What river basins and streamflow gages were
reconstructed?
Streamflow reconstructions were completed for the following 8 rivers
/gages, listed with their period of observed record. See the
Project Map for streamflow
gaging sites, tree-ring sites, and climatic stations that were used
in the project.
A. Colorado at Lees Ferry 1906-1995
B. Salt + Verde + Tonto 1914-2002
C. Gila at head of Safford Valley 1915-2002
D. Green at Green River, UT 1906-1995
E. Colorado near Cisco, UT 1906-1995
F. San Juan near Bluff, UT 1906-1995
G. Salt + Tonto 1914-2002
H. Verde 1914-2002
ï
-
Why should there be a relationship between streamflow and tree
growth in the arid and semi-arid western United States?
Tree growth can be limited if there is not sufficient moisture in
the soil. This can arise from either a lack of precipitation, or
excessive evapotranspiration in warm, dry climates-- or both.
Regionally, precipitation minus evaporation determines the amount of
runoff available in rivers and streams. Hence both trees and streams
can respond to weather and climate patterns that produce sustained
drought condition. In the project study areas, the snow that falls
in the winter prior to a tree's growing season is reflected in the
tree's growth in spring and summer via soil moisture storage.
ï
-
How good are the tree-ring based reconstructions of flow when
compared to the gaged record?
The reconstructions track the timing of high and low flow years
quite well and do a good job of capturing the magnitudes of the
flow, especially in the dry years. The reconstruction models explain
77.4 % of the variance for the Colorado at Lees Ferry (r = 0.88) and
57.8 % of the variance for the Salt-Verde-Tonto (r = 0.76). The
magnitudes of extreme wet years are reconstructed more accurately in
the large Colorado basin than in the Salt-Verde-Tonto basin, in part
because of the latter smaller basin’s “flashier” streamflow.
ï
-
Are the gaged and reconstructed records comparable in terms of mean?
In both basins, the mean of the observed period is higher than the
longterm reconstructed mean. This indicates that, in general, the
20th century has been wetter in both basins than in previous
centuries. (The recent drought years of the late 1990s and early
2000’s were not included in this analysis, however, and their
inclusion would lower the observed-record mean.)
ï
-
What criteria were used to define extreme streamflow episodes in
each basin?
Extreme episodes were based on
thresholds derived
from quantiles of water-year annual discharge (compared to the
mean). High flow years (H) are those with flow > 0.75 quantile
and Low flow years (L) are those with flow < 0.25 quantile. At a
later stage of the analysis, thresholds of > 0.90 and < 0.10 were
also examined in order to assess the most extreme years.
ï
-
What types of synchronous extreme streamflow sceanrios have occurred
in both the observed and reconstructed record and how frequently did
each type occur?
Four synchronous scenarios are possible when comparing the Upper
Colorado and the Salt-Verde-Tonto River systems and are indicated by
a two-letter code. (For consistency, the first letter will always
refer to the Upper Colorado and the second to the Salt-Verde-Tonto):
When the two basins are responding in the same
direction:
HH = High flow (H) in the
Colorado at the same time as high flow (H) in the Salt-Verde-Tonto
LL = Low flow (L) in the
Colorado at the same time as low flow (L) in the Salt-Verde-Tonto
When the two basins are responding in the opposite
direction:
HL
= High flow (H) in the Colorado at the same time as low flow (L) in
the Salt-Verde-Tonto
LH
= Low flow (L) in the Colorado at the same time as high flow (H) in
the Salt-Verde-Tonto
In both the observed and reconstructed records,
HH and LL events were
much more frequent than HL
and HL
events, especially in the long, 444-year reconstructed time series.
