\makeatletter
\@ifundefined{HCode}
{\documentclass[CRBIOL,Unicode,screen,biblatex]{cedram}
\addbibresource{crbiol20260226.bib}
\usepackage[T1]{fontenc}
\newenvironment{noXML}{}{}
\let\citep\parencite
\let\citet\textcite
\def\xcitealp#1#2{\citeauthor{#1}, \citelink{#1}{#2}}
\def\citealt#1#2{\citeauthor{#1}, \citelink{#1}{#2}}
\newcommand*{\citelink}[2]{\hyperlink{cite.\therefsection @#1}{#2}}
\def\defcitealias#1#2{}
\let\citepalias\parencite
\let\citetalias\textcite
\newenvironment{Table}{\begin{table}}{\end{table}}
\def\thead{\noalign{\relax}\hline}
\def\endthead{\noalign{\relax}\hline}
\def\tabnote#1{\vskip4pt\parbox{.92\linewidth}{#1}}
\def\tsup#1{\textsuperscript{#1}}
\def\tsub#1{\textsubscript{#1}}
\RequirePackage{etoolbox}
\def\jobid{crbiol20260226}
%\graphicspath{{/tmp/\jobid_figs/web/}}
\graphicspath{{./figures/}}
\newcounter{runlevel}
\let\MakeYrStrItalic\relax
\def\refinput#1{}
\def\back#1{}
\def\hyphen{\text{-}}
\def\inlinefig#1{\includegraphics{#1}}
\def\0{\phantom{0}}
\def\xsection#1{}
\def\botline{\\\hline}
\DOI{10.5802/crbiol.200}
\datereceived{2026-04-08}
\daterevised{2026-05-26}
\dateaccepted{2026-06-04}
\ItHasTeXPublished
\renewenvironment{quote}{\unskip
\topsep=0pt
\parskip=0pt
\leftskip=0pt
\rightskip=0pt
\parfillskip=0pt
\small\list{}{\unskip
\leftmargin 3.5em
\rightmargin 2.5em
}\unskip\item
\relax
}
{\endlist}
\makeatletter
\g@addto@macro{\UrlBreaks}{\UrlOrds}
\gappto{\UrlBreaks}{\UrlOrds}
\usepackage{hyperref}
\makeatother
}
{\documentclass[crbiol]{article}
\def\CDRdoi{10.5802/crbiol.200}
\let\newline\break
\def\selectlanguage#1{}
\usepackage[T1]{fontenc}
\let\citep\parencite
\let\citet\textcite
\def\citelink#1#2{\citeyear{#1}}
\def\xcitealp#1#2{\citealp{#1}}
\PassOptionsToPackage{authoryear}{natbib} 
\def\href#1#2{\url[#1]{#2}}
\let\refinput\input
\def\xsection#1{}
\def\no{n$^{\mathrm{o}}$}
\newcommand\@coi{}
\newcommand\COI[1]{\gdef\@coi{#1}}
\newcommand\printCOI{\ifx\@coi\@empty\else%
\section*{D\'eclaration d'int\'er\^ets}
\@coi\fi
}
\let\newline\break
}
\makeatother

\usepackage{upgreek}

\COI{The author does not work for, advise, own shares in, or receive
funds from any organization that could benefit from this article, and
has declared no affiliations other than their research organization.}

\begin{document}

%\dateposted{2026-02-16}

\begin{noXML}

\CDRsetmeta{articletype}{review}

\editornote{Article submitted by invitation}
\alteditornote{Article soumis sur invitation}

\title{Integrating intraspecific trait variation and spatiotemporal
variability of selection as levers of action in forest management}

\alttitle{Int\'{e}grer la variation intrasp\'{e}cifique des
caract\`{e}res et la variabilit\'{e} spatiotemporelle de la
s\'{e}lection comme leviers d'action pour la gestion foresti\`{e}re}

\author{\firstname{Fran\c{c}ois} \lastname{Lef\`{e}vre}\CDRorcid{0000-0003-2242-7251}}
\address{INRAE, UR 629 \'{E}cologie des For\^{e}ts M\'{e}diterran\'{e}ennes (URFM), Avignon, France}
\email{francois.lefevre.2@inrae.fr}

\keywords{\kwd{Adaptation}
\kwd{Adaptive strategy}
\kwd{Evolution-oriented forestry}
\kwd{Evolvability}
\kwd{Forest genetic resources}
\kwd{Resilience}}

\altkeywords{\kwd{Adaptation}
\kwd{Strat\'{e}gie adaptative}
\kwd{Sylviculture par et pour l'\'{e}volution}
\kwd{\'{E}volutivit\'{e}}
\kwd{Ressources g\'{e}n\'{e}tiques foresti\`{e}res}
\kwd{R\'{e}silience}}

\begin{abstract} 
Adaptation is a polysemous word referring both to the state of being
adapted and to the selection processes leading to this state. However,
considering population adaptation only as an ultimate achievement in a
static rather than dynamic perspective may lead to a double
misperception:\break (i) that all adapted populations have depleted variation
of adaptive traits and, therefore, low evolvability, and (ii) that
adaptation systematically requires many generations. This static vision
of adaptation as an achievement is particularly inappropriate for
long-lived organisms like trees that continuously respond to selection
and retain high levels of adaptive trait variation. This review
proposes a baseline for a process-based approach of adaptation in
trees, shedding new light on current challenges for forest management
and forest genetic resources conservation in the context of global
change.

A first general section uses the genes--traits--fitness mapping
framework to explore the effect of selection on different types of
traits, depending on their links to other traits and fitness. In a
second section, trait variation and coordination patterns observed in
forest trees are interpreted in terms of biological constraints and
response to selection. A third section investigates, for trees, how the
genes--traits--fitness map varies in space and time with environmental
conditions and developmental stages, resulting in spatiotemporal
variability of selection. A fourth section reviews the impacts of
forestry interventions on trait variation and selection. The conclusive
section illustrates the integration of selection as a dynamic and
partly manageable process in management thinking for forest genetic
resource sustainable use and conservation. 
\end{abstract}

\begin{altabstract}
L'adaptation est un terme polys\'{e}mique qui signifie l'\'{e}tat
d'\^{e}tre adapt\'{e} mais aussi tous les processus de s\'{e}lection
conduisant \`{a} cet \'{e}tat. Consid\'{e}rer l'adaptation uniquement
dans la perspective statique d'\'{e}tat abouti plut\^{o}t que dans une
perspective dynamique peut conduire \`{a} deux id\'{e}es fausses : (i)
que les populations adapt\'{e}es renferment forc\'{e}ment une faible
diversit\'{e} adaptative et sont peu \'{e}volutives, (ii) que
l'adaptation s'inscrit uniquement dans le long-terme de nombreuses
g\'{e}n\'{e}rations. Cette vision statique est particuli\`{e}rement
inappropri\'{e}e pour les esp\`{e}ces \`{a} long cycle de vie comme les
arbres dont les populations s'adaptent en continu et conservent une
grande diversit\'{e} adaptative. Cette revue pose les bases d'une
approche dynamique de l'adaptation et illustre sa pertinence pour
relever les nouveaux d\'{e}fis de gestion foresti\`{e}re dans un
contexte de changement global.

