Assessing the vulnerability of endemic diatomspecies in Lake Baikal to predicted future climate change:a multivariate approach

A N S O N W. M A C K AY *, D . B . R Y V E S w , D . W. M O R L E Y *, D . H . J E W S O N z and P. R I O U A L §

*Department of Geography, Environmental Change Research Centre, Pearson Building, University College London, Gower Street,

London WC1E 6BT, UK, wDepartment of Geography, Loughborough University, Loughborough LE11 3TU, zMagherafelt, Co.,

96 Desertmartin Road, Magherafelt, Co. Derry, BT45 5HE, N. Ireland, UK, §Institute of Geology and Geophysics, Chinese Academy

of Sciences, PO Box 9825, Beijing 100029, China

Abstract

Diatoms in Lake Baikal exhibit significant spatial variation, related to prevailing climate,

lake morphology and fluvial input into the lake. Here we have assessed the threats to

endemic planktonic diatom species (through the development of empirical models),

which form a major component of primary production within the lake. Multivariate

techniques employed include redundancy analysis (RDA) and Huisman–Olff–Fresco

(HOF) models. Our analyses suggest that eight environmental variables were significant

in explaining diatom distribution across the lake, and in order of importance these are

snow thickness on the ice, water depth, duration of days with white ice, suspended

matter in the lake, days of total ice duration, temperature of the water surface in July,

concentration of zooplankton and suspended organic matter. Impacts on dominant

phytoplankton diatom species are highlighted using t-value biplots. Predictions of

future climate change on Lake Baikal are likely to result in shorter periods of ice cover,

decreased snow cover across the lake in spring, increased fluvial input into the lake, and

an increase in the intensification of surface water stratification during summer months.

All these factors are likely to impact negatively on the slow-growing, cold-water

endemics such as Aulacoseira baicalensis and Cyclotella minuta, which currently

dominate diatom assemblages. Instead, taxa that are only intermittently abundant, at

present, in offshore areas (e.g. Stephanodiscus meyerii) are likely to become more

frequent. However, given the climatic gradient across the lake, the timing and extent

of changes in community structure are likely to vary. Moreover, palaeolimnological

records show that Lake Baikal diatom assemblages have been dynamic throughout the

Holocene, with both endemic and cosmopolitan species exhibiting periods of domi-

nance. Effects of climate change on the entire lake ecosystem may yet be profound as the

structure of the pelagic food web may change from one based on endemic diatom taxa

to one dominated by nondiatom picoplankton, and as limnological functioning (e.g.

stratification and mixing) affects deepwater oxygen availability, nutrient cycling and

trophic linkages.

Keywords: endemic diatoms, environmental controls, future climate change, Lake Baikal, multivariate

analysis, multivariate ordination, surface sediments

Received 3 February 2006; revised version received 17 April 2006 and accepted 26 June 2006

Introduction

The magnitude and impact of future climate variability

on freshwater ecosystems is uncertain and this is espe-

cially the case in central Asia, where there is a paucity of

studies compared with, for example, European and

North American regions (e.g. Schindler, 2001; Elliott

et al., 2005). Gaining more knowledge becomes impera-

tive because in recent decades wintertime warming in

central Asia has been greater than any other region ofCorrespondence: Anson Mackay, tel. 1 44 20 7679 0588,

fax 1 44 20 7679 0565, e-mail: amackay@geog.ucl.ac.uk

Global Change Biology (2006) 12, 2297–2315, doi: 10.1111/j.1365-2486.2006.01270.x

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd 2297

the world and this trend is expected to continue, with

winter temperatures in central Asia predicted to

increase by between 2 and 5 1C over the next 50 years

(IPCC, 2001). Uncertainty especially exists with regard

to the impacts of changing climate on planktonic

communities and the pelagic food web. For example,

shifts in dominance between phytoplankton com-

munities are likely to result in shifts between food

sources for zooplankton with unknown impacts on

other organisms higher up the food chain. This may

be due to a number of factors, including changes in

autotroph size (e.g. Moline et al., 2004) or disruption

to trophic linkages during key limnological periods

such as thermal stratification and overturn (Winder &

Schindler, 2004).

Lake Baikal in central Asia has long been recognized

as one of the world’s most remarkable freshwater

ecosystems, especially in terms of its age, depth,

volume, continental setting, long sedimentary archive,

and the high degree of biological endemism that exists

within the lake (Kozhova & Izmest’eva, 1998). Diatoms

contribute a major part of primary production (Fietz

et al., 2005) especially to the spring crop, which is

thought to determine annual productivity in the pelagic

lake (Popovskaya, 2000). Monitoring of diatom crops

has been carried out for many decades now, but these

have been largely restricted to the south basin where

pollution impact is greatest. One study suggests that

during the 1950s, productive years in the lake were

dominated by large celled, endemic species belonging

to the genera Aulacoseria (5Melosira) and Cyclotella

(Popovskaya, 2000), although Synedra tends to bloom

following years of peak Aulacoseira production. In the

1990s, small-celled, cosmopolitan species belonging to

the genus Nitzschia were observed in higher numbers in

some years, which has been attributed to both increased

nutrient input into the lake from economic develop-

ment in the catchment (Bondarenko, 1999) and to global

warming (Popovskaya, 2000). However, diatom-moni-

toring studies over the last decade have shown that

increases in Nitzschia cells have not been maintained in

recent years. Moreover, these monitoring studies show

that trends in the south basin are different from trends

in the north basin, and that species’ responses across the

whole lake are complex. Considerable uncertainty,

therefore, exists about possible drivers influencing

planktonic diatom communities in Lake Baikal.

Certainly in records of recently deposited sediments,

extracted from pelagic regions of the lake, there is little

or no response of diatoms to pollution (Mackay et al.,

1998) but there are lake-wide responses to shifts in

climate associated with the period coincident with

the ‘Little Ice Age’ and subsequent global warming

(Mackay et al., 1998, 2005).

Phytoplankton responses to predicted future climate

variability may be assessed using a variety of modelling

techniques. Recent work by Elliott et al. (2005) linked

together a phytoplankton model with a regional climate

model developed by the Hadley Centre, UK, to estimate

changes in phytoplankton communities and biomass in

an English lake up to 2100. However, given the uncer-

tainties of diatom successional trends across the length

of Lake Baikal and limited climate modelling in this

region of central Asia in general, we have adopted an

alternative approach, by developing empirical models

from spatially derived data of sedimentary diatom

assemblages and environmental data.

In recent years, multivariate techniques have been

used to explore the biogeography of diatom species,

and relationships with their changing environment in

a number of regions prone to intense winter cold (e.g.

Laing & Smol, 2000; Lim et al., 2001; Weckstrom &

Korhola, 2001; Ruhland & Smol, 2002; Ryves et al.,

2002; Bouchard et al., 2004). These studies are based

on multiple sites/lakes spread across gradients of in-

terest and, therefore, the assumption is made that most,

if not all, of the species are cosmopolitan. These data are

then used by palaeolimnologists to reconstruct past

environmental changes (e.g. in northern and Arctic

settings; McGowan et al., 2003; Antoniades et al., 2005;

Solovieva et al., 2005). In Lake Baikal, due to its size and

heterogenous sedimentary environments, the distribu-

tion of diatoms exhibits significant spatial variability

(e.g. Bondarenko et al., 1996; Mackay et al., 2003). How-

ever, many of the diatoms are endemic, and their

ecologies are poorly defined in comparison with some

cosmopolitan taxa.

The principal aim of this study is to assess the

potential threats from future climate change on endemic

diatoms in Lake Baikal, which form a principal compo-

nent at the base of the food chain (Yoshii et al., 1999).

However, this paper is distinctive from other diatom–

environment interaction studies in a number of re-

spects. Firstly, we include information derived from

remote sensing to expand the range of explanatory

variables initially used by Mackay et al. (2003), most

notably data on ice, which influences light transmission

into the epilimnion for photosynthesis. Secondly, many

of the canonical ordination techniques used in previous

studies (including Mackay et al., 2003) rely on w2 dis-

tances (e.g. canonical correspondence analysis; CCA),

which can give undue weight to rare species in the

dataset (Legendre & Gallagher, 2001). As diatom assem-

blages in Lake Baikal tend to be dominated by relatively

few species of abundant phytoplankton, the species

data are, therefore, transformed in order that Euclidean

distance methods, such as redundancy analysis (RDA),

can be used to explore relationships between species

2298 A . W. M A C K AY et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

and environmental variables (Legendre & Gallagher,

2001). The data generated in this study will improve

our knowledge of interactions between Lake Baikal

diatom assemblages and environmental variables that

influence their growth, thus, providing information for

attempts to predict the impacts of future climate change

in an aquatic ecosystem whose biodiversity is of global

significance.

Study area

Lake Baikal (511280–551470N and 1031430–1091580E;

Fig. 1) consists of three basins of different ages, sepa-

rated by interbasin underwater highs: the south and

central basins are separated by the Selenga Delta,

while the underwater Academician Ridge separates

the central and north basins. The age and limnological

0 km 20 40 60 80 100

No

rt

he

rn

Ba

si

n

S o u t h e r nB

a s i n

Mi d

dl

eB

as

in

RiverAngara

RiverSelenga

Maloe More

Selenga DeltaRegion

AcademicianRidge

ListvyankaBay

K h a m a r D a b a nM o u n t a i n R a n g e

South basin

Selenga region

Middle basin

Maloe More

Academician Ridge

North basin

N

E011E501

E011E501

Lake Baikal

MONGOLIA

RUSSIA

CHINA

INDIA

Fig. 1 Map of Lake Baikal showing location of each of the 92 sampling sites used to develop the training set for multivariate analysis.

Geographical locations mentioned in the text are highlighted.

