Application of Streamflow Reconstruction to Water Resources Management - Dendroclimatology - page 238

early example is the routing of Stockton and Jacoby's ( 1976 ) reconstructed flows
through the Colorado River simulation model (Harding et al. 1995 ; Tarboton 1995 ) .
The model simulates operations specific to the Colorado River, including water
allocation, reservoir operations, evaporation, hydropower generation, salinity, flood
control releases, and legal and institutional constraints tied to the Law of the River
(Harding et al. 1995 ) . For a worst-case scenario in that study, the reconstructed flows
for the most severe multiyear drought (1579-1600) were rearranged in decreasing
order (lowest-flow year last) and input into the simulation model. Results indicated
that such a drought has an estimated return period of perhaps 2000-10,000 years,
and would result in Lake Powell being drawn down to dead level storage.
Two recent applications of tree-ring data in river models have utilized the time
persistence properties of the tree-ring data in combination with magnitudes of flow
from the observed flows. Prairie ( 2006 ) , citing reconstructions of Colorado River
flow illustrated in Woodhouse et al. ( 2006 ) , judged that the tree-ring information on
the hydrologic state (wet or dry) is very reliable, but that the magnitudes of recon-
structed flows are too uncertain to justify their use in water management modeling.
He applied a two-stage process to come up with realistic simulations of flow that
took advantage of the perceived strengths of the observed and reconstructed data.
First, the reconstructed annual flows were used in a Markov chain model to generate
the hydrologic state. Second, sequences of annual flows were generated by non-
parametric bootstrapping of the observed flows conditioned on the hydrologic state.
The simulations of annual flow were then spatially and temporally disaggregated
into monthly inputs to a basin-wide decision model. He applied this method using
as input the Colorado River reconstructions of flow at Lees Ferry, Arizona, from
Woodhouse et al. ( 2006 ) . The example effectively demonstrated how annual stream-
flow reconstructions can help determine risk and reliability of various components
of a water resources system.
In the second application drawing on the persistence properties of tree-ring data,
Shamir et al. ( 2007 ) demonstrated how tree-ring information can be incorporated
into a system of hydrologic modeling modules for water-supply risk assessment
in an arid region where the water supply comes primarily from relatively shallow
aquifers (micro-basins) along an ephemeral stream. The method was developed for
the Santa Cruz River, just north of the US-Mexico border in Arizona. The hydrologic
modules included (1) stochastic generation of hourly precipitation, (2) transfor-
mation of the precipitation into daily streamflow, and (3) surface-groundwater
interaction to account for alluvial groundwater recharge. Winter wetness categories
(wet, medium, and dry) were used in the precipitation module. One scenario utilized
a 319-year tree-ring reconstruction of winter precipitation to establish wetness cat-
egories to guide the Monte Carlo precipitation sampling and the generation of daily
flows. Results of the exercise suggested the risk of low water levels in the alluvial
aquifers is greater than indicated by the relatively short instrumental precipitation
and streamflow record.
The more typical direct application of streamflow reconstructions has been to
place statistics of the gauged flow record in a long-term context. Most papers on
streamflow reconstructions include tables summarizing the most extreme conditions