Odekon M. Encyclopedia of paleoclimatology and ancient environments

Подождите немного. Документ загружается.

permitted high-resolution analysis of the geologic record.

Improved methods of measuring paleomagnetism in the cores,

also introduced on board ship in 1978, made it possible to

retrieve the magnetic signal from the cores much more reliably,

with the result that determination of the magnetic stratigraphy

could become a routine matter.

Coincidentally, refinement of methods of mass spectrometry

made it possible to carry out measurements on stable isotopes

on ever-smaller samples. It became feasible to measure isotopic

rations of oxygen and carbon on a massive scale and even to

use these isotopic variations for stratigraphy. Again, this is an

extraordinary example of serendipity in science, where the

DSDP advances in core recovery technology and the advances

in analytical technology combined to open new vistas.

The scientific results of the IPOD phase of the DSDP are sum-

marized in reports in Anonymous (1979), Hay (1988), and Win-

terer (2000). Much of the work planned during IPOD carried

over into the successor Ocean Drilling Program (ODP).

William W. Hay

Bibliography

Anon., 1975. Deep Sea Drilling Project, Legs 1–25. AGI Reprint Series 1.

Falls Church, VA: American Geological Institute, 93pp.

Anon., 1976. Deep Sea Drilling Project, Legs 26–44. AGI Reprint Series 2.

Falls Church, VA: American Geological Institute, 82pp.

Anon., 1979. Deep Sea Drilling Project, Legs 45–62. AGI Reprint Series 3.

Falls Church, VA: American Geological Institute, 77pp.

Emiliani, C., 1954. Temperature of Pacific bottom waters and polar super-

ficial waters during the Tertiary. Science, 119, 853–855.

Emiliani, C., 1981. A new global geology. In Emiliani, C. (ed.), The Ocea-

nic Lithosphere, The Sea, Vol. 7. John Wiley and Sons, pp. 1685–1716.

Hay, W.W., 1988. Paleoceanography: A review for the GSA Centennial.

Geol. Soc. Am. Bull., 100, 1934–1956.

Winterer, E.L., 2000. Scientific ocean drilling, from AMSOC to COM-

POST. In 50 Years of Ocean Discovery. Washington D.C.: National

Academy Press, pp. 117–127.

Cross-references

Carbonate Compensation Depth

Cenozoic Climate Change

Cretaceous Warm Climates

Cretaceous/ Tertiary (K-T) Boundary Impact, Climate Effects

Dating, Magnetostratigraphy

Marine Biogenic Sediments

Methane Hydrates, Carbon Cycling, and Environmental Change

Messinian Salinity Crisis

Neogene Climates

Ocean Anoxic Events

Ocean Drilling Program (ODP)

Ocean Paleocirculation

Ocean Paleotemperatures

Paleoclimate Proxies, an Introduction

Paleogene Climates

Paleoceanography

DELTAIC SEDIMENTS, CLIMATE RECORDS

Introduction

Deltas are low-lying alluvial plains formed near the mouth of a

river, either along a lake or ocean shoreline, and commonly

with a sizeable subaqueous deposit that extends offshore. The

surface morphology of deltas typically consists of multiple

distributary channels that deliver sediment to different portions

of the delta plain and nearshore. At the coast and onto the shelf,

sediment deposition and morphological development in deltas

are largely controlled by waves, tides, and riverine processes

(Galloway, 1975). These factors dominate on shorter timescales

of tens to hundreds of years. However, over millennial and

longer periods it is generally agreed that delta evolution and

sequence formation are controlled by: (a) climate, (b) tectonics,

and (c) sea level. Unfortunately, these controls are not indepen-

dent and so their relative roles in margin sequence develop-

ment, particularly for deltas, remains a primary question in

sedimentary geology.

Unraveling mixed signals of climate, tectonics, and sea level

in deltas remains a difficult task, but the development of secu-

lar paleoclimate records in recent decades has, for the first time,

allowed scientists to more faithfully (re)interpret stratigraphic

and sedimentary records in the context of climate change. Pre-

viously, the impact of climate on delta systems had to be

inferred through somewhat circular and indirect means (i.e., lit-

tle independent confirmation of the presumed climatic forcing).

Although this is also true of tectonics and sea level, the study

of these controls had advanced decades before paleoclimate

research and thus offered better-understood mechanisms and

histories of change. However, it is now recognized that climate

can vary at a much greater magnitude and more rapid rate than

previously considered, and this dynamism prompts reconsi-

deration of the role that climate plays in sedimentary processes

and sequence development of deltas. Furthermore, significant

advances in the numerical modeling of sedimentary and strati-

graphic systems have for the first time allowed conceptual

field-based models to be tested. In the case of climate records

in deltas, this is particularly important because modeling can

help deconvolve multiple forcing mechanisms and complex

system responses.

Deltaic settings have not traditionally been targeted for paleo-

climate reconstructions specifically because of their multiple

environmental controls. Nevertheless, there remains intrinsic

value in studying paleoclimate signatures in deltas because they

serve both as: (a) primary receiving basins for land-derived

erosional and weathering products, and as (b) point-sourced gate-

ways for terrestrial fluxes to the ocean where many paleoclimatic

records are deposited. Thus, the effects of climate change on

entire watersheds are captured in deltaic deposits, which conse-

quently filter climatic signals as they are transferred to marine

repositories. Furthermore, deltaic settings and related marine

deposits are generally important sites of carbon sequestration,

closely tying carbon reservoirs with longer-term (>10

yr) cli-

mate variations.

Influences of climate on deltas

Deltas can largely be considered a function of: (a) riverine

water and sediment discharge to the continental margin, and

(b) the interaction of these fluxes with coastal and marine

processes. In this way, climate plays a major role in both the

downstream forcing of the fluvial system and the modulation

of sea level and coastal weather conditions. As such, there is

strong overprinting of climate on deltaic systems, but precise

signals are difficult to quantify because of complex feedbacks

and secondary influences.

In the terrestrial realm of a delta system (i.e., the catchment

basin), perhaps the most prevalent role that climate plays is via

precipitation. With regard to sediment production, precipitation

DELTAIC SEDIMENTS, CLIMATE RECORDS 265

is a primary control on glacial and fluvial incision, weathering,

and slope failure. However, sediment is readily produced across

a range of climates, from arid to humid and polar to tropical.