L A K E B A I K A L D I A T O M A S S E M B L A G E S 2299

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

characteristics of Lake Baikal has allowed an ecosystem

rich in species endemism to develop, and accordingly

it was designated a World Heritage Site in 1996. The

region around Lake Baikal is one of the most continental

in the world (Lydolph, 1977), with mean daily January

air temperatures of approximately �25 1C, rising

to 119 1C in July. Winters are long, dry and cold, while

summers are relatively short, warm and wet. Lake

Baikal freezes over every year and there is a distinct

north–south gradient in terms of ice duration and

thickness, with the south basin freezing later and thaw-

ing earlier than the north. The lake is distinct from other

deep lakes because its water column is saturated with

oxygen throughout, supporting an extensive, and

almost wholly endemic deep-water fauna. This is due

to both spring and autumnal overturn (i.e. the lake is

dimictic, despite strong stratification at 300 m) and

deep-water renewal from cold surface water intrusions

into deeper waters (Weiss et al., 1991; Wuest et al., 2005).

Methods

Diatom analysis of surface sediments

The dataset used in this study is based on 92 sites from a

training set of 126 surface sediment samples collected

from across the length of Lake Baikal, between 1992 and

1997 (Fig. 1; see Mackay et al., 2003 for full details). The

sedimentary environment of Lake Baikal is heteroge-

neous, and we have considered the impacts of sediment

redeposition, faunal bioturbation and diatom dissolu-

tion on the integrity of sites used in this study. Sites

were discarded if there was (i) visual evidence of

disturbance of the surface sediment–water interface,

(ii) visual evidence of turbidite layers present in the

upper sediments, when cores were opened longitudin-

ally or (iii) very few diatoms present in the surface

sediment (which results in relative abundances not

being representative of likely populations). This was

the case for one site, BAIK111, hence the number of sites

analysed is one less (92) than the 93 sites initially

investigated by Mackay et al. (2003). The surface sedi-

ments are oxygenated, allowing diverse faunal commu-

nities to live at the lake bottom (Martin et al., 2005).

Therefore, it follows that bioturbation will always

be an issue, and is one of the reasons why the strong

seasonal signals in sedimentation are not preserved in

annual laminations. Martin et al. (2005) present evi-

dence that concentrations of animal individuals (and

therefore bioturbation) are higher in shallow water

regions than in the abyssal plains. Moreover, bioturba-

tion probably plays a critical role in the preservation of

diatoms in surface sediments, through physical break-

age and irrigation of upper sediments, both of which

can be expected to enhance silica dissolution (Ryves

et al., 2003). However, we know from our own

high resolution analyses of surface cores that bioturba-

tion does not affect the overall stratigraphy of the

cores, as evidenced from radiometric 210Pb dating

and spheroidal carbonaceous particle analysis (Mackay

et al., 1998).

Given recent debates on the concept of ‘cosmopolitan-

ism’ within diatom research (e.g. Kociolek & Spaulding,

2000; Edlund & Jahn, 2001), diatom taxa were identified

according to a range of published papers and books,

most of which take account of the endemic diatom flora

found in Lake Baikal (e.g. Meyer, 1930; Skvortzov, 1937;

Zabelina et al., 1951; Skabichevskii, 1960; Gleser, et al.,

1988; Khursevich, 1989; Genkal et al., 1992; Flower, 1993;

Kozhova & Kobanova, 1995; Edlund et al., 1996).

Diatom counts were transformed to percentages before

all analyses.

Collation of environmental information

Environmental data in this study come from a number

of sources, including previous Russian studies, our own

measurements of water depth taken during each expe-

dition, and from remote sensing of snow and ice cover-

age across the lake. A description of the collation of

Russian and water depth data has already been detailed

in Mackay et al. (2003), so here we focus on the meth-

odology for the collection of data derived using remote

sensing techniques. Remote sensing is a useful tool in

the study of environmental variables at high spatial and

temporal resolution, especially over large and remote

regions such as Lake Baikal. Snow depth on ice has been

highlighted as one of the most important variables

influencing Lake Baikal diatom species (Mackay et al.,

2003). However, remote sensing of snow thickness/

depth is problematic. The best results are often obtained

from high resolution radar data, combined with an

extensive campaign of measuring actual snow depths

over the scene to calibrate the inferred snow depths (e.g.

Srivastan & Singh, 1991). Owing to the remoteness and

size of Lake Baikal and the dangers involved working

on the lake during the freeze–thaw cycle, it is too

dangerous to collect such ground-truth data. As a

result, this study attempts to investigate the seasonal

dynamics and duration of various ice-cover types on

Lake Baikal based on satellite images in the optical and

infrared spectra. Data for classification of Lake Baikal

ice cover were collated using the AVHRR meteorologi-

cal satellite (Morley, 2005). Similar data were also

collated in the studies of Lake Baikal by Bolgrien et al.

(1995), Le Core (1998), Semovski et al. (2000), Semovski

and Mogilev (2003), Kouraev et al. (in press). Classifica-

tion of snow and ice types on Lake Baikal has been

2300 A . W. M A C K AY et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

previously carried out by Le Core (1998) using the

normalized difference snow index (NDSI) of Riggs

et al. (1994), although this study only exploits general

changes in ice cover characteristics and not, for exam-

ple, ice structure or relative wetness.

Images were chosen between 1996 and 1997, corre-

sponding to the final year in which surface sediments

were taken. AVHRR acquires data at a resolution of

approximately 1.1 km on the ground for each pixel, and

therefore, makes it an ideal tool for studying large sites

(such as Lake Baikal) that span several degrees of

latitude. Initially, the National Oceanographic Atmo-

spheric Administration’s (NOAA) Satellite Active

Archive (SAA) was screened for cloud-free images of

Lake Baikal for 1996–1997, and a total of 34 cloud-free

images of ice cover were obtained (between 10 Decem-

ber 1996 and 2 June 1997). The average time between

images is approximately 5 days, although during May

there are intervals of 10–15 days without cloud-free

images. The AVHRR data were processed using Erdas

Imagine 8.5, and the complete multispectral image was

classified using an unsupervised classification into

groups of pixels using the ISODATA clustering algo-

rithm, which is a centroid-based, iterative approach.

Classes according to coverage type were assigned using

comparisons with Le Core (1998) and by examining the

spectral signatures of certain classes, for example the

relatively higher reflectance of snow compared with

white ice in the visible channel. Pixels were classed into

five categories (open water, white ice, clear ice, wet ice

and snow) and recorded for each diatom sample loca-

tion (Morley, 2005). From these data, the total number of

days of the year when the lake was covered by white

ice, clear ice, wet ice and snow were calculated; a site

location was considered to be covered by a certain type

of ice up until this coverage changes in a subsequent

image. The total time of ice duration was also estimated,

as well as the total time of white ice and snow coverage

together as these are likely to be most limiting to light

penetration. Although only 1 year of data has been used

here, similar patterns have been demonstrated to occur

across the lake in other years covered by the dates of

diatom sample collection (e.g. Bolgrien et al., 1995;

Le Core, 1998; Semovski et al., 2000).

Data analyses

Multivariate datasets, especially those using composi-

tional data, are often analysed using w2 distance metrics

(such as correspondence analysis and the associated

technique of detrended correspondence analysis; DCA),

despite some misgivings with the approach (e.g. Faith

et al., 1987). An underlying issue with w2 distance

methods is that they are overly influenced by rare

species in the dataset, a problem not satisfactorily

addressed by often ad hoc methods of downweighting

included with some analysis programmes (P. Legendre,

personal communication). This is of particular concern

where the aim is to assess the significance of relation-

ships between species and environmental variables

(Legendre & Gallagher, 2001). An alternative approach,

using linear models (Euclidean distance) on suitably

transformed species data, has been recently proposed

(Legendre & Gallagher, 2001) and can be applied to

datasets where gradients of species turnover (as mea-

sured by DCA for example) would imply unimodal

responses, as here. In the absence of clear criteria for

choosing the optimal method a priori, Legendre &

Gallagher (2001) suggest selecting the method that

explains the highest fraction of the species data in direct

ordination. We found a linear model (using RDA) on

square-root transformed relative abundance data (the

Hellinger transformation; Legendre & Gallagher, 2001)

outperformed unimodal models (assessed with CCA),

whether or not species downweighting was applied,

and thus we have adopted this approach here.

All ordination methods were carried out using the

software package CANOCO v. 4.5 (ter Braak &

Smilauer, 2002). Additionally, samples and environ-

mental variables were screened within CANOCO v.

4.5 to determine samples with extreme environmental

values and to exclude those environmental variables

that exhibited high collinearity, which have the effect of

destabilizing the dataset. Relationships between diatom

distributions and environmental variables were initially

examined using the direct gradient technique of RDA,

which constrains ordination axes as linear combinations

of environmental variables. In order to identify a

minimum subset of variables that significantly explain

variation in the diatom data, redundant variables were

removed through a form of step-wise regression (for-

ward selection) together with Monte Carlo permutation

tests (P 5 0.0001; n 5 9999) and associated Bonferroni

corrections (ter Braak & Smilauer, 2002). Separate RDAs

were carried out to determine the explanatory power

of each significant environmental variable without

considering covariables to confirm each was significant

alone. The unique contribution of each of the forward-

selected variables was then assessed through a series of

partial RDAs with the remaining environmental vari-

ables as covariables (Table 1). Significance was again

tested using unrestricted Monte Carlo permutation

tests (P 5 0.0001, n 5 9999 permutations). Significant

positive and negative associations between dominant

planktonic taxa and forward selected environ-

mental variables were further explored using t-value

biplots and associated Van Dobben circles (ter Braak &

Looman, 1994).

