Assessing the vulnerability of Africa's freshwater fishes to climate change … · 2019. 6. 21. · climate change on these species have not been explored on a continent-wide scale.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Contents lists available at ScienceDirect
Biological Conservation
journal homepage: www.elsevier.com/locate/biocon
Assessing the vulnerability of Africa's freshwater fishes to climate change: Acontinent-wide trait-based analysisElizabeth A. Nyboer⁎, Chris Liang, Lauren J. ChapmanMcGill University, Stewart Biology Building, N3/11, 1205 Docteur Penfield, Montreal H3A 1B1, Quebec, Canada
Climate change is a key driver of biodiversity loss across the globe, and freshwater fishes are predicted to beamong the most vulnerable taxa. African freshwater ecosystems are home to one of the most unique and diverseichthyo-faunas on the planet, and freshwater fish species provide essential livelihoods for millions of peopleliving in riparian communities across the continent. Although nearly one sixth of African freshwater fishes havebeen designated as endangered or vulnerable to extinction by the IUCN Red List assessment, the effects ofclimate change on these species have not been explored on a continent-wide scale. In this study, we present thefirst trait-based climate change vulnerability assessment (CCVA) comprising the majority (85%) of Africa'scurrently described freshwater fishes. We assembled data relating to three dimensions of vulnerability includingsensitivity, adaptive capacity, and exposure. In addition, we developed an index of ‘conservation value’ based ontraditional conservation metrics including extinction risk, endemism, and provision of ecosystem services. Wefound that almost 40% of African freshwater fishes are vulnerable to climate change, mostly owing to the manyspecies with highly specialized habitat and life-history requirements, and because of the numerous anthro-pogenic stressors they face. High proportions of species within the Nothobranchiidae and Cichlidae families werefound to be vulnerable. Regions with high frequencies of vulnerable species included the African rift valley lakes,the Congo River drainage, and the coastal rivers of West Africa. Several important data deficiencies wereidentified relating to species' population sizes, genetic variability, and life history traits, and constitute priorityresearch areas for the future. In addition, we highlighted some cases where traditional conservation approachesoverlook species and regions that are predicted to be threatened by climate change.
1. Introduction
Recent assessments of the Earth's climatic patterns indicate that globalsurface temperatures have increased by 0.72–0.85 °C over the last century.In addition, alterations to the world's water cycle have changed pre-cipitation patterns, resulting in a higher frequency of extreme events suchas droughts and floods (Stocker et al., 2013). Ecological stresses associatedwith these climatic changes are important global drivers of species ex-tinctions and biodiversity loss (Thomas et al., 2004; Cahill et al., 2013;Urban, 2015). Climate change effects on species include shifts in distribu-tion (Parmesan and Yohe, 2003; Pecl et al., 2017), changes in habitatavailability (Leadley et al., 2010), modifications to community structure(Walther, 2010), shifts in life-cycle phenology (Yang and Rudolf, 2010),and alterations to population growth trajectories (Martay et al., 2017). Thetask of identifying the species, populations, and regions that will be mostvulnerable to climate change has become a major focus of ecology andconservation biology (Nadeau et al., 2017; Foden et al., 2018).
1.1. Vulnerability of fresh water ecosystems to climate change
Inland aquatic environments are predicted to be among the mostvulnerable to climate change (Dudgeon et al., 2006; Heino et al., 2009;Woodward et al., 2010). Increases in surface water temperature havebeen documented in numerous lakes, rivers, streams, and wetlandsaround the world (Livingstone, 2003; O'Reilly et al., 2003; Adrian et al.,2009). Increases in temperature can lead to higher levels of primaryproduction in lakes reducing water clarity, limiting growth of sub-merged vegetation, altering food-web structure, and disrupting nutrientdynamics (Jeppesen et al., 2009, 2012). Changes in precipitation pat-terns can modify hydrological cycles and increase discharge into lakeand river systems, which can exacerbate eutrophication and waterturbidity through inputs of solutes, pollutants, and fertilizers to surfacewaters (Jeppesen et al., 2009; Knouft and Ficklin, 2017).
These fundamental alterations to ecosystem structure and functioncan affect many different aspects of the biogeography, life-history, and
https://doi.org/10.1016/j.biocon.2019.05.003Received 9 September 2018; Received in revised form 12 February 2019; Accepted 3 May 2019
physiology of freshwater fishes (Myers et al., 2017). A recent worldwideassessment estimated that approximately 25% of freshwater fishes arethreatened with extinction (Collen et al., 2014), confirming that theyare among the most threatened taxa, globally (Olden et al., 2010). Evenwhen the effects of environmental change are not lethal, impacts ongrowth and reproduction can cause significant alterations to fish po-pulations and communities (Ficke et al., 2007; Myers et al., 2017). Fishspecies without dispersal barriers may shift poleward with changingthermal conditions (Sorte et al., 2010; Poloczanska et al., 2013);however, for land-locked fishes inhabiting fresh waters, distributionalshifts are not necessarily possible. Without the option for escape, in-creases in water temperature may alter nutritional requirements, re-duce reproductive output, and cause changes in diet and local habitatuse (Ficke et al., 2007; Myers et al., 2017). The vulnerability of fishpopulations to climate change will therefore depend on their capacityfor dispersal, genetic adaptation, and phenotypic plasticity (beha-vioural and physiological), all of which must be considered acrossmultiple life-history stages (Heino et al., 2009; Brander, 2010;Chessman, 2013).
1.2. African freshwater fishes – integration of climate change intoconservation
According to the Intergovernmental Panel on Climate Change's fifthassessment report (IPCC-AR5), air temperatures across Africa have in-creased by approximately 0.5 °C over the last 50–100 years, and currentclimate change projections show further increases of between 1.3 and4.3 °C across most of the continent by the end of the 21st century (Nianget al., 2014). Increases in air temperature have led to increases insurface water temperatures in many of Africa's fresh water bodies (re-viewed in Ogutu-Ohwayo et al., 2016), and in some cases have causednotable declines in ecosystem productivity (O'Reilly et al., 2003).Africa's freshwaters are home to an estimated 3300 fish species(Lévêque and Paugy, 2010), comprising one of the most species richichthyo-faunas on the planet (Thieme et al., 2005). This fauna isespecially unique because of high proportions of basal and archaicspecies present in many fish communities, and becasue of the im-pressive array of diverse species flocks endemic to multiple lake andriver systems across the continent (Thieme et al., 2005; Lévêque andPaugy, 2010). Notable species radiations have occurred in the Upperand Lower Guinean ecoregions and the Congo River basin (Fig. 1A);however, the most famous and diverse radiation is that of the cichlidfishes (an estimated 2000 species) in the rift valley lakes of East Africa(Roberts, 1975). Africa's fresh waters are often divided into ichthyo-faunal provinces (ecoregions) based on endemism, paleogeography,and physiological or ecological barriers (Roberts, 1975), which are usedin our analysis for interpretive purposes (Fig. 1A).
African freshwater fishes have important socio-economic and cul-tural roles in African societies and are critical to livelihoods of millionsof people across the continent. Many are used in commercial or arti-sanal fisheries (Carr et al., 2013), and fishing pressure has increased byapproximately 42,000 tons per year since 1990 (FAO, 2018). Fishes arealso commonly used in the aquarium trade, and are increasingly usedfor aquaculture to improve food security and reduce pressure on wildfisheries (FAO, 2018). Many of Africa's most important freshwatersystems are threatened by habitat degradation, poor water manage-ment, introduction of invasive species, and/or pollution from agri-culture and urbanization (Thieme et al., 2005). Of the currently de-scribed ~3300 freshwater fish species, more than one sixth (~580species) are estimated to be endangered or vulnerable to extinctionbased on the IUCN Red List assessment. To date, no comprehensiveclimate change vulnerability assessment (CCVA) has been performed onfreshwater fishes of Africa, although some regional reports have beenproduced (Carr et al., 2013, 2014; Sayer et al., 2018).
1.3. Climate change vulnerability assessments
The degree to which species are able to respond to climatic stressorswill depend on the severity of change in their habitat, their biologicaland ecological sensitivities to change, and their ability to track shiftingclimate by colonizing new territory or adapting to novel conditionsthrough physiological or behavioural modifications (Nadeau et al.,2017). Until recently, most large-scale, multi-taxa assessments of spe-cies vulnerability have not accounted for intrinsic, species-specific ef-fects (e.g., niche breadth) that can influence their ability to cope(Williams et al., 2008; Pacifici et al., 2015). Climate change-relatedstresses can be modulated by species' biological, ecological, and phy-siological traits through a diversity of plastic and adaptive responses(Pacifici et al., 2017). Therefore, incorporating ecological and evolu-tionary characteristics with exposure estimates can improve forecasts ofa species' vulnerability to climate change (Williams et al., 2008). Vul-nerability assessments that incorporate species characteristics areknown as trait-based approaches (Williams et al., 2008; Chessman,2013; Pacifici et al., 2015; Foden et al., 2018) and are effective ingenerating novel insights into climate change risk assessments (Böhmet al., 2016) and conservation prioritization (Dickinson et al., 2014).
In general, trait-based CCVAs assemble data within various di-mensions relating to vulnerability including sensitivity, adaptive ca-pacity, and exposure to climate change (Williams et al., 2008; Fodenet al., 2018). This approach is increasingly common for developingpriorities for conservation under climate change (Pacifici et al., 2015),and has been adopted in recent global assessments of corals, amphi-bians, birds (Foden et al., 2013), and reptiles (Böhm et al., 2016). Likeall CCVAs, trait-based approaches have various levels of uncertaintyassociated with their findings, and do not provide empirical predictionsof population range expansion or demographic shifts as some otherapproaches do (Foden et al., 2018). However, this method has gainedtraction recently because it is relatively rapid to perform, accounts forthe effect of species traits, and can be used to assess large numbers ofspecies across broad geographic ranges. As such, trait-based CCVAs arevery useful to lay the groundwork for future research on taxonomicgroups and regions that are understudied (Foden et al., 2018).
In this study we performed a trait-based CCVA of African freshwaterfishes to highlight species and regions that require conservation at-tention under climate change. We incorporated effects of other stres-sors, and compared results to outcomes from traditional conservationmetrics. The objectives were to 1) develop species-specific predictionsabout vulnerability to climatic change by integrating biological traitsand exposure estimates, 2) identify geographic regions of future con-servation priority due to concentrations of climate change vulnerablespecies, and 3) compare species and regions that are climate changevulnerable with those that are considered of high conservation value bytraditional metrics.
2. Methods
2.1. Fish species selection
All African fish species were included in this study if they met thefollowing three criteria: 1) they spend any part of their life-cycle ininland aquatic environments (including diadromous, anadromous, andestuarine species), 2) their distribution ranges are mapped by the IUCN,and 3) their taxonomic classification could be resolved between our twoprimary sources of data, FishBase and the IUCN Red List.