What differs among such environments in terms of climate is

hydrological runoff (=precipitation – evaporation) and thus

the ability to transport sediment to the delta. In other words,

sediment can be produced in many settings but is only tra-

nsported to the delta with sufficient water discharge, which is

primarily controlled by climate. Other important characteristics

of precipitation include its phase (liquid vs. frozen), seasonality,

and episodicity. First, the phase of precipitation controls the

relative role of glacial and fluvial erosion in sediment production,

as well as the timing and nature of water discharge (i.e., the river

hydrograph). Each affects sediment-load grain size and weather-

ing characteristics and, hence, the downstream development of

delta stratigraphy. Second, the seasonality of river discharge

affects the relative interaction of fluvial and marine processes,

whereby strongly seasonal systems are subject to specific periods

of river and marine-dominated conditions during the year. Simi-

larly, the episodic discharge typified by storm-dominated or

semi-arid river systems leaves a strong imprint on deltas, because

sediments are delivered in large but infrequent pulses. Finally,

precipitation exhibits strong secondary control on delta systems

through the type and density of catchment vegetation, which

affects runoff via overland flow and evapotranspiration rates

as well as soil erosion rates. Quantitative modeling efforts have

made progress in capturing the net effect of complex climate, hy-

drological, and sedimentary interplay, with Syvitski et al. (2003)

recently showing that catchment-integrated temperature is a

good predictor of fluvial sediment discharge. This strong rela-

tionship is linked directly with temperature’s influence on preci-

pitation, water phase, evaporation and consequently vegetation

and discharge dynamics.

From the marine side of a delta, climate change plays primary

roles in modulating both eustatic and relative sea level. Probably

the strongest link between climate change and eustatic sea level

has been the orbitally-tuned advance and retreat of continental

ice sheets in the Plio-Pleistocene. The fate of deltaic systems

under these 20–120 m sea level fluctuations has long been con-

sidered in the context of sequence stratigraphy (Vail et al.,

1977), whereby the architecture and distribution of continental

margin deposits has been used to reconstruct records of sea

level change for much of the Phanerozoic. These general seq-

uence models have been rigorously critiqued, though, with

less than ideal results in reconstructing truly eustatic records

of sea level change. However, among the major findings

emerging from these studies are that sea level is not the over-

arching control previously thought and that influences of cli-

mate and tectonics on the terrestrial catchments are often of a

similar magnitude to that for sea level (Shanley and McCabe,

1998). This further illustrates the frustratingly circular process

of reconstructing climatic paleorecords based on stratigraphic

sequences that are simultaneously being interpreted based on

the same sequence-derived paleorecords. Hence, the secular

derivation of climate history from lake, ice, and speleothem

records bodes well for the unbiased interpretation of deltaic

stratigraphic records.

Climatic signal s in deltas

Because climate plays multiple roles in controlling deltaic sys-

tems, its signals and impacts may be looked for across a range

of spatial and temporal scales. Typically, climate impacts on

larger fluviodeltaic systems are expected to be attenuated in

the sizeable catchment, resulting in a weakened and delayed

(10

–10

yr) downstream signal (Castelltort and Van Den

Driessche, 2003). However, recent research has suggested that

even large systems may respond rapidly to climate change

when climate is a dominant control, such as in semi-arid

and sub-tropical monsoon regions (Goodbred, 2003; Thomas,

2003). In smaller fluviodeltaic settings, high-frequency climate

changes (10

1

–10

yr) have a greater chance of influencing the

entire sedimentary system and being preserved in the strati-

graphic record, because the small catchment basin has limited

storage capacity and signals are thus transferred more rapidly

to the delta and shelf.

Climate impacts on deltaic systems may be manifested as

geomorphic, stratigraphic, and sedimentological signals. Along

the landward side of a delta, upstream of major subsidence

(flexure) and accretion, geomorphic features are most likely

to remain exposed and accessible for study. In such fluvially

dominated reaches of a delta system, riverine terrace sequences

and slackwater deposits have been used to reconstruct past

hydrologic conditions. The terraces and slackwater deposits

form at different elevations depending on the height of bankfull

river stage, and hence water discharge. In several systems

recently, the dating of bankfull features using cosmogenic

radionuclides and optically stimulated luminescence have

revealed that their timing and height correlates strongly with

known Quaternary climatic variations (Repka et al., 1997;

Yang et al., 2000). These records indicate that hydrologic con-

ditions in fluviodeltaic systems vary significantly with millen-

nial and orbital-scale climate change. A related manifestation

of changing climate and hydrology is found in river planform

morphology, which may vary in its degree of braiding and/or

sinuosity (Bogaart and van Balen, 2000). Furthermore, river

planforms reflect both sediment load and hydrology as they

are largely controlled by ratios of total load/discharge and bed-

load/suspended load. However, as sediment load can also vary

due to tectonic influences, river planforms are perhaps a more

ambiguous link to climate than terraces and flood deposits.

Climate signals may also be preserved in delta stratigraphy,

which similar to geomorphic signals, generally reflects changes

in hydrology and sediment supply. Unlike upstream geomorphic

records, though, delta sequences commonly preserve a rela-

tively continuous stratigraphic record due to margin subsidence.

Two signals of climate change recognized in delta stratigraphy

include channelized-flood deposits, comprising coarse-grained

sands to cobbles, and the architecture of deltaic strata and their

stacking patterns, which reflect varying sediment inputs due to

changing hydrological conditions. In hemipelagic deposits of

the marine prodelta, climatically driven variation in river dis-

charge are frequently recorded by alternations of bioturbated

hemipelagic layers and organic-rich lithic sapropels reflecting

dry and wet periods, respectively. The sapropels form due to

high river discharge that: (a) supports high primary productivity

and particulate carbon flux under elevated nutrient loading; and

(b) enhances water-column stratification and the development of

suboxic bottom waters, which limits organic matter degradation.

For sedimentological records of climate change, such as micro-

fossils, aeolian flux and grain size, these are also preserved in

delta sequences but in many instances are overwhelmed by flu-

vially derived inputs. However, carbonate ooids that have

formed locally off large sub-tropical river deltas likely reflect

glacial period aridification and decreased fluvial discharge.

Reduced discharge is implicated because modern ooids form

266 DELTAIC SEDIMENTS, CLIMATE RECORDS

only in shallow waters of evaporative high-salinity settings.

Furthermore, formation of carbonates at the rivermouths sug-

gests limited sediment discharge, which would bury carbonate

deposits and preclude their development.

Site cases

Until recently, few studies have specifically investigated climate

signals and impacts in delta systems. The current knowledge

base on this subject comes from several modeling groups, some

field studies, and also by revisiting existing data with a better

understanding of climate change from secular records. Although

the list of site cases discussed below is not exhaustive, there are

two relevant trends. First, the number of modeling studies equals

or exceeds those based on field investigations, and second, the

majority of delta systems in which climate has been investigated

are located in the subtropics (the Rhine/Meuse a notable excep-

tion). Subtropical systems tend to be highly seasonal, receiving

nearly all of their precipitation during the wet summer monsoon

when the intertropical convergence zone (ITCZ) is most proxi-

mal. At orbital timescales, both the position and strength of the

ITCZ varies in response to global boundary conditions (i.e.,

extent of continental ice sheets) and maximum summer insola-

tion, respectively.