L A K E B A I K A L D I A T O M A S S E M B L A G E S 2301

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

Tab

le1

Lim

no

log

ical

,cl

imat

ic,

chem

ical

and

bio

log

ical

dat

au

sed

asp

ote

nti

alex

pla

nat

ory

var

iab

les

for

the

dis

trib

uti

on

of

dia

tom

sin

surf

ace

sed

imen

ts

Cat

ego

ryE

xp

lan

ato

ryv

aria

ble

Co

de

Un

its

Dat

esm

easu

red

Min

imu

mM

axim

um

Mea

n

Sta

nd

ard

dev

iati

on

%ex

pla

ined

sin

gle

var

iab

le

RD

A

P (n5

9999

)

Lim

no

log

yW

ater

dep

th*

Dep

thm

1992

–199

720

.016

78.0

809.

648

0.3

10.1

0.00

01

Tra

nsp

aren

cyin

July

wT

ran

sju

lm

1961

9.0

25.0

21.4

4.6

——

Tra

nsp

aren

cyin

Sep

tem

ber

wT

ran

ssep

m19

612.

011

.06.

02.

0—

—

Alb

edo

inJu

lyw

Alb

edo

%n

/a

5.0

7.0

5.1

0.4

——

Tem

per

atu

reo

fJu

lyw

ater

surf

acez

Tem

pw

s1C

1896

–195

91.

013

.05.

92.

35.

70.

0003

Cli

mat

eA

nn

ual

sola

rra

dia

tio

nw

So

lar

mJm�

2n

/a

4100

.047

00.0

4478

.518

7.6

——

Ab

sorb

edra

dia

tio

nin

July

wA

bso

rbk

calc

m�

2n

/a

13.0

15.0

14.2

1.0

——

Mea

nJu

lyai

rte

mp

erat

ure

wJu

lair

1C

Lo

ng

term

ob

serv

atio

nsw

w9.

015

.012

.21.

7—

—

An

nu

alp

reci

pit

atio

nw

Pre

cip

mm

Lo

ng

term

ob

serv

atio

nsw

w17

5.0

550.

025

7.3

67.7

——

Sn

ow

thic

kn

ess

on

the

lak

ein

Mar

chw

Sn

ow

cm19

72–1

973

2.5

12.5

6.2

3.2

11.0

0.00

01

Ice

thic

kn

ess

inM

arch

§Ic

eth

ick

cm19

72–1

985

65.0

100.

082

.97.

2—

—

Len

gth

of

ice

cov

er§

Icel

eng

thd

ays

Lo

ng

term

ob

serv

atio

nsw

w12

6.0

168.

014

5.3

10.7

——

An

nu

ald

ura

tio

no

fw

hit

eic

eco

ver

age}

Day

sWI

day

s19

96–1

997

076

5317

.17.

80.

0001

An

nu

ald

ura

tio

no

fcl

ear

ice

cov

erag

e}T

clea

rd

ays

1996

–199

70

8516

.222

.6—

—

An

nu

ald

ura

tio

no

fsn

ow

cov

erag

e}T

sno

wd

ays

1996

–199

70

6332

.415

.7—

—

An

nu

ald

ura

tio

no

fw

etic

eco

ver

age}

Tw

etd

ays

1996

–199

710

5732

.011

.0—

—

Ice

cov

erd

ura

tio

n}

Day

sIce

day

s19

96–1

997

8316

813

1.0

24.7

5.7

0.00

05

An

nu

ald

ura

tio

no

fw

hit

eic

e

and

sno

wco

ver

age}

Tw

1s

day

s19

96–1

997

1413

185

.529

.8—

—

Ch

emis

try

N–

NO

3k

Nit

rate

mg

m�

3Ju

ly19

7815

.070

58.1

716

.4—

—

P–

PO

4k

Ph

osp

hat

em

gm�

3Ju

ly19

784.

012

.00

7.73

2.7

——

Su

spen

ded

mat

ter*

*S

usp

Mat

mg

L�

1Ju

ly19

780.

84.

002.

320.

65.

90.

0004

Su

spen

ded

org

anic

C**

Su

spO

rgC

mg

L�

1Ju

ly19

780.

31.

500.

380.

24.

10.

0054

Org

anic

C**

Org

Cm

gL�

1Ju

ly19

781.

11.

701.

330.

2—

—

Org

anic

N**

Org

Nmg

L�

1Ju

ly19

7870

.017

5.00

118.

3824

.2—

—

Org

anic

P**

Org

Pmg

L�

1Ju

ly19

787.

016

.50

9.96

2.1

——

Bio

log

yZ

oo

pla

nk

ton

bio

mas

sin

Sep

tem

ber

wZ

oo

bio

gm�

219

855.

030

.010

.08.

65.

50.

0009

Ori

gin

also

urc

ere

fere

nce

sin

clu

de:

* Dat

afr

om

this

stu

dy

;w A

tlas

Bai

kal

a(1

993)

;z V

erb

olo

vet

al.

(196

5);

§S

him

arae

vet

al.

(199

4);

}M

orl

ey(2

005)

;k V

oti

nse

vet

al.

(197

5);

**T

aras

ov

a&

Mes

cher

yak

ov

a(1

992)

;ww

Lo

ng

term

ob

serv

atio

ns

ov

erth

ela

stce

ntu

ry.

See

tex

tfo

rd

etai

ls.

n/

a,n

ot

avai

lab

le.

2302 A . W. M A C K AY et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

The responses of eight of the main planktonic species

(as untransformed, percentage composition data pre-

sent in at least 10 sites) to the eight significant, modelled

environmental variables were determined by fitting

Huisman–Olff–Fresco (HOF) models (Huisman et al.,

1993) with the software HOF (Oksanen & Minchin,

2002; http://cc.oulu.fi/� jarioksa/). This program fits

the simplest response model from the five defined by

Huisman et al. (1993) to species data using maximum

likelihood estimation and a Poisson error distribution.

These models are a skewed unimodal response (model

V), a symmetrical or Gaussian model (model IV), a

monotonic model with a fixed plateau (model III), a

monotonically increasing or decreasing model (model

II) and finally the null model, which is flat with no

significant trend (model I). This method is particularly

useful when defining the type of species response to an

environmental variable (i.e. if there is a linear, unimodal

or no response). Here, we explore the responses of the

dominant planktonic taxa to each of the significant

modelled variables determined by ordination methods,

and assess the potential impacts of future climate

variability. Gaussian logit regression was used to test

the significance of optima of the HOF models that

exhibited unimodal responses (VI and V) for each of

the common planktonic diatom taxa (GLR version 1.1

Juggins 1994, unpublished program).

Results

Spatial distributions of dominant diatom taxa

Eighty-six diatom species were identified in the surface

sediments from 92 sites. Based on number of occur-

rences and N2 diversity (data not shown), the most

abundant taxa in our dataset were endemic plankton,

although some littoral species were also common,

including Staurosirella pinnata (Ehr.) Williams & Round

and Cocconeis placentula var. euglypta (Ehr.) Grun. Pro-

portional abundances of selected diatom species in the

training-set surface sediments have been mapped to

give an indication of the spatial distribution of species

across the lake (Figs 2a and f). These species have also

been fitted to HOF models: Aulacoseira baicalensis

(Meyer) Simonsen, A. skvortzowii Edlund, Stoermer &

Taylor, Cyclotella baicalensis Skv., Crateriportula inconspi-

cua (Mak. & Pom.) Flower & Hakansson, Cyclotella

minuta (Skvortzow) Antipova., Synedra acus var. radians

Kutz., S. acus var. radians (Kutz). Hust. fo. pusilla,

Stephanodiscus meyerii Genkal & Popovskaya. Mapping

in this way provides a useful, visual guide to species

distribution across the latitudinal gradient encom-

passed by the three basins. It is important to note here

that dissolution processes affect all taxa to different

extents (e.g. Ryves et al., 2003; Battarbee et al., 2005).

This is especially the case for Nitzschia acicularis W.

Smith, a cosmopolitan species commonly found in the

phytoplankton, but which is not recorded in surface

sediments due to dissolution processes that occur

throughout the water column (see Ryves et al., 2003

for full details).

Highest relative abundances of A. baicalensis are

found in the north basin (Fig. 2a), correlated with high

values for snow thickness and white ice formation (see

Fig. 3b), while the area of low abundance in the central

basin (Fig. 2a) occurs in an area of persistent clear ice

(Semovski et al., 2000; Morley, 2005). High values are

also recorded in the south basin, away from the shallow

waters of the Selenga Delta. A. skvortzowii can be found

in surface sediments throughout the lake; valves of both

vegetative cells and resting spores were found in 90 of

92 sites, but only with moderate N2 diversity; data not

shown. However, relative abundances are much greater

in the central region of the south basin than elsewhere

(i.e. relative abundances are very low in the central

region of the central basin, and throughout most of the

north basin; Fig. 2b). C. baicalensis and C. inconspicua

were found in only very small abundances, with no

clear patterns (Figs 2c and d), although it is important to

acknowledge that given the size of cells of the former

species, its contribution to biovolume will be signifi-

cant. C. minuta has a similar distribution to A. baicalensis

although relative abundances in the very north of the

lake are considerably lower (Fig. 2e). S. acus var. radians

fo. pusilla has a much more restricted distribution, only

occurring in the central region of the south basin, with

a couple of occurrences in the central basin (Fig. 2f).

Occurrence of the cosmopolitan species S. acus var.

radians is restricted to the clear ice region of the central

basin (Fig. 2g). There are other occasional occurrences,

mainly in shallower, coastal regions influenced by

fluvial input (e.g. at the Buguldeika Saddle; Fig. 2g).

Taken together, S. acus var. radians agg. is most abun-

dant in the south and central basins, with very little

presence at all in the north basin of Lake Baikal. More-

over, maximum abundance of S. acus coincides with

either circulation currents crossing between the western

shore to the opposite side of the lake in the central

basin, or with the convergence of circulation currents in

the south basin (see fig. 10.3.2 in Shimaraev et al., 1994).

