LTRR-SRP II

THE CURRENT DROUGHTIN CONTEXT:

A TREE-RING BASED EVALUATION OF WATER SUPPLY VARIABILITY FOR THE

SALT-VERDE RIVER BASIN

Dave Meko & Katie HirschboeckUniversity of Arizona - Laboratory of Tree-Ring Research

April 28, 2008Salt River Project, Phoenix, AZ

To update the tree-ring reconstructions of annual streamflow of the Salt-Verde-Tonto Basin through the period of the most recent drought and place it into a long-term, historical context linked to climatic variability

MAIN OBJECTIVE

MAIN PROJECT ACTIVITIES

1. UPDATING TREE-RING CHRONOLOGIES – Field collections and laboratory analysis to develop chronologies in the Salt-Verde basin with data through growth year 2005

2. NEW STREAMFLOW RECONSTRUCTION – Analysis of the new tree-ring chronologies to place the most recent drought in a long-term context

3. EW-LW EVALUATION – Exploration of the seasonal precipitation signal in separate measurement of earlywood and latewood width measurements

4. CLIMATIC ANALYSES – Synoptic dendro-climatology studies of observed record to better interpret the reconstructed record

TREE-RING CHRONOLOGY

UPDATING

Tree-Ring Collections

Douglas-fir at Wahl Knoll site, White Mountains , AZ

Tree-Ring Collections

• Collections at 14 Sites in Fall 2005

• Species:

Douglas-fir ponderosa pine pinyon pine

• Some re-collections, some new collections

• Cores only

-113 -112 -111 -110 -109 -108

33

34

35

36

37Utah Colo

Ariz NM

Little Colorado R.

Colorado R.

Salt R.

Gila R.

Verde R.

Ton

to C

k.

80 km

1

23

4

5

6

7

8

910

11

12

13

14

18

Tree-Ring Sites

Black RiverPine Site

Robinson Mt Site

Link to previous LTRR-SRP I study on joint drought (LL HH) in Salt-Verde and Upper Colorado Basins

Tree Ring Widths – the Basic Data

Site 2 - Black River Pine Core 13B

1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750

HH Years LL Year

Narrow rings in dry years, wide rings in wet years

STREAMFLOWRECONSTRUCTION

PROCESS

-113 -112 -111 -110 -109 -108

33

34

35

36

37Utah Colo

Ariz NM

Little Colorado R.

Colorado R.

Salt R.

Gila R.

Verde R.

Ton

to C

k.

80 km

1

23

4

5

6

7

8

910

11

12

13

14

15

1617

18

19

20

21 22

23

24

25

Tree-Ring Sites

Overview of the Reconstruction Process

Tree Ring Network

Statistical Calibration: regression

Reconstruction Models

Time Series of Reconstructed Streamflow

Observed Streamflow

1920 1930 1940 1950 1960 1970 1980 1990 20000

2

4 Salt+Verde+Tonto, 1914-2007

Year

Flow

(maf

)

1400 1500 1600 1700 1800 1900 2000

100

200

300

400

Year

Flo

w (pct of norm

al)

Reconstructed annual flows, SVT

0.6 0.8 1 1.2 1.4 1.6

1

2

3

y (m

af)

Index

Site2

0.4 0.6 0.8 1 1.2 1.40

2

4

y (m

af)

Index

Site 1

0.6 0.8 1 1.2 1.4

1

2

3

y (m

af)

Index

Site 10

-112 -110 -10832

33

34

35

36

37

R2=0.53RMSE

cv=574.4 kaf

MAE=248.0 kaf

1330-1989 (4 sites)

-112 -110 -10832

33

34

35

36

37

R2=0.69RMSE

cv=466.7 kaf

MAE=179.9 kaf

1451-1982 (10 sites)

-112 -110 -10832

33

34

35

36

37

R2=0.49RMSE

cv=621.5 kaf

MAE=280.6 kaf

1736-2005 (10 sites)

Three Different Models Used(based on different sub-periods)

