Arkansas Water Resources Center

LONG-TERM RECONSTRUCTION AND ANALYSIS OF WHITE

RIVER STREAMFLOW

Research Project Technical Completion Report

Project G-1409-02

Authors

M.K. Cleaveland, D.W. Stahle, J.G. Hehr, Department of Geography, University of Arkansas, Fayetteville, Arkansas 72701

Publication No. PUB-135

June 1991

Arkansas Water Resources Center 112 Ozark Hall

University of Arkansas Fayetteville, Arkansas 72701

LONG-TERM RECOr~STRUCTION AND ANALYSIS

OF WHITE RIVER STREAMFLOW

M. K. Cleaveland, D. W. Stahle, J. G. HehrDepartment of GeographyUniversity of ArkansasFayetteville, AR 72701

Research Project Technical Completion Report

Project G-1409-02

The research on which this report is based was financed in part bythe United States Department of the Interior as authorized by theWater Research and Development Act of 1978 (P.L. 95-467).

Arkansas Water Resources Research CenterUniversity of Arkansas

113 Ozark HallFayetteville, Arkansas 72701

Publication No. 135

June, 1988

Contents of this publication do not necessarily reflect the views andpolicies of the U.S. Department of the Interior, nor does mention oftrade names or commercial products constitute their endorsement or re-commendation for use by the U.S. Government.

The University of Arkansas, in compliance with federal and state lawsand the regulations governing affirmative action and nondiscrimination,does not discriminate in the recruitment, admission and employment ofstudents, faculty and staff in the operation of any of its educationalprograms and activities as defined by law. Accordingly, nothing inthis publication should be viewed as directly or indirectly expressingany limitation, specification or discrimination as to race, religion,color or national origin; or to handicap, age, sex, or status as adisabled Vietnam-era veteran, except as provided by law. Inquiries con-cerning this policy may be directed to the Affirmative Action Officer.

ABSTRACT

LONG- TERM RECONSTRUCTION AND Ar~AL YSIS

OF WHITE RIVER STREAMFLOW

A 281-year reconstruction of White River annual runoff at Claren-don, Arkansas, was developed from a regional average of nine Oklahoma,Missouri, and Arkansas tree-ring chronologies (six post oak, Quercusstellata, and three baldcypress, Taxodium distichum). Inhomogen-e-i-tyof the gaged series was detected with both double mass analysis (usingstate average total annual Arkansas precipitation) and regression(using the regional tree-ring average). Simple regression calibratedthe homogeneous runoff data with the average ring width data from 1930to 1980. Comparing the reconstruction with independent data verifiedthe regression model. Variance of the reconstruction increases signif-icantly during the 20th century, a change that may be caused by climaticshifts or by anthropogenic disturbances in the watershed. Years of sur-plus and deficit runoff are non-randomly distributed in both gaged andreconstructed series. This non-randomness appears to be caused by asignificant tendency for inter-annual persistence of runoff extremes,which may provide a basis for improvement of probabilistic forecasts ofWhite River runoff.

Malcolm K. Cleaveland, David W. Stahle and John G. Hehr

Completion Report to the U.S. Department of the Interior, GeologicalSurvey, Reston, VA, June, 1988.

Climate/Planning/Dendrochronology/Stochastic Hydrology/Paleohydrology/Time Series Analysis/Rainfall-Runoff Pro-cesses/Rivers/Drought/White River/Arkansas/Missouri

Keywords --

TABLE OF CONTENTS

Page

iAbstract.

List of Figures

List of Tables iv

Acknowledgements v

1Introduction

A. Purpose and Objectives 2

3B. StudyArea

5c. Related Research and Activities.

7MethodsandProcedures .

7A. GagingData

13

B. Tree-Ring Data

14c. Calibration and Verification

20Principal Findings and Significance

A. Reconstructed White River Runoff: 201700 to 1980

24B. Analysis of Surplus and Deficit Runoff.

Conclusions 28

30LiteratureCited

34Appendix 1

40Appendix 2

i i

LIST OF FIGURES

Page

Locations of the tree-ring chronologies used inreconstruction of White River Basin annual runoffat Clarendon, Arkanas, (triangle). The six postoak chronologies (circles) are 1) Little MariesRiver, MO, 2) Hahatonka, MO, 3) Democrat Ridge,MO, 4) Neosho River, OK, 5) Roaring River, MO,6) Clayton Ridge, MO, and the three baldcypresschronologies (squares) are 7) Allred Lake, MO,8) Egypt, AR and 9) Black Swamp, AR. The fourlargest impoundments (that affect this study) inthe White River Basin (dashed line) are (A) Beaver,(B) Table Rock, (C) Bull Shoals and (D) Norfolk. .

Figure 1.

4

Figure 2. Double mass analysis of White River cumulative an-nual mean daily discharge at Clarendon, Arkansas,vs. cumulative estimated discharge (see text).Annotations on the graph show the years of largestdischarge (gaged data) and major reservoir closureswith impoundment capacity. 10

Figure 3. Reconstructed (solid line) and observed (dashedline) annual runoff (Jan.-Dec.) of the White Riverat Clarendon, Arkansas, for the calibration andverification periods. The solid horizontal lineis the 1900-1980 gaged mean, and the two horizontaldashed lines are the surplus and deficit runoffthresholds (120 percent and 80 percent of thegaged mean, respectively). 12

Figure 4. Reconstructed annual runoff plotted with a low-pass filter that removes variance at frequenciesof less than eight years (Fritts, 1976). The solidhorizontal line is the 1700-1980 reconstructedmean, and the two horizontal dashed lines are thethresholds used in the analysis of surplus anddeficit runoff (120 percent and 80 percent of themean, respectively .21

iii

ACKNOWLEDGEMENTS

We acknowledge the encouragement and support of Leslie Mack,

Director, Arkansas Water Resources Research Center, statistical assist-

ance by Colm O'Cinneide of the Mathematics Department, and helpful dis-

cussions \,[ith T. J. Blasing, Oak Ridge National Laboratory, and E. R.

Cook, Lamont-Doherty Geological Observatory. Samuel Lehr, Jr., Memphis

District Corps of Engineers, provided the Clarendon hydrological data.

Technical assistance was provided by R. Berg, C. Corke, S. Huntsman, D.

Puckett and K. Thompson. Funded at the University of Arkansas by U.S.

Geological Survey State Water Research Institute Program (Contract No.

14-08-0001-G1409). The National Science Foundation, Climate Dynamics

Program supported development of the tree-ring data used in this study

(grant numbers ATM-8006964, ATM-8120615, ATM-8412912 and ATM-8612343).

v

.INTRODUCTION

The demand for high quality surface and groundwater supplies by

and municipal interests has increased nation-agricultural, industrial

wide, in some cases exceeding existing supplies CU. S. Water Resources

Council, 1978J. The Southcentral United States is experiencing rapid

growth in population and water demand, and available supplies may soon

become inadequate in or near the arid southern Plains or in areas of

intensively irrigated agriculture such as the Grand Prairie of eastern

Arkansas CU. S. Water Resources Council, 1978; Bryant et al., 1985J

Surface water supplies in the Southcentral United States are subject to

substantial interannual variability due to natural fluctuations in cli-

In fact, the Arkansas-White-Red and the Texas-Gulf water resourcemate.

regions CU. S. Water Resources Council, 1978J have been identified as

having two of the three most variable runoff regimes in the 19 subdivi-

sions of the continental United States [Stockton and Boggess, 1979J

Arkansas is particularly well endowed with high quality surface

water resources, and proposals for interbasin transfers both within and

Consideration is being givenfrom Arkansas have generated controversy.

to the transfer of surface water from the White River to augment dwind-

ling groundwater supplies in the Grand Prairie of eastern Arkansas [U.S.

Army Corps of Engineers, n.d.]., where water-intensive rice and soybean

The possible transfer ofsas Agricultural Statistics Service, 1988].

"surplus" water from Arkansas to supplement supplies in Texas has also

been investigated [U. S. Army Corps of Engineers, 1982] and the Dal1.as-

1

.Ft. Worth water supply system will extend eastward to Lamar County,

only 8G km from Arkansas (Dallas Water Utilities, 1986).