Une premi\`{e}re partie pr\'{e}sente un cadre d'analyse g\'{e}n\'{e}ral
des liens entre les g\`{e}nes, diff\'{e}rents types de caract\`{e}res,
et la valeur s\'elective des individus. Dans une deuxi\`{e}me partie,
des patrons de variations et covariations entre caract\`{e}res
observ\'{e}s chez les arbres sont interpr\'{e}t\'{e}s en termes de
contraintes biologiques et de r\'{e}ponse \`{a} la s\'{e}lection. Une
troisi\`{e}me partie \'{e}tudie, chez les arbres, comment les liens
entre caract\`{e}res et valeur s\'elective varient dans l'espace et dans
le temps. Une quatri\`{e}me partie aborde les impacts attendus des
pratiques foresti\`{e}res sur la diversit\'{e} et sur la s\'{e}lection.
La conclusion illustre l'int\'{e}r\^{e}t d'une approche de la
s\'{e}lection comme un processus dynamique, et en partie g\'{e}rable,
dans la r\'{e}flexion sur des strat\'{e}gies de gestion durable et de
conservation des ressources g\'{e}n\'{e}tiques foresti\`{e}res.
\end{altabstract}

%\input{CR-pagedemetas}

\maketitle

\twocolumngrid

\end{noXML}

\defcitealias{Maherali2004}{ibid.}
\defcitealias{Fririon2024}{ibid.}

\section{Introduction}\label{sec1}

\begin{figure*}
\includegraphics{fig01}
\caption{\label{fig1}Three consequences of intraspecific trait
variation within a population of long-lived organisms like trees. At
all times, within-population trait variation has a \textit{mixture
effect} on the population structure, the ecological functions and the
associated communities. In addition, within-population trait variation
allows temporal changes in the phenotypic composition.
\textit{Phenotypic adjustment} is a dynamic change in phenotypes before
reproduction occurs, due to developmental and environmental
(plasticity) effects as well as selection of the adapted phenotypes.
\textit{Genetic evolution} is a dynamic change in genotypes due to
selection and reproduction. Genetic evolution starts at the first
reproduction event. Here, non-overlapping generations are represented
for clarity; phenotypic adjustment and genetic evolution would be
concomitant in overlapping generations.}
\end{figure*}


Intraspecific variation represents in average 25\% of individual trait
variation found within plant communities, and 32\% of the total trait
variation among communities \citep[in a meta-analysis covering
629~plant communities and 36~functional traits]{Siefert2015}. Intraspecific
variation includes within- and between-population 
components.~Within-population trait variation (means, variances and correlations)
has three effects on population \mbox{dynamics} and functions that drive
population adaptation in the short and the long term (Figure~\ref{fig1}).
Firstly, at any time, the complementarity among different phenotypes as
well as phenotype-dependent interactions such as competition or
facilitation mechanisms induce a mixture effect on population
structure, ecological functions and biodiversity of associated
communities \citep[for review]{Hughes2008}. Secondly, within a generation,
phenotypic adjustment is a dynamic process of change in the phenotypic
composition of the population, not only due to \mbox{ontogenesis} and
plasticity, but also to selective elimination of maladapted phenotypes,
extending here the notion introduced 
by \citet{RamirezValiente2025a}. Phenotypic
adjustment continuously modifies the population structure, functions
and biodiversity. Thirdly, across successive generations observed at
the same stage, the evolutionary response to selection is a dynamic
process of change in the genetic composition of the population
resulting from the combination of three mechanisms: phenotypic sorting
of reproducing individuals, intercrossing of these individuals with
variable reproductive success, and heredity mechanisms. Note that the
time scale of genetic evolution can be similar to that of phenotypic
adjustment, when phenotypic selection and reproduction continuously
operate in overlapping generations. Thus, considering within-population
phenotypic and genetic variation in a dynamic perspective in organisms
with long life cycles, the time scales of ecological and evolutionary
mechanisms may overlap \citep{Carroll2007}. 


{Selection first contributes to phenotypic adjustment within
generations and results in \mbox{genetic} evolution across generations. It is
a central {eco-evolutionary} process, whereby changes in the phenotypic
and genetic composition of the population drive the ecological
functions, demographic dynamics and biodiversity of the ecosystem.
These changes can have feedback consequences on selection itself: for
instance when the relative fitness of each individual, which drives
the selection process, depends on the distribution of traits in the
population, or when the intensity of external selection pressures, such
as biotic or abiotic stress, is determined by and affects the
population density. These two cases can be connected to soft and hard
selection mechanisms, respectively \citep[see][for review]{Bell2021}.

Trait variation and selection are tightly linked. The evolutionary
response of a trait to selection depends on (i) its initial variation
in the population; (ii) its correlation to fitness; (iii) its
correlation to other traits also linked to fitness; (iv) its heredity;
and (v) the selection intensity \citep{Endler1986,Lande1979}. Variations and
correlations of traits should be considered in a dual perspective:
trait variation and covariation patterns constrain the potential
response to selection and are themselves impacted by the response to
selection, which \citet{Pigliucci2008} synthesized in the concept of evolvable
evolvability. \mbox{Classically}, trait variations and correlations \mbox{between}
populations are used to infer on past evolutionary response to
selection, e.g.\ local adaptation studies \citep{Kawecki2004}, while
within-population variation patterns are often analyzed in terms of
evolutionary constraints, e.g.\ G-matrix studies \citep{Arnold1992}.
However, microgeographic studies extend the exploration of past
evolutionary response to selection to the within-population scale
\citep{Richardson2014}, and metapopulation studies extend the analysis of
potential future responses to selection to the between-deme scale}
\citep{ArayaAjoy2019}.\looseness=1

There is empirical evidence that trait variation and evolvability are
maintained within populations despite selection, even for traits
tightly related to fitness \citep{Mousseau1987}. This observation might be
surprising at first glance, but several biological and evolutionary
mechanisms can explain the somehow limited impact of selection on
variance and covariance patterns \citep{Arnold2008,Kruuk2008,Pujol2018}. The
variability of selection processes at fine spatiotemporal scales is one
explanation for the persistence of variation throughout evolution.
Indeed, a single trait may be under different types of selection,
which are spatially distributed within a population (in the case of
microgeographic environmental or population structure heterogeneity at
the spatial scale of gene flow), or which operate successively at the
generational time scale. The variability of selection at a larger
scale, between connected populations, may also result in the
persistence of trait variation through balanced effects of migration
and selection \citep[for review]{Lenormand2002}. When selection processes operate at
multiple scales in a hierarchical system, e.g.\ in a metapopulation
combining local adaptation between demes with microgeographic
adaptation within demes, they can interfere in a way that maintains the
adaptive genetic diversity within each deme
\citep{Benthem2020,Cubry2022}. Thus, selection itself deserves to be
considered as a variable and dynamic driver of population dynamics.
Having this perspective in mind, the eco-evolutionary framework
accounts for feedback between the trajectory of a population or a
metapopulation and the trajectory of its selective evolutionary drivers
\citep{Griffith2016,Hanski2012,Lamarins2022}, e.g.\ in a simple case
where density-dependent selection affects population density.