We followed methods developed by Foden et al. (2013) wherebyclimate change vulnerability is assessed based on three vulnerabilitydimensions, namely sensitivity, adaptive capacity, and exposure. In thisframework, sensitivity refers to the inability of a species to persist in its
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
506
current environment if conditions were to change, adaptive capacityrefers to the likelihood that a species can evade environmental threatsthrough dispersal (escape) or micro-evolutionary change (geneticadaptation), and exposure refers to the degree to which climatic para-meters are predicted to change across a species' range. Sensitivity andadaptive capacity were evaluated through analyses of species' life his-tory, ecological, physiological, and genetic traits. Species were rankedas ‘high’, ‘low’, or ‘unknown’ for each trait according to pre-definedqualitative or quantitative scoring regimes. Exposure estimates werebased on downscaled projected changes in temperature and precipita-tion across species ranges, and assigned scores of ‘high’ or ‘low’ basedon thresholds of exposure. A species was considered vulnerable overall ifit scored ‘high’ for sensitivity and exposure, and 'low' for adaptive capa-city.
2.3. Data collection
2.3.1. Trait dataTrait data were collected primarily from the IUCN Red List species
information service (IUCN, 2018) and from FishBase (Froese and Pauly,2018). We used the R packages rfishbase (Boettiger et al., 2017) andrredlist (Chamberlain, 2018) to extract trait data for all species that metour three criteria. Data extractions were performed in R studio v. 3.4.1(R Core Team, 2017). We also integrated data from two IUCN reportsthat investigated climate change vulnerability of freshwater fishes inwestern Africa (Carr et al., 2014) and the Albertine Rift of East Africa(Carr et al., 2013). These studies contain expert-informed data for 517species from West Africa and 551 species from the Albertine Rift (ap-proximately 38% of all species in our dataset), which were used to addmissing information to our data set, and to error check scores. We founda 90% agreement rate (range: 88–100%) between our scores and IUCNexpert-informed scores. Where scores did not match, additional litera-ture searches were performed to confirm accuracy. Where no alter-native information existed, the expert-informed scores were used. Athird report investigating climate change vulnerability of fish species inthe Lake Victoria basin was recently published (Sayer et al., 2018);however, those data were not incorporated as they became availableafter the present analyses were complete. Further details on data col-lection are available in the Supplementary Methods S.1.1.2.
2.3.2. Distribution dataSpecies distribution data were downloaded from the IUCN Red List
spatial data service. IUCN species polygons represent the entire rangeacross which a species is estimated to occur. These ranges represent thebest known estimates of species distribution limits (Foden et al., 2013);however, they often include areas that could not actually be occupiedby the species (e.g., land between two inhabited lakes). Species rangeswere therefore refined by creating a detailed map of freshwater eco-systems across Africa, and clipping the ranges to match this layer. Tocreate the water layer, we combined data from the HydroSHEDS andHydroLAKES databases (Lehner et al., 2008) and the Global Lakes andWetlands Database (GLWD; Lehner and Doll, 2004; SupplementaryMethods S.1.4). For this process, we assumed any freshwater habitatwithin a species range was habitable. While this may have slightlyoverestimated range sizes, this approach was practical for rapid re-finement of generalized range limits, and provided outputs that werereasonable for the spatial resolution of this study. Species ranges andwater data were projected in Africa Albers Equal Area Conic (AAEAC)to account for the curvature of the earth. All range mapping was per-formed in ArcGIS v. 10.5.1 (ESRI, 2017).
2.4. Defining vulnerability dimensions
2.4.1. Sensitivity and adaptive capacityThe sensitivity dimension aimed to quantify the inability of a species
to persist in its current environment should conditions change (Fodenet al., 2013). Sensitivity was split into six trait sets pertaining to: 1)specialized habitat or microhabitat requirements (based on depthranges occupied, the diversity, rarity, and vulnerability of habitatsused, and microhabitat requirements of each species), 2) narrow en-vironmental tolerances (based on historical exposure to precipitationand temperature variability across its range [Supplementary MethodsS.1.4], reliance on seasonal patterns, and sensitivity to sedimentation),3) dependence on interspecific interactions (based on prey specificityand other dependencies), 4) complexity of life history (based on re-quiring particular environmental conditions or social cues to completeits life cycle), 5) rarity (based on abundances, distribution ranges, andpopulation fragmentation), and 6) exposure to other disturbances(based on numbers of non-climate-change threats experienced in itsrange). In this dimension, some ‘traits’ were derived from climatic or
Fig. 1. Map showing (A) the ichthyo-faunal provinces (ecoregions) adapted from Roberts (1975) and Thieme et al. (2005), as well as the major lake and river systemsin Africa, and (B) the distribution of species richness of freshwater fishes in Africa.
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
507
Table1
Des
crip
tion
oftr
aits
ets,
vari
able
s,an
dth
resh
olds
qual
ifyin
gsp
ecie
sas
high
orlo
wfo
rvulnerability
andconservationvalue,
and
the
num
ber
ofsp
ecie
scl
assi
fied
ashi
gh,l
ow,a
ndun
know
nfo
rea
chtr
aita
ndth
resh
old.
Num
bers
and
prop
ortio
nsof
spec
iesq
ualif
ying
forh
ighsensitivity
,low
adaptivecapacity
,hig
hexposure
,hig
hvulnerability
,and
high
conservationvalue
inea
chtr
aits
etar
epr
esen
ted
inbr
acke
tsin
the
first
colu
mn.
Det
aile
dde
scri
ptio
nsof
how
thre
shol
dsw
ere
esta
blis
hed
(with
supp
ortf
rom
the
liter
atur
e)ar
eav
aila
ble
inth
eSu
pple
men
tary
Met
hods
S.1.
3.A
llda
taar
eba
sed
onth
eop
timis
ticsc
enar
iofo
rthe
RCP8
.5cl
imat
ech
ange
emis
sion
ssc
enar
iofo
r20
55.
Sens
itivi
ty(n
=23
75hi
gh;8
4.8%
ofal
lspe
cies
inda
tase
t)
Trai
tse
tVa
riab
leD
escr
iptio
nTh
resh
old/
defin
ition
n
I.Sp
ecia
lized
habi
tat
orm
icro
habi
tat
requ
irem
ents
(n=
727;
26.0
%of
alls
pp.)
1.D
epth
rang
eTa
xon
isre
stri
cted
tosh
allo
wha
bita
tsor
shal
low
dept
hsH
igh
=re
stri
cted
tosh
allo
wde
pths
(<3
m)
78Lo
w=
not
rest
rict
edto
shal
low
dept
hs48
0U
nkno
wn
=no
info
rmat
ion
2235
2.H
abita
tspe
cial
izat
ion
Taxo
nis
aha
bita
tgen
eral
ist
orsp
ecia
list
Hig
h=
rest
rict
edto
1–2
rare
orvu
lner
able
habi
tats
79Lo
w=
inha
bits
dive
rse
orco
mm
onha
bita
ts25
31U
nkno
wn
=no
info
rmat
ion
183
3.M
icro
habi
tat
spec
ializ
atio
nTa
xon
isre
stri
cted
toon
era
reor
vuln
erab
leha
bita
tfor
any
life
cycl
est
age
Hig
h=
spec
ializ
eson
asp
ecifi
cm
icro
habi
tat
652
Low
=al
loth
erta
xa21
41
II.N
arro
wen
viro
nmen
talt
oler
ance
s(n
=14
49;5
1.2%
ofal
lspp
.)4.
Tole
ranc
eto
chan
ges
inpr
ecip
itatio
nH
isto
rica
lvar
iabi
lity
inpr
ecip
itatio
nac
ross
the
taxo
n's
rang
eH
igh
=lo
wes
t25
%pr
ecip
itatio
nva
riab
ility
(≤56
.1m
m)
691
Low
=hi
ghes
t75
%pr
ecip
itatio
nva
riab
ility
(>56
.1m
m)
2102
5.To
lera
nce
toch
ange
sin
tem
pera
ture
His
tori
calv
aria
bilit
yin
tem
pera
ture
acro
ssth
eta
xon'
sra
nge
Hig
h=
low
est
25%
tem
pera
ture
vari
abili
ty(≤
0.66
°C)
682
Low
=hi
ghes
t75
%te
mpe
ratu
reva
riab
ility
(>0.
66°C
)21
116.
Sens
itivi
tyto
incr
ease
sin
turb
idity
:mat
ing
Taxo
n's
mat
ere
cogn
ition
syst
emaff
ecte
dby
chan
ges
intu
rbid
ityH
igh
=re
quir
escl
ear
wat
erfo
rm
ate
reco
gniti
on26
7Lo
w=
allo
ther
taxa
2526
7.Se
nsiti
vity
toin
crea
ses
intu
rbid
ity:f
eedi
ngTa
xon'
sfo
odga
ther
ing
orpr
eyse
lect
ion
affec
ted
bych
ange
sin
turb
idity
Hig
h=
requ
ires
clea
rw
ater
tofin
dfo
od13
7Lo
w=
allo
ther
taxa
2656
8.D
epen
denc
eon
prec
ipita
tion
activ
ated
trig
ger
Taxo
nre
quir
esra
ins
atsp
ecifi
ctim
esto
com
plet
ea
port
ion
ofth
eir
life
cycl
eH
igh
=br
eedi
ngm
igra
tions
trig
gere
dby
rain
s15
4H
igh
=ha
tch
duri
ngra
ins,
requ
ire
rain
sfo
rre
subm
erge
nce
47H
igh
=ju
veni
lem
igra
tions
trig
gere
dby
wat
erle
vel
24Lo
w=
allo
ther
taxa
2586
III.D
epen
denc
eon
inte
rspe
cific
inte
ract
ions
(n=
316;
11.3
%of
alls
pp.)
9.Pr
eysp
ecifi
city
Taxo
nis
apr
eyge
nera
list
orsp
ecia
list
Hig
h=
relie
sex
clus
ivel
yon
one
prey
type
266
Low
=co
nsum
esa
wid
eva
riet
yof
prey
1287
Unk
now
n=
noin
form
atio
n12
4010
.Dep
ende
nce
onot
her
taxa
Taxo
nre
lieso
nan
othe
rtax
onfo
rsom
eas
pect
ofsu
rviv
alH
igh
=de
pend
ento
nan
othe
rsp
ecie
sfo
od,p
rote
ctio
n,or
habi
tat
77Lo
w=
allo
ther
taxa
2716
IV.C
ompl
exity
oflif
ehi
stor
yst
rate
gy(n
=30
7;11
.0%
ofal
lspp
.)11
.Com
plex
ityin
repr
oduc
tive
stra
tegy
Taxo
nre
lies
onpr
ecis
een
viro
nmen
tal/
soci
alcu
es,r
are
habi
tats
for
repr
oduc
tion
Hig
h=
taxo
nre
lies
onse
ason
al,e
nvir
onm
enta
lor
soci
alcu
esfo
rre
prod
uctio
n24
2
Low
=al
loth
erta
xa25
5112
.Com
plex
ityin
earl
ylif
ehi
stor
yre
quir
emen
tsTa
xon'
sla
rvae
oreg
gsre
lyon
prec
ise
envi
ronm
enta
lcu
es,c
ondi
tions
,hab
itats
for
surv
ival
Hig
h=
taxo
nre
lies
onse
ason
alor
envi
ronm
enta
lcue
sor
spec
ific
habi
tats
for
surv
ival
ofeg
gsan
dla
rvae
49
Low
=al
loth
erta
xa27
4413
.Spa
wni
ngcy
cle
Num
ber
and
dura
tion
ofa
taxo
n's
spaw
ning
even
tsH
igh
=sp
awns
only
once
orov
era
shor
ttim
efra
me
(≤2
mon
ths)
70Lo
w=
spaw
nsse
vera
ltim
esor
over
long
ertim
efra
mes
(>2
mon
ths)
312
Unk
now
n=
noin
form
atio
n24
11
V.Ra
rity
(n=
774;
27.7
%of
alls
pp.)
