class: fullbleed, math background-color: black         .absolute.w-100pct.pa-2.center.t-40pct.ba.bw-0.br-0.bg-white-80pct[ ## <b>Predictive modelling of plant communities</b> <br> Challenges for a dynamical approach ] --- layout: true .footer[ <!-- - @DrIsaBlg --> - <i class="fab fa-github"></i>iboulangeat <!-- - isabelle.boulangeat@irstea.fr --> - april 2019, ECOVEG 14, Toulouse <!-- -  --> ] <!-- --> --- class: # Prediction in community ecology .absolute.l-4.t-4[**Why?**] -- .absolute.l-10pct.t-5[climate change<br>habitat fragmentation<br>land-use changes<br>changes in disturbances<br>changes of human practices] -- .absolute.r-6.t-4[**What?**] -- .absolute.l-50pct.t-5[species composition<br>species diversity (taxonomic, functional, phylogenetic)<br>community-level attributes (annual biomass production, canopy height, ...)] --- class: # Modelling approach  --- class: # From explanatory to anticipatory models .absolute.r-2.b-2[Mouquet et al. 2015, JAE] Why ? more data, more tools -- <b>Explanatory</b> : theoretical expectations and tests (hypothetico-deductive) <b>Anticipatory</b> : predictions\* conditionnal to hypotheses (models) <br> \*forecasts, projections -- but necessity for theoretical frameworks to build models in both cases --- class: # Coexistence theories #### What explain the presence of a species in a particular place? --  --  --  --  --  <!-- ========================================================================== --> --- class: title, smokescreen, no-footer background-image: url(abandon_static.png) # Static approaches and their limits --- class: # The SDM framework  --- class: # Explain with SDM  -- .absolute.r-2.b-2[Boulangeat et al. 2012, EcoLet]  --  --- class: # Anticipate with stacked-SDM    -- <br><br><br> Stack of multiple single-species SDM -- <br> -> no species interactions --- class: # Including species interactions 1. Biotic elements as explicative variables (dominant species, community attribute, ...) -- 2. Joint modelling (jSDM) : conditional presence .absolute.r-2.b-2[Pollock et al. 2014, MEE]  --- class: img-left # Integrating stacked-SDM and assembly rules  .absolute.r-2.b-2[D'Amen et al. 2015, JOB and GEB] -- <br>Historical and evolutionary constraints <br>-> potential species pool ("Dark diversity") -- <br>Abiotic constraints <br>-> ex. SDM -- <br>Macroecological constraints <br>-> species richness and/or trait space -- <br>Species interactions (assembly rules) <br>-> random selection, MaxEnt, ... --- class: fit-h1 # Predicting ecological communities : an overview .absolute.r-2.b-2[D'Amen et al. 2017, Biol.Rev.]  <!-- ========================================================================== --> --- class: title, smokescreen, no-footer background-image: url(abandon.gif) # Towards a dynamic perspective --- class: # Limit of a static approach  -- .fixed.bg-white-80pct.t-50pct.l-50pct[->colonisation credit] .fixed.bg-white-80pct.t-80pct.l-50pct[->extinction debt] --- class: # Origin of lags -- #####Rapid <br> global changes   IGBP and resilience center --  --  --- class: fit-h1 # Consequences for multi-species modelling <br> Static approaches **infer** biotic interactions (and resulting community assembly) from species' co-occurrences -- <br><br>-> Hypothesis of equilibrium -- <br> - May be a problem for calibration - Is limiting for prediction of transient states -- <br><br>Dynamic approaches can **simulate** species interactions .absolute.r-2.b-2[Dormann et al. 2018, GEB] <!-- ========================================================================== --> --- class: title, smokescreen, no-footer background-image: url(dandelion.jpg) # Models based on metapopulation dynamics --- class: # From niche to neutral theories  --- class: fit-h1 # Models based on metapopulation dynamics | STATES| **presence**| **absence**| | ------| ---------| --------| | **presence**| 1-p(e) | p(e) | | **absence**| N.p(c) | 1-N.p(c) | .absolute.fr.r-10pct.t-20pct.ofc.w-5-12th[e=extinction ; c=colonisation ; N=regional prevalence] --  --- class: fit-h1 # Models based on metapopulation dynamics  .absolute.r-2.b-2[Talluto et al. 2017, Nature Ecology and Evolution] --- class: # Integrating biotic interactions 1. Biotic elements as explicative variables (dominant species, community attribute, ...) 