WP  2  -  Modeling  non-equilibrium  population  demographic  and  evolutionary  dynamics  and predicting the impact of GC, from local to regional scales.

CC and habitat fragmentation are imposing profound pressures on forest ecosystems to adapt at rates that exceed those associated to past natural climate oscillations. The goal of this WP is to develop scenarios for the adaptive response of tree populations to GC and to identify environmental and  micro-evolutionary  tipping  points.  We  will  rely  on  a  new  generation  of  models  coupling geophysics,  biophysics,  population  dynamics  and  quantitative  genetics. The first  exploratory  step aims to characterize environmental stresses and major tree traits driving population demographic response  to  these  stresses  without  genetic  adaptation.  In  a  second  step  we  will  account  for  the levels of genetic diversity of these traits and for their response to selection.


Objective 1: Environmental tipping points and key traits involved in demographic collapse
The first objective of WP2 is to identify the environmental tipping points beyond which population decline  is  predicted  and  to  identify  the  tree  traits  most  closely  associated  to  such  population collapse. To that aim we will use existing and complementary process-based models (CASTANEA (Dufrene  et  al.  2005);  PHENOFIT  (Chuine  and  Beaubien  2001))  to  investigate  the  interaction between climate, tree physiology and population dynamics from stand to range scale. CASTANEA takes as input fine-scale climatic and soil property data and model ecophysiological processes to simulate  water  and  carbon  balance  and  their  effect  on  population  tree  growth,  reproduction  and mortality  at  stand  level.  CASTANEA  was  parameterized  for  most  of  species  included  in  TipTree (Fagus sylvatica, Quercus robur, Abies alba). PHENOFIT takes as input climatic data and models trees’ phenology, frost and drought damage in order to determine the probability that an average tree survives and produces viable seeds. Using PHENOFIT, (Morin et al. 2007) could identify the biological  processes  (phenology,  drought  or  frost  response)  that  limit  the  potential  distribution  of several  tree  species  at  large  spatial  scales. We  will  use  these  two  kinds  of models  to  perform a sensitivity  analysis  of  tree  population  vital  rates  (growth,  mortality,  reproduction;  CASTANEA)  or species presence (PHENOFIT) to species life history traits. This will be done for different locations across  species  ranges  under  different  climate  change  scenarios  (provided  by  IPCC),  including single or interactive drivers (e.g. a drought following a frost). It will allow characterizing the critical values  of  life  history  traits  leading  to  population  collapse  and  also  the  associated  optimum  trait values. This task will inform WP1 and orientate the choice of target adaptive traits for which levels of genetic diversity and intensity of selection will be assessed. 

Objective 2: Eco-evolutionary dynamics in response to GC and evolutionary tipping points  
Our hypothesis is that the intensity of selection exerted by climate and other aspects of GC on a suite  of  adaptive  traits  can  reduce  populations  to  a  size  where  new  evolutionary  constraints  are generated.  Such  tipping  points  may  either  lead  to  population  extinction  or  to  evolutionary  rescue through rapid adaptation to new environmental conditions. Here, we will develop or adapt integrative models of genetic adaptation to simulate the evolutionary dynamics of tree populations across a few to many (5-100) generations and small to large spatial scales. Our aim is to test the existence and evaluate the magnitude of tipping points related to:

•  Different scenarios of response to climatic-driven selection provided by WP 1 (levels of adaptive diversity and genetic architecture of traits ; intensity of selection for (suites of) traits) •  Different climatic scenarios (as in Objective 1)

•  Different management scenarios (provided by our stakeholders in WP3)

•  Different spatial scales (from stand to landscape and range scales)