14.T
axon
abun
danc
eTa
xon
isra
reor
com
mon
Hig
h=
atle
ast
2of
:rar
e,na
rrow
dist
ribu
tion,
decl
inin
g45
5Lo
w=
atle
ast
1of
:com
mon
,wid
espr
ead,
incr
easi
ng20
35U
nkno
wn
=no
info
rmat
ion
303
15.R
ange
size
:ext
ento
focc
urre
nce
(EO
O)
Tota
lare
aof
the
spec
ies
rang
ehi
gh=
smal
lEO
O(≤
5000
km2 )
350
Low
=la
rge
EOO
(>50
00km
2 )24
4316
.Ran
gesi
ze:a
rea
ofoc
cupa
ncy
(AO
O)
Act
uala
rea
occu
pied
byth
esp
ecie
shi
gh=
smal
lAO
O(≤
500
km2 )
397
low
=la
rge
AO
O(>
500
km2 )
2396
17.P
opul
atio
nfr
agm
enta
tion
Taxo
n's
leve
loff
ragm
enta
tion
Hig
h=
popu
latio
nsfr
agm
ente
dor
very
smal
l47
Low
=po
pula
tions
larg
ean
dco
ntig
uous
1028
Unk
now
n=
noin
form
atio
n17
18
(continuedon
nextpage
)
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
508
Table1
(continued)
Sens
itivi
ty(n
=23
75hi
gh;8
4.8%
ofal
lspe
cies
inda
tase
t)
Trai
tse
tVa
riab
leD
escr
iptio
nTh
resh
old/
defin
ition
n
VI.E
xpos
ure
toot
her
dist
urba
nces
(n=
1012
;36.
2%of
alls
pp.)
18.F
ishi
ngpr
essu
reTa
xon
isha
rves
ted
for
hum
anco
nsum
ptio
nH
igh
=fis
hery
clas
sifie
das
eith
erco
mm
erci
alor
high
lyco
mm
erci
al68
4Lo
w=
not
harv
este
dor
fishe
rycl
assi
fied
asar
tisan
al21
0919
.Oth
erth
reat
sN
umbe
rof
thre
ats
curr
ently
affec
ting
the
taxo
nH
igh
=th
reat
ened
by≥
7ad
ditio
nala
nthr
opog
enic
thre
ats
(top
25%
thre
aten
ed)
135
Low
=th
reat
ened
by<
7ad
ditio
nala
nthr
opog
enic
thre
ats
(low
est
75%
thre
aten
ed)
2658
20.T
hrea
tsin
tens
ified
bycl
imat
ech
ange
Num
ber
ofth
reat
sth
atw
illbe
dire
ctly
inte
nsifi
edby
clim
ate
chan
ge(s
eeTa
ble
S5)
Hig
h=
thre
aten
edby
≥2
clim
ate
chan
ge-in
tens
ified
thre
ats
371
Low
=th
reat
ened
by<
2cl
imat
ech
ange
-inte
nsifi
edth
reat
s24
22
Ada
ptiv
eca
paci
ty(n
=20
00lo
w;7
1.6%
ofal
lspe
cies
inda
tase
t)
Trai
tse
tVa
riab
leD
escr
iptio
nTh
resh
old/
defin
ition
#sp
ecie
s
VII.
Pote
ntia
lfor
disp
ersa
l(n
=12
63;
45.2
%of
alls
pp.)
1.In
trin
sic
prob
abili
tyof
disp
ersa
lTa
xon
has
trai
tsth
atle
adto
low
disp
ersa
lsyn
drom
es(s
eeTa
ble
S6)
Low
=co
mbi
natio
nof
life
hist
ory
trai
tsth
atin
dica
telo
win
trin
sic
disp
ersa
labi
lity
386
Hig
h=
allo
ther
taxa
2407
2.Ba
rrie
rsto
disp
ersa
lTa
xon'
sdi
sper
sali
slim
ited
byec
olog
ical
barr
iers
orst
rong
habi
tata
ffini
ties
asad
ults
Low
=ad
ults
disp
ersa
lres
tric
ted
bybi
ogeo
grap
hic
and
ecol
ogic
alba
rrie
rs53
9
Hig
h=
allo
ther
taxa
2254
3.D
ispe
rsal
ofea
rly
life
stag
esA
bilit
yfo
rta
xon'
seg
gsan
dla
rvae
todi
sper
seLo
w=
eggs
orla
rvae
are
entir
ely
bent
hic
774
Hig
h=
egg
orla
rvae
have
ape
lagi
cst
age
211
Unk
now
n=
noin
form
atio
n18
08
VIII.
Pote
ntia
lto
evol
ve(n
=12
1143
.4%
ofal
lspp
.)4.
Repr
oduc
tive
capa
city
Taxo
n's
mea
nan
nual
rela
tive
fecu
ndity
Low
=lo
wre
lativ
efe
cund
ity(≤
200
eggs
/gra
m/y
ear)
23H
igh
=hi
ghre
lativ
efe
cund
ity(>
200
eggs
/gra
m/y
ear)
77U
nkno
wn
=no
info
rmat
ion
2693
5.Po
pula
tion
grow
thTa
xon'
spo
tent
ialf
orpo
pula
tion
grow
thba
sed
onlif
ehi
stor
ych
arac
teri
stic
s(l
owpo
pula
tion
grow
th=
low
adap
tive
capa
city
;Tab
leS7
)Lo
w=
com
bina
tion
oflif
ehi
stor
ytr
aits
that
indi
cate
low
popu
latio
ngr
owth
1188
Hig
h=
com
bina
tion
oflif
ehi
stor
ytr
aits
that
indi
cate
high
popu
latio
ngr
owth
1550
Unk
now
n=
noin
form
atio
n55
Expo
sure
(n=
1876
high
;67.
2%of
alls
peci
esin
data
set)
Trai
tse
tVa
riab
leD
escr
iptio
nTh
resh
old/
defin
ition
#sp
ecie
s
IX.P
redi
cted
expo
sure
toth
eeff
ects
ofcl
imat
ech
ange
(n=
1876
;67.
2%of
alls
pp.)
1.Ch
ange
sin
mea
nte
mpe
ratu
reA
vera
gech
ange
inm
ean
tem
pera
ture
from
1975
to20
55ac
ross
taxo
n's
rang
eH
igh
=hi
ghes
t25%
(≥2.
84°C
)69
8Lo
w=
low
est
75%
(<2.
84°C
)20
952.
Chan
gesi
nte
mpe
ratu
reva
riab
ility
Ave
rage
chan
gein
tem
pera
ture
AA
D(v
aria
bilit
y)fr
om19
75to
2055
acro
ssta
xon'
sra
nge
Hig
h=
high
est2
5%(≥
0.96
°C)
698
Low
=lo
wes
t75
%(<
0.96
°C)
2095
3.Ch
ange
sin
mea
npr
ecip
itatio
nA
vera
gech
ange
inm
ean
prec
ipita
tion
from
1975
to20
55ac
ross
taxo
n's
rang
eH
igh
=hi
ghes
t25%
(≥5.
84m
m)
682
Low
=lo
wes
t75
%(<
5.84
mm
)19
574.
Chan
ges
inpr
ecip
itatio
nva
riab
ility
Ave
rage
chan
gein
prec
ipita
tion
AA
D(v
aria
bilit
y)fr
om19
75to
2055
acro
ssta
xon'
sra
nge
Hig
h=
high
est2
5%(≥
4.52
mm
)72
5Lo
w=
low
est
75%
(<4.
52m
m)
2068
Vuln
erab
ility
(n=
1142
high
;40.
9%of
alls
peci
esin
data
set)
(continuedon
nextpage
)
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
509
environmental data rather than from empirical data relating directly tospecies ecology or physiology. For example, tolerance to variance intemperature was based on historical environmental variability, andthreat status was based on external anthropogenic stressors.
Adaptive capacity was defined as the likelihood that a species canevade environmental threats through dispersal or micro-evolutionarychange (Foden et al., 2013). Adaptive capacity was comprised of twotrait sets pertaining to: 1) potential to disperse (based on intrinsicqualities that affect dispersal ability in both adult and juvenile lifestages, and on the existence of extrinsic geographical barriers), and 2)potential to evolve (based on reproductive output [i.e., relative fe-cundity] and population growth rate). Species that scored high forsensitivity and low for adaptive capacity were considered to be biologicallysusceptible to climate change. Details on data availability and howvariables within sensitivity and adaptive capacity were scored are avail-able in Table 1 and Table S1. Justification for inclusion of each variable,and special considerations for each trait can be found in SupplementaryMethods S.1.3 and S.1.8.
2.4.2. ExposureExposure was defined as the degree to which climatic parameters are
predicted to change across a species' range. In this study, exposure toclimate change was encompassed by one trait set pertaining to: pre-dicted exposure to the effects of climate change. This trait set was basedon four variables including changes in mean temperature, mean tem-perature variability, mean precipitation, and mean precipitationvariability. Because direct measurements of changes in freshwaterswere rare, we based our estimates on projected changes to surface airtemperature assuming that where these effects were more extreme, sowould be the effects on freshwater ecosystems (O'Reilly et al., 2003;Adrian et al., 2009; Knouft and Ficklin, 2017).
Estimates of climatic changes were derived by subtracting baselinevalues for temperature and precipitation from projected changes.Baseline climate estimates were derived from the WorldClim V1.4 da-tabase (Fick and Hijmans, 2017), and estimates of projected changewere derived from AFRICLIM 3.0 downscaled data (Platts et al., 2015).AFRICLIM incorporates data from two regional climate models (RCMs)and eight general circulation models (GCMs) for increased accuracy ofclimate change estimates across Africa (Platts et al., 2015). Globaltemperature and precipitation projections were downscaled to a re-solution of 10-arcminutes (~20km2 grid cells) using the WorldClimbaseline change factor (Platts et al., 2015). We performed all calcula-tions for two IPCC-AR5 representative concentration pathways (RCP4.5and RCP8.5). Average absolute changes in temperature and precipita-tion were calculated as the difference between average values in 1975(mean of 1961–1990) to 2055 (mean of 2041–2070), and 1975 to 2085(mean of 2071–2100) (Platts et al., 2015). Change in climate variabilitywas estimated through calculations of average absolute deviation(AAD) for the same years. Calculations of AAD were used to estimatetolerance to temperature and precipitation variability across time(months) and space (species ranges). Further details on climate changecalculations can be found in the Supplementary Methods S.1.4. Allcalculations were performed in R studio v. 3.4.1 (R Core Team, 2017).The main results of this paper present the findings from RCP8.5 forprojected changes from 1975 to 2055, but the sensitivity of these resultsto different years and scenarios were determined by comparing thesefindings with results derived from the RCP4.5–2055, RCP4.5–2085, andRCP8.5–2085 projections. These comparisons can be found in theSupplementary Tables and Figures.