Trees, as long-lived and sessile organisms, are particularly prone to
the spatiotemporal variability of selection. Exploring the mechanisms
driving the trajectory of selection, and how forest management
interventions interfere with these mechanisms, helps us to better
understand the trajectories of intraspecific trait variation and
selection. This review consists of four sections. A first general
section uses the genes--traits--fitness mapping framework to explore
the expected impacts of selection on different categories of traits and
syndromes. Using this general framework, a second section specifically
analyzes some typical patterns of trait variation and coordination
observed in trees in the dual perspective of biological constraints and
response to selection. A third section reviews the main drivers of
spatiotemporal variability of natural selection in forest trees. The
fourth section provides an overview of emerging knowledge of the
impacts of forestry interventions on natural and anthropogenic
selection processes. In conclusion, possible uses of this dynamic
vision of selection in forest management and conservation are
illustrated.

\section{Expected response of traits and genes to selection: the
genes--traits--fitness mapping framework}\label{sec2}


The genes--traits--fitness mapping is a helpful conceptual framework to
understand and analyze natural selection processes and their impacts on
traits and genes in a population. The schematic representation in
Figure~\ref{fig2}, here specifically illustrated for trees, extends an
original representation by \citet{Coulson2006} and gets inspiration
from \citet{OddouMuratorio2020}. The map represents the functional
links within and between three hierarchical spaces of variation: the
space of fitness and fitness components, the space of variable traits,
the space of variable genes. 

\begin{figure*}
\includegraphics{fig02}
\vspace*{-5pt}     
\caption{\label{fig2}The genes--traits--fitness map representing
functional links across three hierarchical spaces, for trees
\citep[inspired from][]{Coulson2006,OddouMuratorio2020}. In the space of fitness, tree size is an important
fitness component having a direct impact on survival and reproduction.
Here, the space of variable traits is illustrated with three
hierarchical sublevels where multiple basic traits (e.g.\ water
potential inducing 50\% loss of hydraulic conductivity, or minimum
conductance) determine the physiological functions (e.g.\ resistance to
cavitation, or phenology), which themselves determine the integrated
traits (e.g.\ sensitivity to drought) that are the direct drivers of
fitness components. The integrated traits can result from a single
functional trait or, more generally, from diverse combinations of
functional traits (syndromes). We expect more frequent biological
constraints on basic traits, and more impact of local selection on
integrated traits (see text). Sublevels in the space of variable genes
are not developed in this figure, for simplicity. Environmental
($E$) and ontogenetic ($O$) effects drive the
values of the traits and links. All links do not have the same strength
(line width).}
\vspace*{-5pt}     
\end{figure*}


In the fitness space, survival and reproductive success are the
ultimate fitness components represented in Figure~\ref{fig2}, which can
themselves build upon stage-specific vital rates. For trees, tree size
is a particular trait with a pivotal role on fitness that can be put in
the fitness space (the frontier between fitness components and variable
traits is thin). Indeed, in normal conditions, bigger trees have a
higher probability of surviving to competition (self-thinning) and also
tend to reproduce more, due to higher gamete production and higher
dispersal capacity. Inversely, under some stresses or disturbances,
bigger trees may be disadvantaged. In addition, tree size makes links
between many other traits and fitness, while few traits are directly
linked to survival or reproduction without affecting tree size. In a
sense, the pivotal role of tree size on fitness is analogous to the
role of body size in animals \citep{ArayaAjoy2019}. Still, some traits,
typically those related to defense and resistance mechanisms, may
directly act on survival or reproduction without any effect on tree
size, or with an antagonistic effect on growth {vs.}\ 
survival or reproduction in case of resource allocation trade-offs
\citep{Shipley2006}.


The links between the traits and the fitness components are
responsible for the selection gradients. The space of traits can be
split into several hierarchical sublevels with different links to
fitness \citep{Mousseau1987}. In Figure~\ref{fig2}, three sublevels of traits are
schematized, but the hierarchy might be more complex, or simpler,
depending on the considered set of traits. The integrated traits are
the traits most directly linked to the fitness components, they build
upon functional traits that have more indirect and complex links to
fitness and, at the bottom of the hierarchy, the links of basic traits
to fitness depend on the whole chain of links across sublevels. To
illustrate the concept in forest trees, vigor---defined as
potential growth without stress---, growth sensitivity to
drought or frost tolerance are three typical examples of integrated
traits affecting tree size and sometimes directly affecting survival.
Each of these integrated traits is potentially determined by several
functional traits, such as leaf phenology, which are themselves driven
by multiple basic variable traits, such as chilling and forcing
requirements in the case of phenology. The most basic sublevel would
\mbox{include} the \mbox{variable} parameters of physiological functions, metabolic
pathways, or some variable traits at fine biological level, e.g.\ at the
cellular or tissue level, involved in multiple functions. The
genes--traits--fitness map should be seen as a dynamic connection
network: trait values and links are dynamic, they vary with the
environment and developmental stage, some links are more variable than
others. 

The space of genes controlling the traits could also be split into
several sublevels to represent gene expression and regulation networks,
not represented in this figure for simplicity and because they are out
of the scope of this review. The links between genes and traits
illustrate cases of multiple genes linked to the same trait, which can
be purely additive or include epistatic interactions, as well as
complex pathways leading to the pleiotropic effect of a single gene on
multiple traits. Note that equivalent links also exist between
different sublevels of traits, as well as between traits and fitness:
the concepts of epistasis, i.e.\ interactions among several genes on a
single trait, and pleiotropy, i.e.\ the correlated effect of a single
gene on several traits, could be extended to the effects of traits on
more integrated traits and to the effects of traits on fitness. 