The two  process-based models  CAPSIS-PDG  and  AMELIE  couple  demographic  and  quantitative genetic submodels, and allow studying the rate of multi-trait evolution by including genetic details and  eco-evolutionary  feedbacks.  In  particular,  both  models  can  easily  incorporate  detailed information  about  the  genomic  architecture  of  traits  (provided  by  WP1  and  previous  research projects such as LinkTree). For instance they can represent the number, effect size and linkage of QTLs as well as information on pleiotropy and epistasis. Their flexible nature also enables them to incorporate detailed management options; management scenarios that were already considered in AMELIE include the targeted planting of non-native provenances (assisted migration, see Kuparinen and  Schurr  2007)  and  the  increased  cutting  of  established,  maladapted  trees  (Kuparinen  et  al. 2010).  Apart  from  these  commonalities,  the  two  models  differ  in  key  features:  CAPSIS-PDG includes a detailed ecophysiological submodel (CASTANEA) but is very computer-intensive so that simulations  are  limited  to  small  spatial  and  temporal  scales.  In  contrast,  AMELIE  provides  an aggregated description of the processes underlying demographic variation but can be run at large spatial  and  temporal  scales.  The  joint  use  of  both  models  will  enable  us  to  exploit  their complementary strengths. In particular, physiologically-explicit simulations with CAPSIS-PDG will be used  to  parameterize  the  demographic  component  of  AMELIE  which  will  then  be  used  for  large-scale simulations and management assessments. To demonstrate these complementary strengths, we briefly describe both models below. The  PDG  model  under  the  CAPSIS  platform  (Dreyfus  et  al.  2005)  couples  (1)  a  physiologically process-based  module  (CASTANEA);  (2)  a  quantitative  genetics  module  relating  genotype  to phenotype at some adaptive traits; (3) a demographic module converting tree reserves into seed production  and  tree  mortality  and  modelling  fecundation,  seed  dispersal  and  establishment.  The different modules will be parameterized using published data and results of WP1 (seed and pollen dispersal,  trait  genetic  architecture).  A  first  interesting  property  of  PDG  is  that  individual  fitness dynamically results from the physiological and demographic processes and from the spatio-temporal environmental  variations.  Using  PDG,  we  can  thus  estimate  the  intensity  of  selection  exerted  by different  environment  pressures,  and  thereby  cross  validate  the  experimental  results  obtained  in WP1 (and calibrate AMELIE model). A second interesting property of CAPSIS-PDG is to be able to predict ecosystem services related to carbon and water balances such as carbon sequestration or wood production.  The  individual-based  modelling  framework  AMELIE  (Kuparinen  and  Schurr  2007)  describes  the spatiotemporal  dynamics  of  tree  species  and  their  component  genotypes.  AMELIE  can  mode different  intensities  of  within  population  stabilizing  selection  and  among  population  divergent selection.  AMELIE  is  also  very  flexible  in  describing  demographic  processes  (size-  and  density-dependence  of  recruitment,  growth,  mortality,  seed  and  pollen  dispersal).  The  demographic submodel, will be parameterized using output of CAPSIS-PDG, published data on the study species as well as results of own previous work (e.g. estimates of size- and density dependence of dispersal in  Pinus  halepensis  from  (Schurr  et  al.  2008)).  To quantify  gene flow  by  long-distance  seed  and pollen dispersal, we will use a vector-based approach (Kuparinen et al. 2009; Kuparinen and Schurr 2007;  Nathan  et  al.  2011).  This  vector-based  approach  also  enables  one  to  assess  how  climate change  effects  on  wind-driven  seed  and  pollen  dispersal  may  alter  eco-evolutionary  dynamics (Kuparinen et al. 2009). To forecast eco-evolutionary dynamics at the extent of species ranges, we will  use  statistical  techniques for  aggregating  high-resolution  stand-scale  simulations  of  individual trees/genotypes  (Kuparinen  et  al.  2010).  We  will  take  care  to  quantify  the  forecast  uncertainty resulting from incomplete knowledge of model parameters and relevant processes. Such uncertainty analyses are important for (i) developing robust management strategies in the face of incomplete knowledge, and (ii) identifying which kind of future research is best suited to reduce our uncertainty about eco-evolutionary responses to environmental change (Higgins et al. 2003).