Data extracted from the IUCN Red List, FishBase, WorldClim, andAFRICLIM databases were summarized according to the above trait setsand variables (Supplementary Methods S.1.1). Each species was given ascore of ‘high’, ‘low’, or ‘unknown’ for each variable within the threeTa
ble1
(continued)
Cons
erva
tion
valu
e(n
=17
14hi
gh,6
1.4%
ofal
lspe
cies
inda
tase
t)
Trai
tse
tVa
riab
leD
escr
iptio
nTh
resh
old/
defin
ition
#sp
ecie
s
I.En
dem
ism
(n=
1267
,45.
4%of
alls
pp.)
1.En
dem
ism
Taxo
nis
ende
mic
toa
regi
onH
igh
=en
dem
ic12
67Lo
w=
not
ende
mic
1526
II.IU
CNRe
dLi
stst
atus
(n=
585;
20.9
%of
alls
pp.)
2.Le
velo
fend
ange
rmen
tTa
xon'
slis
ting
unde
rth
eIU
CNvu
lner
abili
tyas
sess
men
tH
igh
=cr
itica
llyen
dang
ered
(CR)
,end
ange
red
(EN
),vu
lner
able
(VU
)58
5Lo
w=
leas
tco
ncer
n(L
C),n
otth
reat
ened
(NT)
1680
Unk
now
n=
data
defic
ient
(DD
)52
8
III.P
rovi
sion
ofec
osys
tem
serv
ices
(n=
712,
25.5
%of
all
spp.
)3.
Fish
ing
Taxo
nis
part
ofan
econ
omic
ally
-impo
rtan
tfis
hery
Hig
h=
fishe
rycl
assi
fied
asei
ther
artis
anal
,com
mer
cial
,or
high
lyco
mm
erci
al68
4
Low
=no
tha
rves
ted
2109
4.O
ther
uses
Taxo
nha
sot
her
cultu
ralo
rec
onom
icus
es(T
able
S9)
Hig
h=
used
for
≥2
othe
rpu
rpos
es69
Low
=us
edfo
r<
2ot
her
purp
oses
2724
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
510
vulnerability dimensions based on pre-defined scoring regimes in eachtrait (details in Table 1, Table S1, and Supplementary Methods S.1.3). Ifa species scored high for even one variable within one trait set in eithersensitivity or exposure, it was given a high score in that dimension, and ifa species scored low for a single variable in adaptive capacity it wasgiven a low score for that dimension. A species was given a high scorefor vulnerability if it had all three of high sensitivity, low adaptive capa-city, and high exposure. Species with high sensitivity and low adaptivecapacity, but low exposure were categorized as ‘high latent risk’. Al-though species in this category are not predicted to experience as muchchange in their physical environment as others, if changes in climateare greater than considered here, their status could change to 'vulner-able'. Species that scored high for both exposure and sensitivity, but alsohave high adaptive capacity were considered ‘potential adapters’. Thesespecies may be able to cope with climate change through dispersal ormicro-evolutionary change. Species that were predicted to be highlyexposed and had low adaptive capacity, but were not highly sensitivemight be naturally resilient to changes in their environment, and wereclassified as ‘potential persisters’. While these categories are not mu-tually exclusive (e.g., the capacity to persist may be adaptive), theyprovide convenient groupings to understand various strategies thatmight be used to cope with environmental change.
For qualitative traits (e.g., taxon requires rains to cue spawningmigrations), ‘high’ or ‘low’ scores depended only on whether a speciespossessed that trait or not. However, continuous traits required selec-tion of arbitrary scoring thresholds (Foden et al., 2013). For such traits,thresholds were chosen based on the distribution of trait values in ourdataset. For example, fishes with estimates of climate change exposurethat fell within the highest 25% of all fishes were given a ‘high’ score,while those in the lowest 75% were given a ‘low’ score. We tested thesensitivity of our analysis to these threshold selections by shiftingthresholds to different cut-off points with lenient scenarios resulting infewer species being given high scores (e.g., the top 15% scored as high)and strict thresholds resulting in more species being given high scores(e.g., the top 35% scored as high). Differences among these three ex-posure thresholds (15%, 25%, and 35%) are compared in the Supple-mentary Tables and Figures. Another source of uncertainty in our dataincluded traits within the sensitivity and adaptive capacity dimensionswhere there were unknown values. To address this uncertainty wecalculated scores for these dimensions under both optimistic and pes-simistic scenarios following Foden et al. (2013). In the optimistic sce-nario all unknowns were coded as low for sensitivity and high foradaptive capacity, and vice versa for the pessimistic scenario. The mainresults of this paper are presented under the lenient/optimistic scenariofor all traits; however, we compared these with results from the strict/pessimistic scenarios, and explored differences in distribution of climatechange vulnerable species in the Supplementary Tables and Figures.Details on various threshold values used under both optimistic andpessimistic scenarios, and information on data availability (numbers ofunknowns) can be found in Table S1. Methods used to deal with ‘un-knowns’ and variations in threshold limits and scenarios are discussedin detail in the Supplementary Methods S.1.6. Discussion on how var-ious thresholds were selected and how unknowns were handled for eachtrait can be found in Supplementary Methods S.1.8.
2.6. Assigning conservation value scores
We included an index of conservation value in this study to demon-strate that the often-subjective judgement calls frequently used to as-sign ‘conservation value’ might result in some vulnerable regions orspecies being overlooked. Conservation value of species was defined inthis study as the relative importance of species based on traditionalconservation prioritization metrics including 1) level of extinction riskas defined by the IUCN Red List, 2) endemism, and 3) potential toprovide ecosystem services through i) fisheries and ii) other cultural oreconomic uses. Species that scored high in any one of these traits were
given a high score for conservation value. Although it may be argued that‘potential to provide ecosystem services’ (i.e., importance to humanuse) is opposed to conservation prioritization, we retain it in our indexof conservation value as many regional conservation plans for Africanfreshwater systems include and/or prioritize management of naturalresources (Phil-Eze and Okoro, 2009; Harrison et al., 2016). However,to test the sensitivity of our results to the inclusion vs. exclusion of‘ecosystem services’ in this index, we performed analyses both ways andcompared results in the Supplementary Tables and Figures.
2.7. Mapping concentrations of climate change vulnerable and highconservation value species
Distributions of species were mapped to identify regions with denseconcentrations of highly climate change vulnerable and high conserva-tion value species. We used univariate maps to show the spatial dis-tributions of species with high scores for vulnerability, conservationvalue, sensitivity, and exposure, and low scores in adaptive capacity. Weused bivariate maps to explore the spatial relationships between biolo-gical susceptibility and exposure, and between climate change vulner-ability and high conservation value.
To produce univariate maps, we created a separate raster layer forevery species in our dataset denoting presence or absence in every 10-min grid cell based on whether any part of their range intersected witha grid cell. We then created a file containing species' binary scores forsensitivity, adaptive capacity, exposure, vulnerability, and conservationvalue under all combinations of scenario (optimistic vs. pessimistic),RCP (4.5 vs. 8.5), year (2055 vs. 2085), and threshold (strict vs. le-nient). Species ranges were then stacked, and each grid cell was as-signed a value indicating the number of species with high scores (or lowscores in the case of adaptive capacity) occupying that grid cell. In theunivariate maps, each grid cell was assigned a colour that graduatedfrom blue to red to signify low to high frequency of species.
To produce bivariate maps, values of species frequencies in theunivariate layers of interest were split into 10 quantiles based on Jenksnatural breaks. These layers were then overlaid, and bivariate plotswere created by assigning each grid cell a score based on the overlap inquantile scores. A bivariate colour scheme was used to classify the gridsuch that cells with low frequencies were given duller colours and cellswith higher frequencies were given darker colours, with tones ap-proaching blue on the y-axis, yellow on the x-axis, and maroon wherethe two traits overlap.
2.8. Trait and family analysis
To determine traits that contributed most to each vulnerability di-mension (sensitivity, adaptive capacity, and exposure) and to conservationvalue we analyzed trait data in two ways. First, we calculated traits thathad the highest proportion of high scores (or low scores in the case ofadaptive capacity) in each category after excluding unknowns, andranked these traits in order of importance. Second, we carried out asensitivity analysis to determine how many species were given a highscore (or low score in the case of adaptive capacity) within each di-mension based exclusively on one trait. This provided an indication ofhow sensitive the analysis was to the specific traits selected, and howremoving a given trait would affect the overall analysis. Traits that gavemore species a high (or low) score in a given category were consideredto be more influential. For families with more than nine species, wedetermined which had the highest proportion of vulnerable species, andassessed the dimensions of vulnerability that were most likely to conferhigh or low scores. We also assessed families most likely to be classifiedas high conservation value. We compared the top 10 ranked families forboth vulnerability and conservation value to assess how families would bedifferently prioritized under these two methods of valuation.
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
511
3. Results
3.1. Data availability and quality
We were able to obtain species ranges for 2984 species; however,taxonomic anomalies, misclassifications, or missing data restricted thenumber of species that could be included in our dataset to 2793. Thisrepresents approximately 85% of the estimated 3300 freshwater fishspecies extant in Africa, and is likely to produce a broadly re-presentative picture of spatial patterns of climate change risk. Speciesdistributions are mapped in Fig. 1B. Data availability of traits rangedfrom 100% of species having data (e.g., climate change projections) toonly 4% (e.g., relative fecundity). Although 4% is quite low, traits wereretained in the analysis if they contributed novel classifications to ourdataset (i.e., if they were not completely redundant with other traits)regardless of the number of species with data. For 19 of the 25 biolo-gical traits quantified, ‘unknowns’ comprised < 10% of the data(Table 1). There were four traits within the sensitivity dimension thathad large data gaps including length of spawning cycle, depth rangeoccupied, prey specialization, and population fragmentation (Table 1).There were two traits within the adaptive capacity dimension that hadlarge data gaps including dispersal of early life stages and reproductivecapacity (Table 1). Sensitivity analyses comparing various thresholds(15%, 25%, 35%), scenarios (optimistic, pessimistic), years (2055,
2085) and projections (RCP8.5, RCP4.5) revealed very little differencebetween RCP8.5 and RCP4.5 for either 2055 or 2085 (Table S10; Fig.S3). There were more differences between the optimistic and pessi-mistic scenarios, especially for the 2055 projections; however, thegeneral spatial patterns remained the same (Table S10; Fig. S3). Ap-plication of different climate change thresholds (15%, 25%, 35%)caused notable difference among distributions of vulnerable species(Fig. S4), indicating that selection of arbitrary thresholds requirescareful consideration. Please see Supplementary Methods S.1.6 andS.1.8 for information on data quality and how we treated unknownsand thresholds in our dataset.