These circulation patterns (which are responsible for the

transport of diatom species from shallow waters into

the pelagic regions of the lake) are dependent on both

whole-lake cyclonic circulation (Verbolov et al., 1965 in

Shimaraev et al., 1994) and local circulation patterns

influenced by fluvial input. S. meyerii (synonymous

with S. binderanus var. baicalensis Popovskaya & Genkal)

is most commonly found in shallow waters of the Maloe

L A K E B A I K A L D I A T O M A S S E M B L A G E S 2303

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

More and the Selenga Delta (Fig. 2h). It is possible that

dissolution processes have less impact at these shallow

sites, although Ryves et al. (2003) in a study in the south

basin, could find no evidence to suggest that water

depth plays a significant role for species highlighted

in this study. It is probable that populations of species

such as S. acus and S. meyerii in the pelagic regions are

sustained by innocula originating from shallower areas

such as the Selenga Delta. This particularly occurs after

ice melt, as water from the Selenga Delta flows down

the west shore of the south basin (Heim et al., 2005).

Multivariate analyses

Environmental variables used in this study are sum-

marized in Table 1. Constrained RDA found eight

variables significant in explaining diatom distributions,

accounting for approximately 44% of the variation in

the diatom dataset (Table 2). In order of importance,

these are snow thickness on the ice, water depth, dura-

tion of days with white ice, suspended matter in the

lake, days of total ice duration, temperature of the water

surface in July, concentration of zooplankton and sus-

pended organic matter. Each variable explains between

approximately 2–5% of the diatom data, independent

of the other seven significant parameters (Table 3).

Together, the unique contribution of these eight vari-

ables is approximately 25% of the total variance in

diatom data, with approximately 19% of the variation

accounted for by covariance between two or more of

the eight variables.

The correlations between these variables and the

distribution of major diatom taxa across the lake (spe-

cies and core locations) can be seen in the RDA biplots

(Figs 3a and b). In terms of species composition, the lake

regions show distinct differences (Fig. 3a). For example,

the shallow water sites towards the centre and north

of Lake Baikal (the Maloe More and the Academician

Ridge) are characterized by the endemic S. meyerii, and

very little snow cover. Sites in the south basin are

dominated by the endemic A. skvortzowii and the

cosmopolitan taxon S. acus var. radians fo. pusilla. These

two taxa are highly correlated with increasing numbers

in zooplankton, and periods of short ice duration on the

Aulacoseira baicalensis (%) Aulacoseira skvortzowii

Cyclotella baicalensis Crateriportula inconspicua

100 %

0 %

100 %

0 %

100 %

0 %

100 %

0 %

(a) (b)

(c) (d)

Fig. 2 Spatial distributions of the percentage abundance of eight of the most abundant diatom phytoplankton found in the full training

set (a–h).

2304 A . W. M A C K AY et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

lake (Fig. 3b). The north basin is most characterized by

A. baicalensis and C. minuta. A. baicalensis is highly

correlated with increasing snow thickness estimates

on the lake, which in turn suggests that this taxon is

more tolerant of low irradiances than other taxa in the

dataset. Culture studies show that this species is also

intolerant of high irradiances (D. Jewson, unpublished

data), which contrasts with reports of Richardson et al.

(2000). However, their estimates of growth were based

on filament length; this can be misleading because with

A. baicalensis longer than normal vegetative cells can be

produced under light stress. Conversely, C. minuta is

highly correlated with ice duration on the lake – (i.e.

longer ice on the lake in spring promotes high abun-

dance of C. minuta during the following autumn turn-

over; Fig. 3b). The shallow waters of the Selenga Delta

are dominated by cosmopolitan plankton, including

S. acus var. radians, Stephanodiscus minutulus (Kutz.)

Cleve & Moller (not shown) and Stephanodiscus parvus

Stoermer & Hakansson (not shown). These latter two

taxa are more commonly found growing during the

summer in more nutrient-rich waters.

Results suggest that climatically controlled variables

(especially during spring, with the presence of ice on

the lake) are important in influencing diatom floristic

variation. Table 4 summarizes the distribution of HOF

models amongst the eight significant environmental

variables. More than half of the models show a signifi-

cant trend. Table 5 shows summary responses of the

eight main taxa to each of the significant environmental

variables, together with the corresponding fitted HOF

model. However, to assess which of these responses are

significant (both positively and negatively) t-value bi-

plots and Van Dobben circles have been constructed

and are shown in Figs 4a and h; ter Braak & Looman,

1994). Species that fall within the Van Dobben circles

are significantly (linearly) correlated (at P 5 0.05) to the

environmental variable, either positively (solid circles)

or negatively (dashed circles).

Snow depth on the lake has a positive, significant

effect on A. baicalensis and C. minuta, but a negative

impact on S. meyerii (Fig. 4a). Thus, we would expect

that, should future climate impacts result in a decline in

snow cover on the lake (see Discussion), then there is

Synedra acus v.radians fo. pusilla

Synedra acus v.radians

Cyclotella minuta (%)

Stephanodiscus meyerii (%)

100 %

0 %100 %

0 %

100 %

0 %

100 %

0 %

(e) (f)

(g) (h)

Fig. 2 (Continued)

L A K E B A I K A L D I A T O M A S S E M B L A G E S 2305

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

C. minuta

A. baicalensis

A. skvortzowii

C. inconspicua S. meyerii

S. acus v. radius

C. baicalensis

S. acus v. radians fo. pusilla

Depth

Snow

Suspmat

Zoobio

DaysWI

DaysIce

C. minuta

A. baicalensis

A. skvortzowii

C. inconspicua S. meyerii

S. acus v. radians

C. baicalensis

S. acus v. radians fo. pusilla

(a) (b)

−1.0−1.0

1.0

1.0−1.0

−1.0

1.0

1.0

Susporgc

Tempws

Fig. 3 Redundancy analysis (RDA) biplots showing (a) species and coded site locations and (b) species, abbreviated significant

explanatory variables and coded site locations. The coded site locations are previously outlined in Fig. 1. There are too many species to

identify individually in one diagram, and so in (b) all species other than the eight highlighted above are suppressed from view.

Table 2 Redundancy analysis (RDA) of species data (86 taxa) with eight forward selected variables using Hellinger’s distance

(Legendre & Gallagher 2001)

RDA Axes

1 2 3 4 Total variance

Eigenvalue 0.210 0.112 0.051 0.029 1.000

P 0.0001

Cumulative % variance of explanatory data 21.0 32.2 37.4 40.3

Sum of all eigenvalues 1.000

Sum of all canonical eigenvalues; P 5 0.0001 0.437

Table 3 Unique contributions to diatom variation of each of the eight significant environmental variables as determined using

RDA, with the remaining seven variables as co-variables, in order

No. Variable Eigenvalue % explained

P value for unique

explanation (n 5 9999)

1 Snow 0.048 4.8 0.0001

2 Depth 0.044 4.4 0.0001

3 DaysWI 0.034 3.4 0.0001

4 SuspMat 0.026 2.6 0.0004

5 DaysIce 0.025 2.5 0.0005

5 Tempws 0.025 2.5 0.0003

7 Zoobio 0.024 2.4 0.0009

8 SuspOrgC 0.018 1.8 0.0054

Total 8 vars 24.4 (combined)

Including covariances, % explained of eight variables 5 43.7% (P 5 0.0001, n 5 9999).

2306 A . W. M A C K AY et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

likely to a be a decline in deep water, pelagic, endemic

diatoms. Depth of water in the lake elicits a similar

response to snow cover (Fig. 4b), but predictions of

future climate change are not likely to have a significant

effect on water depths across the lake in the time scale

of 100 years, as precipitation is greater than evaporation

in all models. Days of white ice duration on the lake

have a positive impact on A. baicalensis but negative

impact on S. meyerii (Fig. 4c; i.e. low irradiances under

the ice favour A. baicalensis). Suspended matter has a

significantly positive impact on A. skvortzowii only (Fig.

4d), which in the south basin is mainly derived from

Selenga River input (Heim et al., 2005). Ice duration (Fig.

4e) on the lake positively affects C. minuta, and this is

likely to be related to the fact that net increase in cells of

this species occurs during autumn overturn (see

’Discussion‘). The temperature of the water surface

during July also impacts significantly on several species

(Fig. 4f). Higher water temperatures have a negative

impact on diatom distribution on both the endemic

Aulacoseria species. However, warmer temperatures

are associated with significantly higher abundances of

S. meyerii, and it may be expected that predicted

changes in climate will result in higher abundances of

this species across the lake. Concentrations of zooplank-

ton, especially the endemic mesozooplankton Epischura

baicalensis Sars. are associated with lower abundances of

C. minuta, because these animals track populations of

C. minuta as a food source (Fig. 4g). Finally, within our

datasets, suspended organic matter in the lake does not

significantly impact on any of the eight species high-

lighted here (Fig. 4h), and only explains under 2% of

total diatom variation (Table 3).

Discussion

Ecological responses of dominant diatom taxa

A. baicalensis is most productive in areas of opaque ice

(e.g. because of snow cover or white ice) rather than

clear ice, which appears to be detrimental to growth.

This implies that increasing light penetration (e.g.

through clear ice) inhibits growth of this taxon. How-

ever, unlike C. minuta, there is only a weak correlation

with ice duration on the lake (Fig. 3b), which will have

implications for spatial responses within Lake Baikal to

future climate change (i.e. between the north and south

basins). In the north basin, ice break-up occurs later in

the spring, and subsequent warming of the lake water

delayed, compared with the south basin. In some years

in the north basin, spring turnover period can be con-

sequently shorter too (Shimaraev et al., 1994). Impor-

tantly, summer stratification is shorter in the north

basin, during which time there is a shift from diatom-

Table 4 Distribution of HOF models amongst the eight significant environmental variables

Model Snow Depth White ice SuspMat Ice dur Tempws Zoobio SuspOrgC Total %

V 1 6 0 1 4 3 0 0 17 4.5

IV 9 7 10 8 6 6 3 9 70 18.5

III 0 0 1 0 0 0 0 0 1 0.2

II 15 16 11 18 14 8 19 13 119 31.4

I 17 13 20 15 20 25 20 19 172 45.4

Note that the majority of the models are linear (II 1 I), 291/379 5 77%.