Apart from the many economic and environmental questions concern-

ing possible interbasin transfers of surplus water, there is uncertainty

about the long-term availability of surplus flow regimes";in Arkansas.

The probable discontinuous nature of surplus flows would impose serious

planning and design constraints on the possible use of this water re-

Because the gaged runoff data is limited to the pastsource component.

century in Arkansas, a thorough investigation of the history and depend-

ability of surplus flows is probably not possible solely on the basis of

the historic record [Rodriguez-Iturbe, 1969J

Purpose and ObjectivesA.

correlated with hydrometeoro-Proxy tree-ring data are often wel

logical variables such as precipitation and temperature, and can there-

fore be useful for developing long-term estimates of specific hydro-

Tree-ring data are particularlylogical variables such as runoff.

suited to the analysis of drought or low flow characteristics because

moisture stress is a fundamental growth-limiting factor which can be

During yearsfaithfully reproduced in properly selected ring width data.

of abundant precipitation, multiple factors such as temperature, competi-

tion, or soil fertility may limit growth in individual trees, usually

creating greater standard errors in the ring width indices derived for

For this reason tree-ring reconstructionsthose years [Fritts, 1976J.

of very wet years usually involve greater error and should be interpreted

cautiously [e.g., Blasing et al., 1988J.

In this paper we use a network of moisture sensitive post oaks

2

'I 1500 50 100

ARKANSAS Km

L---

Locations of the tree-ring chronologies used in reconstructionof White River Basin annual runoff at Clarendon, Arkansas,(triangle). The six post oak chronologies (circles) are 1)Little Maries River, MO, 2) Hahatonka, MO, 3) Democrat Ridge,MO, 4) Neosho River, OK, 5) Roaring River, MO, 6) ClaytonRidge, MO, and the three baldcypress chronologies (squares)are 7) A1lred Lake, MO, 8) Egypt, AR and 9) Black Swamp, AR.The four largest impoundments (that affect this study) in theWhite River Basin (dashed line) are (A) Beaver, (B) Table Rock(C) Bull Shoals and (D) Norfork.

Figure 1.

4

Clarendon gage is far enough above the confluence of the Arkansas and

Mississippi Rivers to be largely unaffected by fluvial damming from

either river CU. S. Geological Survey, 1984J

Intensive logging of the upland oak-hickory and pine forests

occurred during the early twentieth century. These logging operations

and land clearing for agriculture may have affected the discharge, sus-

pended sediments, or bed load of the White River, at least temporarily

Four large-scale impoundments forduring the initial wave of clearing.

flood control, power generation, water supply, and recreation purposes

were constructed in the basin from 1943 to 1980 CU. S. Geological Sur-

1984], and these projects have promoted both the economic develop-

of the central Ozarks and agricultural productivity along the

The volume of high quality surface water stored inlower White River.

these reservoirs is certainly one of the most important resources in

the Ozarks, but the present and future management of these supplies re-

main subject to a conflicting array of public and private pressures.

Related Research and Activities

Properly developed tree-ring chronologies are particulary well

uited as surrogate runoff records because of great age (some species

exceed 1000 years), absolute dating, annual to seasonal resolution,

sensitivity of tree growth to climatic variables that also influence

runoff~ and the wide availability of tree-ring data in the specific

drainage system of interest [Fritts, 1976; Stockton and Boggess, 1980J.

A number of previous studies have employed surrogate or proxy data such

The mostas tree rings to extend relatively short streamflow records.

5

known tree-ring reconstruction of runoff was for theimportant and wel

Colorado River, reported by Stockton (1975). The reconstructed long-

term mean annual runoff in the Colorado River was only about 13.5 maf

between 1564 and 1962, some 2.0 maf year-1 less than the amount allo-

cated in the Colorado River Compact of 1922 [Stockton and Jacoby, 1976]

provide a classic illustration of both the need to consult proxy data

when confronted with short, potentially biased gaging records and the

potential practical importance of tree-ring data.

Other hydrological applications of tree-ring data include a re-

construction of summer streamflow in the Occoquan River, Virginia, which

indicated that critical low flows were more frequent in the reconstructed

data prior to the period of instrumental records [Phipps, 1983J. Cook

and Jacoby [1983J reconstructed summer low flows in the Potomac River,

and for similar reasons concluded that the gaged discharge measurements

for the Potomac are not entirely representative of the last 250 years.

Jones et al. [1984] have demonstrated the hydrological application of

tree-ring data in the British Isles, while Stockton and Fritts [1973J

and Brinkmann [1987J have used tree-ring chronologies to reconstruct

past lake levels

The use of proxy data to investigate long streamflow series in the

Southcentral United States has been limited to early tree-ring studies

by Hawley [1937J in Tennessee. No quantitative estimates of past runoff

have been reported in the White River Basin, although Guyette [1981J has

6

demonstrated significant correlation (r = 0.75) between growth of

eastern red cedar (Juniperus virginiana) and June minimum discharge of

the Gasconade, James, and Current rivers in southern Missouri

vidual red cedar up to 700 years old have been reported from the Ozark

Plateau [Guyette, 1981; Guyette et al., 1980], and hold great promise

as long proxy hydrological series.

METHODS AND PROCEDURES

Gaging DataA.

We selected the U.S. Army Corps of Engineers (CaE) gaging record

at Clarendon for reconstruction because it is the longest in Arkansas

it is believed to provide a reasonable integration of the surface water

supply in the entire White River Basin, the gage has never been moved

and homogeneity of the record does not appear to have been seriously

Clarendon dischargeaffected by post-war reservoir development.

was not available from 1921 to 1930, but a single rating table to con-

vert gage height to discharge for those years was supplied by the Memphis

communication, 1987J. The ratingDistrict CaE [So A. Lehr, Jr., persona

low water) Measurements in 1917--table has the notation "Based on L.W.

High Water 1927"

Correlations between monthly, seasonalized, and annual mean daily

discharge and the regional tree-ring chronologies [Stahle et al., 1985bJ

were used initially to determine which chronologies should be used and

what fraction of the year might be most successfully reconstructed.

Seasonal mean daily discharge for February through July produced the

,

but annual mean daily dischargehighest correlation (r = 0.64, P < 0.001

7

0 1 2 3 4 54

CUM. EST. DISCHARGE (/10

763 -1

m Sec )

Figure 2. Double mass analysis of White River cumulative annual meandaily discharge at Clarendon, Arkansas, vs. cumulativeestimated discharge (see text). Annotations on the graphshow the years of largest discharge (gaged data) and majorreservoir closures with impoundment capacity.

10

L1J

~f- .<I>

c:

L--0

000

-L-E

+

L-

UU

Jc:

:J0

-0'"-<

1>0

III ~

V!<I>

I.. <

I>0>

U

V!

<I>

+

c:c:I..

III <

I>

ro<

I>

I.. <

I><

I>

I-<I>

~.c

I ...

+

-H"""

-0...(;)0<

I>

"C.

Co

iO

>

0 V

! <

I>

-<1>

:2:III

I-

<I>

c: .

.c III

+

-0a-

lii

U~

<

I>~

U

C:

c: 0

-0 <

1>-

c: L-+

ro <

I> ro

+

...<1>

Co

-I..<

I>

(;)1..U

0

I.. +

u<

I>

V!

+

-c:

-.Dc:

<I>

0

.c -

-t- +ro

V!

--<

I>L-

rr'i 5

~

U

IIIU--0+

0

III -

-L-+

<

I>ro

a--t-V

!

c:0+rou...~-rotI..<

I>>"C~

r1ic:

c:0

0--+

+

ro

ro ro

-,I..

