diff --git a/R/analysis/lmer.output-fun.R b/R/analysis/lmer.output-fun.R
index 677701d86e6f526ad5188e9766f8a9345f50bd82..af95d4d33c0d8ba63c6250766747026a54ad4912 100644
--- a/R/analysis/lmer.output-fun.R
+++ b/R/analysis/lmer.output-fun.R
@@ -580,17 +580,18 @@ plot.param <- function(list.res,
'Compet effect x trait',
'Compet response x trait',
'Compet x trait dissimilarity'),
- param.print = 1:5,
- col.names = c('#e41a1c', '#377eb8', '#4daf4a', '#984ea3', '#ff7f00') ,
- col.vec,
- pch.vec,
- names.bio,
- ...){
+ param.print = 1:5,
+ col.names = c('#e41a1c', '#377eb8', '#4daf4a', '#984ea3', '#ff7f00') ,
+ data.type = "all.no.log",
+ col.vec,
+ pch.vec,
+ names.bio,
+ ...){
m <- matrix(c(1:3), 1, 3)
layout(m, widths=c(5.3, 2.39, 2.39 )/10,
heights=c(4)/10)
for (i in traits){
- list.temp <- list.res[[paste("all.no.log_", i ,
+ list.temp <- list.res[[paste(data.type, "_", i ,
"_", model,
sep = '')]]$lmer.summary
param.mean <- list.temp$fixed.coeff.E[param.vec]
@@ -889,6 +890,7 @@ plot.param.biomes.fixed.elli <- function(list.res,
param.print = 1:5,
col.names = c('#e41a1c', '#377eb8', '#4daf4a',
'#984ea3', '#ff7f00') ,
+ data.type = "all.no.log",
col.vec,
pch.vec,
names.bio,
@@ -898,7 +900,7 @@ plot.param.biomes.fixed.elli <- function(list.res,
layout(m, widths=c(4.4, 2, 2 ,5.5)/10,
heights=c(4)/10)
for (i in traits){
- list.temp <- list.res[[paste0("all.no.log_", i ,
+ list.temp <- list.res[[paste0(data.type, "_", i ,
"_", model)]]
## NEED VAR
for (n.vars in seq_len(length(param.vec))){
@@ -968,13 +970,14 @@ format.all.output.lmer <- function(file.name,
list.file.name,
models,
traits = c("SLA", "Wood.density", "Max.height",
- "Leaf.N", "Seed.mass")){
+ "Leaf.N", "Seed.mass"),
+ data.type = "simple"){
files <- c()
for (trait in traits){
for(model in models){
source(model, local = TRUE)
model.obj <- load.model()
- pathout <- output.dir('lmer', model.obj$name, trait, 'simple')
+ pathout <- output.dir('lmer', model.obj$name, trait, data.type)
files <- c(files,file.path(pathout,file.name))
}
}
diff --git a/docs/paper/Makefile b/docs/paper/Makefile
index 8945e998294791a0d6748a94edfe6417e9844b01..9c17428c1d59bad3f9786ade4d679f468fcb463c 100644
--- a/docs/paper/Makefile
+++ b/docs/paper/Makefile
@@ -3,12 +3,19 @@ all: paper.pdf extended_method.pdf extended_data.pdf
paper.pdf: paper.md include.tex refs.bib
pandoc $< --csl=nature.csl --filter pandoc-citeproc --bibliography=refs.bib --template=include.tex --variable mainfont="Times New Roman" --variable sansfont=Arial --variable fontsize=12pt --latex-engine=xelatex -o $@
+paper.docx: paper.md include.tex refs.bib
+ pandoc -s -S $< --csl=nature.csl --filter pandoc-citeproc --bibliography=refs.bib --template=include.tex --variable mainfont="Times New Roman" --variable sansfont=Arial --variable fontsize=12pt -o paper.docx
+
extended_method.md: extended_method.Rmd
Rscript -e "library(knitr); knit('extended_method.Rmd', output = 'extended_method.md')"
extended_method.pdf: extended_method.md include.tex refs.bib
pandoc $< --csl=nature.csl --filter pandoc-citeproc --bibliography=refs.bib --template=include.tex --variable mainfont="Times New Roman" --variable sansfont=Arial --variable fontsize=12pt --latex-engine=xelatex -o $@
+extended_method.docx: extended_method.md include.tex refs.bib
+ pandoc -s -S $< --csl=nature.csl --filter pandoc-citeproc --bibliography=refs.bib --template=include.tex --variable mainfont="Times New Roman" --variable sansfont=Arial --variable fontsize=12pt -o $@
+
+
extended_data.md: extended_data.R
Rscript -e "library(sowsear); sowsear('extended_data.R', 'Rmd')"
Rscript -e "library(knitr); knit('extended_data.Rmd', output = 'extended_data.md')"
diff --git a/docs/paper/abstract_bes.md b/docs/paper/abstract_bes.md
index 390daf62a8cd295544ed712293df067ce37cafe7..0e41964dec09b448b3680b17648835dca0dc26e7 100644
--- a/docs/paper/abstract_bes.md
+++ b/docs/paper/abstract_bes.md
@@ -5,7 +5,7 @@
Sylvie Gourlet-Fleury; Marc Hanewinkel; Bruno Herault; Hiroko Kurokawa;
Yusuke Onoda; Maria Uriarte; Sarah Richardson; Paloma Ruiz;
I-Fang Sun; Goran Ståhl; Nathan Swenson; Jill Thompson; Miguel Zavala;
- Hongcheng Zeng; Jess Zimmerman; Niklaus E Zimmermann; Mark Westoby.