Natural selection directly acts on fitness, with indirect effects on
traits and genes that depend on the network of links connecting them to
fitness. Because the links vary in space and time, selection is a
variable and dynamic process responsible for and impacted by phenotypic
and genetic evolution in the \mbox{population}. 


The quantitative genetics theory tells us that, when a polygenic trait
is under selection, the first response in the space of genes consists
in the emergence of covariations between quantitative trait loci (QTL)
alleles before changes in individual allele frequencies
\citep{Kremer2012,Latta1998}. The contribution of covariance between QTL to the
response to selection starts with two QTL controlling the trait; it
increases with the number of QTL involved and with the intensity of
selection \citep{Kremer2012}. Change in QTL allele frequencies is a rather
long-term process requiring some stability of selection on the QTL over
the evolutionary time scale, even though rapid evolution of allelic
frequencies may occur under strong selection pressure, such as severe
epidemics \citep{Metheringham2025}. Extending these well-known results of
\mbox{quantitative} genetics, we expect a similar mechanism to apply
throughout the genes--traits--fitness map where fitness builds upon
multiple traits through a complex and dynamic connection network. Thus,
we expect that covariations of traits respond rapidly to selection and
that integrated traits are more impacted by selection than basic
traits, because integrated traits have more direct and consequently
more stable links to the fitness components regarding environmental and
ontogenetic effects.

Another expectation is that biological or development constraints are
more often driving the variation of basic traits than integrated
traits, whereas selection is a more important driver of variation for
integrated traits than basic traits (as illustrated in Figure~\ref{fig2}).
Indeed, some basic morphological traits at cellular or tissue levels or
metabolic pathways are so vital that their range of non-lethal
variation is expected to be globally constrained, within and between
populations, and this constrained variation limits their response to
selection. By contrast, integrated traits build upon multiple
alternative functional pathways; their variation is expected to be less
constrained due to functional redundancy among different combinations
of basic or functional traits producing a similar integrated syndrome.
Less constrained, these integrated traits are also tightly linked to
fitness and more directly impacted by selection, resulting in
differentiation between populations and sometimes locally, but not
globally, low levels of variation within populations under selection.
Let us examine these different expectations on trees.

\section{Trait variation and covariation patterns in trees: a matter
of biological constraint or response to selection?}\label{sec3}


The patterns of phenotypic variation observed at different levels of
organization result from the combination of multiple factors:
biological or developmental constraints, phylogenetic legacy from an
ancestor species or population, plasticity, and response to selection.
Similarly, the covariation patterns among traits result from biological
or developmental constraints, phylogenetic heritage, correlated
plasticity and correlated response to selection, which are differently
expressed at interspecific or intraspecific levels, within and between
populations. For trees, common garden experiments are classically used
to avoid confounding plasticity and response to selection in the
analysis of trait variation and covariation patterns. Some methods are
also used to disentangle these effects, or at least to achieve a
conservative test of a response to selection, using {in situ}
observations in the actual environmental conditions
\citep{AcunaMiguez2026,Halliwell2025,SanchezMartinez2024}. 


At the interspecific level, \citet{Maherali2004} reviewed the literature on
two functional traits that contribute to drought resistance, resistance
to xylem cavitation and water transport capacity, among 167~woody plant
species. They found differences in median values between vegetation
types adapted to different drought levels. However, there was still
huge variation among species within each vegetation type, with
considerable overlap across the different vegetation types. This
observation is consistent with the idea that any single functional
trait is not the unique driver of adaptation: adaptation is primarily a
matter of syndromes of multiple traits. In addition, \citetalias{Maherali2004}
{found a strong effect of the phylogenetic legacy on the within-species
variation of resistance to cavitation as well as on the within-species
correlation between the two studied traits, rather than evidence of
convergent adaptive responses to selection. This is consistent with the
idea that biological or developmental constraints on the variation and
covariation of basic or functional traits somehow prevent them from big
changes along species phylogenies. Similarly, in a phylogenetic study
of life history and adaptive traits among Mediterranean pine species,
\citet{Grivet2013} detected almost no adaptive convergence of single traits
across taxa. However, they found three trait correlations that they
interpreted as alternative genetic adaptation strategies to fire: two
negative correlations, between seed dispersal and bark thickness, as
well as between serotiny and maximum life span, illustrating a
classical contrast between r and K-strategies; and a positive
correlation between serotiny and seed dispersal, illustrating a
variation of efficiency in the r-strategy. 


At intraspecific level, provenance trial experiments in trees usually
reveal clinal genetic differentiation patterns for growth-related
traits among populations sampled along environmental gradients, thus
indicating the high evolvability and effective response to selection of
such integrated traits \citep[for review]{Savolainen2007}. We also have
evidence of the response of growth-related integrated traits and
phenology to selection occurring at microgeographic scale within gene
dispersal distance \citep[for review]{Scotti2016}. Evidence of temporal
genetic changes of integrated traits in response to selection across
successive generations of trees is still scarce, but cases of rapid
genetic adaptation have been observed in populations introduced in a
new environment \citep{FallourRubio2009,Peterken2001,Skroppa2010}. At
the bottom of the hierarchical space of traits, globally limited
variation of resistance to cavitation was shown across maritime pine
populations over the species' range, suggesting biological or
developmental constraints \citep{Lamy2014}. 

Generally, the highest values of genetic differentiation among tree
populations are observed for growth, phenology and frost hardiness,
whereas more basic traits linked to leaf morphology and chemistry, wood
anatomy, hydraulic properties or photosynthetic capacity show much
lower differentiation \citep{Alberto2013,Hajek2016}. A similar pattern was also
found in the case of selection at microgeographic scale: that is, a
higher value of differentiation for growth-related traits and phenology
than for leaf morphology and chemistry or water use efficiency}
\citep{Gauzere2020}. These results are consistent with the expectation that
basic traits should be less impacted by selection than integrated
traits due to more distant and flexible links to fitness for basic
traits. Interestingly, in these studies of local adaptation in
different environments, the variation patterns of phenological traits
are highly impacted by recent selection (in the time scale of local
adaptation), as much as or even more than growth-related traits: this
is due to the potentially strong impact of phenology on all fitness
components across different environments. In a rare study assessing
selection gradients within a single population for different types of
traits in two sympatric oak species, \citet{Alexandre2020} found significant
selection gradients for growth traits and defense traits in both
species, for leaf morphology only in one of the species, but not for
phenology in either species: phenology was not tightly linked to
fitness in this particular \mbox{environment}. By contrast, evidence of
spatial variation in selection gradients on phenology was found along
an environmental gradient in beech, showing that phenology was tightly
linked to fitness along this gradient \citep{OddouMuratorio2024}.

