
Background
In 2006, David Meko (University of Arizona Laboratory of TreeRing
Research) developed an extended reconstruction of the Colorado
River at Lees Ferry, from 7622005, for the California Department
of Water Resources. This reconstruction was made possible by
the collection and development of new millenniallength treering
chronologies in the Colorado River basin that incorporated extensive
remnant (dead) wood. These chronologies were developed by two
groups: one led by Meko, and the other by Connie Woodhouse and
Jeff Lukas.
This ~1200year reconstruction for Lees Ferry is very similarduring
the 14901995 period of overlapto the ~500year
reconstruction for Lees Ferry developed by Woodhouse, Gray,
and Meko, since they share a few predictor chronologies. (A
detailed history of reconstructions for the Colorado River basin
is provided in Colorado River Streamflow
– A Paleo Perspective. Charts which compare the Meko reconstruction
with the Woodhouse reconstruction and other Lees Ferry reconstructions
can be found on this page.)
The outstanding feature of the extended Lees Ferry reconstruction
is a sustained dry period in the 1100s, during the Medieval
period, thus the paper describing the reconstruction (Meko et
al. 2007) focuses on this drought.
Reference:
Meko, D.M., Woodhouse, C.A., Baisan, C.A., Knight, T., Lukas,
J.J., Hughes, M.K., and Salzer, M.W. 2007. Medieval Drought
in the Upper Colorado River Basin. Geophysical Research Letters
34, L10705.

Metadata
Observed Record
Location: Colorado River at Lees Ferry,
AZ
USGS Gage: 09380000
Source: US Bureau of Reclamation
Adjustment: "Natural flow"
record extensively adjusted to account for depletions 
Observed
Flows
Period: 19062004
Mean flow: 
15.025 MAF 
Median flow: 
14.850 MAF 
Minimum: 
5.516 MAF 
Maximum: 
25.231 MAF 

Reconstructed
Flows
Period: 7622005
Mean flow: 
14.655 MAF 
Median flow: 
14.827 MAF 
Minimum: 
2.322 MAF 
Maximum: 
24.311 MAF 


Calibration
& Verification
Methods
The Meko Lees Ferry reconstruction employed a novel and complex
reconstruction method. A complete description of the methodology
is provided in Meko et al. (2007), and also in the text
data file for the reconstruction archived at the NOAA's
World Data Center for Paleoclimatology. In brief: 11 residual
treering chronologies were regenerated from the original
ringwidth data to meet uniform standards for minimum series
length and other metrics. Then each chronology was separately
regressed against observed flow, resulting in 11 singlesite
reconstructions of flow. Then these reconstructions were grouped
by their timeperiod coverage into four subsets, so that the
entire 7622005 period was optimally covered. Then within
each subset, a principal components analysis (PCA) was run
on the covariance matrix of the singlesite reconstructions,
and the leading principal component (modes of variability)
was retained and used for the models. These PC1s were then
used as predictors in regressions with observed flow, ultimately
generating four subperiod reconstructions of flow. The final
reconstruction is composed of the flow values from the subset
reconstruction that has the lowest root mean square error
(RMSE) of the subset reconstructions (1,2,3, or 4) which are
available for a given year. Subset reconstruction (1) provides
the values for 7621181 and 2003, (2) provides 11821164,
(3) provides 13652002, and (4) provides 20042005.
Subperiod reconstruction 
Explained Variance
(R2adj)

Reduction of Error
(RE) 
Root Mean Square
Error (RMSE) 
1 (7621181, 2003) 
0.60 
0.58 
2.804 MAF 
2 (11821364) 
0.74 
0.73 
2.256 MAF 
3 (13652002) 
0.77 
0.76 
2.142 MAF 
4 (20042005) 
0.57 
0.54 
2.942 MAF 
Note: Because the final reconstruction over the calibration
period with the observed record (19062004) is derived from
three different models, the R2 shown in the scatterplot (Figure
1) is not identical to the R2 figures given in the table above.
Also, the R2 figures in the table above are adjusted for the
number of predictors in each model.
(For explanations of these statistics, see this
document (PDF), and also the Reconstruction
Case Study page.)

Figure 1. Scatter plot of observed
and reconstructed Colorado River annual flow, 19062004.
Figure 2. Observed (black) and
reconstructed (blue) annual Colorado River annual flow, 19062004.
The observed mean is illustrated by the dashed line. 
LongTerm
Reconstruction
Figure 3. Reconstructed annual
flow for the Colorado River flow (7622005) is shown in blue.
Observed flow is shown in gray and the longterm reconstructed
mean is shown by the dashed line.
Figure 4. The 10year running
mean (plotted on final year) of reconstructed Colorado River
flow, 7622005. Reconstructed values are shown in blue and observed
values are shown in gray. The longterm reconstructed mean is
shown by the dashed line.
^ return to top of page