+ Hongcheng Zeng; Jess Zimmerman; Niklaus E Zimmermann; and Mark Westoby.
% BES-SFE Annual Meeting
Competition is a very important type of ecological interaction, especially in terrestrial vegetation where plants greatly modify the local environment for each other. Competitive interactions influence the growth and survival of individuals, and thereby change community composition over time into the future. However firm generalizations have yet to be established about outcomes of competition among species. Here we show that key species’ traits have consistent influences on growth and competition. Our analysis synthesize individuals tree growth data for more than 3-millions trees across a global set of national forest inventories plus also several large forest-monitoring plots. Some traits have a strong effect on the growth rate of the species. Then traits in part determine the tolerance to competition and the impact of competitor’s on a focal tree. A notable generalization is that trait values that favour tolerance to competition also render species slow growing in absence of competition. There is also a small but detectable benefit in reducing competition from trait-dissimilarity between a focal plant and its competitors. The trait-based picture that emerges is much simpler and more general than a quantification of competition coefficients between each pair of species, which is intractable at the global scale. Our results demonstrate that traits may be used to predict competitive interactions in forests at a large scale. We also anticipate our results to have a profound influence on trait-based-models of community.
diff --git a/docs/paper/extended_data.R b/docs/paper/extended_data.R
index 0363b7a7d5870a8c93a9ee4728bc198c28fc1b38..ffdc81c86c5fa838f8bcf7f9d4081966e50c0450 100644
--- a/docs/paper/extended_data.R
+++ b/docs/paper/extended_data.R
@@ -1,9 +1,15 @@
-## # Methods
+## # Extend data
## ```{r options-chunk}
## opts_chunk$set(dev= c('pdf','svg'), fig.width= 10, fig.height = 5)
## ```
+
+## 
+
+## \newpage
+
+
##+ This deals with some path issues, echo = FALSE, results = 'hide'
git.root <- function() {
system("git rev-parse --show-toplevel", intern=TRUE)
@@ -22,28 +28,6 @@ path.root <- git.root()
source.root("R/analysis/lmer.output-fun.R")
source.root("R/analysis/lmer.run.R")
source.root("R/utils/plot.R")
-library(pander)
-data.set <-read.csv(file.path(path.root, 'output', 'data.set.csv'))
-dat.2 <- t(data.set[, -(1:2)])
-colnames(dat.2) <- data.set$set
-var.names <- rownames(dat.2)
-var.names[1] <- '# of trees'
-var.names[2] <- '# of species'
-var.names[3] <- '# of plots/quadrats'
-var.names[4] <- '% of angiosperm'
-var.names[5] <- '% of evergreen'
-var.names[6] <- '% cover Leaf N'
-var.names[7] <- '% cover Seed mass'
-var.names[8] <- '% cover SLA'
-var.names[9] <- '% cover Wood density'
-var.names[10] <- '% cover Max height'
-rownames(dat.2) <- var.names
-dat.2 <- as.data.frame(dat.2)
-dat.2$Var <- var.names
-rownames(dat.2) <- NULL
-dat.2 <- dat.2[, c(15, 1:14)]
-##+ TableE1_Data, echo = FALSE, results='asis', message=FALSE
-pandoc.table(dat.2[1:10, ], caption = "Data description")
##+ ComputeTable_Effectsize, echo = FALSE, results = 'hide', message=FALSE
@@ -51,15 +35,13 @@ list.all.results <-
readRDS.root('output/list.lmer.out.all.NA.no.log.rds')
library(pander)
-mat.param <- do.call('cbind', lapply(c('Wood.density', 'SLA', 'Max.height'), extract.param, list.res = list.all.results, model = 'lmer.LOGLIN.ER.AD.Tf.r.ecocode.species'))
+mat.param <- do.call('cbind', lapply(c('Wood.density', 'SLA', 'Max.height'),
+ extract.param, list.res = list.all.results,
+ model = 'lmer.LOGLIN.ER.AD.Tf.r.ecocode.species'))
colnames(mat.param) <- c('Wood.density', 'SLA', 'Max.height')
-effect.size.mat <- rbind(abs(mat.param[1,]),apply(abs(mat.param[3:4,]),2,sum), mat.param[5,])
-row.names(effect.size.mat) <- c('Direct trait effect', 'Effect/response', 'Limiting similarity')
-
-##+ Table1_Effectsize, echo = FALSE, results='asis', message=FALSE
-pandoc.table(effect.size.mat, caption = "Effect size")
-
-## The full effect size without summing *effect* and *response* is in the Table 2.