Cases of GxE interaction are frequently observed in forest trees,
making plasticity itself a variable trait. \citet{Alia2024} used a
provenance/progeny test in two contrasted environments to assess the
evolvability of different traits in 11~maritime pine populations: the
plasticity of water-use efficiency (WUE) showed the highest
evolvability, even higher than growth, while mean WUE showed the lowest
evolvability. Using a provenance/clone experiment in two contrasted
environments with the same species,
\citet{RamirezValiente2025b} showed that
populations adapted to a mild climate had higher growth plasticity than
populations adapted to arid conditions. Analyzing dendrochronological
data between and within successive generations within an introduced
cedar population initially planted in a dry eroded land,
\citet{FallourRubio2009} detected a significant change in the growth
plastic response to summer drought across generations, but no
generational change in mean growth. In this case, growth plasticity
increased from the first to the third generation, while drought stress
progressively decreased due to soil restoration under the development
of the new forest ecosystem. All these results show that plasticity
traits can be highly responsive to selection, at least as much as
integrated traits and more than basic traits. However, this is not
always the case: in a common garden experiment of oak populations
sampled along an altitudinal gradient, \citet{Soularue2023} observed a
genetic cline on budburst date but not on its plastic response to
temperature.


Trait correlations also vary at intraspecific levels and, therefore,
potentially respond to selection \citep[for review]{Climent2024}.
Intraspecific trait correlations between and within populations are
generally different \citep{AcunaMiguez2026,
Alia2024,Prada2016}. Between-population
trait coordination potentially indicates a correlated response to
selection, i.e.\ the emergence of adaptive trait syndromes as bases of
intraspecific ecotypes \citep{Blasini2021}. Within-population trait
coordination can also result from a correlated response to selection
within populations, when several trait syndromes increase fitness as
observed in a beech population at the southern \mbox{margin} of the species
distribution range \citep{Bontemps2017}. Trade-offs between growth vigor in
favorable conditions and stress resistance traits are commonly observed
at both scales, between populations \citep{Rehfeldt2001} and within
populations \citep{Fririon2023}. Note that having a similar trade-off within
and between populations is necessary but not sufficient to provide
evidence of a biological ``cost'' of resistance such as a resource
allocation balance. For instance, at the between-population level,
\citet{Modrzynski2002} observed in a phytotron experiment that Norway spruce
provenances from high elevations were more tolerant to drought, but
grew less in good conditions, than provenances from low elevation.
Rather than representing a ``cost of drought tolerance'', this
difference could be explained by a difference in phenology, because in
this species high elevation provenances start growing earlier in the
common environment and develop deeper root systems before the drought
occurs. At within-population level, adaptive trade-offs also simply
result mathematically from some types of correlated selection, such as
threshold selection on a linear combination of multiple traits, i.e.\ 
index selection (Figure~\ref{fig3}). For trees in particular, trade-offs
emerging from selection are expected between traits that jointly
contribute to tree size. Indeed, tree size can be seen as a selection
index being the sum of annual growths in variable conditions, which
mathematically results in a trade-off after selection between the
different traits that contribute to increase tree size depending on
different annual conditions, as classically observed between vigor and
stress tolerance traits (Figure~\ref{fig3}). More generally, when fitness is
expressed as a weighted sum of multiple traits, selection is expected
to generate trade-offs among these traits, depending on their relative
contributions to fitness. Thus, considering that trait contributions to
fitness are represented through the functional links on the
genes--traits--fitness map, this representation could also serve to
predict trade-offs among traits that could emerge from selection.


\begin{figure*}
\vspace*{-4pt}     
\includegraphics{fig03}
\vspace*{-4pt}     
\caption{\label{fig3}Phenotypic trade-offs emerging from index selection.
(a) X\tsub{1} and X\tsub{2} are two independent traits, in a standardized
scale, contributing positively to a weighted selection index with
relative contributions 0.6 and 0.4, respectively: 
$\mathrm{index}=0.6\mathrm{X}_1+0.4\mathrm{X}_2$. The rate of selection on the
index is 0.6 (the 60\% lowest values are eliminated). In grey are the
values before selection, the white ``${+}$'' is the initial population mean
before selection. In red are the values of selected individuals, the
green ``${+}$'' is the mean after selection. With these values, selection
generates a negative phenotypic correlation (no reproduction considered
here) ${-}$0.49 within the selected population. Considering the situations
before and after selection as two populations, note the opposite
(positive) sign of the between-population correlation caused by the
increase of both traits means in the selected population.
(b)~Values of the correlation (trade-off) within the selected
population as a function of (i) the weighting of the traits in the
index and (ii) the rate of selection. Note the important effect of the
weighting. For instance, tree size is the sum of annual growths in good
years (related to vigor) and annual growths in stressful years (related
to stress tolerance): selection on tree size creates a trade-off
between vigor and stress tolerance, and the value of the negative
correlation depends on the frequency and intensity of stress events
that determine the relative contributions of both traits to the index.}
\vspace*{-4pt}     
\end{figure*}


\section{Spatiotemporal variability of selection in trees}\label{sec4}


Selection pressures, and therefore the traits under selection, vary in
time and space within tree \mbox{populations}. Within a forest, the external
causes of tree mortality change from year to year driven by the annual
climatic fluctuations and other disturbances \citep{PetitCailleux2021}. Thus,
living trees in the forest can be seen as survivors to multifactorial
and stepwise selection: e.g., a rare severe frost event occurring in an
arid site will select the frost tolerant among drought tolerant trees,
and the genetic impact of the selection event on frost tolerance may
last over several generations. Selection pressures on trees also depend
on the environment \citep{Alia2014}, and high within-population
micro-environmental heterogeneity results in the interference between
multiple selection processes spatially distributed within gene
dispersal distance \citep{Brousseau2015,Gauzere2020}.


In a stable environment, the forest stand dynamics itself act as an
internal driver of spatial and \mbox{temporal} variability of selection, for
two reasons. Firstly, trees' adaptive capacities and their relations to
fitness vary across developmental stages \citep{Alia2014,Pardos2014,Peltier2020},
making selection gradients variable throughout individual lifespan. The
consequences of such developmental effects on the spatiotemporal
variability of selection might differ between even-aged forests, with
synchronous developmental stages, and uneven-aged forests, where
various developmental stages coexist, forming an irregular stand
structure. Secondly, stand growth and demography dynamics generate
temporal changes in the intensity of density-dependent selection and
other soft selection mechanisms \citep{Godineau2023}. They also induce
temporal changes in the level of stresses like drought that increases
with stand leaf area \citep{Breda1995} or in the intensity of pests and
parasites pressures that vary positively or negatively with host
density \citep{Knight2013,Sholes2008}. 