Table 5 HOF responses for each of the common planktonic taxa in Lake Baikal

Snow Depth White ice SuspMat Ice dur Tempws Zoobio Susp OrgC

Aulacoseira baicalensis II1 V* II1 IV* II1 IV* I IV*

A. skvortzowii IV* II1 IV* V II� V II1 I

Stephanodiscus meyerii II� IV* II� II� V* I II� II�Cyclotella minuta II1 V* II1 I II1 IV* II� I

Synedra acus v. radians II� I II� IV* IV* I I IV*

Crateriportula inconspicua I I I II1 I I II� I

S. acus v. radians f. pusilla I II1 IV* IV IV* II� II1 II�Cyclotella baicalensis I II1 I I II1 IV* I I

Where unimodal models are apparent (i.e. IV and V), significant optima determined using GLR are indicated with an asterisk.

L A K E B A I K A L D I A T O M A S S E M B L A G E S 2307

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

(a) (b)

(c) (d)

(e) (f)

(g) (h)

2308 A . W. M A C K AY et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

dominated phytoplankton to dominance by autotrophic

picoplankton, especially the cyanobacteria Synechocystis

limnetica Popovskaya (Popovskaya, 2000; Fietz et al.,

2005).

A. skvortzowii populations can develop both in spring

in pelagic regions of the lake, and during autumn

turnover in shallow-water bays (e.g. Bolshiye Koty

and Barguzin Bay; Fig. 1) (Kozhova & Izmest’eva,

1998). According to the HOF models, growth of A.

skvortzowii under increasing snow and ice cover is

inhibited when compared with the deep-water ende-

mics, A. baicalensis and C. minuta. Thus a combination of

environmental factors, including earlier ice out time

and warmer waters in the south basin (as suggested

by RDA, Fig. 3b) are likely to be responsible for the

spatial distributions observed here.

Like A. baicalensis, C. minuta is exclusively pelagic.

Although this species can be found growing under the

ice in spring (when it usually auxosporulates; Kozhova

& Kobanova, 1995), the main period of net increase in

cell numbers occurs during autumn turnover when

nutrients become available again and concomitant

water transparency is low (confirmed by RDA and the

high correlation with increased suspended matter in

the water column; Fig. 3b). High relative abundance of

this species may be related to periods of warmer waters

and a dominance of autumnal over spring productivity,

due to either longer summers (Bradbury et al., 1994) or

restriction of the size of the spring crop relative to the

autumn crop, due to long, cold winters, with concomi-

tant increased ice duration on the lake. The latter reason

has been invoked by Mackay et al. (2005) to explain the

dominance of this species in the sediment record during

the period commonly referred to as the ‘Little Ice Age’

(approximately AD 1200–1840). At this time, there was a

weakening of the North Atlantic Oscillation (NAO;

Shindell et al., 2001) and the expansion of the Siberian

High (SH) across central Asia (Krenke & Chernavskaya,

2002). Our data suggest, therefore, that factors that

promote a decline in ice and snow cover during spring

(or variables that are linearly related to these factors),

will also have a negative impact on populations of

C. minuta the following autumn.

In palaeoecological studies, abundance of both forms

of S. acus is interpreted as being associated with in-

creased concentrations of silica, related to increased

catchment run-off and river discharge (Kilham &

Kilham, 1990; Bradbury et al., 1994). The distribution

of S. acus observed here reflects these parameters. For

example, biogeochemical silica mass balance (Callender

& Granina, 1995) shows that the production of diatoms

in Lake Baikal is supported mostly by remineralization

processes in the water column, accompanied by signifi-

cant contribution from fluvial inputs of dissolved silica,

especially from the Selenga River which flows into the

central and south basins. Very low relative abundances

of S. acus agg. in the north basin are likely, therefore, to

be related to increased snow and ice cover (as suggested

by RDA, Fig. 3b), together with lower concentrations of

silica input via rivers than in the other two basins

(Shimaraev et al., 1994). It is also worth noting that the

valves of S. acus suffer considerable breakage and

dissolution in the water column at the surface sedi-

ment–water interface (Ryves et al., 2003; Battarbee et al.,

2005). It is likely, therefore, that the distribution of this

species is under-represented in the sediment record.

In agreement with the Kilham & Kilham (1990)

hypothesis for Si : P ratios for cosmopolitan diatoms in

African Great Lakes (i.e. Synedra4Nitzschia4Stephano-

discus), Bradbury et al. (1994) suggested from palaeoe-

cological studies that phosphorus concentrations are

likely to be important for growing populations of

S. meyerii. This will occur when nutrient dynamics are

dominated by circulation, promoting a long period of

spring overturn and large release of nutrients from

bottom sediments. A recent study by Muller et al.

(2005) demonstrated that flux rates for phosphorus

(between the surface sediment–water interface) were

significantly higher in the Maloe More than elsewhere

in the lake during the period investigated (1994–2003),

agreeing with the assertion by Bradbury et al. (1994) that

high phosphorus supply is an important factor for this

species. S. meyerii exhibits a linear response to snow

depth over the gradient sampled, with highest abun-

dance at minimum snow depths (Fig. 4a; Table 5). RDA

also shows that S. meyerii is found in highest abun-

dances in shallow regions of Lake Baikal with low levels

of snow cover when the lake is frozen. The Maloe More

is subjected to high winds (Shimaraev et al., 1994), while

the region is also semi-arid, receiving less than 200 mm

of precipitation annually compared with the 300–

400 mm falling annually in the south basin (Atlas

Baikala, 1993). Both these factors result in low levels

of snow accumulation on the ice of frozen Maloe More

during spring. This fits with the assumption of

Bradbury et al. (1994) that this species is indicative of

Fig. 4 t-Value biplots, showing relationships between the eight highlighted species and each of the significant explanatory variables

(a 5 Snow; b 5 Depth; c 5 DaysWI; d 5 SuspMat; e 5 DaysIce; f 5 Tempws; g 5 Zoobio; h 5 SuspOrgC). Species that fall within the Van

Dobben circles are significantly (linearly) correlated (at P 5 0.05) to the environmental variable, either positively (solid circles) or

negatively (dashed circles).

L A K E B A I K A L D I A T O M A S S E M B L A G E S 2309

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

a dry climate (less snow cover) dominated by water

column circulation and nutrient regeneration.

Future impacts of climate change on Lake Baikal diatomassemblages

In continental regions in the northern hemisphere, ice

cover is an important variable in determining function-

ing of lake ecosystems. This is especially the case with

regard to the interactions between diatom communities

and ice and snow cover in locations vulnerable to global

warming, such as the Arctic (e.g. Douglas et al., 1994;

Sorvari et al., 2002; Smol et al., 2005), and high altitude

regions such as the European Alps (e.g. Lotter & Bigler,

2000; Ohlendorf et al., 2000) and in the Canadian Cor-

dillera (e.g. Karst-Riddoch et al., 2005). Ice cover plays a

major role in the stratification of the water column,

which in turn has an influence on lake mixing and

nutrient cycling processes (Shimaraev et al., 1994). A

recent study of 26 sites around the northern hemisphere

has shown that ice cover on lakes and rivers has been

responding to global warming since the mid 19th

century, with net reduction in ice duration through

combinations of later ice formation and earlier ice

break-up (Magnuson et al., 2000). One of the case

studies used in this investigation was an ice phenology

record from Lake Baikal, near the mouth of the Angara

River (Fig. 1). This record has been monitored continu-

ously since 1869, and its relationship to northern hemi-

sphere (NH) temperatures and large-scale climate

indices such as the NAO and the Arctic Oscillation

(AO) have been examined in detail (e.g. Livingstone,

1999; Shimaraev et al., 2002; Todd & Mackay, 2003).

These studies are important as they provide plausible,

mechanistic links between regional climate and its

impact on biology, especially the primary producers

within Lake Baikal. Todd & Mackay (2003) have argued

that, because some modes of NH variability are them-

selves predictable, then aspects of ice cover change on

Lake Baikal ought be predictable too and, therefore, it

should be possible to forecast some of the changes in the

biology that may occur. More recently, Panagiotopoulos

et al. (2005) detail a decline in the strength of the

Siberian High since the 1970s, which is linked to in-

creases in precipitation in central Asia. This decline is

projected to continue at least until approximately 2100,

and while it has not been possible to determine the

cause of the decline, future climate general circulation

model predictions suggest a negative correlation

between increases in greenhouse gases (GHGs) and

the intensity of the SH. Here, we will discuss some of

the possible impacts of future climate change on the

diatom populations in Lake Baikal as determined from

our multivariate approach.

Future climate impacts will result in:

(i) Shorter ice duration: Using scenarios of anthropo-

genic greenhouse gas emissions, model predictions of

climate variability over the next 100 years show pro-

nounced warming during winter in the Lake Baikal

region (IPCC, 2001). For example, under conditions of

a 1% annual increase in CO2, model simulations suggest

that, during the period 2071–2100, mean winter tem-

perature changes over the Northern Asian region range

from 1 4.8 to 1 9.3 1C, relative to the 1961–1990 period

(IPCC, 2001). Todd & Mackay (2003) calculated that,

assuming a linear dependence of Lake Baikal ice

duration on winter temperature, ice duration in the

south basin may decrease by a further 15–28 days

by 2071–2100. This decrease is rather less than that

estimated by Shimaraev et al. (2002) who calculated

that ice duration may decrease by as much as 2 months

in the south basin. Our analyses suggest that shorter

ice duration will favour growth of species which pre-

sently develop in shallow and coastal water regions

(e.g. A. skvortzowii, S. meyerii and S. acus; Fig. 4e),

whose cells are then transported into the open

water through circulation processes in the lake. Con-

versely, populations of the pelagic, endemic species

C. minuta are likely to decline, as ice duration gets

shorter (Fig. 4e).