+row.names(mat.param) <- c('Direct trait effect', 'Mean competition',
+ 'Competitive effect', 'Competitive response',
+ 'trait similarity')
##+ Table2_Effectsize, echo = FALSE, results='asis', message=FALSE
-pandoc.table(mat.param, caption = "Full effect size")
+pandoc.table(mat.param, caption = "Standaridized parameters estimates presented in Fig 2.")
diff --git a/docs/paper/extended_data.Rmd b/docs/paper/extended_data.Rmd
index 6d7f7af33331cbcb382a2f2c28f15528ff5c3c14..00ea37e41341ab0e633f1f63691edb1ee4e0e693 100644
--- a/docs/paper/extended_data.Rmd
+++ b/docs/paper/extended_data.Rmd
@@ -1,9 +1,15 @@
-# Methods
+# Extend data
```{r options-chunk}
opts_chunk$set(dev= c('pdf','svg'), fig.width= 10, fig.height = 5)
```
+
+
+
+\newpage
+
+
``` {r This deals with some path issues, echo = FALSE, results = 'hide'}
git.root <- function() {
system("git rev-parse --show-toplevel", intern=TRUE)
@@ -24,29 +30,6 @@ path.root <- git.root()
source.root("R/analysis/lmer.output-fun.R")
source.root("R/analysis/lmer.run.R")
source.root("R/utils/plot.R")
-library(pander)
-data.set <-read.csv(file.path(path.root, 'output', 'data.set.csv'))
-dat.2 <- t(data.set[, -(1:2)])
-colnames(dat.2) <- data.set$set
-var.names <- rownames(dat.2)
-var.names[1] <- '# of trees'
-var.names[2] <- '# of species'
-var.names[3] <- '# of plots/quadrats'
-var.names[4] <- '% of angiosperm'
-var.names[5] <- '% of evergreen'
-var.names[6] <- '% cover Leaf N'
-var.names[7] <- '% cover Seed mass'
-var.names[8] <- '% cover SLA'
-var.names[9] <- '% cover Wood density'
-var.names[10] <- '% cover Max height'
-rownames(dat.2) <- var.names
-dat.2 <- as.data.frame(dat.2)
-dat.2$Var <- var.names
-rownames(dat.2) <- NULL
-dat.2 <- dat.2[, c(15, 1:14)]
-```
-``` {r TableE1_Data, echo = FALSE, results='asis', message=FALSE}
-pandoc.table(dat.2[1:10, ], caption = "Data description")
```
@@ -55,18 +38,15 @@ list.all.results <-
readRDS.root('output/list.lmer.out.all.NA.no.log.rds')
library(pander)
-mat.param <- do.call('cbind', lapply(c('Wood.density', 'SLA', 'Max.height'), extract.param, list.res = list.all.results, model = 'lmer.LOGLIN.ER.AD.Tf.r.ecocode.species'))
+mat.param <- do.call('cbind', lapply(c('Wood.density', 'SLA', 'Max.height'),
+ extract.param, list.res = list.all.results,
+ model = 'lmer.LOGLIN.ER.AD.Tf.r.ecocode.species'))
colnames(mat.param) <- c('Wood.density', 'SLA', 'Max.height')
-effect.size.mat <- rbind(abs(mat.param[1,]),apply(abs(mat.param[3:4,]),2,sum), mat.param[5,])
-row.names(effect.size.mat) <- c('Direct trait effect', 'Effect/response', 'Limiting similarity')
-```
-
-``` {r Table1_Effectsize, echo = FALSE, results='asis', message=FALSE}
-pandoc.table(effect.size.mat, caption = "Effect size")
+row.names(mat.param) <- c('Direct trait effect', 'Mean competition',
+ 'Competitive effect', 'Competitive response',
+ 'trait similarity')
```
-The full effect size without summing *effect* and *response* is in the Table 2.
-
``` {r Table2_Effectsize, echo = FALSE, results='asis', message=FALSE}
-pandoc.table(mat.param, caption = "Full effect size")
+pandoc.table(mat.param, caption = "Standaridized parameters estimates presented in Fig 2.")
```
diff --git a/docs/paper/extended_data.md b/docs/paper/extended_data.md
index 570a8364cc7eeacc5deb9860c540be038c40d9ba..61924fbe90b764aadd31c39c508e4fc5c3ed6374 100644
--- a/docs/paper/extended_data.md
+++ b/docs/paper/extended_data.md
@@ -1,4 +1,4 @@
-# Methods
+# Extend data
```r
@@ -6,120 +6,33 @@ opts_chunk$set(dev= c('pdf','svg'), fig.width= 10, fig.height = 5)
```
+
+\newpage
-------------------------------------------------------------
- Var Sweden NVS US Canada NSW
--------------------- -------- ------ ------- -------- ------
- # of trees 202605 53775 1371107 495008 906
- # of species 26 117 493 75 101
-# of plots/quadrats 23487 1415 59840 15995 63
- % of angiosperm 0.2702 0.9401 0.6331 0.3438 0.9989
- % of evergreen 0.7295 0.991 0.3727 0.6485 0.9238
- % cover Leaf N 0.9971 0.9911 0.9231 0.9949 0
- % cover Seed mass 0.9971 0.9995 0.9918 0.9958 1
- % cover SLA 0.9971 0.9986 0.9134 0.9945 0
-% cover Wood density 0.9967 0.9959 0.9443 0.9943 0.9923
+---------------------------------------------------------------
+ Wood.density SLA Max.height
+-------------------------- -------------- -------- ------------
+ **Direct trait effect** -0.1366 0.06761 0.07342
- % cover Max height 0.9791 0.9999 1 0.9986 1
-------------------------------------------------------------
+ **Mean competition** -0.2153 -0.186 -0.227
-Table: Data description (continued below)
+ **Competitive effect** -0.04333 0.06167 -0.03017
-
-----------------------------------------------------------
- France Swiss Spain BCI Paracou Japan Fushan
--------- ------- ------- ------ --------- ------- --------
- 184316 28404 418805 27123 46404 4630 14701
+ **Competitive response** 0.04628 -0.04506 -0.1399
- 127 62 122 239 712 140 72