Nile delta (North Africa)

The Nile delta is located on the arid north coast of Africa, but the

river’s headwaters and main discharge source lie in East Africa

where runoff is seasonally derived from the summer SW Asian

monsoon. As such, water and sediment discharge to the delta

have varied significantly with orbitally-tuned changes in mon-

soon precipitation. These variations are recorded in prodelta tur-

bidites of the Nile system, which show alternating hemipelagic

marls and organic-rich laminites that correlate with stable-

isotope proxies for freshwater input. The individual marl/sapro-

pel couplets track the 41-kyr precessional cycle, with packages at

the 100-kyr and 400-kyr eccentricity cycles (Postma, 2001).

These specific turbidite sequences are described from late

Miocene to early Pliocene deposits. Younger records of climatic

control have been recognized from fluvial terraces and channel

planform along the Nile’s middle reaches (Williams et al.,

2000). Although these records are upstream of the delta, the

changes in water and sediment discharge are almost certainly

recorded at the margin but are yet to be described.

Ganges-Brahmaputra (South Asia)

The Ganges and Brahmaputra rivers drain the foreslope and

backslope of the Himalayan front range, respectively. The two

rivers coalesce near the Bengal margin to form the world’ s lar-

gest subaerial delta and deep-sea fan system. The rivers are also

solely driven by the wet summer monsoon, when over 80% of

water and 95% of sediment discharge occurs in only four months.

Given global significance of the Himalayan/Tibetan uplift and

its associated drainage systems, many portions of the Ganges flu-

vial delta system have been independently investigated in the

past 15 years, revealing important variations in glacial activity,

alluvial fan and floodplain development, delta evolution, and

regional oceanography.

Taken together, the observed patterns show the immense

Ganges dispersal system to respond to multi-millennial-scale

(<10

yr) climate change in a system-wide and largely contem-

poraneous manner, and that major sedimentary signals are

transferred rapidly from source to sink with little apparent

attenuation (Goodbred, 2003). Some of the major signals of cli-

mate change recorded in the delta stratigraphy include: (a) a

thick (20–30 m) coastal mangrove facies that was deposited

during rapid sea level rise in the early Holocene, and (b) a mas-

sive stored sediment volume stored during this period that

reflects a sediment discharge at least 250% higher than the

world’s largest modern load of 1 billion tonnes (Goodbred

and Kuehl, 2000). This high discharge period corresponds to

the early Holocene hypsithermal, when peak regional insola-

tion supported a stronger than present monsoon and associated

precipitation. In contrast, signals recorded in the delta during

the Last Glacial Maximum (18–20 ka) reflect aridification and

very low river discharge. Carbonate ooids, typical of shallow,

evaporative tropical waters, formed on the Bengal shelf at this

time directly adjacent to the rivermouth (Wiedicke et al., 1999).

Similarly, reconstructed planktonic foraminifera assemblages

from the northern Bay of Bengal (Ganges discharge basin)

indicate peak surface-water salinities corresponding to the

LGM and apparently arid conditions (Cullen, 1981). These

records from the delta and margin correlate well with findings

from the catchment basin, and reflect the overall dominance

of climate change in controlling this large fluviodeltaic system

during the Quaternary.

Colorado, Brazos, and Trinity deltas (south-central

North America)

These three rivers comprise a series of small catchments drain-

ing the semi-arid to moderately humid hills of central Texas,

USA and discharging to the Gulf of Mexico. Presently, the riv-

ers are driven by episodic floods and are forming small deltas

at best. However, studies of the Texas shelf have revealed

expansive fluviodeltaic sequences formed during a lower stand

of sea level, and that they extend laterally and coalesce into an

alongshelf composite delta system (Rodriguez et al., 1998).

The shelf delta and its connection to upstream alluvial valleys

are significantly larger than can be explained by modern

hydrology and discharge, indicating an important change in cli-

mate or drainage basin character. The region is not tectonically

active, but recent paleosol studies in the catchment reveal per-

iods of significantly higher rainfall that are roughly correlable

with the delta formation (Stiles et al., 2003). Although a defini-

tive connection between the large Texas shelf deltas and a pre-

viously more humid climate has not been established, available

data and examples from other similar systems imply a probable

and significant linkage.

Rhine-Meuse delta (north-central Europe)

The Rhine-Meuse river drains central Europe and discharges

to the southern North Sea coast. Much of the research on this

system during the past decade has had a focus on understand-

ing the signatures of climate and eustasy contained in the late

Quaternary stratigraphy and geomorphology (Törnqvist, 1998;

Veldkamp and Tebbens, 2001 ). Being a temperate river system,

climatic forcings have probably not varied as dramatically

as in the sub-tropical systems, but nonetheless the magnitude,

timing, and nature of discharge appear to have played a role

in shaping the fluviodeltaic system upstream of the hinge line.

Terrace development in this river-dominated reach of the system

is recognized as a primarily climate-controlled process through

both modeling and field studies. However, the downstream

impacts of such hydrological changes in the Rhine-Meuse

DELTAIC SEDIMENTS, CLIMATE RECORDS 267

system appear to be insufficient to overcome the dominant

influence of glacioeustasy during the Quaternary. There are

likely climate impacts at the delta and margin but these

are indistinguishable at the current resolutions of data and

modeling.

Modeling studies

Climate has long been recognized as an important control on

fluviodeltaic systems, but it has been difficult to confidently

link observed stratigraphic and sedimentological signatures to

climatic forcing. Such limits of field investigations are reflected

in the relatively restricted results discussed above. However,

exponential increases in computing power have allowed major

advances to be made in the numerical modeling of fluvial and

deltaic systems. Several current models offer the opportunity

to compare results of system forcing by any combination of

climate, eustasy, and tectonics, thereby helping to elucidate

the specific character of climate impacts on stratigraphic

sequences. In addition to modeling of the Rhine-Meuse

mentioned above, the Niger delta sequence has also been inves-

tigated using an industry-based model (Van der Zwan, 2002).

Investigation of the Niger was aimed at testing the impact of

presumed Milankovitch-scale variations in sediment discharge

over the Neogene. The orbitally linked forcing of regional cli-

mate and sed iment occurred via global climate, insolation-

control of monsoon strength, and vegetation. Results suggest

that sediment supply does vary at both the large-scale (Ma)

and Milankovitch-scale time frames under climatic forcings,

and that there was a modest im pact on stratigraphy. However,

under icehouse con ditions with rapid glacioeustatic sea level

change, fluctuations in climate-forced sediment supply were

greatly overshadowed. These findings echo those of other

modeling studies in that climate is often found to be an impor-

tant, but secondary, control on fluviodeltaic sequences.

Furthermore, most modeling studies strongly suggest that cli-

matic signals at the margin greatly lag the actual forcing in

the catchment basin (Castelltort and Van Den Driessche,

2003). This notion conflicts with recent field evi dence from

several of the sub-tropical rivers discussed above. Much of

these discrepancies likely arise fr om our limited knowledge

of appropriate input values for pre-modern climate, riverine,

and coastal conditions.