Tree-ring sites not in model

Tree-ring sites in model*

• Tree-ring sites have variable time coverage

• Uniform time coverage required for a model

Sub-period reconstructions ultimately blended into final time series of reconstructed streamflow

RESULTS OF THE NEW

RECONSTRUCTION

Annual Reconstructed Flows, 1330-2005

2002 and 1996 have the lowest reconstructed annual flows in the entire record (28% and 30% of normal* respectively)

•Maximum number of consecutive years below normal = 5 (in 1590s and 1660s)

•Longest stretch of consecutive years below normal in recent interval of 1914-2005 is 4 years (in 1950s)

Plotted as % of normal* *normal = defined as 1914-2006 median of observed flows

1400 1500 1600 1700 1800 1900 2000

100

200

300

400

Year

Flo

w (

pct

of n

orm

al)

Reconstructed annual flows, SVT

Median

2002 baseline

5-yr dry spells

Close up of cores from two different trees at Site 10, located near Flagstaff:

// = false ring bands

“Missing” Rings (locally absent on tree where cored)

This core is missing the year 2002 (27 of the 30

trees at this site had no 2002 ring)

This core has a very narrow 2002 “micro-

ring”

(only 3 trees at this site had a

2002 ring)

Missing-Ring Percentage Through Time

2002 was unprecedented for frequency of missing rings

How unusual is such a high % of missing rings?

1400 1500 1600 1700 1800 1900 20000

20

40

60

80

Rin

g M

issi

ng (

%) Missing Rings 2002

1400 1500 1600 1700 1800 1900 20000

200

400

600

800

Year

Num

ber

of C

ores Sample Size

2002

Variations in Time-Averaged Flows

1400 1500 1600 1700 1800 1900 2000

50

100

150

200

Ending year of 6-yr period

Flo

w (

pct

of n

orm

al)

Reconstructed, SVT, 6-yr running mean

80% CI

1999-2004 Baseline

Plotted as % of normal* *normal =median of all 6-year running means

• 14 distinct prior occurrences of flow as low as 1999-2004 average

• 1- 3 occurrences in each century

• Most severe conditions at ~1590 and ~1670

Variations in Length of Intervals Between High Flow / Wet Years

High flows and large floods can occur during periodsof drought and low flows

1950 &1951 1953-1956 narrow rings

1952 wide ring

Wide rings can occur within otherwise

narrow-ring sequences

Floods / High Flows & Reconstructed Flows

High flow / flood “wet years” are tracked reasonably well by Verde River tree-ring reconstruction

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

1925 1935 1945 1955 1965 1975 1985 1995 2005

Inst

anta

neou

s P

eak

Flo

w

(cm

s)

0

50

100

150

200

250

300

Mea

n A

nnua

l Flo

w (c

ms)

Instantaneous Peak Discharge

Observed Annual flow

Reconstructed Annual Flow

Note vertically exaggerated scale on this axisVerde River Basin Comparison:

Observed, Reconstructed, & Instantaneous Peak Flows

Length of Intervals Between Wet YearsBased on Observed Flows, 1914-2007

• Interval longer in 1950s than during recent drought period• If not for mildly wet 1952, the earlier interval would have

been 25 years• Median interval is 2 years in the observed record

1920 1930 1940 1950 1960 1970 1980 1990 2000

500

1000

1500

2000

2500

3000

3500

4000Widest gap = 13 yr

Water Year

Flo

w (

kaf)

Recent gap = 9 yr

“Wet Year” =flow above 75th percentile

Length of Intervals Between Wet Years Based on Reconstructed Flows, 1330-2005

Longest interval = 22 years (1382-1403)

Recent interval = 12 years (1993-2004)

1950s interval = 12 years (1953-1964)

10 intervals ≥ 12 years Median interval is 3 years

1400 1500 1600 1700 1800 1900 2000

500

1000

1500

2000

2500

3000

3500

4000Widest gap = 22 yr

Water Year

Flo

w (

kaf)