-0.D

N

ro-~roU

-00I..<I>

a-

~0\.- 0)II!C\0a.0~ co1"'\.0+***-;to\0.0 ***0\0.0..,.01.Il"\N

I

com.NI!) 0a:>0\~INIn0\~***N\D.0If)0\~Ia1'"\0\~

~~~

0)UIc0\0. 010.0+***0<0.0

~

~IC"\

0\~I0Il\0\~ ***co,....0-~.0--I

10-.t.IC'I

-.taro0\~IN~0\~ ***r-t').0

0>UI

C-0\.0 co0.0+**00~.0 **""N ,0-N0-~I000-~ **0'1~.0***0It).a -.tL!"\.01-.t

0~.'"NI

**~N.0

.c0

0\(X

) N

0\ (J\

~

~

I I

0 0

"" 0

0\ 0\

~

~ ~0-.~N..,.It)""'

11

.11 .

0 CO

0

0\ .C

O..-0

11 0\

+- --

, C

O

I.c

Q)O

\O+

- L

-1"\-;:,

0\E

-,..-

V>

-uro

;:, -,:,

-,:, Ll-O

C

C

Q)

roro

-J.0\

L 11

-,:, N

Q)

0 cO

\a..C

O

ro-ro

III 0\

IL

Q)

-C

00

-0 0

LJ L

,-,:, 0\

Q)

Q)

L-E

III

-0 0

L <

!>

-,:, -,:,

L LJ

IIIQ

) Q

) 0

-,:,Q

) ul-+

- 0

L C

C

-'I-

Q)-':'

ro L

L C

U

Q

)'I-

Q)

ro -a.

0 'I-

'I-'1---

C111-

Q)

C

Q)

Q)

-,:, Q

) O

)Q)

Q)

+--

~L+

-V>

III+

-0)

IIILJ Q

)Q

) L

1II.a-,:,

-III -

'I- C

III

'I- ro

L +

-0

C

Q)

Q)

CQ

) E

.a Q

)III

Q)

E-

III ~

C

;:,

U0

+-

Q)

c--Q

) Q

) 'I-

.a ~

Q

) 'I-

L. +

- >

Q

)01111

Q)

-0'1-00

.a +

- U

coco -

-,:,0\0\ Q

) 111'1-

Q)--

U

0 0

+-

C

a.III"

Q)

C;:,Q

)Q)

L.>

0-,--

Q)IlC

U

I-':'LL

'l-Nct:

-roLL

'l-CO

L

00 -0\

ro-,:,1-1-

-,:, a.

C

11 E

ro-,:,-,:, Q

) ,\0

0cc

.ccr-- U

-,:,roro +

-00\Q

) -,:,-

LL--

LL 0

roQ)Q

) .00'

'I-;:,

Q)

Q) 11

'I- <

!>

III0-+

-+-0

LJ+-

>IIIV

>V

>C

O+

- -t-

-LJLJO

\III+--

C+

- -Q

)-L 0

c+-+

- +

-;:,LI-Q

) C

c,

IIILJ -,:,

-Q)Q

)LCQ

) Q

)U

--Q)O

LU

+-

-UU

>-

-.;:,'I---O

+-E

+

a.'I-'I-'I-cro;:'III",EQ

)'I-'l-O>

E-O

OO

Q)Q

)(.)L-+-

.UU

OO

LJQ)+

-rooU

U

ilia.+-

IIIC

+

-.aOIllA

UO

CC

IIIO

--O

OQ

) Q

)LC-+

-+

t--,:,.c0

1IIro+

-+-

Q)+

-L -

-roroCL

L+-+

-Q

)--O)-IIIQ

)Cro

LQ)Q

)-ro- ro+

-LLLU

Ia. 'l-U

III0

L L

00 -

UO

O-,:,-':':I:

'l-CU

UQ

)Q)

C-O

-

Q)

--+-O

C-O

"'-O-C

C--U

-O)III-O

OO

a.OO

roroQ)+

--III... .

-IIIIII+-+

-.~U

III

Q)O

OO

O+

-LLIIQ);:,

L-ro

ro Q

) 0

L -,:,

+-

0) vi

vi vi

vi;:'Q

)Q)C

~

Q)O

Q)

:?:C-C

-01- Q

:ZQ

:C-C

-C-C

-

ro.aU-':'Q

) 'l-0).c..et***

* **

0 -T"'""'"""

1900 1920 1940

YEAR1960 1980

Figure 3. Reconstructed (solid line) and observed (dashed line)annual runoff (Jan.-Dec.) of the White River at Clarendon,Arkansas, for the calibration and verification periods.The solid horizontal line is the 1900-1980 gaged mean,and the two horizontal dashed lines are the surplus anddeficit runoff thresholds (120 percent and 80 percent ofthe gaged mean, respectively).

12

Tree-Ring DataB.Nine high quality tree-ring chronologies were selected from the

50 now available in the Southcentral United States [Stahle et al.

1985a, 1985bJ on the basis of proximity to the White River Basin,

tal length of record~ and correlation with White River discharge.

nine chronologies are based on two species, post oak and baldcypress,

from well-drained upland and poorly drained wetland habitats, respect-

ively, and both species exhibit strong sensitivity to drought during

and before the growing season [Stahle and Hehr, 1984; Stahle et al.,

The direct correlation between post oak growth and moisture1985a].

anomalies is consistent with the xeric nature of their upland sites,

and it has been known for more than half a century that the moisture

signal in tree growth can often be maximized by selecting native trees

from these well drained upland positions [Douglass, 1920]. The direct

correlation between the radial growth of swamp-grown baldcypress was

discovered more recently [Bowers, 1973; Stahle et al., 1985a, 1988J, and

extends the range of drought sensitive tree species into a distinctive

and widespread bottomland environment.

Each tree-ring chronology represents a mean value function of the

detrended ring width measurement series available for each year from

Chronology30 to 50 trees per site, usually with two radii per tree.

development started with the absolute crossdating of each radius [Stokes

and Smiley, 1968] and the measurement of each dated ring to 0.01 mm.

The series of annual ring width measurements were then detrended and

transformed to dimensionless indices using the ARSTAN program [Cook,

13

Holmes et al., 1986]. This procedure removes biological growth

trends related to increasing tree age [Fritts, 1976], and the flex-

ibility of the spline curves fitted to the measurement series was

strictly controlled to avoid removing more long-term variance than ab-

Low order serial correlation present in the annualsolutely necessary.

ring width series of most trees was largely removed from each tree-ring

chronology using autoregressive (AR) modeling procedures [Cook, 1985J

Finally, it was necessary to remove some remaining long-term variance

trend in the derived chronologies, which appears to be due, in part, to

changing chronology sample size and an age-related decline in growth

vigor of oaks [e.g., Stahle and Cleaveland, 1988; Blasing et al., 1988].

When the nine residual series were averaged, this regional average

series had weak serial correlation (r-l =-0.13), apparently due to re-

The aver-inforcement of weak persistence in the separate chronologies.

age was modeled as an AR(3) process to derive a serially random predictor

chronology for calibration [Meko, 1981J. Serial correlation in unmodeled

~-ring time series appears to arise primarily from biological factors

(e.g., food storage, crown area, root area) [Fritts, 1976], but some per-

To enhance reconstructionsistence may also be due to climatic forcing.

fidelity in the frequency domain, the autoregressive properties of the

Clarendon runoff series were added to the serially random tree-ring based

estimates to complete the reconstruction [r4eko, 1981J (see below).

Calibration and Verificationc.

An empirical approach was used to identify the best predictor var-

iables and time interval to calibrate the tree-ring and annual runoff

14

series. The tree-ring and runoff data were both prewhitened prior to

calibration, and marginally significant first-order serial correlation

= 0.22, P < 0.10) was modeled as an AR(l process and removed from

the runoff series. Principal components analysis [Cooley and Lohnes,

1971] of the nine chronology network was performed and the amplitude

series of the first two eigenvectors (with eigenvalues> 1.0, that re-

tain 44.6 percent and 16.5 percent of the variance in the tree-ring

data set, respectively) were entered into stepwise multiple regression

with annual runoff from 1930 to 1980 [Draper and Smith, 1981J. This

approach explained less variance in the gaged data than bivariate re-

gression between the gaged runoff series and an average of the nine

tree-ring chronologies. In addition, loadings on the second eigenvector

all negative for post oak chronologies and positive for baldcypress

chronologies, suggesting that physiological or ecological differences

unrelated to hydrometerological conditions may be involved in the tree-

ring variance associated with the second eigenvector. For these reasons,

the regional average of the nine chronologies was used to reconstruct

annual runoff.

In an attempt to further assess the homogeneity of the gaged data,

and to select the most reliable subperiod for calibration, the tree-ring

data and the annual runoff data were entered into linear regression for

four subperiods, 1900-1929,1930-1951,1952-1980, and 1930-1980. These

subperiods were selected in light of the apparent inhomogeneity in the

Clarendon data before 1930, and the possible impact of Bull Shoals and

other large impoundments on the tree growth -runoff relationship after

15

1951. The coefficients and statistics computed in these four regres-

sian analyses are listed in Table 1, and the tree-ring data explain

the most runoff variance for the period from 1930 to 1951 and the least

from 1900 to 1929. The regression slope and intercept

Table

1) com-

puted during these two subperiods are significantly different (P < 0.05;

Steel and Torrie, 1980; SAS Institute Inc., 1985) which, with double

Also,mass analysis, suggests that the gaged series is heterogeneous.