Conclusions

Deltas are heterogeneous systems controlled by a suite of

processes, which altogether makes it challenging to isolate cli-

mate signals and their impacts. Nevertheless, climate is well

recognized as a primary control of deltaic systems, and the

ongoing reconstructions of secular climate records will allow

better linkages to be made between climate and the stratigraphic

record. If recent trends in both field and modeling efforts con-

tinue, then our understanding of climate’s role in controlling

river deltas and seaward basins is almost certain to advance

quickly with the application of available paleoclimate records.

Ultimately, our understanding of such systems will be suffi-

ciently advanced that new information about climate change

and its impacts can be determined by specifically targeting

the complex, but high-resolution, archives contained in deltaic

sequences.

Steven L. Goodbred, Jr.

Bibliography

Bogaart, P.W., and van Balen, R.T., 2000. Numerical modeling of the

response of alluvial rivers to Quaternary climate change. Glob. Planet.

Change, 27, 147–163.

Castelltort, S., and Van Den Driessche, J., 2003. How plausible are high-

frequency sediment supply-driven cycles in the stratigraphic record?

Sediment. Geol., 157,3–13.

Cullen, J.L., 1981. Microfossil evidence for changing salinity patterns in

the Bay of Bengal over the last 20 000 years. Palaeogeogr., Palaeocli-

matol., Palaeoecol., 35, 315–356.

Galloway, W.E., 1975. Process framework for describing the morpholo-

gic and stratigraphic evolution of deltaic depositional systems. In

Broussard, M.L. (ed.), Deltas, Models for Exploration. Houston, Texas:

Houston Geological Society, pp. 87–98.

Goodbred, S.L., Jr., 2003. Response of the Ganges dispersal system to cli-

mate change: A source-to-sink view since the last interstade. Sediment.

Geol., 162,83–104.

Goodbred, S.L., Jr., and Kuehl, S.A., 2000. Enormous Ganges-Brahmaputra

sediment discharge during strengthened early Holocene monsoon.

Geology, 28, 1083–1086.

Postma, G., 2001. Physical climate signatures in shallow- and deep-water

deltas. Glob. Planet. Change, 28,93–106.

Repka, J.L., Anderson, R.S., and Finkel, R.C., 1997. Cosmogenic dating of

fluvial terraces, Fremont River, Utah. Earth Planet. Sci. Lett., 152,

59–73.

Rodriguez, A.B., Anderson, J.B., and Bradford, J., 1998. Holocene tidal

deltas of the Trinity incised valley; analogs for exploration and produc-

tion. Trans. Gulf Coast Assoc. Geol. Soc., 48, 373–380.

Shanley, K.W., and McCabe, P.J. (eds.), 1998. Relative role of Eustasy, Cli-

mate, and Tectonism in Continental Rocks, Special Publication 59.

Tulsa OK, USA: SEPM (Society for Sedimentary Geology), 234pp.

Stiles, C.A., Mora, C.I., Driese, S.G., and Robinson, A.C., 2003. Distin-

guishing climate and time in the soil record; mass-balance trends in Ver-

tisols from the Texas coastal prairie. Geology, 31, 331–334.

Syvitski, J.P.M., Peckham, S.D., Hilberman, R., and Mulder, T., 2003. Pre-

dicting the terrestrial flux of sediment to the global ocean: A planetary

perspective. Sediment. Geol., 162,5–24.

Thomas, M.F., 2003. Late Quaternary sediment fluxes from tropical water-

sheds. Sediment. Geol., 162,63–

81.

Törnqvist, T.E., 1998. Longitudinal profile evolution of the Rhine-Meuse

system during the last deglaciation; interplay of climate change and

glacio-eustasy? Terra Nova, 10,11–15.

Vail, P.R., Mitchum, R.M., Jr., Todd, R.G., Widmier, J.M., Thompson, S.,

III, Sangree, J.B., Bubb, J.N., Hatleilid, W.G., 1977. Seismic stratigra-

phy and global changes of sea level. In Payton, C.E. (ed.), Seismic

Stratigraphy - Application to Hydrocarbon Exploration, Memoir 26.

Tulsa, OK: American Association of Petroleum Geologists, pp. 49–212.

Van der Zwan, C.J., 2002. The impact of Milankovitch-scale climatic for-

cing on sediment supply. Sediment. Geol., 147, 271–294.

Veldkamp, A., and Tebbens, L.A., 2001. Registration of abrupt climate

changes within fluvial systems: Insights from numerical modelling

experiments. Glob.Planet. Change, 28, 129–144.

Wiedicke, M., Kudrass, H.-R., and Hübscher, C., 1999. Oolitic beach bar-

riers of the last glacial sea-level lowstand at the outer Bengal shelf. Mar.

Geol., 157,7–18.

Williams, M.A.J., Adamson, D., Cock, B., and McEvedy, R., 2000. Late

Quaternary environments in the White Nile region, Sudan. Glob. Pla-

net. Change, 26, 305–316.

Yang, D., Yu, G., Xie, Y., Zhan, D., and Li, Z., 2000. Sedimentary records

of large Holocene floods from the middle reaches of the Yellow River,

China. Geomorphology, 33,73–88.

Cross-references

Continental Sediments

Cyclic Sedimentation (Cyclothems)

Glacial Eustasy

Hypsithermal

Last Glacial Maximum

Monsoons, Quaternary

Paleohydrology

Sapropels

Sea Level Change, Last 250 Million Years

268 DELTAIC SEDIMENTS, CLIMATE RECORDS

Sea Level Change, Post-Glacial

Sea Level Change, Quaternary

Sedimentary Indicators of Climate Change

DENDROCLIMATOLOGY

Introduction

Dendroclimatology is the branch of dendrochronology that uti-

lizes absolutely dated and annually resolved growth layers in

woody plants, such as trees, for the reconstruction and analysis

of past climate variability. The same underlying principles of

dendrochronology (described in Dating, dendrochronology)

still apply. At the core of dendroclimatology lies the central

tenet of all dendrochronological applications; the accurate

crossdating of multiple tree ring samples that are all uniformly

controlled by climate, such that the same signature of narrow

and wide (or dense and less dense) rings can be identified in

all of the sampled trees across time and space. The stronger

the common “signal” between trees (i.e., the greater the degree

to which narrow or wide rings are consistently narrow or wide),

the stronger will be any reconstruction of climate that is

derived from them. Methodologies have been developed for

testing signal strength in chronology indices, and for calibra-

tion and verification of climate/tree-growth models, and these

will be discussed in the ensuing sections.