3.2. Summary of vulnerability dimensions
Unless otherwise stated, the following summaries present data de-rived from the RCP8.5 emissions scenario at a 25% exposure thresholdfor the year 2055, and used the optimistic/lenient scenario for all traits.Percentages were calculated as the proportion of all species (n = 2793)in the study. We identified 2375 freshwater fishes (85.3%) that scoredhigh for sensitivity to climate change under the optimistic scenario(Fig. 2, Table 1, Fig. S1AB [distribution], Fig. S6, Table S11 [by fa-mily]), and 2698 (96.6%) under the pessimistic scenario (Fig. 2, TableS1). We found that 753 species (27%) qualified as highly sensitive be-cause of a single trait, the most common being low tolerance to
Fig. 2. Numbers of species with high scores (or low in thecase of adaptive capacity) for each cimate change vulner-ability dimension and for conservation value under the(A) optimistic scenario and (B) pessimistic scenario. Forboth scenarios, we have provided counts and percentagesof all species that qualify for each vulnerability dimension(high sensitivity, low adaptive capacity [Ad. Cap.], highexposure), for combinations of dimensions (potentialpersisters, potential adapters, potentially at risk), forthose that are vulnerable overall, and for those those thatare of high conservation value. We have also providednumbers of species classified as ONLY highly sensitive, oflow adaptive capacity, highly exposed, and those that didnot receive any ‘vulnerable’ score in any dimension. Allvalues contained within boxes (not including conserva-tion value) add to 100%. All values are based on theRCP8.5 emissions scenario for 2055 at the 25% exposurethreshold.
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
512
temperature and precipitation change (24.7% and 24.4%, respectively),intense fishing pressure (24.5%), and high microhabitat specialization(23.3%; Fig. 5). The majority of species possessed multiple traits thatcategorized them as highly sensitive: 54.5% (1523 species) possessedbetween two and five traits, and a small proportion (3.5%, 99 species)possessed six or more. Other traits that contributed highly to sensitivityincluded low population abundance (16%, 455 species), narrow spatialrange (14.2%, 397 species), and exposure to climate change ex-acerbated threats (13.3%, 371 species). Apart from fishing, commonthreats included sedimentation (35%), dams and water diversion(29%), deforestation (14%), and mining effluents (13%, Table S5).
In our analysis, 71.6% (2000) of freshwater fishes were deemed tohave low adaptive capacity under the optimistic scenario (Fig. 2,Table 1, Fig. S1CD [distribution], Fig. S6, Table S11 [by family]). Thisincreased to 97.5% (2724) under the pessimistic scenario (Fig. 2, TableS1). There were 1258 species (45%) that were given a low score foradaptive capacity because of a single trait, the most common of whichwas having poor potential for population growth (Fig. 5). The traits thathad the largest influence on adaptive capacity included dispersal of earlylife history stage (27.7%), physical barriers to dispersal (19.2%), andlow estimated population growth rates (42.5%). There were 1620 fishspecies (58%) that possessed between two and three traits that in-dicated low adaptive capacity with a small proportion (1%, 32 species)possessing four traits.
Finally, we found that 67.2% (1876) of freshwater fish species werepredicted to be highly exposed to climate change (Fig. 2, Table 1, Fig.S1EF [distribution], Fig. S6, Table S11 [by family]). Under the RCP8.5-2055 scenario at a 25% threshold, there were 919 fish species (33.0%)that were given a high score for exposure because of just one trait, themost common of which was exposure to high levels of mean tempera-ture change (Fig. 5). We compare numbers of exposed species underdifferent thresholds (15%, 25%, and 35%) in Table S1 and comparevulnerability distributions in Fig. S4.
3.3. Summary of climate change vulnerability
We identified 1142 freshwater fishes (40.9%) as highly vulnerable
to climate change under the optimistic scenario (Fig. 2, Table 1, Fig.S1GH [distribution], Fig. S6, Table S11 [by family]), and 1759 (63.0%)under the pessimistic scenario (Fig. 2). An additional 613 (21.9%) wereidentified as of ‘high latent risk’ under the optimistic scenario, whichincreased to 874 (31.3%) under the pessimistic scenario (Fig. 2). Wecompare numbers of vulnerable species under different scenarios (op-timistic, pessimistic), years (2055, 2085), and emission scenarios(RCP4.5, RCP8.5) in Table S10. We compared differences in con-centrations of vulnerable species under optimistic vs. pessimistic sce-narios in Fig. S2, among years and RCPs in Fig. S3, and among variousexposure thresholds (15%, 25%, and 35%) in Fig. S4.
3.3.1. Vulnerable regionsBivariate maps of biological susceptibility vs. exposure highlighted re-
gions of overall climate change vulnerability across Africa (Fig. 3AB). Inthese maps, regions highlighted in blue represent high concentrations ofspecies that have high sensitivity and low adaptive capacity but that are nothighly exposed to climate change (i.e., high latent risk; Fig. 2). Regionswith high concentrations of species of high latent risk included outlyingtributaries of the Nile and Congo rivers, Lake Tanganyika, and the Nilo-Sudan, Northern Africa, and Southern Temperate ecoregions (Fig. 3A).When analyzed by proportion, regions of high latent risk included much ofthe Congo and Limpopo river basins, and Lake Tanganyika (Fig. 3B).Regions where species have low sensitivity and high adaptive capacity, butare expected to be highly exposed (i.e., potential persisters/adapters) arehighlighted in yellow. High concentrations were found in the main bran-ches of the Congo and Nile rivers, many of the rivers and lakes within theZambezi ecoregion, and the coastal rivers of the Lower Guinea ecoregion(Fig. 3A). When analyzed by proportion, Lake Victoria also stood out(Fig. 3B). Areas with high frequencies of climate change vulnerable species(indicated in maroon tones; Fig. 3AB) included many of the coastal andinland rivers and lakes in the Upper Guinea and Nilo-Sudan ecoregions, inparticular the Volta and Niger river drainages. The rift valley lakes (Vic-toria and Malawi), the Okavango delta, and tributaries of the ZambeziRiver were also found to be at risk (Fig. 3A). When analyzed in terms ofproportions, regions within the Sahara desert and the Orange River alsostood out as vulnerable (Fig. 3B).
Fig. 3. Bivariate maps showing regions where there is high overlap in the (A) total count and (B) percent of species that are highly biologically susceptible (i.e., scoredhigh for sensitivity and low for adaptive capacity) and highly exposed to climate change. Distributions are based on the optimistic scenario for the RCP8.5 emissionsscenario and the 25% exposure threshold for the year 2055.
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
513
3.3.2. Vulnerable familiesFamilies that contained the highest percentage of climate change
vulnerable species included the Nothobranchiidae (61% of allNothobranchiids), Cichlidae (58%), and families within the orderSiluriformes (the catfishes) including Bagridae (56%), Malapteruridae(47%), Clariidae (38%), and Mochokidae (38%; Table 2; Fig. S6).Species within the family Eleotridae (sleeper gobies) and Gobiidae (truegobies) had 50% and 43% vulnerable species, respectively. For theNothobranchidae, the most common traits for low adaptive capacityincluded barriers to dispersal (161 species, 76% of nothobranchids),and limited dispersal of early life history stages (74 species, 35%), andfor sensitivity it was microhabitat requirements (188 species, 88%) andstrong dependence on environmental triggers (43 species, 20%). Forcichlids, the most common traits for low adaptive capacity included poordispersal of early life-history stages (249 species, 27% of all cichlids)and low population growth rates (429 species, 47%), and for sensitivityit was disruption of mate-finding due to turbidity (249 species, 27%),highly specified diets (188 species, 20%), and low abundances (219species, 24%). Both Eleotridae and Gobiidae families had low adaptivecapacity due to low population growth rates (5 species, 42% and 6species, 29%, respectively), and high sensitivity due to micro-habitatrequirements (4 species, 33% and 12 species, 57%). The four mostvulnerable catfish families all had low adaptive capacity because of theirlow population growth rates (287 species, 63% of vulnerable catfishspecies), and were sensitive because of low tolerance to temperaturechanges (90 species, 31%), dependence on microhabitats (71 species,25%), and fishing pressure (80 species, 28%).
3.4. Summary of conservation value
We identified 1714 (61%) freshwater fish species of high conserva-tion value under the optimistic scenario based on the traditional metricsof endemism, threat status, and ecosystem services provided (Fig. 4AB,Table 1, Fig. S1IJ [distribution], Fig. S6, Table S11 [by family]), and2232 (80%) species that qualified under the pessimistic scenario (Fig. 2,Table S1). Removing 'provisioning of ecosystem services' from ourcalculations of conservation value reduced the number of species to 1454(52.1%) representing a loss of 260 species from this index (Fig. S5).
There were 973 (35%) species that scored high for conservation valuebecause of one trait (Fig. 5). Twenty percent (571 species) scored highfor conservation value exclusively because of endemism, followed byfishing (207 species, 7%), IUCN threat status (171 species, 6%), andother cultural and economic uses (24 species, 0.8%; Fig. 5). The ma-jority of species in our dataset (2167 fish, 78%) possessed between twoto three traits that scored high for conservation value.
3.4.1. Regions and families of high conservation valueRegions with highest concentrations of species of high conservation
value include the rift valley lakes in eastern Africa (Victoria,Tanganyika, and Malawi), the Congo River and its tributaries, and theNiger River delta (Fig. 4A). When analyzed by proportions of fishspecies, the Orange River drainage, regions of the northern Saharadesert, rivers and lakes in Eastern Coastal ecoregion, and some parts ofthe Great Lakes ecoregion also stood out as high conservation value(Fig. 4B). When ‘provisioning of ecosystem services’ was removed fromthe conservation value index, regions of the Congo and Orange riverswere lost, but the rift valley lakes and many of the coastal West Africanrivers were retained (Fig. S5). Families with many species of highconservation value included the Cichlidae (87% of all cichlids), the No-thobranchiidae (75%), the Mastacembelidae (67%), and families withinthe order Siluriformes including Bagridae (67%), Claroteidae (54%),and Mochokidae (52%; Table 2). For cichlids, the most common traitsthat gave high scores for conservation value included endemism (738species, 87% of all cichlids) and ‘other cultural and economic uses’ (477species, 52%), primarily the aquarium trade. These species are alsoimportant targets of artisanal fisheries (383 species; 41%). For notho-branchiids, the most common traits included ‘other cultural and eco-nomic uses’ (147 species, 69%), again mostly for the aquarium trade,endemism (98 species, 46%), and IUCN threat status (81 species; 38%).