(ii) Reduced snow cover extent: Snow cover extent in

central Asia is controlled by a number of factors, most

notably the availability of precipitation (Clarke et al.,

1999) linked to changes in e.g. Atlantic sea surface

anomalies and the associated NAO (Ye, 2001). Snow

cover on frozen Lake Baikal is the strongest single

variable in explaining variation in diatom assemblages

(Tables 1 and 3). Light is important because it has a

direct influence on algal photosynthesis (and primary

production) and provides the energy needed for con-

vective mixing to keep diatom cells suspended in the

photic zone (Granin et al., 2000). On Lake Baikal, depth

of surface snow often exceeds 10 cm, which results in

light becoming limiting for diatom cells in the water,

and mixing processes to weaken, causing heavier dia-

toms to sink out of the photic zone (Granin et al., 2000).

A. baicalensis is adapted to low light conditions and,

therefore, snow cover offers this species a competitive

advantage over other taxa, which partly explains its

dominance in the north basin of Lake Baikal. However,

should snow cover impede too much light, then growth

of this taxon too is inhibited (Mackay et al., 2005).

Regional analysis of snow cover extent indicates that

between the 1970s to the middle of the 1980s snow

cover over Eurasia was more extensive than between

late 1980s to late 1990s (Robinson & Frei, 2000), while

model predictions of future global warming suggests

that there will be less overall winter snow accumulation

2310 A . W. M A C K AY et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

(Barnett et al., 2005). Our models suggest a decline in

snow cover associated with global warming will have a

negative impact on deep-water pelagic species such as

A. baicalensis and C. minuta, while promoting net growth

of more opportunistic, high-irradiance tolerant species

(both cosmopolitan and endemic), including S. meyerii

(Table 5; Figs 3b and 4a).

(iii) Increased amounts of precipitation: Shimaraev et al.

(2002) report that over the last 100 years, there has been

a positive relationship between increasing temperatures

and hydrological inflow into the lake, which also corre-

late to observed trends in increasing levels of precipita-

tion in central Asia (Aizen et al., 2001). As precipitation

levels at the latitude of Lake Baikal are estimated to

increase by 5–10%, Shimaraev et al. (2002) further esti-

mate that water inflow via Lake Baikal tributaries will

increase by 5–10% too. It is likely, therefore, that there

will be an increase in more generalist species with

wider tolerances being transported into the pelagic

regions of the lake. For example, t-value biplots (e.g.

Fig. 4d) highlight that increases in suspended matter are

significantly associated with the endemic A. skvortzowii,

although resuspension of this taxon from littoral

regions during autumn turnover will be important

too. Increased fluvial input into Lake Baikal will also

result in increased silica and nutrient input into the

lake, which in turn will encourage growth of species

such as S. acus even in the north basin, because of its

higher requirement for silica (Bradbury et al., 1994).

There is considerable palaeolimnological evidence that

S. acus was in the past a dominant component of the

diatom flora across the length of Lake Baikal, linked to

prevailing warmer and wetter climates (e.g. Bradbury

et al., 1994; Karabanov et al., 2000; Mackay et al., 2005;

Morley, 2005). Other cosmopolitan species, such as

Asterionella formosa Hassall, A. subarctica (O. Muller)

Haworth, A. granulata (Ehrenberg) Simonsen, which

can be relatively abundant locally in coastal areas,

would also likely benefit from such changes, and evi-

dence is growing that these increases may already be

underway (e.g. Kozhova & Izmest’eva, 1998; Mackay

et al., 1998).

(iv) An increased period of summer stratification: Pre-

dicted increases in temperature are also likely to have

an important impact on the length (i.e. longer, and

increased stability) and depth of summer and autumn

surface-water stratification. An increase in the stratifi-

cation period will have a negative impact on the success

of the resting stages produced by A. baicalensis and C.

minuta, as they slowly sink through the water column to

avoid warmer conditions and nutrient depletion (D.

Jewson & N. Granin, unpublished data). If turnover

during autumn is delayed, then the resting cells of these

species will sink further in the water column during

stratification, making their resuspension less likely,

resulting in a smaller number of cells being entrained

back into the photic zone for the following years in-

oculum. Thus, under a warmer climate, populations of

autumnal taxa such as the slow growing C. minuta will

decline.

We acknowledge, however, that there is a consider-

able amount of uncertainty, especially with respect to

the extent of increasing summer and autumnal stormi-

ness in the Lake Baikal region, which may actually

increase resuspension of cells from deeper waters or

from littoral regions. Another major impact resulting

from increased stratification will be a shift in balance

between large phytoplankton (i.e. diatoms) and auto-

trophic picoplankton (especially Synechocystis limnetica;

Fietz et al., 2005). Picoplankton dominate Lake Baikal

stratified waters, and because compared with diatoms

relative cell-size is small, mass growth occurs due to

high rates of nutrient assimilation and high rates of cell

division (Popovskaya, 2000). The distribution of pico-

plankton in the lake varies greatly with region, thereby

again ensuring a complex lake-wide response to global

warming (Fietz et al., 2005). Nevertheless, such shifts are

likely to have important implications for trophic inter-

actions in the Lake Baikal aquatic ecosystem, especially

between primary producers and consumers. Our ana-

lyses suggest that increases in zooplankton biomass are

linked to a decline in C. minuta, and a relative increase

in coastal species, including A. skvortzowii and S. acus v.

radians fo. pusilla. We are not at a stage yet to assess how

such changes will impact on complex consumers such

as the endemic gammarids (e.g. Macrohectopus branickii),

or how these changes will be propagated up the food

chain. However, the effects are likely to be important for

many endemic fish and may have consequences at all

trophic levels over time, even affecting the endemic

freshwater seal, Phoco sibirica (Yoshii et al., 1999). Thus,

although the magnitude and nature of the impacts of

such changes on Lake Baikal’s biodiversity and ecolo-

gical functioning are uncertain, there may be major

consequences for the conservation and economic

value of the lake. For example, a recent study on

another deep rift-lake rich in biodiversity, Lake

Tanganyika, has shown that recent climate warming

has strengthened lake stratification, which has had

a negative impact on nutrient recycling, causing a

decline in algal biomass and subsequent fish stocks

(O’Reilly et al., 2003).

Confounding factors and limitations of study

Linked to variations in diatom assemblages caused by

climate change, are confounding factors such as poten-

tial increases in pollutant input. Pollutant input occurs

L A K E B A I K A L D I A T O M A S S E M B L A G E S 2311

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

especially through poorly regulated land-use in the

catchment of Lake Baikal, most notably around the

southern basin and the Maloe More region, west of

the central basin (Fig. 1). For example, Popovskaya

(2000) highlights how in the shallow waters of the

Selenga Delta region, standing crops of phytoplankton

are approximately eight times higher in the 1980s than

they were in the 1960s, and that this figure is still

increasing. These large increases are attributed to

anthropogenic eutrophication of the Selenga River

and associated shallow waters of the Selenga Delta.

Evidence of eutrophication is also borne out by the

palaeolimnological record; increases in relative

abundances of small centric taxa more commonly asso-

ciated with nutrient rich waters (e.g. S. minutulus

and A. formosa) were found in a short core extracted

from the shallow waters of the Selenga Delta (fig. 11 in

Mackay et al., 1998). Recent attempts have been

made to model the impact of anthropogenic disturbance

on the lake’s biota using an ecosystem disturbance

model (Silow et al., 2001). Initial results suggest all

components of the planktonic communities become

disturbed, especially under-ice diatom communities,

together with an overall increase in nutrient concentra-

tion and increase in summer phytoplankton biomass.

These changes are similar to those predicted by

future climate variability and, therefore, it is now

important that refined models are developed to take

account of interactions between localized eutrophica-

tion and climate change impacts on the lake’s

ecosystem.

In Lake Baikal, diatom growth is only limited by Si

every third or fourth year, when there is a high abun-

dance of A. baicalensis (so-called Melosira years). While

the specific causes for Melosira years are still uncertain,

their occurrence is influenced by the interaction bet-

ween its life cycle and factors that influence population

growth, such as nutrients and grazing. We acknowl-

edge, therefore, the likelihood that our interpretations

underestimate the full complexity in the Lake Baikal

ecosystem. For example, we have not considered re-

sponses to Si in our training set, which other diatom-

based studies have found to be important (e.g. Ruhland

& Smol, 2002). Furthermore, phosphate and nitrate are

drawn down (by nondiatomaceous phytoplankton)

each year to undetectable concentrations in the upper

layers during summer stratification. Changes in future

climate will undoubtedly affect the availability and use

of nutrients, such as phosphate and nitrate, by aquatic

algae in the lake, and these responses have not as yet

been taken into account. This is partly because our

knowledge of the interactions between Lake Baikal

phytoplankton and nutrients is still relatively poor,

but also because we have not yet been able to compile

modern values for many of these nutrients, including

Si. Moreover, although diatoms responses will vary

spatially between the three lake basins, a single lake-

wide response is unlikely, confounded with the knowl-

edge that for some species in our dataset (e.g. cosmo-

politan taxa such as S. acus), we will not have captured

full environmental gradient responses.

As highlighted above, dissolution of diatom valves is

a significant process in both the water column and the

surface sediment–water interface (Ryves et al., 2003).