+ **trait similarity** 0.04945 0.08251 0.06553
+---------------------------------------------------------------
- 17611 2597 36462 2033 2157 318 623
-
- 0.7414 0.3631 0.3469 0.9974 1 0.7352 0.9204
-
- 0.285 0.5526 0.8163 0.7778 0.8345 0.7037 0.7527
-
- 0.9886 0.9923 0.9936 0.9283 0.7252 0 0.8034
-
- 0.9916 0.9924 0.9901 0.9352 0.5577 0.9702 0.6441
-
- 0.9921 0.9674 0.9731 0.9302 0.733 0.997 0.9997
-
- 0.989 0.9503 0.9895 0.9307 0.735 0.997 0.9935
-
- 0.9999 0.9973 1 0.9542 0.635 0.9985 0.9585
-----------------------------------------------------------
-
-Table: Table continues below
-
-
--------------------
- Luquillo Mbaiki
----------- --------
- 14011 17600
-
- 82 204
-
- 399 989
-
- 1 0.9948
-
- 0.9904 0.7257
-
- 0.9928 0.4019
-
- 0.9928 0.01403
-
- 0.9928 0.4019
-
- 0.9928 0.4702
-
- 0.9928 0
--------------------
-
-
-
-
-
--------------------------------------------------------------
- Wood.density SLA Max.height
-------------------------- -------------- ------- ------------
- **Direct trait effect** 0.1366 0.06761 0.07342
-
- **Effect/response** 0.08961 0.1067 0.17
-
- **Limiting similarity** 0.04945 0.08251 0.06553
--------------------------------------------------------------
-
-Table: Effect size
-
-The full effect size without summing *effect* and *response* is in the Table 2.
-
-
---------------------------------------------------------
- Wood.density SLA Max.height
-------------------- -------------- -------- ------------
- **Tf** -0.1366 0.06761 0.07342
-
- **sumBn** -0.2153 -0.186 -0.227
-
- **sumTnBn** -0.04333 0.06167 -0.03017
-
- **sumTfBn** 0.04628 -0.04506 -0.1399
-
- **sumTnTfBn.abs** 0.04945 0.08251 0.06553
---------------------------------------------------------
-
-Table: Full effect size
+Table: Standaridized parameters estimates presented in Fig 2.
diff --git a/docs/paper/extended_method.Rmd b/docs/paper/extended_method.Rmd
index 3780bb78945deeba5f039ec12f9c066bd4eb0c0b..b1dfc3a51888ca9d02d61d5a85d22772f36922a5 100644
--- a/docs/paper/extended_method.Rmd
+++ b/docs/paper/extended_method.Rmd
@@ -3,6 +3,7 @@
# Data
## Tree plot data
+
We collected data reporting individuals tree growth and allowing to
compute competitors local abundance in their neighbourhood. The
objective was collate data sets spanning most of the forest biomes of
@@ -11,31 +12,31 @@ of the traits of interest (SLA, wood density, and maximum height) (in
tropical forest it was one of the main limitation). We collected two
main type of data: (1) national forest inventory (NFI) where tree above a given threshold are sampled in a network of small plots covering the country (with a sampling generally on a regular grid); (2) large permanent plot in size ranging from 0.5ha to 50ha where the x-y coordinates of all tree above a given threshold are recorded (here after refered as *xy* data type). These large plots are generally located in tropical climate but some plots are also located in temperate climate. Tree growth are
estimated through diameter at breast height ($D$) multiple-census (the only exception being French
-data where growth is estimated with short tree core) and the species of each individuals is recorded. From diameter growth we computed basal area growth between to census as $D_2^2/4 \times \pi - D_1^2/4 \times \pi$. We excluded trees with too extreme positive or negative diameter growth value following criterias developed in one of the site [@condit_mortality_1993] and implemented in the R package [CTFS R](http://ctfs.arnarb.harvard.edu/Public/CTFSRPackage/). The minimum $D$ for measurment of tree range between 1cm to 10cm (expected the Swiss NFI with 12cm), to allow comparison between data set we restricted our analysis for tree greater than 10cm of $D$ (thus only the Swiss data exceed this threshold). See the Table S1 for more details on the data. We excluded from the analysis any plots with harvesting during the growth measurement period, that was identified as a plantation, or overlapping the forest edge.
+data where growth is estimated with short tree core) and the species of each individuals is recorded. From diameter growth we computed basal area growth between to census as $D_2^2/4 \times \pi - D_1^2/4 \times \pi$. We excluded trees with too extreme positive or negative diameter growth value following criterias developed in one of the site [@condit_mortality_1993] and implemented in the R package [CTFS R](http://ctfs.arnarb.harvard.edu/Public/CTFSRPackage/). The minimum $D$ for measurment of tree range between 1cm to 10cm (expected the Swiss NFI with 12cm), to allow comparison between data set we restricted our analysis for tree greater than 10cm of $D$ (thus only the Swiss data exceed this threshold). See the Table M1 and Table M2 for more details on the data. We excluded from the analysis any plots with harvesting during the growth measurement period, that was identified as a plantation, or overlapping the forest edge.