2. Regroup species 3. Multi-species model --- class: # Integrating biotic interactions Community level: 2 ~species .absolute.r-2.b-2[Vissault et al., in revision]  --   --  --  --  -- --- class: # Predictions: a simulation approach .absolute.r-2.b-2[Vissault et al., in revision] <br>  --  <!-- ========================================================================== --> --- class: title, smokescreen, no-footer background-image: url(melezein_ubac_freaux2.JPG) # Dynamic vegetation models (DVM) --- class: # The interest of DVM .absolute.r-2.b-2[Snell et al. 2014, Ecography] 1. Multi-species! -- 2. Dynamic (simulation models) : transient states and lags -- 3. Process-based interactions : potential to predict non-analog communities -- 4. Multi-scale : individual to landscape processes -- 5. Spatially explicit -- <br><br> Processes of interest: reproduction, establishment, growth, mortality --- class: compact # Dynamic vegetation models  **Forest gap models (stand models)** <br>JABOWA (Botkin 1972), FORET (Shugart 1884), ZELIG (Smith 1988), SORTIE (Deutschmann 1997) <br>Aim : optimize **wood production** for harvest <br>Principle : indiv. based models based on **competition for light** -- <br><br> **Forest Landscape Models (FLM)** <br> LANDIS (He 1999), LANDCLIM (Schumacher 2004), TreeMig (Lischke 2006), LANDIS II (Scheller 2007)  <br>Aim : account for landscape processes (fire, seed dispersal) <br>Principle : upscaling stand models - Cohorts, height classes, PFT, representative cells, ... - Aggregate spatial and temporal scales <br> --- class: compact, fit-h1 # Dynamic vegetation models  .absolute.r-2.b-5[Quillet et al. 2010, Env. Rev.] **Dynamic Global Vegetation models (DGVM)** <br> LPJ (Sitch 2001), MC1 (Bachelet 2001) <br>Aim : simulate **Net Primary Productivity** <br>Principle : **Photosynthesis** ~ light + CO2 + temperature -- <br><br> **Combined DGVM and forest models** <br> LPJ-GUESS (Smith 2003), LM3-PPA (Weng 2015) <br>Aim : improve the simulation of transitions between biomes <br>Principle : model coupling (Hybrid-DGVM) May include: - height-structured competition for light - within-PFT variation --- class: # A scaled overview  --- class: compact, img-left # Beyond forest focus: an exemple with FATE-HD FATE-HD among DVMs <br><br>  .absolute.r-2.b-2[Boulangeat et al. 2014, GCB] -- A landscape model dealing with **forest and non-forests** - Response to climate (via Habitat model) - Vegetation diversity (PFG) - Simpified population dynamics (competition for light, dispersal, demography) - Semi-quantitatif (easy to parameterize) - Disturbances (fire, grazing, mowing) --- class: # FATE-HD model .absolute.r-2.b-2[Boulangeat et al. 2014, GCB]  --  .absolute.r-4.t-3[Germination <br> Recruitment<br> Growth<br>Survival<br>Fecundity] --- class: # FATE-HD model .absolute.r-2.b-2[Boulangeat et al. 2014, GCB]  .absolute.r-4.t-3[Germination <br> Recruitment<br> Growth<br>Survival<br>Fecundity]  --  --  --- class: # FATE-HD model .absolute.r-2.b-2[Boulangeat et al. 2014, GCB]      --- class: # Parameterization : functional Groups  .absolute.r-2.b-2[Boulangeat et al. 2012, GCB] --- class: # Is it enough to represent biodiversity?  --  --- class: # Is it enough to represent biodiversity? <br><br>  --- class: # Parameterization : data  .absolute.r-2.b-2[Boulangeat et al. 2014, Ecography] --- class: # FATE-HD vs SDM (at equilibrium) Refine PFGs distribution inside their habitat limits  --- class: # Scenarios  .absolute.r-2.b-2[Boulangeat et al. 2014, Ecography] --- class: # Scenarios  .absolute.l-4.t-30pct[Intensification]  <!-- .fixed.b-2.l-4[<video height="400" autoplay><source src="intens.mp4" type="video/mp4"></video>] --> .absolute.r-5.t-30pct[Abandonment]  <!