As suggested above, climate may influence tree growth

across space and time and can result in a high degree of com-

mon variability expressed in the pattern of annual growth rings

in trees across broad regions (Douglass, 1919; Fritts, 1976;

Hughes et al., 1982; Cook and Kairiukstis, 1990). The tree ring

chronologies with the highest degree of strength in their com-

mon signal tend to be from regions where climate is at its most

limiting to growth, and is most regionally coherent, and where

other non-climatic factors such as competition and disturbance

are at a minimum. Trees growing in regions with little climate

variability and optimum conditions for growth are less likely to

produce the high variability in annual growth necessary for

robust reconstructions of climate. It is for this reason that den-

droclimatologists seek out trees from those areas where trees

are most likely to be limited by climatic factors such as tem-

perature or drought, normally at the limits of their ecological

range (see Dating, dendrochronology).

Describing tree -growth /climate relationships

The response of tree growth to climatic forcing involves the

interaction of many complex factors. Due to this complexity

there are no suitable theoretical models to allow for the extrac-

tion of a climatic signal from tree rings based on ecophysiolo-

gical or mechanistic studies alone (Fritts, 1976; Hughes et al.,

1982; Cook and Kairiukstis, 1990). Dendroclimatologists there-

fore rely upon semi-empirical relationships between tree

growth and climate for calibrating the climate response of a

given chronology (Fritts, 1976; Guiot et al., 1982). The multi-

variate statistical analyses used include principal component or

eigenvector analysis, and canonical correlation and regression

(Fritts et al., 1971; Fritts, 1976, 1991; Guiot et al., 1982).

Theprocedure that estimates the statistical growth-environment

relationship is called calibration. The tree-growth/climate

relationships revealed in this manner, called response functions,

may not necessarily reflect a directly coupled response. Rather,

they may involve long chains of interacting cause and effect

linkages that ultimately effect growth in the trees, resulting

in enhanced or suppressed growth for a particular year. The

climate response is evident because the elements of weather

and climate affect the microclimates for individual trees, and

thereby the physiological processes that in turn control growth

(Gates, 1968a,b; Fritts et al., 1971; Fritts, 1976 ; Salisbury

and Ross, 1992). When a sufficiently strong relationship

has been established between tree growth and some para-

meter of climate (e.g., temperature or precipitation), the

statistically-calibrated relationship can be used to generate a

transfer function for the reconstruction of climate from tree

growth for the period prior to the available climatic data set

(Lofgren and Hunt, 1982).

For calibration, standardized tree ring indices (typically, ring

width or maximum latewood density) are compared with the

longest available instrumental records of climate, from meteor-

ological stations that most closely represent conditions at the

site. Long climatic time-series are required for a proper and

robust statistical comparison (Fritts, 1976; Hughes et al., 1982;

Cook and Kairiukstis, 1990). However, baseline weather data

for a given tree site are usually lacking, and the nearest meteor-

ological stations may be many kilometers from the location of

the trees. This problem is compounded by the search for the

oldest and least disturbed trees available, which are likely to

be in areas remote from most weather stations, and at substan-

tially higher elevations. Hughes et al. (1978) illustrated how

the percentage of chronology variance explained by climate

and prior growth, respectively, could change from 34% and

31% to 45% and 28%, respectively, when rainfall data from

sites 60 km and 6 km distant were used. To some extent, this

problem may be ameliorated by analyzing broad scale para-

meters such as sea level pressure (SLP) and sea surface tem-

perature (SST) anomaly patterns, which may in turn influence

local conditions at the site (e.g., Douglas, 1973;D’Arrigo

et al., 2003). In most instances, temperature exhibits far less

spatial variability than precipitation. It may therefore be

more accurately represented by relatively distant instrumental

records.

It is desirable to have monitored radial growth in conjunction

with daily or even sub-daily environmental conditions over suc-

cessive seasons at important dendroclimatic sites. This allows

for a more mechanistically based climate/growth model. Such

models are actively being pursued though at present they are

limited to open-canopy, arid-site conifers with wide rings and

uncomplicated growth structure (e.g., Fritts and Shaskin,

1994). Application of such mechanistic models to more mesic,

closed-canopy forest environments is not yet viable.

In traditional dendroclimatological studies, tree growth for

an entire season, as determined from the mean value function

for each annual ring, is usually compared against monthly or

seasonal climate parameters for the determination of the most

significant factors affecting growth at the site (Fritts, 1982).

In reality, climatic conditions on daily or even shorter time-

scales may have a significant impact on the annual growth ring.

Transient climate features such as severe frosts, very high tem-

peratures, heavy snowfall, or windstorms may impact on

growth for a given year, t. In most instances, however, such

short-term factors can only be modeled when they leave

obvious features in growth rings (e.g., traumatic resin ducts,

frost-rings, false-rings), and monthly and seasonal data are

more reliably modeled.

DENDROCLIMATOLOGY 269

Some important assumptions and limitations exist for

dendroclimatic reconstructions, which must be carefully

considered:

1. It is assumed that the Uniformitarian Principle applies; i.e.,

the physical and biological processes that link the present

environment with present variations in tree growth are the

same processes that will have been in operation throughout

any segment of the pre-calibration period (Fritts, 1976).

2. Similarly, the climatic conditions that produced past anoma-

lies in tree growth are assumed to be analogous to the same

processes enacting upon tree growth during the calibration

period.

3. It is usually assumed that the systematic relationship

between the limiting climatic parameter and the biological

responder (tree growth) can be approximated by a linear

mathematical expression. Lofgren and Hunt (1982) note,

however, that in the case of transfer functions, the predic-

tand set need not be a linear function of climate; e.g., the

logarithm of a precipitation record could be used and there-

fore the linearity constraint is not so severe a restriction as it

might first appear.

4. The data are assumed to be normally distributed when para-

metric statistics are tested for significance, as the effects of

outliers and any other non-normal data may distort the testing

of significance (Fritts, 1990). Non-normal data can be nor-

malized prior to regression through several techniques out-

lined in Fritts (1976) and Cook and Kairiukstis (1990).

5. The observations must also be assumed to be independent of

each other for most statistical testing procedures. A serial-

dependence or autocorrelation among successive observa-

tions, or among observations in space, may seriously reduce

the number of degrees of freedom used for statistical signifi-

cance testing. The effects of autocorrelation may be

removed by prewhitening the data through autoregressive

modeling (e.g., Cook et al., 1992). Alternatively, prior

growth can be included in the regression as a predictor vari-

able, lagged up to m prior years.

It is clear that the assumptions listed above are not completely

met by the climate/growth relationship used for calibration and

climate reconstruction procedures. They may instead be

regarded as comprising the limitations of the system (Fritts

et al., 1971; Lofgren and Hunt, 1982) and, within these limita-

tions, some percentage of the variability of climate will be

explained by a dendroclimatic reconstruction. The quality and

the length of both the tree ring and climatic data play a crucial

role in the validity of the resulting models, and results derived

from their calibrations and climate reconstructions should be

verified, as discussed below.