{In the context of climate change, the temporal horizon of climatic
scenarios matches the lifespan of trees. Disturbances and uncertainties
are key drivers of selection on this time scale. Climate change
projections consist in two phases: a first phase of predicted climatic
transition with annual fluctuations and disturbance regime shifts in
the next 30 to 50~years, followed by a second phase with increased
uncertainties due to divergent model predictions \citep{IPCC2021}. Most
of the current forest trees will experience the first phase of change
in selection pressures, possibly survive until the more uncertain
second phase, and some of them might even survive beyond the longest
climate projections. In other words, shifts in selection processes are
expected at the same velocity or even faster than the response of tree
populations to selection, and the spatiotemporal variation of selection
processes is highly uncertain. 


Beyond the effects of the mean climatic trend on the dominant
selection pressures, such as the global increase of drought and heat
stress impacts on forests \citep{Allen2010}, disturbance regimes on
European forests have already shifted towards higher frequency and
intensity since 1950 \citep{Patacca2023}. \citet{Thom2017} showed that small
scale hazards, when surviving trees remain at a distance reachable by
gene flow, are expected to exert a selection at population or
metapopulation levels and simultaneously offer opportunities for the
establishment of selected progenies, which can ultimately foster
adaptation, while large-scale disturbances are expected to impede
forest adaptation. \citet{Peltier2019} detected a detrimental impact of
repeated droughts on growth decline in ponderosa pine, with regional
differences of sensitivity among populations. \citet{Seidl2022} used the
adaptive cycle framework to explore post-disturbance reorganization of
forest ecosystems, depending on the changes in the structure and
composition of the ecosystem: this approach could probably be extended
to the reorganization of phenotypic trait variation and genetic
diversity.}\looseness=-1

\section{Impacts of forestry interventions on trait variation and
selection}\label{sec5}


Intentionally or not, silviculture and forest planning interventions
are expected to drive trait variation, not only through the control of
the genetic composition of plantation forests but also through multiple
impacts on selection in naturally regenerated forests: (i)~by inducing
or modifying genetic drift and gene flow that interfere with selection;
(ii) by changing the environmental conditions that determine the type
and intensity of natural selection; (iii) by generating additional
anthropogenic selection 
\citep{Aitken2013,Lefevre2004,Lefevre2014,Savolainen1992}. Currently,
the quantitative assessment of these potential impacts mostly relies on
modeling work, as empirical assessment is just emerging based on
ex-post reanalysis of long-term monitoring experiments and new
experimental designs. 


Simulation studies on different types of forests show that the large
population size and important gene flow capacities of trees generally
prevent significant genetic drift due to management. In the case of
large populations, as commonly observed for temperate tree species,
classical silvicultural strategies are not expected to reduce the
number of reproducing trees as much as required to induce genetic
drift \citep{Godineau2023}. Even in the case of highly diverse tropical
forests with low population density for each species (a few adult trees
per hectare), \citet{Degen2006} showed that the frequency and intensity of
selective logging have an impact on tree growth and population
demography, without a significant reduction of gene diversity or an
increase in inbreeding.

Intentional assisted gene flow is an option to rescue populations that
suffer demographic and genetic decline under strong selection. The
expected benefits include demographic support with pre-adapted
genotypes, genetic enrichment with new alleles of interest and
reduction of inbreeding depression, while the main associated risks are
outbreeding depression and introduction of undesired alleles
\citep[for review]{Aitken2013}. In particular, this option is envisaged to foster forest
adaptation to climate change \citep{Browne2019,Devresse2025}. Practical
experiments are still rare and recent, lacking long term monitoring of
potential risks associated to the introduction of long distance
material that could bring in undesired heritable characteristics linked
to the target traits of interest \citep{Young2020}. When assisted gene
flow would consist in introducing distant genetic material into a small
local population, a strategy to reduce the risks of maladaptive genetic
swamping and of further drop in effective population size is to use a
highly genetically diverse donor gene pool \citep{Lefevre2004}. Assisted
gene flow is also \mbox{envisaged} in the case of emerging pests and diseases
like the ash dieback in Europe \citep{SemizerCuming2021}. In this case,
we expect lower risk of maladaptation or outbreeding depression when
the introduced resistant genotypes come from the surroundings of the
declining stand. 


\citet{Kramer2008} made a distinction between three categories of traits
with regard to natural and anthropogenic selection: (i) traits under
natural selection with little consideration in management, such as
flowering phenology when it does not induce major defaults; (ii) traits
selected in forestry with little adaptive consequences, such as spiral
grain; and (iii) traits under both natural and anthropogenic selection,
such as growth-related traits and resistance traits. The
genes--traits--fitness map could help identify complex links between
natural and anthropogenic selection through their respective target
traits. All traits that are somehow linked to tree size fall in the
category of traits submitted both to natural and anthropogenic
selection. For these traits, management interventions interfere with
natural selection through different mechanisms. Understanding these
mechanisms can help to solve potential trade-offs between short-term
objectives, such as reducing competition or stress levels, and
long-term objectives, such as fostering genetic adaptation. 


In the short term, increasing morphological or functional
intraspecific trait variation by planting genetic mixtures can increase
forest productivity and resilience due to positive mixture effects,
e.g.\ reduced competition for resources, and phenotypic adjustment. In a
40-year provenance mixture experiment of Norway spruce, \citet{Pretzsch2021}
showed that increased provenance diversity around trees reduces
competition and increases tree growth and stand growth in the same
magnitude as in species mixtures (${+}$28\%). Similarly, at more juvenile
stages, the reduction of competition and increased tree growth in
clonal mixtures compared to monoclonal stands in 9-year-old loblolly
pine experiments (${+}$4--5\%) was attributed to morphological trait
variation \citep{Carter2025}. Thus, genetic mixing induces positive mixture
effects and phenotypic adjustment. However, the ultimate consequences
on genetic evolution are context dependent, because of antagonistic
impacts on the response to selection. Indeed, genetic mixing increases
the variance of the trait under selection on the one hand, but reduces
the intensity of one potential mechanism of selection, competition, on
the other hand, and the balance between these antagonistic effects is
contingent to the local conditions. 


Favoring tree species mixtures is an option to increase forest
resilience. In such a situation, the interference of species diversity
with the potential evolution of each species is questionable. Using a
demo-genetic modeling approach in a climate change scenario,
\citet{Devresse2025} showed that beech and fir are not expected to have
slower genetic evolution in mixture than in monospecific stands,
because both species evolve jointly and the interspecific mixture does
not induce sufficient genetic drift to hamper the response to
selection. However, the introduction of other pre-adapted species may
over-compete the local species and preclude their evolution.