-- .fixed.b-2.r-4[<video height="400" autoplay><source src="abandon.mp4" type="video/mp4"></video>] --> .absolute.r-2.b-2[Boulangeat et al. 2014, Ecography] --- class: # Changes in regional diversity  .absolute.r-2.b-2[Boulangeat et al. 2014, Ecography] --- class: # Changes in regional diversity .absolute.r-2.b-2[Boulangeat et al. 2014, Ecography]  -- .absolute.l-3.t-5[additive effects] -- .absolute.r-3.t-5[multiplicative effects] --- class: # Change in local diversity  .absolute.r-2.b-2[Boulangeat et al. 2014, Ecography] --- class: # Diversity decomposition - change in **composition** - change in **abundances** .absolute.r-2.b-2[Boulangeat et al. 2014, Ecography] -- \\[ ^2D = ^0D \times EF \\] -- \\( ^0D = \\) richness -> beta = compositional turnover \\( EF =\\) evenness factor -> beta = abundances re-arrangement --- class: fit-h1, no-footer # Changes over time, elevation and diversity dimensions  .absolute.r-2.b-2[Boulangeat et al. 2014, Ecography] --- class: fit-h1 # Main highlights about the changes in functional diversity .absolute.r-2.b-2[Boulangeat et al. 2014, Ecography] -- <br> **Effet of human land-use:** Heterogeneity (beta) in average decreases in case of tree colonization but increases in case of habitat loss **Effect of different PFG assemblages:** The diversity response depends on elevation **Interaction between drivers:** Multiplicative effects are found when land abandonment is combined with climate change --- class: # Limits of DVMs -- 1. Difficult to parameterize 2. Difficult to evaluate error propagation 3. Limited to communities for which there is enough knowledge about constitutive species -- Mechanisms with limited knowledge - mortality and climate change induced mortality - climate change impact on growth or water-use efficiency - seeds (germination) - phenology and shifts - soil processes <!-- ========================================================================== --> --- class: title, smokescreen, no-footer background-image: url(polarbear.png) # Challenges and perspectives --- class: fit-h1 # Scaling issues -- - **space** : modelling dispersal and landscape disturbances -- - **time** : from daily to annual processes -- - **organisational levels** : represent the whole communities <br>(individuals->cohorts->populations->PFG->PFT->community attributes) -- .absolute.r-2.b-30pct[_Biodiversity vs explicit processes,<br> a necessary trade-off_]  .absolute.r-2.b-2[Gallien et al. 2010, DID] --- class: # Modelling dispersal #### Upscaling from individual processes to landscape levels indiv. demography -> pop. demography -> metapopulation dynamics -- 1. Continuity in space and time for model calibration <br> -> inverse modelling/metamodel, resampling, remote sensing data, ... -- 2. Resolution vs extent <br> -> heterogeneity and spread within a cell see Snell et al. 2014, Ecography -- 3. Dispersal kernels <br> -> A trait-based parameterisation see Tamme et al. 2014, Ecology <br> -> Landscape properties that affect seed dispersal <br> -> Vegetative reproduction --- class:w # Species interactions - Dimension reduction <br>-> Functional groups, using networks to reduce the dimension of interactions -- - Ontology <br>-> A mix between functional groups and age classes? -- - Phenology <br>-> May impact competition for light -- - Interactions (competition and facilitation) via soil ressources (incl. water) <br>-> necessary to improve herbaceous species modelling -- - Trophic interactions <br> ex. plant-pollinisator <br> ex. effect of grazers on seed dispersal and nutrient enrichment --- class: # Response to global changes <br> **Climate change** : impacts on demography <br> -> hybrid approach or inverse modelling? -- **Human practices** : retroactions <br> -> a prospective approach combining stories and simulations? --- class: # Conclusion #### Necessary trade-offs - Species number (or ecological scale) vs explicit processes - Resolution vs extent (space and time) #### Solutions - Multi-scale model integration see Clark et al. 2009, Ecological Monograph ; Talluto et al. 2016, GEB - Hybrid approaches (implicit and explicit processes) see Gallien et al. 2010, DID --- class: no-footer background-image: url(marais_acide_lautaret.JPG) # Thanks  <br><br><br><br><br><br><br><br><br> Slides can be found at http://iboulangeat.github.io/Slides/ <br> Contact: isabelle.boulangeat@irstea.fr