Calibration of the climate response

The word calibration in dendroclimatology refers to the fitting of

statistical models that can be applied to one or more predictor

variables in order to estimate or reconstruct one or more pre-

dictand (Fritts, 1990). The term response function refers to the

weights or coefficients of the statistical model that describes

the manner in which trees respond to climate (Fritts et al.,

1971; Fritts, 1976). The response function does not measure

the climate-growth response directly but rather the effectiveness

of a particular statistical model at predicting the element of tree

ring variation forced by external factors (Hughes et al., 1982).

Climate parameters are often highly intercorrelated, making

classical multiple regression analyses difficult to apply directly

for response function analysis. Instead, the principal compo-

nents of climate are often used in order to provide orthogona-

lized data sets for use in regression (Fritts et al., 1971; Fritts,

1976; Guiot et al., 1982; Briffa and Cook, 1990). Additional

complications may arise from the autoregressive properties of

the tree ring series, where the growth in a particular year is

not only dependent upon conditions in the season of growth

but also in one or more previous seasons (Fritts et al., 1971;

Fritts, 1976). Some form of prior growth parameter is often

employed through lagging the tree ring series up to m prior

years. Alternatively, both the tree ring time-series and the cli-

mate data can be prewhitened through autoregressive modeling

to remove the effects of autocorrelation prior to response func-

tions analysis (e.g., Cook et al., 1992).

There are several response function procedures in use at pre-

sent. The primary differences relate to the manner in which

relatively unimportant predictor PCs are removed from consid-

eration during regression screening (Briffa and Cook, 1990).

Blasing et al. (1984) noted that this is not always a straightfor-

ward procedure. Decisions regarding the number of climatic

variables to include, confidence limits, and the number of

eigenvectors to allow as candidate predictors in the regression,

can affect the response function in unpredictable ways, leading

to errors in interpretation. Blasing et al. (1984) recommended

using the correlation function as an initial, interpretive guide

prior to response function analysis, since the correlation func-

tions are easier to understand, easier to replicate, and more dif-

ficult to alter subjectively. They also demonstrated how prior

tree growth variables used in the regression can mask true cli-

matic effects, and noted the usefulness of the correlation func-

tion for detecting such masking. The effect of prior growth

variables can lead to mistakenly attributing real climatic effects

to prior growth. They should therefore be considered separately

from the response functions.

The simple correlation functions between the tree ring

chronologies and the climate variables are used to interpret

the climate response. Both the tree rings and the climate vari-

ables (where warranted) are first prewhitened as an order-p

autoregressive (AR) process of the form:

i¼1

ti

þ e

ð1Þ

where X

is the time series being modeled, the f

are the p auto-

regressive coefficients, and e

is the series of random shocks or

white noise residuals unexplained by the AR(p) model (Box

and Jenkins, 1970). The e

of the tree rings and climate variable

are used to develop the regression model and subsequent cli-

mate reconstructions, with persistence due to climate added

back (where warranted).

Reconstructing climate

Transfer functions are statistically derived equations that relate

two sets of variables for which a causal relationship can be

described, but for which the derivation of an analytical expres-

sion cannot be achieved (Fritts et al., 1971; Lofgren and Hunt,

1982). In the case of dendroclimatology, one set of variables

consists of the tree ring indices, and the other the climatic para-

meter of interest (e.g., temperature or precipitation). The cali-

brated function of tree growth with the appropriate climate

parameter can then be used to transfer the annual patterns of

tree growth into yearly estimations of the climate variable, for

270 DENDROCLIMATOLOGY

periods of time prior to the recording of that variable (Fritts,

1976; Lofgren and Hunt, 1982).

For many studies, PC regression analysis can be used to

transform tree ring indices into estimates of a single climate

variable (i.e., temperature). Prior to regression, the tree ring

chronologies and climate variables are pre-whitened as order-

p AR processes, as per Equation (1). The reconstruction of cli-

mate (Y ) is therefore accomplished by estimating the regression

equation over the calibration period as:

Y ¼ Xb þ E ð2Þ

where matrix X is the tree ring predictor set, b is the column

vector of standardized regression coefficients used to estimate

climate from tree rings, and E is the column vector of errors

due to unexplained variance in the regression model. b can be

estimated by ordinary least squares (OLS) as:

b ¼ðXTX Þ

1

 X

ð3Þ

where the superscript T denotes the transpose of matrix X and

1

denotes the inverse of X

X. In the form of normalized

departures, estimates of climate from tree rings (Ŷ) can be

obtained by multiplying X by b as follows:

Y ¼ Xb ð4Þ

When PCA is performed on the tree ring chronologies, values

of X are multiplied by the eigenvector loadings (E), producing

a new set of predictors:

U ¼ XE ð5Þ

These are the PC amplitudes or scores. Prior to regression, each

amplitude is normalized by the square root of its respective

eigenvalue. The yearly variations in each amplitude reflect

the relative importance of its eigenvector in describing the

overall pattern of variance seen in the original data.

At this stage, a subset of amplitudes that explain most of the

original variance in each data set is selected for regression ana-

lysis, after deleting the higher order eigenvectors. The eigenva-

lue-1 criterion (e.g., Guttman, 1954) is used here for screening

predictors, deleting those eigenvectors with eigenvalues less

than 1.0. This selection criterion is conservative in that it only

retains around 20–40% of the eigenvectors, which may explain

60–70% of the total variance in the original data (Cook et al.,

1994). In this way, the number of candidate predictors is

reduced and, therefore, so is the potential for artificial predict-

ability that might arise from the inclusion of too many predic-

tors in the model (Davis, 1976; Lofgren and Hunt, 1982).

Once a subset of the predictor eigenvectors is selected, the

predictand is regressed on selected predictor PCs, and the stan-

dardized regression coefficients are nothing more than the sim-

ple correlations between the two. The best-order regression

model is selected by the minimum AIC criterion (Akaike,

1974), and the regression model is then:

Y ¼ U

þ E ð6Þ

where the prime denotes the reduced-order model of PC ampli-

tudes and regression coefficients selected by the AIC. The

matrix B

of regression coefficients can then be expressed in

terms of the original predictors as

b ¼ E

ð7Þ

This back-transformation to coefficients on the original predic-

tors eliminates the interpretational disadvantage of working with

regression weights of the PC amplitudes (Cook et al., 1994).

Verification of results

The statistical verification of climatic reconstructions is an inte-

gral part of dendroclimatology (Fritts, 1976; Hughes et al.,

1982; Cook and Kairiukstis, 1990). Both response function

and transfer function procedures can be verified through statis-

tical analyses. The climate response equations are verified by

comparing estimated values of tree growth with the actual

values over an independent period (Briffa, 1984; Briffa and

Cook, 1990). For this purpose, the climatic data are split into

two equal segments (Figure D31). Identical predictor-selection

criteria are used for both time periods to enable a more direct

comparison of results, as recommended by Briffa and Cook

(1990). Standardized periods of analysis, each of sufficient

length to minimize sample-to-sample variability in the PC load-

ing patterns (i.e., usually greater than 25–30 years) are also

used. The problem of determining the confidence limits, both

on R

and on the original climate predictors, is due to the

fact that the selection of final PC predictors is an a posteriori

decision, with an unclear definition of what the correct degrees

of freedom are (Briffa and Cook, 1990).