+
+For each individual tree we computed local abundance of competitors of each species as the sum of basal area ($m^2.ha^{-1}$) in tree neighborhoud. For *xy* data type we defined the neighborhoud as 15m radius, as the radius of maximum interaction found in previous studies generally ranged between 10 and 20 m [@uriarte_neighborhood_2004; @lamanna_functional_2014]. To avoid edge effect we excluded trees less than 15 m from the edge of the plot. For the NFI type data the neighborhood was computed as the basal area of each competitor per species over the whole plot as the coordinates of trees in the plot were generally not available. The size of the plot where largest trees are sampled vary between 10 and 25m in radius. In the FIA data the four subplots are 7.35 m in radius, but we grouped the four subplots (that are less than 20 m away) in single estimate of the local competitors abundance after initial exploration of the data. The range of neighborhood used in the competition analysis is thus between 10 and 25 m radius (but most plots were between 10 and 15m radius). To account for variation of the abiotic conditions within the $xy$ plot we divided the plot in 20x20m quadrats.
-For each individual tree we computed local abundance of competitors of each species as the sum of basal area ($m^2.ha^{-1}$) in tree neighborhoud. For *xy* data type we defined the neighborhoud as 15m radius, as the radius of maximum interaction found in previous studies generally ranged between 10 and 20 m [@uriarte_neighborhood_2004; lamanna_functional_2014]. To avoid edge effect we excluded trees less than 15 m from the edge of the plot. For the NFI type data the neighborhoudwe computed the basal area of each competitor per species over the whole plot as the coordinates of trees in the plot were generally not available. The size of the larger plot where largest trees are sampled vary bewteen 10 and 25m. In the FIA data the four subplots are 7.35 m in radius, but we grouped the four subplots (that are less than 20 m away) in single estimate of the local competitors abundance after initial exploration. The range of neighborhood used in the competition analysis is thus between 10 and 25 m radius. To account for variation of the abiotic conditions with the $xy$ plot we divided the plot in 20x20m quadrats.
+Based on plot latitude and longitude, we extracted mean annual temperature (*MAT*) and mean annual sum of precipitation (*MAP*) from the [worldclim](http://www.worldclim.org/) data base [@hijmans_very_2005]. Based on MAT and MAP we classified plots in biomes using modified Whittaker biomes[@ricklefs_economy_2001] after digitising the biomes climatic boundaries. In addition we used local ecoregion classification to group data in similar abiotic conditions[^Kopper] to have a fine definition within each data set.
-Based on plot latitude and longitude, we extracted mean annual temperature (*MAT*) and mean annual sum of precipitation (*MAP*) from the [worldclim](http://www.worldclim.org/) data base [@hijmans_very_2005]. Based on MAT and MAP we classified plots in biomes using modified Whittaker biomes[@ricklefs_economy_2001] after digitising the biomes climatic boundaries. In addition we used the Kopper-Geiger climatic zones to have a finer definition of the climatic zones[@kottek_world_2006].
+[^Kopper]:Or we can us the Kopper-Geiger climatic zones to have the same the climatic zones in all countries[@kottek_world_2006].
## Traits data
-We used functional traits extracted from existing data base for each species. We focused on wood density, species specific leaf area ($SLA$) and maximum height because these traits have prevously been show to be related to tree competitive interaction [@wright_functional_2010; @uriarte_trait_2010; @ruger_functional_2012; @kunstler_competitive_2012; @lasky_trait-mediated_2014]. The traits were either extracted from the [TRY](http://www.try-db.org/) data base [@kattge_try_2011] for most temperate forest and local data based collected on site for most tropical forest and then missing species were filled with TRY (see Table S1). Several of the NFI data provided height measurement that allowed to compute the maximum height per species (as the 99% quantile) which are more reliable than the maximum height reported by the TRY data base. We estimated maximum height for each species for the most NFI data (France, US, Spain, Switzerland; for Sweden we used the estimate from the French data and for Canada we used the estimate from the US data). For the few missing data we extracted measurement from the TRY database. For the other data set we used estimate from the local traits data base or TRY. Because in our approach we need to account for the traits of all competitors species present in the neighbourhood, when species traits data was missing we used the genus mean and if no genus data was available the mean of the species present in the Country. To avoid to analysis data were the proportion of missing traits was too high, we restricted our analysis to plots were the percentage of basal area of tree with no species level trait data were less than 10% and 5% for tree with no genus level trait data.
+
+We used functional traits extracted from existing data base for each species. We focused on wood density, species specific leaf area ($SLA$) and maximum height because these traits have prevously been show to be related to tree competitive interaction [@wright_functional_2010; @uriarte_trait_2010; @ruger_functional_2012; @kunstler_competitive_2012; @lasky_trait-mediated_2014]. The traits were either extracted from the [TRY](http://www.try-db.org/) data base [@kattge_try_2011] for most temperate forest or local data based collected locally for most tropical forest and then missing species were filled with TRY (see Table S1). Several of the NFI data provided height measurement that allowed to compute the maximum height per species (as the 99% quantile) which are more reliable than the maximum height reported by the TRY data base. We estimated maximum height for each species for the most NFI data (France, US, Spain, Switzerland; for Sweden we used the estimate from the French data and for Canada we used the estimate from the US data). For the few missing data we extracted measurement from the TRY database. For the other data set we used estimate from the local traits data base or TRY. Because in our approach we need to account for the traits of all competitors species present in the neighbourhood, when species traits data was missing we used the genus mean and if no genus data was available the mean of the species present in the Country. To avoid to analysis data were the proportion of missing traits was too high, we restricted our analysis to plots were the percentage of basal area of tree with (i) no species level trait data were less than 10% and (ii) tree with no genus level data was less than 5%.