Dissolution causes differential preservation of diatom

species in Lake Baikal – for example, C. minuta is

relatively resistant to dissolution, whereas other taxa,

including N. acicularis, S. meyerii and S. acus are very

susceptible to dissolution (Battarbee et al., 2005).

N. acicularis is a common diatom found in Lake Baikal,

especially in the south basin. However, because

virtually no valves preserve in the surface sediments,

we are unable to model impacts of climate change on

this species, and we acknowledge that this is a potential

weakness in our methodology. Nevertheless, all other

taxa are represented, and although differential dissolu-

tion means we cannot provide absolute quantitative

estimates of how diatom assemblages will change

in the future, our qualitative interpretations remain

robust.

In this study, we have demonstrated that spatial

differences in limnological characteristics linked to pre-

vailing climate have a significant impact on the distri-

bution of both endemic and cosmopolitan diatoms in

Lake Baikal. We have modelled these relationships

using multivariate techniques. We have then used these

models to predict impacts of future climate change on

diatoms in this globally significant ecosystem. Our

analyses indicate that the responses to future climate

change are complex, both in terms of impacts on

individual species, but also on basin-wide differences

between the main regions of the lake. Overall, we

predict that by 2100, a decline in ice duration, decreased

snow cover on the lake, increased fluvial input and

increased stratification of the upper waters during

summer months will result in an increase in endemic

and cosmopolitan diatoms which at present dominate

shallow-water, offshore regions. However, these in-

creases will occur at the expense of slow-growing,

cold-water pelagic endemics. Ongoing monitoring of

phytoplankton communities across the lake will pro-

vide strong evidence to test our predictions from sur-

face sediment diatom assemblages and quantitative

data on population change. If our predictions are cor-

rect, such programmes will act as vital early warning

indicators of ecological change in Lake Baikal, and their

continuation needs to be supported by the international

scientific community.

2312 A . W. M A C K AY et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

Acknowledgements

We would like to thank the various funding agencies that havehelped contribute to this work, especially UK NERC (GR3/10529; NER/S/A/2001/06430), EU FPV Continent (EVK2-CT-2000-0057) and The Royal Society. We thank Nick Granin at theLimnological Institute, Irktusk, with help in accessing relevantRussian information. We thank Pierre Legendre for advice on theuse of multivariate statistics. Finally, we are grateful for thecomments from three reviewers, which have helped to improvethis paper.

References

Aizen EM, Aizen VB, Melack JM et al. (2001) Precipitation and

atmospheric circulation patterns at mid-latitudes of Asia.

International Journal of Climatology, 21, 535–556.

Antoniades D, Douglas MSV, Smol JP (2005) Quantitative esti-

mates of recent environmental changes in the Canadian High

Arctic inferred from diatoms in lake and pond sediments.

Journal of Paleolimnology, 33, 349–360.

Atlas Baikala (1993) Baikal Atlas. Federal Service of Surveying

and Cartography, Moscow (in Russian).

Barnett TP, Adam JC, Lettenmaier DP (2005) Potential impacts of

a warming climate on water availability in snow-dominated

regions. Nature, 438, 303–309.

Battarbee RW, Mackay AW, Jewson D et al. (2005) Differential

dissolution of Lake Baikal diatoms: correction factors and

implications for palaeoclimatic reconstruction. Global and

Planetary Change, 46, 75–86.

Bolgrien DW, Granin NG, Levin L (1995) Surface temperature

dynamics of Lake Baikal observed from AVHRR images.

Photogrammetric Engineering and Remote Sensing, 61, 211–216.

Bondarenko NA (1999) Floral shift in the phytoplankton of Lake

Baikal, Siberia: recent dominance of Nitzschia acicularis.

Plankton Biology and Ecology, 46, 18–23.

Bondarenko NA, Guselnikova NE, Logacheva NF et al. (1996)

Spatial distribution of phytoplankton in Lake Baikal, Spring

1991. Freshwater Biology, 35, 517–523.

Bouchard G, Gajewski K, Hamilton PB (2004) Freshwater diatom

biogeography in the Canadian Arctic Archipelago. Journal of

Biogeography, 31, 1955–1973.

Bradbury JP, Bezrukova YV, Chernyaeva GP et al. (1994) A

synthesis of post-glacial diatom records from Lake Baikal.

Journal of Paleolimnology, 10, 213–252.

Callender E, Granina L (1995) Biogeochemical silica mass bal-

ance in Lake Baikal, Russia. In: Proceedings 8th International

Symposium of Water–Rock Interaction, pp. 341–344. Balkema,

Rotterdam.

Clarke MP, Serreze MC, Robinson DB (1999) Atmospheric con-

trols on Eurasian snow extent. International Journal of Climatol-

ogy, 19, 27–40.

Douglas MSV, Smol JP, Blake Jr W (1994) Marked post-18th

century environmental change in high-arctic ecosystems.

Science, 266, 416–419.

Edlund MB, Jahn R (2001) Report of a workshop on ‘Biogeogra-

phy and endemism of diatoms’. In: Proceedings 16th Interna-

tional Diatom Symposium (ed. Economou-Amilli A), pp.

575–587. Koeltz Scientific Books, Koenigstein.

Edlund MB, Stoermer EF, Taylor CM (1996) Aulacoseira skvortzo-

wii sp. nov. (Bacillariophyta), a poorly understood diatom

from Lake Baikal, Russia. Journal of Phycology, 32, 165–175.

Elliott JA, Thackeray SJ, Huntingford C et al. (2005) Combining

a regional climate model with a phytoplankton community

model to predict future changes in phytoplankton in lakes.

Freshwater Biology, 50, 1404–1411.

Faith DP, Minchin PR, Belbin L (1987) Composition dissimilarity

as a robust measure of ecological distance. Vegetatio, 69, 57–68.

Fietz S, Kobanova G, Izmest’eva L et al. (2005) Regional, vertical

and seasonal distribution of phytoplankton and photosyn-

thetic pigments in Lake Baikal. Journal Plankton Research, 27,

793–810.

Flower RJ (1993) A taxonomic re-evaluation of endemic

Cyclotella taxa in Lake Baikal, Siberia. Nova Hedwigia, 106,

203–220.

Genkal SI, Glezer ZI, Davydova NN et al. (1992) Diatom algae of

the USSR: relict and modern, Vol. II (ed. Makarova IV),

pp. 7–48. Nauka, St. Petersburg (in Russian).

Gleser SI, Makarova, Moisseeva AI et al. (eds) (1988) The Diatoms

of the USSR, Fossil and Recent, Vol. 2. Nauka, St. Petersburg (in

Russian).

Granin NG, Jewson DH, Gnatovsky RY et al. (2000) Turbulent

mixing under ice and the growth of diatoms in Lake Baikal.

Verhein International Vereinigung Limnologie, 27, 1–3.

Heim B, Oberhaensli H, Kaufmann H (2005) Variation in Lake

Baikal’s phytoplankton distribution and fluvial input assessed

by SeaWiFS satellite data. Global and Planetary Change, 46, 9–27.

Huisman J, Olff H, Fresco LFM (1993) A hierarchical set of

models for species response analysis. Journal of Vegetation

Science, 4, 37–46.

IPCC (2001) Climate Change 2001: The Scientific Basis. Working

Group 1 contribution to the IPCC Third Assessment Report

(Summary for Policy Makers) (http://www.ipcc.ch).

Karabanov EB, Prokopenko AA, Williams DF et al. (2000) A new

record of Holocene climate change from the bottom sediments

of Lake Baikal. Palaeogeography, Palaeoclimatology and Palaeoe-

cology, 156, 211–224.

Karst-Riddoch TL, Pisaric MFJ, Smol JP (2005) Diatom responses

to 20th century climate-related environmental changes in

high-elevation mountain lakes of the northern Canadian

Cordillera. Journal of Paleolimnology, 33, 265–282.

Khursevich GK (1989) Atlas of Stephanodiscus and Cyclostephanos,

Bacillariophyta, from the Upper Cenozoic deposits of the

USSR. (ed. Velichkevich FY), 167 pp. Nauka & Teknika, Minsk

(in Russian).

Kilham P, Kilham SS (1990) Endless summer, internal loading

processes dominate nutrient cycling in tropical lakes. Fresh-

water Biology, 23, 379–389.

Kociolek JP, Spaulding SA (2000) Freshwater diatom biogeogra-

phy. Nova Hedwigia, 71, 223–241.

Kouraev A, Semovski SS, Shimaraev MN et al. (in press)

Observations of Lake Baikal ice from satellite altimetry and

radiometry. Remote Sensing of Environment.

Kozhova OM, Izmest’eva LR (1998) Lake Baikal: Evolution and

Biodiversity. Backhuys Publishers, Leiden.

Kozhova OM, Kobanova GI (1995) On the morphology and

systematics of centric diatom algae of Lake Baikal. In: Ecologi-

L A K E B A I K A L D I A T O M A S S E M B L A G E S 2313

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

cal Problems, Vol. 2 (ed. Kozhova O.M.), pp. 51–52. Nauka,

Novosibirsk (in Russian).

Krenke AN, Chernavskaya MM (2002) Climate changes in the

preinstrumental period of the last millennium and their man-

ifestations over the Russian Plain. Isvestiya, Atmospheric and

Oceanic Physics, 38, S59–S79.

Laing TE, Smol JP (2000) Factors influencing diatom distribu-

tions in circumpolar treeline lakes of northern Russia. Journal

of Phycology, 36, 1035–1048.

Le Core HL (1998) Use of numerical models and satellite data to study

physical processes in Lake Baikal. Unpublished PhD Thesis,

University of Leicester, UK.

Legendre P, Gallagher ED (2001) Ecologically meaningful trans-

formations for ordination of species data. Oecologia, 129,

271–280.