3.5. Comparison of climate change vulnerability and conservation value
Of the 1142 species that were found to be vulnerable to climatechange under the optimistic – RCP8.5 – 2055 scenario, 294 (26%) wereoverlooked by the conservation value index. Bivariate maps highlightregions where indices of climate change vulnerability and conservation
Table 2Numbers and proportions of species within families containing ≥9 species that are highly vulnerable and of high conservation value. Highlighted cells show the top 10families for conservation priority (CP) based on vulnerability to climate change (vul), conservation value (CV), and a combination of the two (combined). Thesevalues are based on the optimistic scenario of the RCP8.5 emissions scenario for 2055.
Family (Order) CP: vul CP: CV combined Total # vul % vul# high
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
514
value are at odds, and where there is overlap in the two indices(Fig. 4CD). Regions where species are highly vulnerable to climatechange, but are not of high conservation value (highlighted in yellow)included many of the inland rivers in the Nilo-Sudan ecoregion, pri-marily the inland portions of the Niger River and areas surroundingLake Chad, some tributaries of the Nile river, and tributaries of theZambezi and Okavango rivers in southern Africa (Fig. 4C). When ana-lyzed by proportion, the Orange and Limpopo rivers in southern Africaalso stood out (Fig. 4D). Noteworthy areas with low climate changevulnerability but high conservation value (highlighted in blue) includedLake Tanganyika, the Orange and Limpopo rivers and their tributaries,and several lesser tributaries of the Congo River (Fig. 4C). When ana-lyzed by proportion the entire Congo River drainage and several inlandrivers in the western portion of the Nilo-Sudan ecoregion also stood out
(Fig. 4D). Areas containing the greatest numbers and proportions offishes that are both climate change vulnerable and of high conservationvalue (indicated in maroon) included the coastal lakes and rivers in theUpper and Lower Guinea ecoregions (e.g., the Niger River delta and theVolta River drainage), the majority of the Congo river drainage, some ofthe East African rift valley lakes (Victoria and Malawi), and the Oka-vango delta (Fig. 4C). When analyzed by proportion some regions of inthe Sahara desert also stood out. When ecosystem services were re-moved from the conservation value index the Orange River was nolonger highlighted in blue in the bivariate map; however, there are fewother noteworthy differences (Fig. S5). Family-level analysis revealed adifferent order of conservation prioritization based on climate changevulnerability vs. conservation value (Table 2).
Fig. 4. Univariate maps showing the distribution of the (A) number and (B) proportion of species that are of high conservation value, and bivariate maps showing (C)total count and (D) percent of regions where there is high overlap of species that are highly climate change vulnerable and of high conservation value. Distributionsare based on the optimistic scenario for the RCP8.5 emissions projection and the 25% exposure threshold for the year 2055.
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
515
4. Discussion
This study presents the first comprehensive trait-based CCVA ofAfrica's freshwater fish species. Results indicated that a high proportionof species are likely to be negatively affected by climate change, with41% of species qualifying as highly vulnerable under an optimisticscenario using the RCP8.5 projection for 2055. This proportion isslightly higher than those found in regional assessments of freshwaterfishes in the Albertine rift (31%; Carr et al., 2013) and West Africa(39%; Carr et al., 2014), and substantially higher than proportionsdocumented in global assessments of amphibians (22%), corals (15%),birds (24%; Foden et al., 2013), and reptiles (22%; Böhm et al., 2016).Although direct comparisons are limited by variation among studies'scoring systems, it is interesting that African freshwater fishes have oneof the highest vulnerability percentages, supporting predictions thatthis group is highly susceptible to environmental change (Dudgeonet al., 2006).
This study offers several important contributions to nascent con-servation strategies for this fauna. First, it identifies species and regionsthat are highly vulnerable to climate change, and pinpoints biologicaland ecological characteristics that most contribute to this vulnerability.Second, it highlights knowledge gaps that hinder our understanding ofspecies vulnerability. Third, it identifies species and regions where
typical conservation metrics may under- or over-emphasize conserva-tion priority, and provides information necessary to incorporate vul-nerability into conservation planning.
4.1. Vulnerable species and families
Targeted conservation interventions require species- and region-specific information on threats to ecological systems (Abrahms et al.,2017). We identified regions with high frequencies of vulnerable spe-cies, and we identified species and families that fall within categories ofvulnerable, high latent risk, potential persisters, or potential adapters(Tables S11 and S12). Species of high latent risk are considered to bebiologically susceptible, but are not predicted to be highly exposed toclimate change. However, if climate change proceeds at a more extremepace than predicted, they could become vulnerable in the future. Thesespecies should be closely monitored and re-assessed as new projectionsemerge. Species that are potential adapters or potential persisters re-present fishes whose environments are likely to change, but who havethe capacity to avoid, adapt, or cope with these changes. These fishesmay avoid extinction or extirpation if they are able to use their survivalcapacities as predicted. Vulnerable species are very likely to be nega-tively affected by climate change, and are therefore of highest priorityfor climate change mitigation and conservation action. Here we cannot
Fig. 5. Summary of the number of species qualifyingfor high vulnerability exclusively due to single traitwithin the dimensions of (A) sensitivity, (B) adaptivecapacity, and (C) exposure, and for (D) conservationvalue. Using Panel A (Sensitivity) as an example, thisshows that SV5 (Tolerance to temperature change)qualified ~200 species as highly sensitive that did notscore high for any other Sensitivity variable. The in-terpretation is that, if SV5 were removed from the‘Sensitivity’ dimension, those ~200 species would nolonger qualify has highly sensitive, and therefore notqualify has highly vulnerable overall. Traits are rankedby importance. Data are based on an optimistic sce-nario for unknown trait values under the RCP8.5emission scenario at a 25% exposure threshold for2055.
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
516
provide lengthy descriptions of families or species falling into thesecategories; however, family-level details are provided in the Supple-mentary Tables S11 and S12, and species-level details are available inthe online data archive associated with this publication.
4.2. Species traits' contributions to vulnerability
The contributions of biological traits to vulnerability were highlyvariable among species and families, reflecting the diversity of lifehistory strategies and ecological niches occupied by fishes across Africa.Microhabitat specialization was a very important sensitivity trait.Temporary pools (e.g., in flooded forests) emerged as an importantmicrohabitat for many species, along with cool, clear rainforeststreams, and shallow, fast flowing creeks, mirroring findings of Carret al. (2013, 2014). High dependence on microhabitats indicates thatspecies have narrow environmental tolerances and are likely to havelocally-adapted phenotypes (Hannah et al., 2014; Nadeau et al., 2017).Adverse effects on microhabitats are likely to expose such species toeffects of climatic change. In addition, many species were classified assensitive based on small population sizes, narrow distributions, or de-clining abundances. In general, species that have small population sizesare expected to be vulnerable to stressors as they are less likely tocontain resilient individuals or to produce offspring with advantageousgenetic combinations for coping with novel climatic scenarios (Sgròet al., 2011). Carefully monitoring rare species, identifying key causesof population decline, and taking steps to mitigate those causes canreduce their sensitivity to climate change.
Barriers to dispersal were important factors leading to low adaptivecapacity in many fish species. Freshwater fishes are embedded withinterrestrial landscapes that limit dispersal within and among drainagebasins. (Olden et al., 2010; Bush and Hoskins, 2017). Climate changehas been shown to shift environmental conditions across landscapes,and these shifts are expected to progress with time (Parmesan andYohe, 2003). Species that have restricted capacities for dispersal will beunable to make range shifts that match these changing conditions, andare likely to be of high extinction risk (Bush and Hoskins, 2017). Forexample, lakes Malawi and Tanganyika have shorelines comprised ofrocky, swampy, and sandy patches that act as ecological islands, manyof which host their own species of small, specialized Cichlid fishes(Lowe-McConnell, 1969). These species might not have the ability totraverse inhospitable terrain, and are therefore likely to be vulnerableto climate change. Maintaining intact habitat can improve the adaptivecapacity of such species.
Low population growth rate also emerged as an important barrier toadaptive capacity. Population growth rate was inferred from life historytraits assuming that r-selected species with fast life-cycles will havehigher population growth (and therefore better adaptive capacity).Although there is mixed evidence for this pattern in fishes (Pinsky andByler, 2015), generally, species with shorter generation times andhigher fecundity have faster adaptive responses to environmentalchange (Lande, 1993; Williams et al., 2008; Pacifici et al., 2017). Thatbeing said, traits that are favourable to one aspect of adaptation may beharmful for another. For example, higher relative fecundity may beadvantageous for faster rates of evolutionary adaptation, but thesesame species are often total spawners making them more susceptible tounpredictable seasonal conditions (Hughes, 2017). Furthermore, spe-cies with fast generation times tend to be smaller-bodied and shorter-lived, two traits that reduce capacity for dispersal (Comte and Olden,2018). The ways that various life history traits affect the capacity ofspecies to adapt to climate change are likely to be highly context spe-cific, and more research is required to determine whether predictablepatterns exist.
4.3. Non-climate change environmental stressors
Freshwater fishes are exposed to a wide array of anthropogenic
stressors in their natural habitats, including overexploitation, habitatdegradation, and water pollution, among others (Dudgeon et al., 2006).Of the threats examined in this study, fishing was dominant with over1000 species targeted in commercial or artisanal fisheries across thecontinent. Expansion of urban and industrial areas, land conversion foragriculture, natural resource extraction (e.g., mining), and introductionof invasive species also have widespread impacts on African aquaticenvironments. Although ‘exposure to stressors’ is not an intrinsic spe-cies trait, it was included in this study because many stressors are likelyto be intensified by climate change; this variable thus provides im-portant information for predicting vulnerability (Foden et al., 2018;Xenopoulos et al., 2005; Ficke et al., 2007). In addition, species alreadyaffected by other stressors are likely to be using physiological, genetic,and behavioural resources to cope, and might therefore have acuteresponses to the additional impacts of climate change (Staudt et al.,2013).
4.4. Identifying knowledge gaps and research needs
One of the limitations of trait-based CCVAs is the paucity of data forsome traits (Pacifici et al., 2015). In this analysis 19% (203) of climatechange vulnerable species were classified as data deficient according tothe IUCN Red List, including several species in the Bagriidae, Notho-branchiidae, Clariidae, and Cichlidae families (Table S13). In this study,there were large knowledge gaps for several traits, especially thosepertaining to species' adaptive capacity. For example, only 4% of specieshad data on relative fecundity, and estimates of natural mortality andindividual growth rate were so rare (1% and 2%, respectively) thatthese traits did not meet the inclusion criteria for this analysis (seeMethods). In addition, measures of genetic diversity, connectivityamong populations, and population size, growth, and structure areabsent for the majority of freshwater fish species in Africa. If this in-formation were available it could help to understand past and futurepopulation trajectories (Foden et al., 2018), and could lead to differentclassifications of vulnerability. For instance, if a highly sensitive specieswas found to have high genetic variability despite low population size,this could alter its vulnerability ranking. Ecological traits pertaining tothe senstivity dimension were better represented in this analysis; how-ever, there were still significant data gaps for depth range, spawningcycle, and population fragmentation. Of these, basic life-history in-formation, population size, and spawning traits stood out as importantfor understanding climate change effects, and could help in assessingspecies' vulnerability.
