
Salt-Verde River system of Arizona. The motivation was SRP's dependence on
imported Colorado River water from the Central Arizona Project (CAP) to buffer
against reduced runoff on the Salt and Verde Rivers in times of drought. Results
emphasized that severe drought years and clusters of drought years are strongly
coherent across the two river systems, and suggest that drought-induced shortages
in local Arizona water supply are unlikely to be offset by excessive runoff in the
upper Colorado.
8.3.2 Probabilistic Interpretation of Streamflow Reconstructions:
Example for the Colorado River
Streamflow reconstructions are statistical estimates of what the flow might have
been in any given year, and the uncertainty associated with these estimates is usually
displayed in the form of error bars. The user of the reconstruction can directly refer