# Analysis
-The general approach . The approach follows the approach of neighbourhood modelling[@canham_neighborhood_2006] which estimate the effect of different species of neighbouring competitors $c$ on the growth of focal species $f$ by estimating a competition parameters $\alpha$ representing the effect of the local abundance (basal area) on the growth of the focal species. The estimation relies on analysing tree individual growth (in term of their basal area growth) in function of tree size and the abundance of the competitive species in its neighbourhood. After exploring different equation we selected the equation \label{G1}, which provides good fit and can be linearised in log-log.
+The approach follows the neighbourhood modelling framework[@canham_neighborhood_2006] which estimate the effect of different species of neighbouring competitors $c$ on the growth of focal species $f$ by estimating a competition parameters $\alpha$ representing the effect of the local abundance (basal area) on the growth of the focal species. The estimation relies on analysing tree individual growth (in term of their basal area growth) in function of tree size and the abundance of the competitive species in its neighbourhood. After exploring different equation we selected the equation 1, which provides good fit and can be linearised in log-log.
\begin{equation} \label{G1}
-G_{s,f,p,i,t} = G_{max} \times D_{i,t}^{\gamma} \times exp(\sum_{c=1}^{C_p} {\alpha_{c,f} \times B_{c}}).
+G_{s,f,p,i} = G_{max} \times D_{i,t}^{\gamma} \times exp(\sum_{c=1}^{C_p} {\alpha_{c,f} \times B_{c}}).
\end{equation}
+where:<!-- **need to decide if we use $- \alpha$ or $\alpha$ ???**-->
-**need to decide if we use $- \alpha$ or $\alpha$ ???**
-
-where:
-
-- $G_{s,f,p,i,t}$ is the basal area growth, of the individual $i$ on plot $p$ at time $t$ and species $f$ from set $s$,
-- $\log{G_{max}}$ is the basal area growth with no competition,
+- $G_{s,f,p,i}$ is the basal area growth, of the individual $i$ on plot $p$ and species $f$ from set $s$,
+- $G_{max}$ is the basal area growth with no competition,
- $\gamma$ represents the diameter $D$ power with a random focal species $f$ effect, and
-- $B_c$ is basal area of the competitor species $c$. Here if the competitive parameters $\alpha_{c,f}$ is negative this represents competition, if positive facilitation.
+- $B_c$ is basal area of the competitor species $c$. Here if the competitive parameters $\alpha_{c,f}$ is negative this represents competition, if positive facilitation. $C_p$ is the number of competitor species on plot $p$.
The model was fitted as:
@@ -50,49 +51,42 @@ We then included the effect of trait in $G_{max}$ and $\alpha$ as previously don
\end{equation}
where:
-- $m_0$ is mean Gmax, including a plot (20x20m quadrats were used in large $xy$ plots) $p$ random effect to account for the variation of the local abiotic conditions, a random focal species $f$ effect, and a random set effect $s$ to account for the difference between data set.
+- $m_0$ is mean Gmax, including a plot or quadrats for $xy$ data $p$ random effect (20x20m quadrats were used in large $xy$ plots) to account for the variation of the local abiotic conditions, a random focal species $f$ effect, and a random set effect $s$ to account for the difference between data set.
- $m_1$ is the slope for the link between $G_{max}$ and the trait $t_f$.
\begin{equation} \label{alpha}
-\alpha_{c,f}= c_0 + c_r \times t_f + c_e \times t_c + c_l \times \vert t_c-t_f \vert
+\alpha_{c,f}= \alpha_0 + \alpha_r \times t_f + \alpha_e \times t_c + \alpha_s \times \vert t_c-t_f \vert
\end{equation}
-
where:
-- $c_0$ is the mean competition effect, and including a random focal $f$ species effect,
-- $c_r$ is the slope for the link between competition and the trait $t_f$ of the focal species, this thus the competitive response part in the framework of the workshop,
-- $c_e$ is the slope for the link between competition and the trait $t_c$ the trait of the competitor trees, this thus the competitive effect part in the framework of the workshop,
-- $c_l$ is the slope for the link between competition and absolute distance of the trait $\vert{t_c-t_f}\vert$ , this thus the absolute distance/limiting similarity part in the framework of the workshop.
+
+- $\alpha_0$ is the mean competition effect, including a random focal $f$ species effect,
+- $\alpha_r$ is the slope for the link between competition and the trait $t_f$ of the focal species, this thus the competitive response in Fig. 1 and include a random focal $f$ species effect,
+- $\alpha_e$ is the slope for the link between competition and the trait $t_c$ the trait of the competitor trees, this the competitive effect in Fig. 1,
+- $\alpha_s$ is the slope for the link between competition and absolute distance of the trait $\vert{t_c-t_f}\vert$, this the trait similarity in Fig. 1.