In fact, no HL
events at all occurred in the reconstructed record, and only 2
LH
events occurred. In the observed record, only 3
HL
events and no LH
events occurred. In order to examine some LH-like
scenarios in the observed record, the Colorado threshold was relaxed
to < 0.50, yielding LH
events. Due in part to the quantile method, the number of
LL events tends to be
counterbalanced by the number of HH
events, but overall – in both the observed and reconstructed records
– LL events are more frequent
occurrences than HH events. A
year-by-year listing of the HH,
LL, HL
and LH
events, color coded by each year’s streamflow threshold (>0.90, >
0.75, < 0.25, < 0.10) is presented in Appendix 5 of the
Final Report.
ï
-
What are the implications of the project results?
A widely held working hypothesis is that the Upper
Colorado River Basin (UCRB) can serve as a buffer to compensate for
extreme low flow in the Salt-Verde-Tonto Basin during drought
periods. One reason this is plausible is because of the size of the
UCRB basin and the overall magnitude of its discharge. It has also
been assumed that streamflows in the two river systems are
relatively independent of each other due to a difference in the
climatic regimes influencing each basin.
While the above presumptions sound logical, our analysis did not
substantiate them. We found that the simple correlation between the
annual flow values of the two 444-year reconstructed time series was
r = .599 ( r
2 = 0.359), which is a statistically significant
correlation, given the large sample size. Far more compelling with
respect to extreme events, however, was the overwhelming dominance
of HH and LL events in comparison to extremes where the basins
responded in the opposite direction.
These results suggest that annual streamflow variability in the
Salt-Verde Basin – especially extreme streamflow – is not
independent of annual streamflow variability in the Colorado Basin.
Therefore it is not unlikely that severe drought in one basin will
be accompanied by severe drought in the other basin, such as has
occurred during the western drought scenario of the late 1990s and
early 2000’s. It should be noted that the immense water supply of
the large Upper Colorado basin may allow it to continue to serve as
a buffer even when it is in a low flow scenario because a little
flow can go a long way in a smaller watersheds such as the
Salt-Verde-Tonto, however ever-increasing demand on the water supply
of the UCRB by many stakeholders is likely to reduce the size of
this buffer in the future.
ï
- Is the
observed frequency of occurrence of HH and LL years greater than can
be expected by chance?
What’s the
probability of different lengths of LL and HH sequences ?
The probability of the
observed numbers of joint HH and LL years is less than one in a
billion if the extreme streamflow in the two basins were indeed
independent of one another!
For different lengths of LL and HH sequences, there is > 10% chance
that a single year could be an extreme LL year or an extreme HH
year. There is also a 10% chance that 2 LL years (or HH years) could
occur in any given 4-year period and a > 5 % chance that 2 years out
of a moving 3-year window will be extreme LL years (or HH years).
ï
-
What are the implications of the observed tendency for extreme years
to occur in sequences or clusters?
It is possible that if the number of wet extreme years is about
equal to the number of dry extreme years, that the two extremes
could “cancel each other” on a year-to-year basis such that there
would be little long-term stress on water supply operations.
However, because of the clustering tendency of extreme events, it is
more probable that episodes of sustained drought or sustained high
flow will persist, placing more of a burden on water systems
management and operations. In other words, reservoir storage can
buffer water supplies in the Upper Colorado River Basin and
Salt-Verde-Tonto region, but supplies will be increasingly strained as droughts extend over multi-year
periods.
ï
-
How can the probability of joint multi-year drought occurrence in
the two basins be assessed?
The results of the Monte Carlo analysis are summarized in this
Table,
which lists for each length of moving average the number of LL
events in the reconstructed flows and threshold number of LL events
with empirical non-exceedance probability 0.50, 0.95, 0.99 and 0.999
by chance. For example, for the 7-year moving average, the Lees
Ferry reconstruction and SVT were simultaneously below their drought
thresholds in 62 years, which is far more than expected by chance:
the Monte Carlo tally indicates the probability is of fewer than 50
LL events if the flows on the Colorado and the SVT are unrelated.
The expected number of chance events, given by the probability point
is only 27. In summary, none of the observed LL counts (column
labeled ) has a greater than 0.001 probability of occurring by
chance alone. Since there were only 1000 simulations done for the
analysis, it is also true no simulation had a count of LL events as
large as for the reconstructed series themselves.