Thinning regimes, i.e.\ frequency, intensity and type of thinning
(selective from below, from above or not selective), are generally
aimed at reducing competition. They have multiple impacts on selection.
Demo-genetic models coupling forest dynamics and genetics are useful
for disentangling these impacts. If they were strictly non-selective,
some intensive thinning regimes currently used in forestry could
drastically reduce natural density-dependent selection mechanisms, i.e.\ 
substitute selective self-thinning with non-selective thinning,
resulting in a drastic reduction of the evolutionary rate of
growth-related traits across generations (${-}$84\%) \citep{Godineau2023}.
However, in practice, thinning is often selective from below, in which
case the forester selects in the same direction as natural selection
does, in favor of trait values that promote tree size, and the
substitution of natural selection with anthropogenic selection could
even result in increased adaptive evolutionary rate (${+}$30\%) compared to
no management \citep{Fririon2024}. Note that selective thinning partly from
above, sometimes used to obtain an economic revenue to pay for the
intervention, always opposes natural selection with possible
long-lasting consequences on the maintenance of maladapted genotypes in
the forest. 


Under drought stress conditions, density reduction is also used to
alleviate stress, providing a short-term functional benefit for the
current tree generation. In this case, the short-term benefit is
associated with reduced selection intensity for drought tolerance, due
to reduced drought mortality (hard selection mechanism) and reduced
contribution of drought tolerance to tree size (soft selection
mechanism through competition and self-thinning), slowing down genetic
adaptation. A strategy to avoid this short vs.\ long-term
trade-off{---}reducing stress in the short term
hampers genetic adaptation in the 
long-term{---}consists in leaving more room for natural selection in the juvenile
stages and reducing stress in later stages \citepalias{Fririon2024}. Thinning
regimes in harvested stands can also result in enhanced growth of the
remaining trees. This leads to shorter rotation between successive
harvests and, therefore, shorter generation time \citep{Kramer2008},
which interfere with the velocity of genetic evolution. 

\section{Conclusion for research and forest management}\label{sec6}

\subsection{Extended views on the genes--traits--fitness mapping
conceptual framework}


Here, I used the genes--traits--fitness mapping conceptual framework,
focusing on the spaces of variable traits and fitness, to explore the
joint dynamics of trait variation and selection in trees. The schematic
representation on Figure~\ref{fig2} only shows part of the picture;
other aspects that are not represented in this figure need to be
considered. Regarding the space of traits, the multiple effects of the
environment appear independently on this representation, which does not
account for coordinated environmental effects on multiple traits and
links. Missing coordinated environmental effects does not facilitate
the exploration of the role of plasticity in the joint dynamics of
trait variation and selection  \citep{Chevin2013b,Chevin2025}, even
though we have empirical evidence that plasticity traits are tightly
linked to fitness. Obviously, the extension of the representation of
the space of variable genes, through an analog dynamic connection
network with links between environmental and ontogenetic effects on
gene expressions and pathways, would deserve exploration. 

Another aspect hardly represented on a diagram is the spatiotemporal
variability of selection mechanisms: their integration through time and
space still remains to be analyzed. For instance, in even-aged forests,
multiple stepwise selection \mbox{mechanisms} globally result in a composite
selection process from one generation to the next. The integrated
\mbox{composite} selection relates to the genes and traits spaces through an
integrated genes--traits--fitness map. Obtaining this integrated map,
some genes and traits may appear to be tightly linked to the composite
fitness for two reasons: either they are tightly linked to one
important selection mechanism in particular, or they have consistent
and robust links with multiple selection mechanisms. These two
situations are not mutually exclusive. In the second case, we might
even imagine some genes or traits poorly linked to each selection
mechanism, but consistently linked to all: these genes or traits would
be strongly involved in and impacted by the response to composite
selection. Looking at between-population adaptation, \citet{Yeaman2016}
detected cases of convergent selection on specific gene alleles
between two distantly related species, lodgepole pine and interior
spruce. At within-population scale, \citet{Scotti2023} found genetic
signature of microgeographic selection in different tree species,
without evidence of convergent selection across species so far. These
results suggest the possibility of strong links between genes and
fitness, and particularly robust links in case of convergent selection
in different species. In plants, some types of genomic elements playing
a key role in phenotypic variation, notably some transcription factors
\citep{Engelhorn2025}, could be good candidates for genes having an important
role in composite selection processes.

Nevertheless, although incomplete, the present review already provides
indications for applied management.

\subsection{A dynamic vision of adaptation in trees}

Trees are submitted to temporally variable, successive, selection
pressures during their lifespan. Tree populations cover large areas due
to high dispersal capacities, and microgeographic environmental
heterogeneities make selection spatially variable within populations.
Selection pressures, selection intensities and selection gradients (the
links of traits to fitness) vary in space and time, even at fine
scales, driven by the local environmental conditions, which consist of
external biotic and abiotic factors, combined with internal factors
such as population structure or developmental stages. Therefore, local
adaptation in trees should be seen as a dynamic process of response to
multiple selection pressures with spatiotemporal dependencies, rather
than an achieved long-term response to a single selection driven by
mean local conditions. Within-population trait (co)variation is
simultaneously a constraint for and the result of selection; it
continuously varies within and across generations of trees. Considering
the spatiotemporal variability of selection helps us to better
understand the causes and consequences of trait variation patterns. The
interdependent dynamics of trait (co)variation and selection are
impacted by developmental factors, population dynamics, external
environmental factors, and forest management interventions. 

Selection is expected to have more impact on integrated traits and
trait syndromes than on the variation of functional and basic traits.
This flexibility across the traits--fitness map allows for rapid
response to selection without severe erosion of the variation of
functional and basic traits, which are the basic bricks of the capacity
of response to variable selection pressures. An illustration of this
flexibility was given above with the southern beech population adapted
to drought through multiple trait syndromes conferring drought
resistance while maintaining the variation of each trait involved
\citep{Bontemps2017}. 


Forest management interventions have multiple impacts on the
spatiotemporal variability of selection, which can ultimately result in
reinforcement or impairment of natural selection, and eventually create
additional anthropogenic selection. Understanding these impacts is
strategic for the design of management options, whether it be for
forestry or conservation purposes, particularly in the context of
environmental changes and uncertainties. In this context, natural
selection should be considered as a lever of action in forest
management and forest genetic resource conservation, as illustrated
below in two situations. This is not an exhaustive list of situations
where evolution-oriented forestry would be beneficial \citep{Lefevre2014}.