When the equation \label{alpha} is developed in the competition index of equation \label{G1} the parameters are directly related to community weighted means of the different traits variables.
\begin{equation} \label{alpha}
-\sum_{c=1}^{C_p} {\alpha_{c,f} \times B_{c}} = c_0 \times B_{tot} + c_r \times t_f \times B_{tot} + c_e \times B_{t_c} +c_l \times B_{\vert t_c - t_f \vert}
+\sum_{c=1}^{C_p} {\alpha_{c,f} \times B_{c}} = \alpha_0 \times B_{tot} + \alpha_r \times t_f \times B_{tot} + \alpha_e \times B_{t_c} +\alpha_l \times B_{\vert t_c - t_f \vert}
\end{equation}
-
Where:
+
- $B_{tot} = \sum_{c=1}^{C_p} {B_{c}}$,
- $B_{t_c} = \sum_{c=1}^{C_p} {t_c \times B_{c}}$,
- and $B_{\vert t_c - t_f \vert} = \sum_{c=1}^{C_p} {\vert t_c - t_f \vert \times B_{c}}$.
-All explicative variables and the response variable were standardized by Std. Dev. so the parameters are effect size (in addition the trait, the basal area growth and the diameter were centred to facilitate convergence). The model were fitted using $lmer$ in [lme4](http://cran.r-project.org/web/packages/lme4/index.html).
-
-
-**NEED TO EXPLAIN THAT WE HAVE TWO MODELS ONE WITH FIXED BIOMES AND ONE WITH RANDOM ECOREGION?**
+All explicative variables and the response variable were standardized by Std. Dev. so the parameters are effect size (in addition the trait, the basal area growth and the diameter were centred to facilitate convergence). The model were fitted using $lmer$ in [lme4](http://cran.r-project.org/web/packages/lme4/index.html). We fitted two version of this model. In the first version all parameters $m_{0 to 1}$ and $\alpha_{0,r,e,s}$ were estimated with and ecoregion random effect to account for the variation between data set and different abiotic region. In the second version, we repeated the same analysis but including different fixed estimates of these parameters per biomes to explore variation between biomes[^options].
-fit with random kopper zones
-or
-random kopper zones and biomes fixed need to try
-
-explain multi traits as well
+[^options]: There is different options here, that can still changes slightly, I will run a R-inla estimation (Bayesian faster than stan or jags to decide that).
+**\color{red}TO BE DONE** Because traits similarity may be more strongly related to multi-traits distance than a single trait distance we also explored a model with all three traits and a traits distance based on the euclidean distance of the three traits (standardized). This model expand the equation 1 with a parameter $m_1$, $\alpha_e$ and $\alpha_r$ per trait and an effect of multi-traits similarity (with the parameter $\alpha_s$).
# Table and Figures
```{r kable, echo = FALSE, results="asis"}
library(xtable)
dat <- read.csv('../../data/metadata/sites/sites_description.csv')
-print(xtable(dat), scalebox = '0.3')
-#, format = 'markdown', caption = "Table S1. Description of all data set used in the analysis.")
+print(xtable(dat, caption = "Table M1. Description of all data set used in the analysis."), scalebox = '0.3')
```
@@ -127,4 +121,32 @@ a tropical forest. Ecological Applications 12:1344–1363.
20. Wiser, S.K., Bellingham, P.J. & Burrows, L.E. (2001) Managing biodiversity information: development of New Zealand’s National Vegetation Survey databank. New Zealand Journal of Ecology, 25, 1–17.
21. https://nvs.landcareresearch.co.nz/
+\newpage
+
+
+```{r kable2, echo = FALSE, results="asis"}
+library(pander)
+data.set <-read.csv(file.path('../../output', 'data.set.csv'))
+dat.2 <- t(data.set[, -(1:2)])
+colnames(dat.2) <- data.set$set
+var.names <- rownames(dat.2)
+var.names[1] <- '# of trees'
+var.names[2] <- '# of species'
+var.names[3] <- '# of plots/quadrats'
+var.names[4] <- '% of angiosperm'
+var.names[5] <- '% of evergreen'
+var.names[6] <- '% cover Leaf N'
+var.names[7] <- '% cover Seed mass'
+var.names[8] <- '% cover SLA'
+var.names[9] <- '% cover Wood density'
+var.names[10] <- '% cover Max height'
+rownames(dat.2) <- var.names
+dat.2 <- as.data.frame(dat.2)
+dat.2$Var <- var.names
+rownames(dat.2) <- NULL
+dat.2 <- dat.2[, c(15, 1:14)]
+pandoc.table(dat.2[1:10, ], caption = "Table M2. Data description")
+```
+
+
# References
diff --git a/docs/paper/extended_method.md b/docs/paper/extended_method.md
index e4f01d2e75c2729044a6d189fa1ca772857a099a..a1c324985101d503f23b06f9761ca29f46eac24b 100644
--- a/docs/paper/extended_method.md
+++ b/docs/paper/extended_method.md
@@ -3,6 +3,7 @@
# Data
## Tree plot data
+
We collected data reporting individuals tree growth and allowing to
compute competitors local abundance in their neighbourhood. The
objective was collate data sets spanning most of the forest biomes of
@@ -11,31 +12,31 @@ of the traits of interest (SLA, wood density, and maximum height) (in
tropical forest it was one of the main limitation). We collected two
main type of data: (1) national forest inventory (NFI) where tree above a given threshold are sampled in a network of small plots covering the country (with a sampling generally on a regular grid); (2) large permanent plot in size ranging from 0.5ha to 50ha where the x-y coordinates of all tree above a given threshold are recorded (here after refered as *xy* data type). These large plots are generally located in tropical climate but some plots are also located in temperate climate. Tree growth are
estimated through diameter at breast height ($D$) multiple-census (the only exception being French
-data where growth is estimated with short tree core) and the species of each individuals is recorded. From diameter growth we computed basal area growth between to census as $D_2^2/4 \times \pi - D_1^2/4 \times \pi$. We excluded trees with too extreme positive or negative diameter growth value following criterias developed in one of the site [@condit_mortality_1993] and implemented in the R package [CTFS R](http://ctfs.arnarb.harvard.edu/Public/CTFSRPackage/). The minimum $D$ for measurment of tree range between 1cm to 10cm (expected the Swiss NFI with 12cm), to allow comparison between data set we restricted our analysis for tree greater than 10cm of $D$ (thus only the Swiss data exceed this threshold). See the Table S1 for more details on the data. We excluded from the analysis any plots with harvesting during the growth measurement period, that was identified as a plantation, or overlapping the forest edge.