regression results indicate that post-war reservoir construction in

basin has perturbed the natural relationship between climate and runoff.

The regression model from the 1930 to 1980 period explains 50 percent of

the annual runoff variance and was used to derive the transfer function

to reconstruct runoff from 1700 to 1980 for the following reasons

there is no statistical difference between the regression coefficients

calculated for 1930 to 1951 and 1952 to 1980; ii) serious inhomogeneity

is not apparent in the discharge data after 1930; (iii) although the ex-

plained variance is maximized from 1930 to 1951, a sample size of only

22 years may not be adequate to insure a stable regression relationship

It should also be noted that calibrations based on separate averages

of the upland post oak and bottomland cypress chronologies each explained

38 percent of the annual runoff variance from 1930 to 1980, twelve per-

cent less than was explained by an average of both species. This is con-

sistent with the assumption that runoff from the Ozark Plateau and Western

Lowlands is reflected primarily by the post oak and baldcypress chronolo-

gies, respectively, and that each region can contribute independently to

White River discharge measured at Clarendon.

16

The transfer function used to reconstruct White River annual run

off was

(2)

where Yt is reconstructed runoff for year t in km3 year-l and Xt is the

regional average of the nine tree-ring chronologies for year t. The

-1standard errors of the slope and intercept are 6.95 and 6.98 km3 year

respectively. The AR(l) persistence model determined for the gaged

runoff series (AR(l) coefficient = 0.22) was then added to the estimated

series. The addition of persistence changes the reconstructed mean

slightly, so the reconstruction was adjusted to maintain the equality

of the observed and reconstructed means during the calibration interval

(1930-1980).

To evaluate the accuracy and stability of the reconstruction, the

subperiod calibrations based on 1930 to 1951 and 1952 to 1980 (with co-

efficients that are not statistically different from the 1930 to 1980

transfer function) were also used to reconstruct annual flow during the

alternate period for which statistically independent gaged runoff data

is available (1952-1980 and 1930-1951, respectively). Several statist-

ical comparisons were made between the gaged and reconstructed runoff

values during the verification periods (Table 1). The correlations and

first difference correlations are both strongly positive and highly

significant (Table 1

,

demonstrating strong covariance of observed and

reconstructed series outside the period in which regression forces an

optimum relationship. The sign tests [Conover, 1980] indicate signif-

icant skill in reconstructing the direction of departures from the mean

17

(P<0.10) and paired !-tests [Steel and Torrie, 1980] reveal no signif-

icant difference between the average of reconstructed and observed run-

off (Table 1).

The final verification test used was the reduction of error (RE)

statistic which compares the actual and estimated runoff during the

verification period with the actual mean runoff during the calibration

interval, and is a measure of information gained by using the regression

estimates of runoff rather than simply the mean of runoff during the

calibration interval [Gordon, 1982; Blasing et al., 1988]. Values of

the RE theoretically range from -w to +1.0, and any positive value is

considered significant ( P < 0.05, N>10) [Gordon and LeDuc, 1981]. The

RE statistics calculated on the actual and reconstructed runoff data

The positive RE statistics indicate that the re-are +0.38 and +0.60.

construction is contributing unique paleohydrological information.

The reconstruction has also been compared with the independent

Tablegaged data from 1900 to 1929 that was not used in any calibration

1, Fig. 3). Although this early runoff data may be systematically biased

relative to the post-1929 data, it can still be useful for independent

All verification tests are passed,verification of the reconstruction.

although the correlations are lower than found for 1930-1951 and 1952-

1980, and the RE is low, but still positive (Table 1)

The descriptive statistics in Table 2 indicate that the reconstruc-

tion reproduces the mean and variance properties of the independent run-

off data reasonably well, but examination of Fig. 3 reveals specific

The reconstructed runoff serieslimitations of the regression estimates.

18

L-a>>~a>-t-.c:3:Q

).c-t-"f-0"f-"f-0c:::JL-

CO

::Jc:c:CO

"1;JQ

)-t-U::JL--t-IIIc:0Ua>L-

"1;Jc:C

O

"1;Ja>>L-a>III

.Q0"f-0IIIU-t-III .

-IIIL- C

Oa>

III-t-c:u

CO

CO

.Y.

L- L-

co~.cU

~

c:-0C

O"1;J

U

c:-a>-t-L-III

CO

-t-uC

O-t--t-U

>C

O

.NW~I-

"1:J

I~oc "0

I~0::l~ l~ "0

i~Ct:

ro+

-

.--IX)

-'" .--IX)

N-co ~\C)InLroQ)

>--0LQ

).0E::I;z:

coccoOJ

~ 0\N.10Nr-..0\.It)N IC

\IC

\.\0N ,...0-0

IC\

NNN.l-N

roc0-t-ro>Q)

"0"0Lro"0Cro-t-<

n

0\\0.,'"IX).0..- t"\

IX).IX)

..-\0.t--0\0\a.--

It!E:JEXIt!::;::

-10.~~ -\D.-.t-.t

0\\0.U"\

-q-

'"0.a)10 r-..I').U

'\U

'\

roE;:]Ec~ t'"\1t\.~~ ~I().I()("I.a..- ,...N.a~ ~\0.\0

(\J(\J

C<

II (\J

0) -

C

"1;:1(\J

<II

~

~

CX

)

0.ti)ti)

IX)

0.0\""IX

>r---.r---

It"I

0.It)"'3"

00.00I/)19

~.q-.10N 0\r-...\0N~«).IC

\N a)1"\.oq-N co'o;t.r---N

.0U

IC

~

0\'"

0..0\

0N ~1'"'\.0\

N

.0U

IC

~

N~

N

..a a

"""

.aUI

c~

t--

t-- -

..~

0~

.QInc:

* \0

ti"\ -..ti"\

0ti"\

c0+-roLro>"-0+-cQ)

U"-"-Q)

8

***aN.ac0-t-It!Q)

LL0UIt!

LQ)

(/)

NN.0 '"a.aI'"0. U"\

IX>.0 co~.0III!II!Q)

c~~(.f)

a)'"".0I ,..."".0

I0\Ii).0 ~\("\.0IIX

>~.InIn0+L;]~

U\I1Q

)>

- InQ)

>- IIIOJ

>- C/)

Q)

>-U)

Q)

>--roEL-

azca+-~.aL-

+-U)

c

~0.01\

c..,+-cI1J

.U

.I'+

-aL-

.0I1J C

<X

) .Q

)O>

OO

>-- .U

10V""e+

-vc...Y

.Ozc..**

I1J.DU

*

does not fully reproduce the extremes apparent in the gaged data, par-

ticularly positive extremes. The worst estimated annual runoff value

is 1927, which is the largest annual runoff amount ever measured in the

White River Basin. This indicates that the tree-ring estimates of the

magnitude of high runoff periods contain the greatest errors, probably

due largely to inability of trees to respond linearly to very wet con-

ditions [Fritts, 1976]. Fortunately, the point at which estimation er-

rors become large appears to be well above the surplus threshold set

at 120 percent of the mean (Fig. 3). This indicates that the recon-

struction should be useful for investigations of the history and timing

of surplus flows defined conservatively as less than 130 percent of the

long-term mean. Of course, the surplus issue also involves interest

in the absolute magnitude of surplus flows, but reconstruction errors

associated with the largest runoff amount Fig. 3) indicates that the

reconstruction should be interpreted cautiously in terms of the abso-

lute magnitude of surplus flows

PRINCIPAL FINDINGS AND SIGNIFICANCE

A.

The reconstructed annual runoff for the White River at Clarendon

from 1700 to 1980 is presented in Fig. 4 and Appendix 2. The descript-

ive statistics for the gaged and reconstructed series are presented in

Table 2. The variance and range statistics are highest for the gaged

data, illustrating the underestimation of runoff extremes by the recon-

struction. The skewness and kurtosis of the gaged runoff are also both

larger than for the reconstruction, bat both series approximate a normal

20

14-.cO

+.> V1COV1QJ+.>

or- S- CQJ .cQJUQJ+.>UCC S-CO or- QJ QJ

or- c 0-S- +.>co..- a>COQJOO

+.>S-V1CCO"OQJO C>NV1CO0 or- QJES-C+.>Q) 0 or- Cs-.c..-QJ

U+.>"O"OS-CO or- QJ QJ

..c: c0-+.>0V1

V1I1jOS- "ONQJQJ -

+.>.c ..-I-CO°r-+.>4-4- c4-

000V1- N CV1 \0 or- ~COr"'-.S-S-0-0'\0

c:+.>3: or-0 "OU

..-V1 3: or-+.>+.>4-

CO+.> Q)or- QJ"o

..c:s-.c+.>U-+.>"O°r-- C3: "oco

V1C"OS-I1jV1Q)CO ~+.> QJ +.>~CO-0 11jS-

~+.>QJ~o-.c E V1

C"I4- or- "0 4-4-QJQJO0 +.>CCUV1 0~ CO ~ or- -S-..c:S-V1>,+.> +.> >,..-..-V1..-Q)COV1CCO>~ V1 0 C or-CQJUCO+.>C..-QJ UCO S-Q)QJ