The reliability of a particular model is assessed directly by

calculating a number of verification statistics that measure the

degree of similarity between independent estimates of climate

from the model, and the corresponding instrumental data from

independent time periods (Gordon, 1982; Fritts et al., 1990).

Verification procedures may employ several statistics: para-

metric statistics that involve assumptions of the underlying

probability distributions of the data, and are therefore sensitive

to violations of these assumptions, and nonparametric statistics

that are insensitive to such violations.

Following are several verification statistics that are com-

monly used: the Pearson product-moment correlation (R

the sign test (ST), the reduction of error test (RE), the product

means test (PM), and the coefficient of efficiency (CE). Full

descriptions of these tests can be found in the following: Fritts

et al., 1990; Cook et al., 1992, 1994; Salinger et al., 1994.

The R

statistic is one of the simplest and most commonly

used verification statistics. It measures the relative varia-

tion (covariance) between two data sets, and is completely

insensitive to differences in the mean and variance between

the two (Fritts et al., 1990). Since it reflects the entire spectrum

of variation (both high and low frequencies) its value can be

affected by trends in the two data sets. This is resolved by cal-

culating a new correlation coefficient from the first differences

of the two data sets, thereby measuring only the high-frequency

variance in common. The data sets are assumed to be normally

distributed, and the variance is assumed to be linearly related

(Fritts, 1976; Fritts et al., 1990).

The ST is a nonparametric measurement of reliability

that counts the number of times the signs of the departures

from the sample means agree or disagree, and offers the advan-

tage of measuring the associations at all frequencies. Results

are based on two possible outcomes: (+1) when the estimate

and observations agree (i.e., are on the same side of the depen-

dent data mean) and (1) when they disagree (Fritts et al.,

1990). A successful ST is indicated when the signs of the esti-

mates are correct more often than would be expected from

random numbers.

DENDROCLIMATOLOGY 271

The RE statistic offers a highly sensitive measure of reliabil-

ity. It is similar to the explained variance statistic obtained with

the calibration of the dependent data, and values range from

negative infinity to a maximum of 1.0, which indicates a per-

fect estimation (Fritts et al., 1990). The RE is calculated by

dividing the sum of the squared differences between actual

and estimated values by the sum of the squared differences

between actual values and the mean of the calibration period,

then subtracting that value from 1.0. The significance of RE

cannot be tested, though positive values indicate some degree

of model skill. Errors are unbounded, so that even one extreme

error in an otherwise nearly correct set of estimates may result

in a negative RE and unfounded rejection of the model. While

RE does estimate the explained variance with considerable

accuracy (Gordon and LeDuc, 1981), the size of the sample

being analyzed (n

) may markedly affect its significance. For

example, with n

= 20, an RE of zero was roughly equivalent

to the 0.95 confidence limit, but for n

= 10 or 15 an RE statis-

tic significantly different from zero would have to be greater

than 0.25 and 0.12, respectively (Fritts et al., 1990).

The CE was introduced into dendroclimatology by Briffa

et al. (1988) and is very similar to the RE, with values ranging

from negative infinity to a maximum value of 1.0. The only

difference is that the CE statistic uses the mean of the verifica-

tion period in the denominator, instead of the mean of the cali-

bration period. Consequently, if the verification and calibration

means are identical, then CE = RE. However, when the means

are not equal, RE will be greater than CE by a factor related to

that difference (Cook et al., 1994). Since the value of the CE

statistic will usually be less than for RE, it is an even more dif-

ficult test to pass. Although there is no significance test for CE,

any value greater than zero indicates some degree of model

skill (Cook et al., 1994).

The PM test takes into account both the sign and magnitude

of the departure from the calibration average, and is computed

from the product of the actual and estimated annual departures

from the mean, with positive and negative means summed

separately (Fritts, 1976). The product is positive when the

departure sign is estimated correctly, and negative when incor-

rectly estimated. If randomly guessed, the means of the positive

Figure D31 Tree ring reconstruction of temperature anomalies from Interior British Columbia (bottom), showing calibration and verification

statistics and comparisons between actual and reconstructed temperatures (top). The full calibration period explains 53% of the variance in the

original temperature series. Calibrating with the early (1912–1951) and late periods (1952–1991) independently, explains 60% and 50%,

respectively, reflecting better fidelity in the early half of the record (data courtesy of R.J. Wilson).

272 DENDROCLIMATOLOGY

and negative products are approximately the same (ignoring

their signs); while for a correct reconstruction, the mean is lar-

ger for the positive products (Fritts, 1976). The difference

between positive and negative products is tested with the t sta-

tistic, and is considered statistically significant when the value

of t is higher than the expected value at the appropriate confi-

dence level. A mean positive product that is significantly larger

indicates a tendency for both actual and estimated departures

from the average value to be large when the sign is correctly

estimated, and small when the sign is incorrectly estimated.

The latter case indicates that approximately average conditions

both occurred and were estimated, thereby indicating the cor-

rectness of the reconstruction in spite of its incorrect sign

(Fritts, 1976).

Reconstructions of climate from tree rings

Through his research on trees of the arid American southwest

(see Dating, dendrochronology), A.E. Douglass came to under-

stand the connection between climate and tree growth. In one

of his earliest attempts to reconstruct climate, Douglass

(1909), in his paper “Weather cycles in the growth of big trees,”

described the connection between available moisture, food

supply and annual incremental growth, and extrapolated long

records of Ponderosa pine as proxy records of rainfall for

northern Arizona. All of the methodological applications were

done manually; curve fitting, standardization and chronology

building, and all correlation analyses were done without the

benefit of computers. This, of course, limited the numbers of

trees used and the numbers of calculations made. For the next

few decades, much work was done in this manner and climate

reconstructions were laboriously produced.

Harold C. Fritts introduced a more biological, plant physiol-

ogy-based research focus to the field of dendrochronology in

the 1950s and 1960s (e.g., Fritts and Fritts, 1955; Fritts,

1956, 1960a). He employed the use of dendrographs that mea-

sured daily changes in stem growth to help in understanding

the climate tree growth relationship in greater detail. He also

introduced high-level statistical methods and the use of main-

frame computing to process large sets of data (e.g., Fritts,

1960b, 1962, 1963). In addition, his work introduced the use

of canonical regression and large regional networks of tree ring

chronologies for the spatial reconstruction of climate (e.g.,

Fritts et al, 1971). These were highly innovative and important

works, made possible by the use of computers, which led to

major advances in the methodology used for reconstructing cli-

mate from tree rings. The list of researchers who made signifi-

cant contributions to the field of dendroclimatology, both

before and after the work of Fritts, is impressive, but goes far

beyond the scope of this chapter. Some important highlights

are noted below.