Recent gap 12 yr

“Wet Year” = flow above 75th percentile

1400 1500 1600 1700 1800 1900 2000

500

1000

1500

2000

2500

3000

3500

4000Widest Pap = 22 yr

Water Year

Flo

w (

kaf)

Gage = SVTN

t=676

Nw

=141

Time Series of Reconstructed Flows, 1330-2005, with Wet Years Flagged

0 1000 2000 3000 4000 50000

0.2

0.4

0.6

0.8

1

Flow (kaf)

F(x

)

CDF of Reconstructed Flows

0.75 quantile 1914-2005: 1710 kaf 1330-2005: 1596 kaf

1914-2005

1330-2005

0 5 10 15 20 250

0.2

0.4

0.6

0.8

1

Gap (yr)

F(x

)

CDF of Gap Between Wet Years

------------------------------------Period M

edia

n0.

90q

Max

1914-20051330-2005

23

89

1222

1914-2005

1330-2005

EARLYWOOD-LATEWOOD EVALUATION

Earlywood / Latewood Evaluation

Ring width can be partitioned into parts formed early and late in the growth year

Studies have shown some success at inferring summer rainfall variations

from latewood width

-113 -112 -111 -110 -109 -108

33

34

35

36

37Utah Colo

Ariz NM

Little Colorado R.

Colorado R.

Salt R.Gila R.

Verde R

.

Ton

to C

k

80 km

Latewood Precipitation Signal – Wahl Knoll Example

2 4 6 8 10 12

ONDJFMAMJJAS

Number of Months

End

ing

Mon

th

0

0.05

0.1

0.15

0.2

0.25

0.3

Signal Strength by Seasonal Grouping

Prism Precipitation

Wahl Knoll Latewood Index

1897-2005

Input-OutputModel

• Signal confined mainly to summer

• Maximum of ~30% of variance of July or July+August precipitation explained by modeling

Testing for Latewood Signal of Summer Rainfall

• Total width had signal for annual precipitation, but no signal for summer precipitation

• Latewood width had a weak but significant signal for summer precipitation

SUMMARY: Results encouraging, but summer precipitation signal in partial ring widths is too weak to expect useful reconstruction of summer monsoon variability from this limited site coverage

THE CLIMATIC CONTEXT OF RECENT DROUGHTS& HIGH FLOW EPISODES

The “Big Picture” Global Climate Context

•Recent drought: mean NH temperatures near record highs

•1950s drought: mean NH temperatures near middle of long-term warming trend

•Wet late 1970s to early 1980s: mean NH temperatures higher

•Wet period 1915-20: mean NH temperatures low

(not shown here) severe tree-ring drought of 1899-1904: mean NH temperatures very low

1880 1900 1920 1940 1960 1980 2000

-2

0

2

T

(de

g F

) NH Mean Annual Temperature

1880 1900 1920 1940 1960 1980 2000

10152025

Flo

w (

maf

) Annual FlowsColorado RLees Ferry

1880 1900 1920 1940 1960 1980 2000

2

4

Flo

w (

kaf) Annual Flows

Salt+Verde+Tonto

NH Temperature Data from NASA/GISS; data are departures from 1951-80 mean based on GHCN met stations

Upper Colorado Flowsnatural flows for Lees Ferry from USBR

Salt-Verde-Tonto Flows from USGS

Horizontal lines for flows are at medians

Link to LTRR-SRP- I

Project

Updated LL and HH years exhibit

anomaly patterns similar to those of the earlier study

700 mb composites of new LL and HH years for Dec-Feb

Low FlowYears in Both SVT & UCRB

High FlowYears in Both SVT & UCRB

Floods / High Flows & Reconstructed Flows

High flow / flood “wet years” are tracked reasonably well by Verde River tree-ring reconstruction

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

1925 1935 1945 1955 1965 1975 1985 1995 2005

Inst

anta

neou

s P

eak

Flo

w

(cm

s)

0

50

100

150

200

250

300

Mea

n A

nnua

l Flo

w (c

ms)