4- .cO-"OOO+'>V1Q) 00 QJ+'>V1O'\CS-U QJ -or-~ or- I ..S-UO"OC+.>COQJCOV1QJr"'-.V1QJc~-~E00-UQJQJ"oQJQJS-.c c:~4-+.>0+.>

a:<tUJ>

.-:to

QJ!-~0)"-U-

21

The gaged and reconstructed mean runoff for the perioddistribution.

1930 to 1980 are less than the reconstructed long-term mean from 1100 to

1980,

but these differences are not statistically significant [Steel

and Torrie, 1980].

Examination of Fig. 4 suggests a long-term trend in annual runoff

from 1800 to 1900, but there is no significant linear trend from 1700

There does appear to be a substantial increase in runoffto 1980.

variance around 1900, which is statistically confirmed by an F-test on

the ratio of reconstructed variances from 1700 to 1899 and 1900 to 1988

(P<0.05) [Steel and Torrie, 1980J. The four lowest, and two of the

four highest reconstructed annual runoff values occur in the twentieth

Assuming that the ratio of actual to reconstructed runoffcentury.

variance is time stable, the White River appears to have experienced

more variable runoff during the twentieth century than over the preced-

ing 200 years. This apparent change in runoff variability may be due

to an actual climate change [e.g., Kutzbach, 1970], may reflect anthro-

pogenic disturbances to the remnant old-growth forests sampled, or may

reflect large scale anthropogenic disturbances to the watershed (e.g

regional land clearing, acid rain deposition, or CO2 fertilization).

Efforts to detrend the variance of the tree-ring time series could also

cause an increase in twentieth century variance [Blasing, et al., 1988]

but our variance detrending was cautious and is probably not responsible.

The filtered reconstruction (Fig. 4) suggests that prolonged (5-

low and high runoff departures tend to alternate in an os-to la-year

cillatory manner, but these oscillations are too irregular for direct

22

tially (suggested by Fig. 3), and should provide some insight into the

secular variability of surplus flows in the White River system.

If the intervals between surplus flows are randomly distributed,

they should approximate an exponential distribution, and this hypothesis

can be tested with the Lilliefors criterion [Conover, 1980]. The dis-

tribution of intervals between surplus years (~ 120 percent of the mean)

the test of randomness for gaged runoff from 1900 to 1980 (P<0.05)

and for reconstructed runoff from 1700 to 1980 (P<O.Ol The recon-

structed data also fail when tested from 1900 to 1980 (P<0.01 Inspec

tion of test results indicates that the gaged series fails Lilliefors

test primarily because surplus runoff events tend to cluster into suc-

cessive years (high incidence of one year intervals between surplus run-

Three consecutive years of surplus occur three times (1927-1929

1949-1951, and 1973-1975). In the reconstruction, four consecutive

years of surplus occur 1774-1777, 1891-1894), and six of nine years are

estimated to have had surplus runoff from 1774 to 1782.

The longest interval without surplus runoff in the gaged series was

11 years, occurring from 1957 to 1968, and ten consecutive years, from

1935 to 1945! were also recorded (Fig. 4). The reconstructed runoff

series indicates that prolonged periods of 27 years 1717 to 1744),20

years (1823 to 1843), and 11 years (1811 to 1822) without surplus runoff

have occurred in the White River Basin since 1700. The underestimation

of actual runoff amounts by the reconstruction is a potential problem to

a threshold analysis of surplus or deficit flow, but does not appear to

be a serious limitation to this study because (i estimation error

25

uted when compared to an exponential distribution with Lilliefors test

(P<O.Ol). The clustering of low runoff years also appears to explain

this non-randomness~ with several examples of successive drought years

lasting from two to five years in the period from 1700 to 1980. During

the driest periods the reconstruction indicates that deficit flows oc-

curred in as many as six of seven years from 1868 to 1874, and six of

10 years from 1785 to 1794. The recurrence of these historic dry peri-

ods over the White River Basin would no doubt place severe strain on

the highly developed surface water supply system, even though this sys-

tern has been designed and managed with severe short-term drought as a

primary consideration. On the positive side, the longest interval be-

tween deficit runoff was seven years in the observed data (1947 to 1954)

and 12 years in the reconstructed data prior to the twentieth century

1708 to 1720 and 1841 to 1853) (Fig. 4). Most of these periods without

deficit flow were characterized by a high incidence of surplus runoff.

The non-random interannual distribution of surplus and deficit run-

off events in the White River Basin appears to be a product of large

scale climatic variability. Some of this variability may eventually be

tied to more slowly changing boundary conditions of the atmosphere such

as sea surface temperatures, or the El Nino/Southern Oscillation. If

such associations can be demonstrated, they may permit some improvement

in long-term hydrological forecasts once a change in the related boundary

condition is detected. In lieu of a better understanding of the atmo-

spheric conditions se$ponsibleforextended periods of surplus or deficit

runoff in the White River Basin, we have attempted to identify statistical

27

data indicate that surplus and deficit flows are not randomly distrib-

uted through time. This implies that regimes of unusually low or high

flows can become established and persist for two or more years, as has

been witnessed during the twentieth century (e.g., low runoff was reo

corded for three consecutive years in 1900-1902, 1954-1956, and 1963-

1965; high runoff occurred three consecutive years in 1927-1929, 1949-

1951, and 1973-1975). Periods as long as 27 years without surplus run-

off occur in the reconstructed record. Non-random occurrence of sur-

plus and deficit runoff years may also imply a systematic component to

the atmospheric conditions that govern discharge in the White River and

elsewhere in the Southcentral United States (e.g., Stahle and Cleave-

land,1988). This interannual persistence of low and high runoff re-

gimes, and the presence of spectral peaks in the 14.0 to 18.67 year

period range, were both also detected in independent climate and tree-

ring data sets from Texas [Stahle and Cleaveland, 1988J and suggest a

large-scale macroclimate control. If the physical mechanism (or mech-

anisms) responsible for this apparent persistence of climate indices

and runoff in the Southcentral United States can be identified, it

could lead to improved forecasting of runoff extremes

This study demonstrates that tree-ring data can be useful for ap-

plied hydrological analysis in the Central United States, including

detection of inhomogeneity in gaging records. With the extensive net-

work of existing chronologies, and the development of longer red cedar

and baldcypress chronologies, there is considerable potential to extend

these hydrological applications in both time and space.

29

LITERATURE CITED

Arkansas Agricultural Statistics Service, Arkansas AgriculturalStatistics 1987, Ark. Agricultural Experiment Station, Fayetteville,AR, 1988.

Arkansas Legislature, Act 1051 of 1985, An Act to Establish a Mech-anism to Determine the Requirements of the Water Users of This State;And for Other Purposes, State of Arkansas 75th General Assembly, LittleRock, 1985.

Blasing, T.J., D. W. Stahle and D. N. Duvick, Tree-ring based recon-struction of annual precipitation in the South-Central United Statesfrom 1750 to 1980. ~~ Resour. ~., 69,163-171,1988.

Bowers, L. J., Tree-ring dating of the bald cypress (Taxodiumdistichum [L] Rich.) in the Lower Mississippi Valley, M.S. thesis,Ark. State Univ., Jonesboro, 1973.

Bryant, C. T., A. H. Ludwig and A. E. Morris, Ground-water problems inArkansas, ~ Resources Investigation Report 85-4010, U. S. GeologicalSurvey in cooperation with the Ark. Dept. of Pollution Control and Ecol-ogy and the Ark. Soil and Water Conservation Comm., Little Rock, 1985.