\subsection{Forest management with natural regeneration under
climate change}

In naturally regenerated production forests facing new stresses, e.g.\ 
climate change, there is a need to find a compromise between a
short-term objective to enhance the production of the current stand
under pressure and a longer-term objective to foster genetic
improvement for stress tolerance in the next generation. Natural
selection can definitely help to find such a compromise. Simulations of
a regular Douglas fir stand under drought stress provided a first
estimate of the expected genetic gain in productivity that can be
obtained through natural selection on vigor and drought tolerance:
${+}$13\% gain per generation on mean stem diameter at the final cut in the
simulated context \citep{Fririon2024}, which is in the same order of
magnitude as the average gain expected through breeding
\citep{Jansson2017}. A multi-objective strategy that enhances short-term
productivity and fosters long-term adaptation can be designed through a
thinning regime that allows for natural selection to operate in the
juvenile stage before reducing stress, and thus reducing natural
selection, at a later stage \citep{Fririon2024}. To speed up adaptation, it
is also proposed to use more dynamic silviculture, i.e.\ to reduce the
rotation time and, therefore, accelerate generation turnover. This
action on the denominator of the ``gain per generation'' assumes that
most of the natural selection operates at the juvenile stage, so that
reducing the generation time does not significantly decrease the total
gain (the numerator). However, this assumption may be questionable.
Firstly, reducing generation time may result in enhanced selection for
vigor and reduced selection for stress resistance, due to fewer or less
intensive stress events and to the trade-off between vigor and
resistance traits, which might, at the end result, in faster but lower
adaptation. This is expected to happen in forests currently under
moderate and rare stress, whereas the initial assumption that
sufficient selection for stress tolerance is achieved in shorter time
remains valid in forests under heavy stress. Secondly, considering the
huge uncertainty about future selection pressures and adaptation needs,
reducing the duration of exposure to natural selection also reduces the
chance of multiple selection criteria to combine within each
generation. This results in sequential multi-trait selection, i.e.\ 
sparing different selections across different generations, rather than
multi-trait index selection within each generation. Will one selection
process be more efficient than the other at the end? In any case,
multi-trait index selection offers more \mbox{insurance} through time.
Considering these different aspects together, reducing generation time
may not be much effective to accelerate the adaptation of forests that
are currently under moderate stress and will suffer a heavy shift in
the stress regime; it will be more effective and efficient in forests
already under strong selection. The effects of thinning on the
selection intensity for traits of interest, on the one hand, and on the
generation time, on the other hand, ultimately combine in higher or
lower evolutionary rates per generation. A compromise to combine
short-term stress reduction and faster evolution towards stress
tolerance can be reached by adjusting the dates, intensities and types
of successive thinning, taking into account the local context of
natural selection.

\subsection{Dynamic conservation of forest genetic resources}

Regarding the conservation of forest genetic resources, priority is
given to dynamic {in situ} strategies whenever possible,
allowing the conserved populations to evolve in their original
environment \citep{EUFORGEN2021}. Relocation is sometimes envisaged for some
marginal populations that have developed specific adaptive
characteristics of interest but considered locally at risk due to
severe selection pressure. The so-called dynamic {ex situ}
approach consists in moving these marginal populations of interest to
a safer place and letting them dynamically evolve there. However, it
should be clear from this review that conserving a specific adaptation
in an evolving population cannot be achieved without conserving the
selection mechanisms that create and maintain this adaptation. If the
new site where the population is transferred is less selective, the
next generation of trees will lose part (possibly a large part) of
their ancestral adaptive specificities and rapidly evolve towards new
characteristics fitting the new environment, as previously illustrated
with the cases of rapid adaptation of introduced populations 
(Section~\ref{sec2}). Therefore, if the conservation objective 
is to preserve the
original adaptive specificities, dynamic {ex situ}
conservation is not the most effective strategy. In a dynamic
approach, not only genetic diversity but also selection itself should
become a target of conservation. Before taking a decision of
relocation, the risk of local extinction should not be overestimated.
Having a reasonable amount of dieback is a rather positive sign of
ongoing adaptation, given that the threshold of population viability is
not passed, of course. An alternative option overlooked for these
extreme populations developing specific adaptive characteristics under
severe stress could be to ensure sufficient variation to feed selection
and rescue the population at risk where it is, i.e.\ using genetic
mixing strategies like provenancing or evolutionary rescue
\citep{Hoffmann2021}. This would be a type of assisted gene flow strategy,
not introducing pre-adapted material (that may not be available for
such marginal situations) but introducing the least-maladapted material
with the highest genetic diversity possible, as a demographic supply,
with the hope that enough offsprings will reach the level of adaptation
required. For species lacking such marginal populations of interest in
conservation programs, installing highly diverse genetic mixtures in
stressful marginal areas surrounding the species range would not
necessarily be an absurd option to consider and assess, aiming to
achieve conservation with genetic improvement by natural selection.
This would be a dynamic {ex situ} approach in which genetic
mixtures would be installed not in sites with less selection but in
sites with specific selection of interest, despite the evident
associated risk for management and probable additional costs. This
would mean looking for the effectiveness of the conservation rather
than just its efficiency, i.e.\ emphasizing the level of adaptation
potentially conserved despite the cost rather than emphasizing the
chance of conservation success with a reduction of adaptive capacity
(which is another type of cost).

\subsection{Back to the original definition of a resilient approach
in management of natural resources}

In the context of climate change, forest managers classically identify
some specific predictable adaptation goals, such as climate adaptation.
A further goal is to integrate the evolvability, or adaptive potential,
as an additional objective to cope with unpredictable future needs. In
addition to these goals, integrating the variability of selection into
forest management thinking, and considering the selection process as an
objective per se, would further help to develop a flexible adaptive
strategy, taking the benefit of natural selection processes rather than
\mbox{counteracting} them, to drive the forests towards multiple \mbox{adaptation}
\mbox{requirements.} This environmental stewardship approach in management
\citep{Mathevet2018} would also be a resilient approach, as defined by
\citep{Holling1973}: 

\begin{quote}
The stability view emphasizes the equilibrium, the
maintenance of a predictable world, and the harvesting of nature's
excess production with as little fluctuation as possible. ({\ldots}) A
management approach based on resilience, on the other hand, would
emphasize the need to keep options open, the need to view events in a
regional rather than a local context, and the need to emphasize
heterogeneity. Flowing from this would be not the presumption of
sufficient knowledge, but the recognition of our ignorance; not the
assumption that future events are expected, but that they will be
unexpected. The resilience framework can accommodate this shift of
perspective, for it does not require a precise capacity to predict the
future, but only a qualitative capacity to devise systems that can
absorb and accommodate future events in whatever unexpected form they
may take.
\end{quote}

\section*{Disclaimer}

Intentionally, the author did not use AI for this review, neither for
the literature search and analysis nor for the writing and
illustrations, leaving perfectibility as an opportunity for the reader
to develop her/his own improved personal thoughts on the topic. 

\printCOI 

\back{}

\printbibliography
\refinput{crbiol20260226-reference.tex}

\end{document}