Geoscience Reference
In-Depth Information
In the initial meeting, paleoscientists learned of the key questions Denver Water
hoped to address with tree-ring reconstructions and their concerns regarding the use
of proxy streamflow data from tree rings. In turn, they endeavored to convey the sci-
ence behind the reconstructions and the limitations of the proxy data. In subsequent
meetings, a number of issues surfaced. These included: the availability and reliabil-
ity of estimated natural flows, which would be used to calibrate models; the problem
of disaggregating reconstructed annual values into the daily values needed for water
system model input; and the ability of the tree-ring reconstructions to match the
extreme low flows, critical for drought assessment and planning.
Some specific challenges emerged. To begin with, Denver Water wanted to know
precisely how well the drought years of 2000-2002, and 2002 in particular, could be
reconstructed with tree rings. The extreme low-flow year of 2002 was treated as a
test case for assessing the skill of the tree-ring reconstructions. A related challenge
was to evaluate the replication of other droughts in the twentieth century by the tree-
ring reconstructions, and to explore techniques to better match these extremes. With
regard to the uncertainty in the estimated flows, although it is known that recon-
structed values more closely match the gauge values in very dry years compared to
Water wanted a demonstration of this in order to more closely assess the accuracy
of the dry year reconstructed values. Finally, besides estimates of past water supply,
Denver Water was interested in the possibility of estimating demand, using an index
based on water usage, primarily in summer.
After preliminary proof-of-concept reconstructions were generated, work was
begun to update western Colorado tree-ring chronologies to include the 2002 ring,
and then to develop updated water year streamflow reconstructions through 2002
for Denver Water's three upper Colorado River basin gauges. Results indicated that
tree growth in 2002 very well reflected 2002 low flows, and subsequent recon-
structions provided estimates that closely matched the values of the 2000-2002
flows. After updating and expanding tree-ring collections for the South Platte River
basin and generating streamflow reconstructions for those gauges through 2002,
reconstructions resulted that were of similar high quality, especially in the replica-
tion of the 1950s and recent droughts. Along with the reconstructions, work was
begun to explore approaches to refine estimates of low flow and to better describe
the uncertainty related to the tree-ring-based estimates of flow. One approach was
the development of an 'ensemble' reconstruction, in which reconstruction models
were calibrated on numerous subsets of calibration years (Webb and Woodhouse
2003). When the ensemble members were plotted (30-40 members for the South
Platte River), it was possible to see years in which all solutions converged (typ-
ically the driest years) and years in which the estimates were more variable. In
several cases early in the gauge record, tree-ring estimates for a particular year all
indicated dry conditions while the gauge value indicated an average or wet year.
Denver Water personnel suggested errors in the estimation of the natural flows
could be the cause of the mismatch in those years, something the paleoscientists
had not considered. The range of scenarios presented by this ensemble approach
provided a means to explore uncertainties related to statistical modeling and the