Burnash, R. J. C. and R. L. Ferral, Contamination in the hydrologicdata base, In~. Engineering Foundation~. .Q!!. Improved HydrologicForecasting-Why~~, Am. Soc. of Civil Engineers, New York, 1980.

Conover, W. J., Practical Nonparametric Statistics, Second Edition, JohnWiley, New York, 1980.

Cook, E. R., A time series analysis approach to tree-ring standardization,Ph.D. dissertation, Univ. of Arizona, Tucson, 1985.

Cook, E. R. and G. C. Jacoby, Potomac River streamflow since 1730 as re-constructed:.by tree rings, ~. ~. ~. Meteorol., ~, 1659-1672, 1983.

Cooley, W. W. and P. R. Lohnes, Mu1t~iate Data Analysis, John Wiley,New York, 1971.

Currie, R. G., Evidence for 18.6-year signal in temperature and droughtconditions in North America since A.D. 1800, J. Ggopbisical ~., ~,11055-11062,1981.

30

Currie, R. G., Periodic (18.6-year and cyclic (11-yeari induced droughtand flood in western North America, ~. Geophysical~., 890, 7215-7230,1984.

Dallas Water Utilities, Water Supply Sources (map), City of Dallas,Texas, 1986.

Douglass, A. E., Evidence of climatic effects in the annual rings oftrees, Ecology, 1,24-32,1920.

Draper, N. R. and H. Smith, Applied Regression ~nalysis, Second Edition,John Wiley, New York, 1981.

Gordon, G.A., Verification of dendroclimatic reconstructions, in Climate~~-~~, Cambridge Univ. Press, Cambridge, 1982.

Gordon, G. A. and S. K. LeDuc, Verification statistics for regressionmodels, in Preprints Seventh.f.2.!!f. .Q.!!. Probability~ S!at1stics1D-Atmospheric Sciences, Am. Meteorological Soc., Boston, 1981.

Guyette, R. P., Climatic patterns of the Ozarks as reconstructed fromtree-rings. M.S. thesis, Univ. of Missouri, Columbia, 1981.

Guyette, R. P., E. A. McGinnes, Jr., K, Evans and G. Probasco, A climatehistory of Boone County, Missouri, from tree-ring analysis of eastern

red cedar, ~ ~~, ~ 17-28, 1980.

Hawley, F. M., Relationship of southern cedar growth to precipitationand runoff, Ecology, ~, 398-405, 1937.

Holmes, R. L., R. K. Adams and H. C. Fritts, Tree-ring chronologies ofwestern North America: California, eastern Oregon and northern GreatBasin with procedures used in the chronology development work, includ-ing Users Manuals for Computer Programs COFECHA and ARSTAN, ChronologySeries VI, Lab. of Tree-Ring Res., Univ. of Ariz., Tucson, 1986.

IMSL Inc., ~ Library Reference Man~, Edition~, Vols. 1-4, IMSLInc., Houston, Texas, 1982.

Jenkins, G. M. and D. G. Watts, Spectral Analysis~lli Applications,Holden-Day, San Francisco, 1968.

Jones~ P. D.~ K. R. Briffa and J. R. Pilcher~ Riverflow reconstructionfrom tree rings in southern Britain~ l. Climat. ~, 461-472~ 1984.

Karl, T. R., L. K. Metcalf, M. L. Nicodemus and R. G. Quayle, Statewideaverage climatic history Arkansas 1891-1982, Historical Climatology

31

Series 6-1, Nat. Climatic Data Ctr., Ashville, NC, 1983.

Kohler, M. A., On the use of double-mass analysis for testing theconsistency of meterological records and for making required adjust-ments,~. ~. Meteorological~., 1Q, 188-189, 1949.

Kutzbach, J. E., Large-scale features of monthly mean Northern Hemi-sphere anomaly maps of sea level pressure, Monthlv Weather Rev., 98,708-716,1970. "~"V"'J- -

Meko, D. M., Applications of Box-Jenkins methods of time series anal-ysis to the reconstruction of drought from tree rings, Ph.D. disserta-tion, Univ. of Arizona, Tucson, 1981.

Phipps, R. L., Streamflow of the Occoquan River in Virginia as recon-structed from tree-ring series, ~~. ~., ~, 735-743, 1983.

...Rodriguez-Iturbe, I., Estimation of statistical parameters of annualriver flows, Watg~ Besources Res., ~, 1418-1421, 1969.

Statisti~s, Version .?:. Edition,SAS Institute Inc., SAS User's Guide:SAS Inst. Inc., Cary~C~-~.-

Stahle, D. W. and J. G. Hehr, Dendroclimatic relationships of post oakacross a precipitation gradient in the south-central United States,A!LQals Assoc. Am. Geographers, Ii, 561-573, 1984.

Stahle, D. W., M. K. Cleaveland and J. G. Hehr, A 450-year drought re-construction for Arkansas, United States, Nature, 16,530-532, 1985a.

Stahle, D. W., J. G. Hehr, G. G. Hawks, Jr., M. K. Cleaveland and J.R. Ba1dwin, ~-~ chronologies.fQ!'- ~ southcentral United States,Tree-Ring Lab. and Off. of the State Climatologist, Dept. of Geography,Univ. of Arkansas, Fayetteville, 1985b.

Stahle, D. W. and M. K. Cleaveland, Texas drought history reconstructedand analyzed from A.D. 1698 to 1980,~. Climate, 1,59-74, 1988.

Stahle, D. W., M. K. Cleaveland and J. G. Hehr, North Carolina climatechanges reconstructed from tree rings: A.D. 372 to 1985. Science,~,

1517-1519, 1988.

Stockton, C. W., Long-term streamflow records reconstructed from treerings, Papers Qf ~ Lab. Qf Iree-~~. No.5, Univ. of Ariz. Press,Tucson, 1975.

32

Stockton, C. W. and H. C. Fritts, Long-term reconstruction of waterlevel changes for Lake Athabasca by analysis of tree rings. WaterResour. Bull., 1, 1006-1027, 1973.

Stockton, C. W. and G. C. Jacoby, Long-term surface-water supply andstreamflow trends in the Upper Colorado River Basin, Lake Powell Res.~. ~.,~, 1976. --

Stockton, C. W. and W. R. Boggess, Geohydrological implications ofclimate change on water resource development, Contract Report DACW72-78-C-0031 , for U. S. Army Coastal Engineering Res. Center, FortBelvoir, Virginia, C. W. Stockton & Associates, Tucson, 1979.

Stockton, C. W. and W. R. Boggess, Augmentation of hydrologic recordsusing tree rings, in~. Engineering Foundation~. ~ ImprovedHydrologic Forecasting--~~~, Am. Soc. of Civ. Engs., New York,1980.

Stokes, M. A. and T. L. Smiley, ~_lntroduction.!Q ~-~ Dating,Univ. of Chicago Press, Chicago, 1968.

U. S. Army Corps of Engineers, Six-state High Plains-Ogallala Aquiferregional resources study: A Report to the U. S. Dept. of Commerce andthe High Plains Study Council, Summary Report, Water Transfer Element,Southwestern Division, Dallas, TX, 1982.

U. S. Army Corps of Engineers, Draft Arkansas State Water Plan for theUpper White River Basin, Little Rock District, Little Rock, 1988.

U. S. Army Corps of Engineers, The Eastern Arkansas Region comprehen-sive study, Memphis District, Memphis, TN. (due March 1990), n.d.

U. S. Geological Survey, Water resources data Arkansas water year 1983,~~ ~ Report AR-83-1, U.S.G.S. Water Resources District, LittleRock, 1984.

U. S. Water Resources Council, The Nation's Water Resources 1975-2000,Second National ~~ Assessmen~ ~ Q.. i~~r Resources Council,Vol. 1, Summary, U. S. Government Printing Office, Washington, D.C.,1978.

33

xa:zlJJ{l.{l.<

.~IUQ)

VI

1-'+-Q)

0)L-10

.cUIn-0>

-

100I::10Q

)::;:

10::JI::I::cC"0C10>

-.ci-I::.QL,In10InI::10.:.:L-cC,c0"0I::Q

)L-10

UQ)

.Dco~ ..-I

~ ~ ~ ~ ~~ ~ ~I

~ j~

~

aaaIt)N 00..,.~..,.

00rrINN 00-0\N 00--.;toN 00""~..- 00IC

\\0 aa0- aa0\\0 0010I!"I~ 00..-0\.-.,

aa0\0\..-000\- 00q-\0-

C)

C)

~t'"\N 00a)\0'" 00~0\'" 00IX

>t-.N 000\r--

00If\"""

00f"'."" aa1'"\1'"\

00roN 000\N 000r--

000\V~.-00\.- 00b0N

000t-~ aaro"Ot

'" 00InIn.- 00,...0..- 00.q-t'"\

00q-U

"\

00'"'" 00..-It"\

00It),...

00~0\-.t

aa-.t", ~N00\.--

00N\0 00a:>t--

00~0.-- 000\D000-.:t~ 00..-N""~ 00"fIf)\0 000\..".II')

aa~'"r- 00""r--

00It)10 00-It\ 00N~ 000CO

Ii)""001~ aa~""'o;t

00-~'" 00-0"" aa-Nco 00<'-I

0\tr\

00IC\

0t")

aa.-NN 00tt'\0 aar--10 000\q- 000\I') 00N~ 000Ii)Nq-00\- 000'""'

000\0IX)

00t--N\0 00<X

)0\N 00N,...0;1-

00N0N 00f"'".."..-- 0000\~ 00r-..-"" 00~""""00Nr---

00U"\

0\

00(X)

0~ 00""~-.t"

~00\~34

0000\~ aacoNN 00tt)..;r~ 00C

X)

CX

).-- 00~C

O~ 00Ir\f""..'" 00..-f'-..N aart\0\\0 00IC

"\\0'O

;f"

C>

C>

a>r-r- 00aJN~ 000NN1000\- aa\0.-\0

aaNa.-- 000<X

)

00It)0.-- 00~\0I")00~<X

>\C

\

00r--r--It) 000\~-.r

00NNN 00a:>~r- 00.-'""N 00\D\D 001"\N~to-.001- 00a:>r-.qo

00q-..-..- 00NIt) 000r--00It)l'"'-I'"'-

00NCO

r---

000\, N.--

aaIt)""'It) aaN\0~ 00, , 00r--.0N 000\0\NIX

)00\- 00-10N 00""~..;t

00If)If)N 000\0~ 00"I"\D aart"\'" 00U

)r-..

00~~..- 000(X)

N0010(X)

(X)

00CX

>r---~ 00\00~ aar-\0'""'

0\00\~ 000N 00t-O-

N

aa\0\0~ 00U'\

0\ 00IX)

0t'"I

00""0~ 000\\0 000\NN0C>

.--.tN

'\

00-'""" 00N0\N 00Q)--.t

00to~""0-01..- 000'" 00""N

00CO

VIC\

00CO

CO

000r- 0C>

\D0\"" aat')'".q-

00\t"\~~ aa\0Q

)

00NIl'\N 00It)\DNaaU

"\IX

)- 00'"\0N 0aoq-'""'IC

'\