Instantaneous Peak Discharge

Observed Annual flow

Reconstructed Annual Flow

Note vertically exaggerated scale on this axisVerde River Basin Comparison:

Observed, Reconstructed, & Instantaneous Peak Flows

0

500

1000

1500

2000

2500

3000

3500

4000

4500

1925 1935 1945 1955 1965 1975 1985 1995 2005

Ins

tan

tan

eo

us

Pe

ak

Flo

w

(cm

s)

Verde River Basin Instantaneous Peak Flows Classified by Synoptic Type

Water Year

Synoptic Convective Tropical Storm

Jan 1993

Mar 1938

Feb 1995

Dec /Jan 2005

Winters of 1978-80

Dec /Jan 1952

Feb 1927 TS NormaSep 1970

1950s convective dominance

Analyzing the reconstruction synoptically:

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

1925 1935 1945 1955 1965 1975 1985 1995 2005

0

50

100

150

200

250

300

Instantaneous Peak Discharge

Observed Annual flow

Reconstructed Annual FlowSynoptic Convective Tropical Storm

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

1925 1935 1945 1955 1965 1975 1985 1995 2005

0

50

100

150

200

250

300

Instantaneous Peak Discharge

Observed Annual flow

Reconstructed Annual Flow

Both reconstructed & observed annual flows track the magnitude of

the instantaneous peaks best during SYNOPTIC

(winter) events

Analyzing the reconstructionsynoptically:

“Recent Drought” Low flow years in

SVT Dec – Feb

1950s Drought Low flow years in

SVT Dec – Feb

“Recent” High flow years in

SVT Dec – Feb

Synoptic Circulation Patterns for SVT

1950s pattern vs. “Recent Drought” pattern Recent High Flow Year pattern

Verde Basin study: tree-ring record is a good indicator of winter storm

track activity

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

1925 1935 1945 1955 1965 1975 1985 1995 2005

Inst

anta

neo

us

Pea

k F

low

(c

ms)

0

50

100

150

200

250

300

Mea

n A

nn

ual

Flo

w (c

ms)

Instantaneous Peak Discharge

Observed Annual flow

Reconstructed Annual Flow

Water Year

Atmospheric Circulation Anomaly Patterns (700 mb)associated with Dry and Wet Intervalsin the Verde Basin

“Current” Drought

1950s+Drought

1978-1980

1996 & 2002

1993

1953-1964

Synoptic Atmospheric Circulation Patterns Linked to Dry and Wet Intervals

Importance of BLOCKING

circulation anomaly patterns:

Blocking leads to the PERSISTENCE of circulation features that produce EXTREMES

drought /

Are streamflow variations cyclic?

Spectrum, 1330-2005 Reconstruction

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

200

400

600

800

1000

1200

1400

BW

Frequency (yr-1)

Rel

ativ

e V

aria

nce

Spectrum

AR(1) Null Continuum95% CI

ENSO footprint?

SUMMARY & CONCLUSIONS

Reconstruction Model Summary

• Ring widths of the new collections have a strong annual runoff signal

•Subset models blended together yield a streamflow reconstruction covering 1330-2005

• The reconstruction explains 49- 69% of the variance of the annual flows

Extreme Single-Year Summary

• The reconstructed 1996 value was the 2nd lowest reconstructed flow since 1330

• The reconstructed 2002 value was the LOWEST reconstructed flow since 1330

• From tree’s perspective 2002 was a year like no other: 60% of 300+ cores were missing the 2002 ring!

CONCLUSIONS

1) Single-year intensity: drought in recent years unsurpassed in long-term tree-ring record (i.e., 1996, 2002)

2) Multi-year intensity: 14 distinct prior occurrences of flow as low as 1999-2004 average

CONCLUSIONS

3) Several intervals between “drought relieving” wet years were longer than any observed in the instrumental record

4) Winter storm track position key factor in drought signature (1950s vs. recent drought)

Final Report will be posted at:

http://fp.arizona.edu/kkh/srp2.htm