The results strongly support a linkage in multi-year drought
occurrence in the two basins.
ï
-
How have the numbers of LL and HH years varied from
century-to-century?
APPENDIX 5 (pdf) display a century-by-century record
of extreme streamflow events in the Upper Colorado and the
Salt-Verde-Tonto River basins. The 19th century experienced more
LL years than any other in the
record (22 events) and the 18th and 20th centuries (prior to 1965)
experienced the most HH years
(14 events).
|
Century |
# LL Years |
# HH Years |
|
1520-1599 |
11 |
11 |
|
1600s |
11 |
8 |
|
1700s |
13 |
14 |
|
1800s |
22 |
11 |
|
1900 -1964 |
8 |
14 |
ï
-
What atmospheric circulation patterns lead to the LL, HH, LH and HL
scenarios?
Seasonal composites of upper level (500 mb geopotential height)
circulation anomalies for each type of extreme event scenario (based
on the observed record) are shown in these two
FIGURES. The
characteristic circulation pattern for LL events is
higher-than-normal upper level pressure over the west in early
winter (Oct -Dec) and over the North Pacific ocean storm track
region in mid- to late winter (Jan - Mar). The inverse of this
pattern leads to HH events. LH and HL scenarios arise when the
Pacific storm track appears to shift to an anomalous poleward (HL)
or equatorward (LH) location.
ï
-
Are there any linkages of these anomaly patterns to climate-scale
driving mechanisms such as sea surface temperature anomalies, El
Niño, La Niña, etc.?
Preliminary examination of El Niño, La Niña influences and ocean
indices such as the Pacific Decadal Oscillation (PDO), and the
Atlantic Multidecadal Oscillation (AMO) suggest linkage to some –
but not all LL years in the observed record (see
FIGURES). A
University of Arizona M.S. thesis is in preparation to examine this
in more detail.
ï
-
Is the recent drought in the Salt-Verde Basin unprecedented in
severity?
The annual flows of the basin decreased beginning in water-year 1994
and culminated in single-year flows for 2000 and 2002 lower than
any previously experienced in the observed record (FIGURE,
top left). The 5-year, 11-year and 15-year running means reached
their recent low points in the periods ending with water year 2004.
As a 5-year running mean, the recent drought is about as severe
as the lowest-flow period in the 1950s. The same is true of the
11-year running mean, suggesting that the period commencing with the
decline in water year 1994 and continuing through water year 2004
ranks with the driest conditions in the entire gaged record. When
the running mean is extended to 15 years (FIGURE,
lower right), the recent drought no longer ranks among the most
severe; the reason for this is the wet sequence of years in the
early 1990s enters the moving average. But up to an averaging period
of 11 years, it appears the recent drought is at least comparable in
severity to any earlier drought in the gaged record.
The tree-ring reconstruction for the Salt+Verde+Tonto (SVT) ends in
1988, and so does not cover the recent drought. Nevertheless that
reconstruction does sample the 1950s drought, and because the 1950s
drought was characterized by flow departure of roughly the same
magnitude as the recent drought, we can use the lowest reconstructed
flows of the 1950s to indirectly evaluate the relative severity of
the recent drought in the context of the reconstruction to A.D.
1199.
A plot of 11-year running means of the SVT reconstruction with the
baseline marked as the low point in the 1950s suggests that the
current drought was exceeded in severity several times in the past
800 years (FIGURE). Eight
distinct periods before the start of the gaged record show lower
11-year mean flow than the lowest reconstructed value of the 1950s.
The most severe of the tree-ring droughts was in the late
1500s, during the well-documented “mega-drought” of North
America, when 11-year average flow is reconstructed about 100 cfs
below the lowest flows of the 1950s.
BOTTOM LINE: The recent drought, while severe, does not
appear to be unprecedented when viewed in a multi-century context.
ï
|
|