~~~~ 00~U"\

~

00IX)

IX>

00\D0000\r-r 0aNN- 00Il"\\0rn 00\0U

'\N 000LC

\N 00~\D 00C

O\0 00.qo\0 00(X

)(X

)ti"\

00-.;t0\'"" C

>C

>1"\Ir\ti"\

N~0\~

NX0zW0..0..c(

~~.,..,

'f-'f-0C:J~10:JCC<~Q

)i-(J:JLi-V

Ic0(JQ

)~..V

I10V

IC10~L<..C0"0CQ

)L10

U ..--~ILroQ)

>-.-INQ)

.QcoI-

~I ~ ~

111

4 ~ 4

1

00r--

"':t1'"'1."':t

N NIC).0\N (X)

0'1.t")N "=

tr-...<X

>..- r-..N.I!\N NIC

).0\- ""'10.-""'

I!'\~.I!'\~ It)0.0"" IC

\I")..--N

""'Lf\.~""'

~0.NN 00).Il"I"" r---.\0I'f)

0\~.I!'\N ~\0.Y

0-

n r-..N.NI<

)

\0\().\0N 00\.~N0~,...

r--0.IX

)N

N.--.r---N tl'\N.aN 10~.0\N 0\t"-.\0N (X

)I!\.r---N ..-0\., N 00.It)~ Nl!).r-..N0N, ~ 1010.~.-- \0cr-.cr-.-

a.-.N.-'"'".0\N -.;tl().a:>C

\j

0~.0rt"\

0\N.\00-

cor---.0\N r-..

~.U\

~ r---.:t".a)C'4

.-IC\.N0'""r--

~ t")0.~N

It)It).t")~ ~N.NN C

'I\0.t'"\t'"\

~10.0.q-

Ii)N.aIi) ~.-.0\'""'

t'"\-.t.\0..- 1'\0\., ~C

>~r---~ I<

)0.Q

)N <X

)-.t.r---

40

~-.t.U"\

N \00.0N Q)

N.t--..- ~f"'I.NN a~.t'"\t'"\

t'"\N.t--N 0\\0.NI'")r-..0.tt)tt)

N.--.I")N .-~.NNa~r-..-

<X

)a.10N Nro-.ro-~ N\0. IX

)t"\.\DN ..-..-...-""'~~.o:tN 0\0.NN 0\tl"\.cotl"\

0\0r---- ""N.q-N \0~-.""t'"\

~~.\0~ ~.-.N~ r--';:t.U"\

N r-..\0.0""<

X>

Q).\{'\

.- N~.1'"1N ~~.Lr\tl"\

0"-

N.ti"\ti"\

ar-r-..~ LC

'I..-.l""-N r--.~.t'--N

r--~.<

X)

C'J

~~.~N 0\0.0\- r--;;t-.Q)

N t'").q-.NN \0\0.f'"\f'"\

coo;t.0NNr--.0\..- 1"'\N.to-..1"'\

0\ti").~ti")

0a)r-

0'1N.It")t') r--0.0)N roro.0N0\..-.f'--N ..-<X

).<X

)..- 00\.(X

)N \00.NI"')\0IX

).0\..- N0.0N ~\0.0\~00\r---

\01""\.coN\0<X

:).oq-

N 11"\\0.1"'\1"'\

""r--.0"" aIt"\.a)N 0\~.It)I<)

\D<X

).NNo,:t

tc).o,:t.- N\0.0.--

1"\N.C

X)

..-aaIX>

yo-

NQ).t'"\

~ Lr\\D.'"'i"

IX)

It'\.r""-oo-

""~.IX)

Nt'"\(X

).10..- N,...«)N IC

\-.:t.-.:tN r---N.r---N(X

)t--.t--N 0\\0.I!'\o;t

0~CX

)~

ro0\.roN \0~.0\Nr<)

co...-r<)

0~.IX>

N mIX).('.I

N oq-\0.aI") r-..0\.IrI~~0-.NN ('I0.0\('I

ro-I'").-.:tI'")

0N<X

>..-

Nt'"\.r--

\0.-.0rt"\

a~.LC\

N ~~.'""~ coco.0('oJ

a)0.I!)N \0a:>.~N 0'110.,...N t'"\oq-.~a1'"\co..-

..-It').a1'"\

NX0ZW0-0-ct:

.,-.0"-IL.roQ

)>

-

""'E~'f-'f-0C:J~ro:JcC~"0Q)

of-U:JL.of-1/1C0UQ

)~,1/1ro1/1cro~L.ct:,c0"0CQ

)L.roU..--INQ

)

.QroI-

11

~ rol

11

o:tl

~j ~

NIC\.0\- 1""\\D..--1""\

IC\

~.IC\

~ It)0.0N I!'\1""\.~"-,

(X)

0'1.""'N .q.t--.IX

)~ r N.If)N.".t'"\..".N N1("\.0\Na0r--..-

r--N.Nt'i

\0I!\.\0N 00\.IrI""'r---~.101<

\

~'".U'\

N ~\D.YO

-

I")tl'I~..-tl'I

":t-o.NN 00\.IrI1"1

a.-r-..- r--0.C

ON

N11").r---

N0\t"-o\0N IX

)IC

\.r--N ~0\.r--N a0.~~N~.r---N '"N.aN 10oq-.0\N\0\0.~..- \D0'\.0'\.-0Nr-..