One of the more important reconstructions of climate from

tree rings was a reconstruction of streamflow from the Color-

ado River (Stockton, 1976; Stockton and Jacoby, 1976) shown

in Figure D32. This work was important because it demon-

strated that the estimates for overall streamflow of the Colorado

River, upon which western states water rights were calculated,

had substantially overestimated the average annual amount of

water by more than 4 million acre feet (maf). The official mea-

surements of 17.3 maf were dominated by data from the early

1900s, which Stockton’s work showed to encompass the wet-

test 30-year period of the past 500 years. The overall average

based on the long tree ring reconstruction was 13.5 maf, and

clearly illustrated the pitfalls of making decisions based on

short-term data.

With the development of microcomputing in the 1970s and

1980s it became possible to generate large numbers of iterations

of data, making it reasonable to reconstruct climate from multi-

ple well-replicated tree ring chronologies. This has led to the pre-

sent day scenario where synoptic scale reconstructions of

climate can be made from hundreds of tree ring chronologies

built from thousands of trees, using desktop computers. A recent

example of such spatial reconstruction work is Cook et al. (1999)

who reconstructed drought indices for the contiguous USA

based on a gridded network of 425 moisture-sensitive tree ring

chronologies that were calibrated and verified against 286 grid-

points of PDSI (Palmer drought severity index). The reconstruc-

tion spanned the period from 1700–1978. This work has recently

been expanded to include 835 tree ring series, and an increased

overall length of reconstruction back to 1

AD for the western

USA. The chronologies and PDSI grid points now extend into

Canada and Mexico, giving far greater scope to the analyses of

past drought. A drought area index (Figure D33) for the western

USA shows evidence of a prolonged period of drought that

lasted for much of the 10th–13th centuries, coinciding with the

Medieval Warm Period (MWP). A current, prolonged drought

in the western USA can now be better assessed within the

Figure D32 Colorado River streamflow reconstruction based on tree ring sites from the river’s catchment. The grey shaded area shows the period

that the Colorado River Treaty was based on, which turned out to be the wettest period of the past 500 years. More water was divided between

states than existed for the long-term mean.

DENDROCLIMATOLOGY 273

context of the natural variability for the past 2,000 years. The

current drought is shown to be orders of magnitude less than

the MWP drought, suggesting dire consequences to a water-hun-

gry society if MWP-like conditions returned. Much of these data

and their reconstructions, as well as numerous other climate

reconstructions from tree rings can be downloaded from the

NOAA paleoclimatology website and the International Tree-

Ring Data Bank (http://www.ncdc.noaa.gov/paleo/ ).

Tree rings and global warming

Tree rings have proven to be one of the most important proxies of

climate in the global warming debate. This is due to their ability

to reproduce high-resolution, absolutely dated records of climate

over broad geographic areas for the past millennium and beyond.

This work has been largely centered on tree ring sites from

the boreal forests where temperature is the most limiting factor

to growth. Early work in Alaska by Giddings (1938, 1943)

showed the potential for working with northern treeline trees

and explored their relationship to climate. Jacoby et al. (1985)

followed up on this research, producing reconstructions of

degree-days in Alaska. Meanwhile some of the first attempts to

reconstruct large scale Arctic and Northern Hemisphere tem-

peratures based on networks of temperate to sub-Arctic trees

were being undertaken (e.g., Jacoby and D’Arrigo, 1989;

D’Arrigo and Jacoby, 1993 ). Much recent attention has been

focused on a multi-proxy reconstruction of Northern Hemi-

sphere temperature by Mann et al ( 1998 ), which included tree

ring series and showed the recent warming trend to be

unmatched for the past 1,000 years. Esper et al. (2002) used net-

works of tree ring sites and novel standardization procedures (see

Dating, dendrochronology) to produce a Northern Hemisphere

temperature reconstruction that placed the recent warming

within the scope of the natural variability of climate for the past

1,000 years, and suggested a MWP warming that rivaled the cur-

rent temperature increase. D’Arrigo et al. (2006) provide a recent

version of a Northern Hemisphere temperature reconstruction

that includes newly collected data and different methodologies

from previous versions. While the scientific debate surrounding

global warming goes on, tree rings remain among the most

important proxies used for paleoclimate research.

Brendan M. Buckley

Bibliography

Akaike, H., 1974. A new look at the statistical model identification. IEEE

Trans. Automat. Contr., AC-19, 716–723.

Blasing, T.J., Solomon, A.M., and Duvick, D.N., 1984. Response functions

revisited. Tree-Ring Bull., 44,1–15.

Box, G.E.P., and Jenkins, G.M., 1970. Time Series Analysis: Forecasting

and Control. San Francisco, CA: Holden-Day.

Briffa, K.R., 1984. Tree-Climate Relationships and Dendroclimatological

Reconstruction in the British Isles. PhD thesis, University of East

Anglia, Norwich, England.

Briffa, K.R., and Cook, E.R., 1990. Methods of response function analysis.

In Cook, E.R., and Kairiukstis, L.A. (eds.), Methods of Dendrochronol-

ogy: Applications in the Environmental Sciences. Dordrecht: Kluwer

Academic, pp. 240–247.

Briffa, K.R., Jones, P.D., Pilcher, J.R., and Hughes, M.K., 1988. Recon-

structing summer temperatures in northern Fennoscandinavia back to

A.D. 1700 using tree-ring data from Scots pine. Arct. Alp. Res., 20,

385–394.

Cook, E.R., and Kairiukstis, L.A. (eds.), 1990. Methods of Dendrochronol-

ogy: Applications in the Environmental Sciences, (1st ed.). Dordrecht:

Kluwer Academic, 394pp.

Cook, E.R., Bird, T., Peterson, M., Barbetti, M., Buckley, B., D’Arrigo, R.,

and Francey, R., 1992. Climatic change over the last millennium in

Tasmania reconstructed from tree rings. The Holocene, 2, 205–217.

Cook, E.R., Briffa, K.R., and Jones, P.D., 1994. Spatial regression methods

in dendroclimatology: A review and comparison of two techniques. Int.

J. Climatol., 14, 379–402.

Cook, E.R., Meko, D.M., Stahle, D.W., and Cleaveland, M.K., 1999.

Drought reconstructions for the continental United States. J. Clim., 12

(4), 1145–1162.

D’Arrigo, R., Wilson, R., and Jacoby, G., 2006. On the long-term context

for late 20th century warming. J. Geophys. Res., 111, D03103, doi:

10.1029/ 2005JD006352.

Figure D33 Tree ring based reconstruction of drought indices for the western USA depicted as drought area indices (DAI), expressed as

percentage of area in drought with a PDSI of <2. This is a smoothed plot that accentuates decadal scale variability. Of particular interest is a

“Medieval Dry Period” of extended drought from the mid 10th to late 13th centuries that is unmatched in intensity for the past 1,200 years (data

courtesy of E.R. Cook).

274 DENDROCLIMATOLOGY