Lim DSS, Kwan C, Douglas MSV (2001) Periphytic diatom

assemblages from Bathurst Island, Nunavut, Canadian High

Arctic: an examination of community relationships and habitat

preferences. Journal of Phycology, 37, 379–392.

Livingstone DM (1999) Ice break-up on southern Lake Baikal and

its relationship to local and regional air temperatures in Siberia

and to the North Atlantic Oscillation. Limnology and Oceano-

graphy, 44, 1486–1497.

Lotter AF, Bigler C (2000) Do diatoms in the Swiss Alps reflect

the length of ice cover? Aquatic Sciences, 62, 125–141.

Lydolph PE (1977) Climates of the Soviet Union. World Survey of

Climatology, Vol. 7. Elsevier, Amsterdam. pp. 91–115.

Mackay AW, Battarbee RW, Flower RJ et al. (2003) Assessing the

potential for developing internal diatom-based inference mod-

els in Lake Baikal. Limnology and Oceanography, 48, 1183–1192.

Mackay AW, Flower RJ, Kuzmina AE et al. (1998) Diatom

succession trends in recent sediments from Lake Baikal and

their relationship to atmospheric pollution and to climate

change. Philosophical Transactions of the Royal Society of London,

353, 1011–1055.

Mackay AW, Ryves DB, Battarbee RW et al. (2005) 1000 years of

climate variability in central Asia: assessing the evidence using

Lake Baikal diatom assemblages and the application of a

diatom-inferred model of snow thickness. Global and Planetary

Change, 46, 281–297.

Magnuson JJ, Robertson DM, Benson BJ et al. (2000) Historical

trends in lake and river ice cover in the northern hemisphere.

Science, 289, 1743–1747.

Martin P, Boes X, Goddeeris B et al. (2005) A qualitative assess-

ment of the influence of bioturbation in Lake Baikal sediments.

Global and Planetary Change, 46, 87–99.

McGowan S, Ryves DB, Anderson NJ (2003) Holocene records of

past effective precipitation in West Greenland. The Holocene,

13, 239–249.

Meyer KI (1930) Introduction into the algae flora of Lake Baikal.

Bulletin of the Moscow Society of Naturalists, 39, 179–396 (in

Russian).

Moline MA, Claustre H, Frazer TK et al. (2004) Alteration of the

food web along the Antarctic Peninsula in response to a

regional warming trend. Global Change Biology, 10, 1973–1980.

Morley DW (2005). Reconstructing past climate variability in

continental Eurasia. Unpublished PhD Thesis, University of

London, UK, 370 pp.

Muller B, Maerki M, Schmid M et al. (2005) Internal carbon

and nutrient cycling in Lake Baikal: sedimentation, upwelling

and early diagenesis. Global and Planetary Change, 46,

101–124.

Ohlendorf C, Bigler C, Goudsmit GH et al. (2000) Causes and

effects of long periods of ice cover on a remote high Alpine

lake. Journal of Limnology, 59, 65–80.

Oksanen J, Minchin PR (2002) Continuum theory revisited: what

shape are species responses along ecological gradients? Ecolo-

gical Modelling, 157, 119–129.

O’Reilly CM, Alin SR, Plisnier PD et al. (2003) Climate change

decreases aquatic ecosystem productivity of Lake Tanganyika,

Africa. Nature, 424, 766–768.

Panagiotopoulos F, Shahgedanova M, Hannachi A et al. (2005)

Observed trends and teleconnections of the Siberian High:

a recently declining center of action. Journal of Climate, 18,

1411–1422.

Popovskaya GI (2000) Ecological monitoring of phytoplankton

in Lake Baikal. Aquatic Ecosystem and Health Management, 3,

215–225.

Richardson TL, Gibson CE, Heaney SI (2000) Temperature

growth and seasonal succession of phytoplankton in Lake

Baikal, Siberia. Freshwater Biology, 44, 431–440.

Riggs G, Hall D, Salomonson V (1994) A snow index for the

Landsat Thematic Mapper and MODIS. In: Proceedings of

IGARSS ‘94 – International Geoscience and Remote Sensing Semi-

nar, pp. 1–10. ESA Publications, Noordwijk.

Robinson DA, Frei A (2000) Seasonal variability of northern

hemisphere snow extent using visible satellite data.

Professional Geography, 52, 307–315.

Ruhland KM, Smol JP (2002) Freshwater diatoms from the

Canadian Arctic treeline and development of paleolimnologi-

cal inference models. Journal of Phycology, 38, 249–264.

Ryves DB, Jewson DH, Sturm M et al. (2003) Quantitative and

qualitative relationships between planktonic diatom commu-

nities and diatom assemblages in sedimenting material and

surface sediments in Lake Baikal, Siberia. Limnology and

Oceanography, 48, 1643–1661.

Ryves DB, McGowan S, Anderson NJ (2002) Development and

evaluation of a diatom-conductivity model from lakes in West

Greenland. Freshwater Biology, 47, 995–1014.

Schindler DW (2001) The cumulative effects of climate warming

and other human stresses on Canadian freshwaters in the new

millennium. Canadian Journal of Fisheries and Aquatic Sciences,

58, 18–29.

Semovski SV, Mogilev NY (2003) The use of satellite imagery

in investigating the dynamics of the ice and snow cover Lake

Baikal. Nordic Hydrology, 34, 33–50.

Semovski SV, Mogilev NY, Sherstyankin PP (2000) Lake

Baikal ice: analysis of AVHRR imagery and simulation of

under-ice phytoplankton bloom. Journal of Marine Systems,

27, 117–130.

Shimaraev MN, Kuimova LN, Sinyukovich VN et al. (2002)

Manifestation of global climatic changes in Lake Baikal during

the 20th century. Doklady Earth Sciences, 383A, 288–291.

Shimaraev MN, Verbolov VI, Granin NG et al. (1994) Physical

Limnology of Lake Baikal. A Review (eds Shimaraev MN, Okuda

S), 80 pp. BICER Publishers, Irkutsk and Okayama.

2314 A . W. M A C K AY et al.

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315

Shindell DT, Schmidt GA, Mann ME et al. (2001) Solar forcing

of regional climate change during the Maunder Minimum.

Science, 294, 2149–2152.

Silow EA, Baturin VA, Stom DJ (2001) Prediction of Lake

Baikal ecosystem behaviour using an ecosystem disturbance

model. Lakes and Reservoirs: Research and Management, 6,

33–36.

Skabichevskii AP (1960) Planktonic diatoms of the freshwaters of the

USSR: Systematics, Ecology & Distribution. MSU, Moscow (in

Russian).

Skvortzov BV (1937) Bottom diatoms from Olkhon Gate of Baikal

lake, Siberia. Philippine Journal of Science, 62, 293–377.

Smol JP, Wolfe AP, Birks HJB et al. (2005) Climate-driven regime

shifts in the biological communities of arctic lakes. Proceedings

of the National Academy of Sciences, 102, 4397–4402.

Solovieva N, Jones VJ, Nazarova L et al. (2005) Palaeolimnological

evidence for recent climatic change in lakes from the northern

Urals, arctic Russia. Journal of Paleolimnology, 33, 463–482.

Sorvari S, Korhola A, Thompson R (2002) Lake diatom response

to recent Arctic warming in Finnish Lapland. Global Change

Biology, 8, 153–163.

Srivastan KS, Singh RP (1991) Microwave radiometry of snow-

covered terrains. International Journal of Remote Sensing, 12,

2117–2131.

Tarasova EN, Mescheryakova AI (1992) Modern State of the

Hydrochemical Regime of Lake Baikal. Nauka, Novosibirsk (in

Russian).

ter Braak CJF, Looman CWN (1994) Biplots of reduced-rank

regression. Biometrical Journal, 36, 983–1003.

ter Braak CJF, Smilauer P (2002) Canoco for Windows 4.5.

Biometrics, the Netherlands.

Todd MC, Mackay AW (2003) Large-scale climate controls on

Lake Baikal ice cover. Journal of Climate, 16, 3186–3199.

Verbolov VI, Sokolınikov VN, Shimaraev MN (1965) Hydrome-

teorological Conditions and Heat Balance of Lake Baikal. Nauka,

Moscow.

Votinsev KK, Mescheryakova AI, Popovskaya GI (1975)

Organic Matter Turnover in Lake Baikal. Nauka, Novosibirsk

(in Russian).

Weckstrom J, Korhola A (2001) Patterns in the distribution,

composition and diversity of diatom assemblages in relation

to ecoclimatic factors in Arctic Lapland. Journal of Biogeography,

28, 31–45.

Weiss RF, Carmack EC, Koropalov VM (1991) Deep-water

renewal and biological production in Lake Baikal. Nature, 349,

665–669.

Winder M, Schindler DE (2004) Climate change uncouples

trophic interactions in an aquatic ecosystem. Ecology, 85,

2100–2106.

Wuest A, Ravens T, Granin NG et al. (2005) Cold intrusions

in Lake Baikal: direct observational evidence for deep-water

renewal. Limnology and Oceanography, 50, 184–196.

Ye HC (2001) Quasi-biennial and quasi-decadal variations in

snow accumulation over northern Eurasia and their connec-

tions to the Atlantic and Pacific Oceans. Journal of Climate, 14,

4573–4584.

Yoshii K, Melnik NG, Timoshkin OA et al. (1999) Stable isotope

analyses of the pelagic food web in Lake Baikal. Limnology and

Oceanography, 44, 502–511.

Zabelina MM, Kiselev IA, Proshkina-Lavrenko AI et al. (1951)

Diatom Algae (ed. Proshkina-Lavrenko AI), 108 pp. Sovetsya

Nauka, Moscow (in Russian).

L A K E B A I K A L D I A T O M A S S E M B L A G E S 2315

r 2006 The AuthorsJournal compilation r 2006 Blackwell Publishing Ltd, Global Change Biology, 12, 2297–2315