Clearly, missing data can introduce uncertainty into interpretationsfor conservation goals and may bias our understanding of which traitsconfer vulnerability, or which species are the most vulnerable. Thefindings of this study should therefore not be interpreted as precisepredictions, but rather viewed as a forecasts of plausible vulnerabilitypatterns across continental Africa informed by the best available datafor broad range of species (Foden et al., 2018). By providing such anoverview this study can be used to guide future research. For example,one way forward is to implement ecological surveys designed to collectdata on traits that are relevant for predicting species’ climate changevulnerability. From the results of this and other CCVAs, these includebasic life-history traits (e.g. individual growth rate, age at maturity,relative fecundity), information on reproductive behaviour (e.g.,spawning cycles), and estimates of species’ population sizes, populationconnectivity, and dispersal ability (Foden et al., 2013; Carr et al., 2013,2014; Böhm et al., 2016; Hare et al., 2016). Molecular data on popu-lation structure and genetic variation within populations can provideestimates of local adaptation, gene pool mixing, historical populationgrowth, and other population processes (Foden et al., 2018). Surveysand collections should focus on species groups that are both data de-ficient and likely to be climate-change vulnerable (list available inTable S13); however, imputation techniques can also be used to reliablyestimate values among closely related species where traits have a strong
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
517
phylogenetic signal (Penone et al., 2014). Further suggestions for futurestudy are available in the Supplementary Materials S.1.9.
4.5. Conservation planning
Resources for the conservation of natural systems are limited, soefforts to protect biological diversity require prioritization (Arponen,2012). Traditional indices of species' conservation value commonlyincorporate estimates of economic or socio-cultural value, phylogeneticdistinctiveness, endemism, extinction risk (e.g., IUCN status), andrarity, among others (Reece and Noss, 2014; Capmourteres and Anand,2016). However, such value judgements are often subjective (Arponen,2012), and assessments that ignore climate change might overlook re-gions or species that are at risk from long-term environmental shifts.Because our understanding of how climate change affects species isrelatively recent, it is not often incorporated into conservation plans(Stewart et al., 2018), and the most recent continent-wide conservationassessment of African freshwater ecosystems only peripherally in-tegrates climate change in its prioritization scheme (Thieme et al.,2005). The main accomplishment of the current study is to provide thefirst comprehensive CCVA summarizing current knowledge on themajority (85%) of African freshwater fish species. These findings can beused to guide ecosystem monitoring efforts and experimental or field-based research. This work also demonstrates that some conservationmetrics may overlook regions that are likely to be vulnerable to climatechange, which is an important consideration as new conservation plansare developed.
4.5.1. Comparison of climate vulnerability to traditional conservation valuemetrics
Areas of high conservation value (based on extinction risk, en-demism, and provision of ecosystem services) did not always align withareas predicted to be highly affected by climate change (e.g., the Niger,Nile, Zambezi, and Okavango rivers and their tributaries, and areassurrounding Lake Chad). These regions might be ignored by traditionalconservation assessments that do not incorporate climate change vul-nerability. Species and families that are both highly climate changevulnerable and of high conservation value, and the regions in which theyare concentrated, deserve particular conservation attention to bothmitigate current threats and to plan for future climate change adapta-tion interventions. Although the measures included in the conservationvalue index in this study may be somewhat subjective, such findings arelikely to hold true regardless of what conservation priorities are used, asdemonstrated by the sensitivity analysis performed by removing someaspects of conservation value from our index (Fig. S5).
4.5.2. Contribution to conservation managementConservation plans require knowledge on the major biological and
ecological constraints of species and families, (e.g., distribution, criticalhabitat), information on their conservation status, and key threats theyface (Stewart et al., 2018). Understanding how species are groupedaccording to their relative risk to climate change and knowing wherethese species are concentrated can be used to direct ecosystem mon-itoring, and can help to identify conservation strategies that benefitmultiple species (Foden et al., 2018). Having this information acces-sible in a cohesive dataset, such as the one produced for this study(Nyboer et al., 2019), can facilitate creation of localized, adaptiveconservation strategies.
For species and families that are vulnerable to climate change,conservation actions should include preserving and restoring habitats,targeting research to fill data gaps, and carrying out regular monitoringof vulnerable populations. For species not highly climate change vul-nerable, conservation should focus on preserving current habitats asthey are, because these species are unlikely to need to move if theirenvironments are kept intact (Stewart et al., 2018). In addition, moredirect management approaches will also be necessary. Policies that aim
to improve connectivity of fragmented landscapes, preserve habitatsthat reduce impacts of climate change for a diversity of species (e.g.,forests, heterogeneous regions), and reduce existing threats (e.g., re-ducing harvest rates, protecting riparian areas) are proactive ap-proaches for conserving biodiversity under climate change (Mantyka-Pringle et al., 2016; Pletterbauer et al., 2018).
Effects of climate change are a moving target, and our under-standing of global warming is ever-changing. Traditional approaches ofstatic reserve designs need to be adjusted to incorporate the shiftingnature of climatic space under future conditions (Williams et al., 2008;Webster et al., 2017). Protected areas might not contain species afterconditions within reserve boundaries change (Abrahms et al., 2017).Conservation managers will need to design and implement adaptableconservation plans (as opposed to static reserve designs) that can beupdated as new estimates of vulnerability emerge and shifts in biodi-versity concentrations occur (Williams et al., 2008). Recent approachesto conservation suggest developing networks of landscapes that containa diversity of species and environmental conditions that promote con-nectivity and increase the likelihood of adaptive species responses(Webster et al., 2017).
Future conservation plans will be specific to each region, but shouldtry to (i) enhance access to important, rare, and vulnerable habitatssuch as migratory routes and spawning grounds, (ii) preserve connectedhabitats or build connectivity of fragmented landscapes (i.e., maintaincorridors; reduce water abstraction) to provide routes for dispersal orgene flow, (iii) focus on regions that can reduce the effects of climatechange (i.e., forests) or maintain habitats in zones that are likely to beless affected by future climatic changes, (iv) preserve areas with highhabitat heterogeneity that contain climate refuges and a diversity ofniches to sustain genetic and species diversity and the capacity formany species to adapt, (v) engage in habitat restoration for areas thatare already highly threatened and that also contain species that arelikely to be vulnerable, and (vii) reduce impacts of other stressors byimplementing protective land-use policies, using selective fishingmethods to target less vulnerable species, altering fishing regulations tomaintain genetic or phenotypic diversity in harvested populations, andsuppressing invasive species introductions.
Acknowledgements
This research was funded by the Natural Sciences and EngineeringResearch Council of Canada (PGS-D3-426309) to E.A.N. and by NaturalSciences and Engineering Research Council of Canada Discovery Grant(RGPPIN/06675-2015), Canada Research Chair funds for L.J.C.).
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.biocon.2019.05.003.
References
Abrahms, B., DiPietro, D., Graffis, A., Hollander, A., 2017. Managing biodiversity underclimate change: challenges, frameworks, and tools for adaptation. Biodivers. Conserv.26, 2277–2293.
Adrian, R., O'Reilly, C.M., Zagarese, H., Baines, S.B., Hessen, D.O., Keller, W., Livingstone,D.M., Sommaruga, R., Straile, D., Van Donk, E., et al., 2009. Lakes as sentinels ofclimate change. Limnol. Oceanogr. 54, 2283–2297.
Arponen, A., 2012. Prioritizing species for conservation planning. Biodivers. Conserv. 21,875–893.
Böhm, M., Cook, D., Ma, H., Davidson, A.D., García, A., Tapley, B., Pearce-Kelly, P., Carr,J., 2016. Hot and bothered: using trait-based approaches to assess climate changevulnerability in reptiles. Biol. Conserv. 204, 32–41.
Brander, K.M., 2010. Impacts of climate change on fisheries. J. Mar. Syst. 79, 389–402.Bush, A., Hoskins, A.J., 2017. Does dispersal capacity matter for freshwater biodiversity
climate change cause extinction? Proc. R. Soc. B Biol. Sci. 280, 20121890.Capmourteres, V., Anand, M., 2016. Conservation value: a review of the concept and its
vulnerable: climate change vulnerability and human use of wildlife in Africa'sAlbertine Rift. In: Occasional Paper of the IUCN Species Survival Commission No. 48.IUCN, Gland, Switzerland and Cambridge, UK.
Chamberlain, S., 2018. Package ‘rredlist. https://github.com/ropensci/rredlist.Chessman, B.C., 2013. Identify species at risk from climate change: traits predict the
Dudgeon, D., Arthington, A.H., Gessner, M.O., Kawabata, Z.I., Knowler, D.J., Lévêque, C.,Naiman, R.J., Prieur-Richard, A.H., Soto, D., Stiassny, M.L.J., Sullivan, C.A., 2006.Freshwater biodiversity: importance, threat status and conservation challenges. Biol.Rev. 81, 163–182.
ESRI, 2017. ArcGIS Desktop: Release 10. Environmental Systems Research Institute,Redlands, CA, USA.
Fick, S.E., Hijmans, R.J., 2017. Worldclim 2: New 1-km spatial resolution climate surfacesfor global land areas. Int. J. Climatol. 37, 4302–4315.
Ficke, A.D., Myrick, C.A., Hansen, L.J., 2007. Potential impacts of global climate changeon freshwater fisheries. Rev. Fish Biol. Fish. 17, 581–613.
Foden, W.B., Butchart, S.H.M., Stuart, S.N., Vié, J.C., Akcakaya, H.R., Angulo, A.,DeVantier, L.M., Gutsche, A., Turak, E., Cao, L., et al., 2013. Identifying the world'smost climate change vulnerable species: a systematic trait-based assessment of allbirds, amphibians and corals. PLoS One 8, e65427.
Foden, W.B., Young, B.E., Akcakaya, H.R., Garcia, R.A., Hoffmann, A.A., Stein, B.A.,Thomas, C.D., Wheatley, C.J., Bickford, D., Carr, J.A., et al. (2018). Climate changevulnerability assessment of species. Climate Change. doi.org/https://doi.org/10.1002/wcc.551.
Food and Agriculture Organization (FAO) of the United Nations, 2018. The State of WorldFisheries and Aquaculture 2018 - Meeting the Sustainable Development Goals.(Rome, Italy).
Froese, R., Pauly, D., 2018. FishBase. World Wide Web electronic publication. www.fishbase.org.
Hare, J.A., Morrison, W.E., Nelson, M.W., Stachura, M.M., Teeters, E.J., Griffis, R.B.,Alexander, M.A., Scott, J.D., Alade, L., Bell, R.,.J., et al., 2016. A vulnerability as-sessment of fish and invertebrates to climate change on the Northeast U.S.Continental Shelf. PLoS One e0146756, 11.
Harrison, I.J., Brummett, R., Stiassny, M.L.J., 2016. The Congo River Basin. In: Finlayson,C.M., Milton, G.R., Prentice, R.C., Davidson, N.C. (Eds.), The Wetland Book II:Distribution, Description and Conservation. Springer Science, Netherlands, pp. 1–18.