<X

)r-..0'\N ~.1(\~ 0..-.N..- r-.~.coN ..-\t"\.,...N0~.0tt"\

0\N.\0y-IC\

IC\.0\N oq-

U\.C

X)

Nati"\r- tt)0.oq-N

14')N.014')

-.:t..-.0\t"\

""'.q-.\0~ ""0\.r-yo-

N\0.t'"\"" ""\0.0oq-

a).q-.r--~ It)It).""..- q-'".'"'"t'10.Q

)Na.q-r-..-40

0'"."""" t"\N.r--.N 0\\0.('-It'i0\N.r .- t--I").NN.--0\.NN ~"t".I!'IN \00.0Nr--0.ti)ti)

N.-.trINaIl"\t----

CX

)a.\0N Nt"-ot"-o--

N\0.--..- CX

)tr\.\0N .-.-..-t'"\

0\0.NN 0\tI'"\.a>tI'"\

0\0\.'=t"

Ntt)N.--rN \0~

..""'

1""\

0\0r-

r--\0.0~1""\~.\01""\

0\.-.Nq-

r-...t-oLC

)Noq-oq-.LC

"\tl'\

~N.tr\r<

'I

a)0).LC'I

y-

Ntl"\.1"'1N\C

\~.t--N I"'---.N0r-r-

t").q-.NN \0\0.t"\t"\

CX

)~.0~0\0.0\.- r-~.0\Nt'--.-.C

X)

N U"\

U"\..-.N'"N.,...'" 0\ti").~ti")

0CO

r-~ Nr-.Q)

w--

0\('i.Ir\~ r---0.C

X)

N a:>a:>.0N.--C

X>.CX

>.--

00\.IX)

N \00.('it4')

.-\0.0\.- Q

).--.t--N("I0.0("Ia~t"-~ \0co.0).-

0\-.;t.IC\

t<)

\0cx:>.NN \0r<\.C

X)

Ntl)r--.0tl)

0I!'I.a)Nt'"\N.(X

)~ \0<X

).~N U"\

\D.1"\1"\

'"10.0..-00IX:)

q-I!"\.q-

N0\.tr\..- ~10.q--- a>I!'\.r-.

""..q-.CX

)N~.q-..q-N t--N.t--Nt")(X

).\0..- N"".«)Na:>l-.l-N 0\\0.IC

\-.t0~IX

)

a)0\.a)N \0~.0\N r---0\.~0-0'1IX

).NN -.t\D.0t")1"'\ro.~1"'\

0~.<X

)NN0.0\N ,...~.~~0NC

O~ oq-0\.NN

~.q-.'""'~ <X

)co.aN\D~.01"'\

0~.It)N!'"\~...-..- ('J,...r--\Dco.~N 0\\0.r--N..-Ir\.~ ro0.U

"'\N01"\a:>~

,;

.I

I

\D fi"\ ~ N a fi"\ U"\ 0 CX) 0\ ~ 0\ .-a1 \D co .--, CX) r.-"\ .-0\ , U"\ \D \D

0\ ~ .--\D 0\ 0\ ~ , U"\ , 0\ \0 .--, r.-"\ N r.-"\.-- N N ~ N r.-"\ N N N r.-"\ N

, ~ a r.-"\ CX) U"\ ~ U"\ ~ , N .-U"\ .-I U"\ ~ IC'I N a IC'I .-r.-"\ .--\D N U"\ 0\ ~

CX) r.-"\ co , .--0\ 0\ co N CX) IC'I co N , \D

N N .-r.-"\ .--N r.-"\ N r.-"\ N N N N N

.--N a a IC'I , N r.-"\ ~ CX) N ~ \0 fi"\~I ~ ~ r: ~ ~ ~ ~ 0. ~ ~ ~ ~ ~ ~'1 IC'I ~ , ~ CX) IC'I 0\ r.-"\ IC'I r.-"\ 0\ t'"\ co , N .--r.-"\ N .--N r.-"\ N r.-"\ N N t'"\ N .--

0\ , co N a CX) 0 co IC'I r.-"\ N 0\ co ~

,j ~ 0. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ r: ~"""'I a 0 , \D N r.-"\ N , IC'I .--0\ , .--~

r.-"\ N r.-"\ r.-"\ N N fi"\ N N .--N fi"\

.--co N ~ .--\0 , IC'I , t'"\ .--0\ co ~

",' ro. ~ ~ ~ r: ~ 0. ~ ~ ~ ~ ~ ~ ~~1 IC'I , \D IC'I a 0\ 0\ .--N co ~ fi"\ .-co

N .--N N r.-"\ .--N r.-"\ .--r.-"\ ~ N t'"\ fi"\

N \D .--IC'I , 0\ fi"\ .--\D \D N 0 .--a1 \D IC'I , CX) 0\ ~ 0\ 0 N 0 t'"\ co N CX)

, fi"\ CX) CX) , 0\ lrI t'"\ t'"\ N r.-"\fi"\ r.-"\ N .--r.-"\ r.-"\ fi"\ N.-- N .--N fi"\

, \0 0\ .--a a 0 CX) , .-U"\ .-, lrI

.--1 ~ ~ ~ ~ ~ ro. r: ~ ~ ~ ~ ~ ~ ro.,. , r.-"\ 0 \D CX) 0\ 0\ co a IC'I CX) ~ U"\ a a

r.-"\ N r.-"\ .--r.-"\ fi"\ t'"\ .-r.-"\ .--fi"\ .-N ~

lrI IC'I r.-"\ IC'I IC'I 0\ \D IC'I .--\0 .-U"\ ~ NI a .-r.-"\ , r.-"\ 0\ , ~ \0 r-- 0 ~

N ~ \0 a 0\ \D .--co CX) ~ r.-"\ r-- \0 a 0

N N r.-"\ .--r.-"\ ~ ~ r.-"\ N N N .--r.-"\ N

."0 .--fi"\ .--N IC'I .--~ .--~ 0\ t'"\ co r.-"\ .--(1) I 0\ ~ \D IC'I .--\D fi"\ \D ~ \D .--~ a ~:J ~ c: IC'I fi"\ .--, .--~ 0 \D 0\ IC'I co U"\ \0 \0

N r.-"\ .--N fi"\.-- N ~ .--fi"\ N Nof-c:0U N 0\ N IC'I .--\D 0 N N , \0 \0 \D N 0\I co ~ IC'I a CX) , cr, a 0\ 0\ \D \0 IC'I 0\ \0

~ a .-N, a , \D \0 .--IC'I CX) , t'"\ , IC'I 0 fi"\

INN N .--N N N t'"\ N ~ N t'"\ N r.-"\ ~N

(1)Q) "0

-ro a a a a a a 0 a a a 0 a 0 a 0.a u ~ lrI \D , CX) 0\ 0 .--N fi"\ ~ U"\ \D , coro Q) CX) CX) CX) CX) CX) CX) 0\ 0\ 0\ 0\ 0\ 0\ 0\ 0\ 0\I- 0 ~ ~ ~ .--~ .--~

41

~ ~--