Heino, J., Virkkala, R., Toivonen, H., 2009. Climate change and freshwater biodiversity:detected patterns, future trends and adaptations in northern regions. Biol. Rev. 84,39–54.
Hughes, P.W., 2017. Between semelparity and iteroparity: Empirical evidence for acontinuum of modes of parity. Ecol. Evol. 7 (20), 8232–8261. https://doi.org/10.1002/ece3.3341. Published 2017 Sep 7.
International Union for the Conservation of Nature (IUCN), 2018. The IUCN red list ofthreatened species. http://www.iucnredlist.org, Version 2018-1.
Jeppesen, E., Kronvang, B., Meerhof, M., Sondergaard, M., Hansen, K.M., Anderen, H.E.,Lauridsen, T.L., Liboriussen, L., Bekilogu, M., Ozen, A., et al., 2009. Climate changeeffects on runoff, catchment phosphorus loading and lake ecological state, and po-tential adaptations. J. Environ. Qual. 38, 1930–1941.
Jeppesen, E., Mehner, T., Winfield, I.J., Kangur, K., Sarvala, J., Gerdeaux, D., Rask, M.,Malmquist, H.J., Holmgren, K., Volta, P., et al., 2012. Impacts of climate warming onthe long-term dynamics of key fish species in 24 European lakes. Hydrobiologia 694,1–39.
Knouft, J.H., Ficklin, D.L., 2017. The potential impacts of climate change on biodiversityin flowing freshwater systems. Annu. Rev. Ecol. Evol. Syst. 48, 111–133.
Lande, R., 1993. Risks of population extinction from demographic and environmentalstochasticity, and random catastrophes. Am. Nat. 142, 911–927.
Leadley, P., Pereira, H.M., Alkemade, R., Fernandez-Manjarrés, J.F., Proença, V.,Scharlemann, J.P.W., Walpole, M.J., 2010. Biodiversity Scenarios: Projections of 21stCentury Change in Biodiversity and Associated Ecosystem Services, Technical Series.132. Secretariat of the Convention on Biological Diversity, Montreal, Canada, pp. 50.
Lehner, B., Doll, P., 2004. Global Lakes and Wetlands Database. World Wildlife Fund,Washington, DC, USA.
Lehner, B., Verdin, K., Jarvis, A., 2008. New global hydrography derived from spaceborneelevation data. Eos. Trans. AGU 89, 93–94.
Lévêque, C., Paugy, D., 2010. Animal resources and diversity in Africa, freshwater fishesin Africa. In: Encyclopedia of Life Support Systems (EOLSS). Eolss Publishers, Paris,France Developed under the auspices of the UNESCO.
Livingstone, D.M., 2003. Impact of secular climate change on the thermal structure of alarge temperate central European lake. Clim. Chang. 57, 205–225.
Lowe-McConnell, R.H., 1969. Speciation in tropical freshwater fishes. Biol. J. Linn. Soc. 1,
F., Bunn, S.E., Rhodes, J.R., 2016. Prioritizing management actions for the con-servation of freshwater biodiversity under changing climate and land cover. Biol.Conserv. 197, 80–89.
Martay, B., Brewer, M.J., Elston, D.A., Bell, J.R., Harrington, R., Brereton, T.M., Barlow,K.E., Botham, M.S., Pearce-Higgins, J.W., 2017. Impacts of climate change on na-tional biodiversity population. Trends. Ecography 40, 1139–1151.
Myers, B.J.E., Lynch, A.J., Bunnell, D.B., Chu, C., Falke, J.A., Kovach, R.P., Krabbenhoft,T.J., Kwak, T.J., Paukert, C.P., 2017. Global synthesis of the documented and pro-jected effects of climate change on inland fishes. Rev. Fish Biol. Fish. 27, 339–361.
Niang, I., Ruppel, O.C., Abdrabo, M.A., Essel, A., Lennard, C., Padgham, J., Urquhart, P.,2014. Africa. In: Barros, V.R., Field, C.B., Dokken, D.J. (Eds.), Climate Change 2014:Impacts, Adaptation, and Vulnerability. Part B: Regional Aspects. Contribution ofWorking Group II to the Fifth Assessment Report of the Intergovernmental Panel onClimate Change. Cambridge University Press, Cambridge, United Kingdom and NewYork, NY, USA, pp. 1199–1265.
Nyboer, E.A., Liang, C., Chapmpan, L.J., 2019. Data from: Assessing the vulnerability ofAfrica's freshwater fishes to climate change: A continent-wide trait-based analysis.Dryad Digital Repositoryhttps://doi.org/10.5061/dryad.8fp561k.
Ogutu-Ohwayo, R., Natugonza, V., Musinguzi, L., Olokitum, M., Naigaga, S., 2016.Implications of climate variability and change for African lake ecosystems, fisheriesproductivity, and livelihoods. J. Great Lakes Res. 42, 498–510.
Pacifici, M., Visconti, P., Stuart, H., 2017. Species' traits influenced their response torecent climate change. Nat. Clim. Chang. 7, 205–208.
Parmesan, C., Yohe, G., 2003. A globally coherent fingerprint of climate change impactsacross natural systems. Nature 421, 37–42.
Pecl, G.T., Araujo, M.B., Bell, J., Blanchard, J., Bonebrake, T.C., Chen, I., Clark, T.D.,Colwell, R.K., Danielsen, F., Evengard, B., et al., 2017. Biodiversity redistributionunder climate change: impacts on ecosystems and human well-being. Science355, 1–9.
Penone, C., Davidson, A.D., Shoemaker, K.T., DiMarco, M., Rondinini, C., Brooks, T.M.,Young, B.E., Graham, C.H. AND Costa, G.C. (2014). Imputation of missing data inlife-history trait datasets: which approach performs the best? Methods Ecol. Evol.. 5,961–970.
Phil-Eze, P.O., Okoro, I.C., 2009. Sustainable biodiversity conservation in the Niger Delta:a practical approach to conservation site selection. Biodivers. Conserv. 18, 1247.
Pinsky, M.L., Byler, D., 2015. Fishing, fast growth and climate variability increase the riskof collapse. Proc. R. Soc. B Biol. Sci. 282, 20151053.
Platts, P.J., Omeny, P.A., Marchant, R., 2015. AFRICLIM: high-resolution climate pro-jections for ecological applications in Africa. Afr. J. Ecol. 53, 103–108.
Pletterbauer, F., Melcher, A., Graf, W., 2018. Climate change impacts in riverine eco-systems. In: Schmutz, S., Sendzimir, J. (Eds.), Riverine Ecosystem Management.Aquatic Ecology Series, vol 8 Springer, Cham, Switzerland.
Poloczanska, E.S., Brown, C.J., Sydeman, W.J., Kiessling, W., Schoeman, D.S., Moore,P.J., Brander, K., Bruno, J.F., Buckley, L.B., Burrows, M.T., et al., 2013. Global im-print of climate change on marine life. Nat. Clim. Chang. 3, 919–925.
R Core Team, 2017. R: A Language and Environment for Statistical Computing. https://www.R-project.org/.
Reece, J.S., Noss, R.F., 2014. Prioritizing species by conservation value and vulnerability:a new index applied to species threatened by sea level rise and other risks in Florida.Nat. Areas J. 34, 31–45.
Roberts, T.R., 1975. Geographical distribution of African freshwater fishes. Zool. J.Linnean Soc. 57, 249–319.
Sayer, C., Maiz-Tomé, L., Darwall, W.R.T., 2018. Freshwater Biodiversity in the LakeVictoria Basin: Guidance for Species Conservation, Site Protection, Climate Resilienceand Sustainable Livelihoods. IUCN, Gland, Cambridge.
Sgrò, C.M., Lowe, A.J., Hoffmann, A.A., 2011. Building evolutionary resilience for con-serving biodiversity under climate change. Evol. Appl. 4, 326–337.
Sorte, C.J.B., Williams, S.L., Carlton, J.T., 2010. Marine range shifts and species in-troductions: comparative spread rates and community impacts. Glob. Ecol. Biogeogr.19, 303–316.
Staudt, A., Leidner, A.K., Howard, J., Brauman, K.A., Dukes, J.S., Hansen, L.J., Paukert,C., Sabo, J., Solórzano, Luis A., 2013. The added complications of climate change:understanding and managing biodiversity and ecosystems. Front. Ecol. Environ. 11,494–501.
Stewart, B.A., Ford, B.M., Van Helden, B.E., Roberts, J.D., Close, P.G., Speldewinde, P.C.,2018. Incorporating climate change into recovery planning for threatened vertebratespecies in southwestern Australia. Biodivers. Conserv. 27, 147–165.
Stocker, T.F., Qin, D., Plattner, G.-K., Alexander, L.V., Allen, S.K., Bindoff, N.L., Bréon, F.-M., Church, J.A., Cubasch, U., Emori, S.P., et al., 2013. Technical summary. In:Stocker, T.F., Qin, D., Plattner, M., G.-K. (Eds.), Climate Change 2013: The PhysicalScience Basis. Contribution of Working Group I to the Fifth Assessment Report of theIntergovernmental Panel on Climate Change. Cambridge University Press,Cambridge, UK, and New York, NY, USA.
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520
Thieme, M.L., Abell, R., Stiassny, M.L.J., Skelton, P., Lehner, P., Teugels, C.G., Dinerstein,E., Toham, A.K., Burgess, N., Olson, D., 2005. Freshwater Ecoregions of Africa andMadagascar: A Conservation Assessment. Island Press, Washington, DC.
Thomas, C.D., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J., Collingham, Y.C.,Erasmus, B.F.N., Ferreira de Siqueira, M., Grainger, A., Hannah, L., 2004. Extinctionrisk from climate change. Nature 427, 145–148.
Walther, G.R., 2010. Community and ecosystem responses to recent climate change.Philos. Trans. R. Soc. B 365, 2019–2024.
Webster, M.S., Colton, M.A., Darling, E.S., Armstrong, J., Pinsky, M.L., Knowlton, N.,Schindler, D.E., 2017. Who should pick the winners of climate change? Trends Ecol.
Evol. 32, 167–173.Williams, S.E., Shoo, L.P., Isaac, J.L., Hoffmann, A.A., Langham, G., 2008. Towards an
integrated framework for assessing the vulnerability of species to climate change.PLoS Biol. 6, e325.
Woodward, G., Perkins, D.M., Brown, L.E., 2010. Climate change and freshwater eco-systems: impacts across multiple levels of organization. Philos. Trans. R. Soc. B 365,2093–2106.
Xenopoulos, M.A., Lodge, D.M., Alcamo, J., Märker, M., Schulze, K., Van Vuuren, D.P.,2005. Scenarios of freshwater fish extinctions from climate change and water with-drawal. Glob. Chang. Biol. 11, 1557–1564.
Yang, L.H., Rudolf, V.H.W., 2010. Phenology, ontogeny and the effects of climate changeon the timing of species interactions. Ecol. Lett. 13, 1–10.
E.A. Nyboer, et al. Biological Conservation 236 (2019) 505–520