+data where growth is estimated with short tree core) and the species of each individuals is recorded. From diameter growth we computed basal area growth between to census as $D_2^2/4 \times \pi - D_1^2/4 \times \pi$. We excluded trees with too extreme positive or negative diameter growth value following criterias developed in one of the site [@condit_mortality_1993] and implemented in the R package [CTFS R](http://ctfs.arnarb.harvard.edu/Public/CTFSRPackage/). The minimum $D$ for measurment of tree range between 1cm to 10cm (expected the Swiss NFI with 12cm), to allow comparison between data set we restricted our analysis for tree greater than 10cm of $D$ (thus only the Swiss data exceed this threshold). See the Table M1 and Table M2 for more details on the data. We excluded from the analysis any plots with harvesting during the growth measurement period, that was identified as a plantation, or overlapping the forest edge.
+
+For each individual tree we computed local abundance of competitors of each species as the sum of basal area ($m^2.ha^{-1}$) in tree neighborhoud. For *xy* data type we defined the neighborhoud as 15m radius, as the radius of maximum interaction found in previous studies generally ranged between 10 and 20 m [@uriarte_neighborhood_2004; @lamanna_functional_2014]. To avoid edge effect we excluded trees less than 15 m from the edge of the plot. For the NFI type data the neighborhood was computed as the basal area of each competitor per species over the whole plot as the coordinates of trees in the plot were generally not available. The size of the plot where largest trees are sampled vary between 10 and 25m in radius. In the FIA data the four subplots are 7.35 m in radius, but we grouped the four subplots (that are less than 20 m away) in single estimate of the local competitors abundance after initial exploration of the data. The range of neighborhood used in the competition analysis is thus between 10 and 25 m radius (but most plots were between 10 and 15m radius). To account for variation of the abiotic conditions within the $xy$ plot we divided the plot in 20x20m quadrats.
-For each individual tree we computed local abundance of competitors of each species as the sum of basal area ($m^2.ha^{-1}$) in tree neighborhoud. For *xy* data type we defined the neighborhoud as 15m radius, as the radius of maximum interaction found in previous studies generally ranged between 10 and 20 m [@uriarte_neighborhood_2004; lamanna_functional_2014]. To avoid edge effect we excluded trees less than 15 m from the edge of the plot. For the NFI type data the neighborhoudwe computed the basal area of each competitor per species over the whole plot as the coordinates of trees in the plot were generally not available. The size of the larger plot where largest trees are sampled vary bewteen 10 and 25m. In the FIA data the four subplots are 7.35 m in radius, but we grouped the four subplots (that are less than 20 m away) in single estimate of the local competitors abundance after initial exploration. The range of neighborhood used in the competition analysis is thus between 10 and 25 m radius. To account for variation of the abiotic conditions with the $xy$ plot we divided the plot in 20x20m quadrats.
+Based on plot latitude and longitude, we extracted mean annual temperature (*MAT*) and mean annual sum of precipitation (*MAP*) from the [worldclim](http://www.worldclim.org/) data base [@hijmans_very_2005]. Based on MAT and MAP we classified plots in biomes using modified Whittaker biomes[@ricklefs_economy_2001] after digitising the biomes climatic boundaries. In addition we used local ecoregion classification to group data in similar abiotic conditions[^Kopper] to have a fine definition within each data set.
-Based on plot latitude and longitude, we extracted mean annual temperature (*MAT*) and mean annual sum of precipitation (*MAP*) from the [worldclim](http://www.worldclim.org/) data base [@hijmans_very_2005]. Based on MAT and MAP we classified plots in biomes using modified Whittaker biomes[@ricklefs_economy_2001] after digitising the biomes climatic boundaries. In addition we used the Kopper-Geiger climatic zones to have a finer definition of the climatic zones[@kottek_world_2006].
+[^Kopper]:Or we can us the Kopper-Geiger climatic zones to have the same the climatic zones in all countries[@kottek_world_2006].
## Traits data
-We used functional traits extracted from existing data base for each species. We focused on wood density, species specific leaf area ($SLA$) and maximum height because these traits have prevously been show to be related to tree competitive interaction [@wright_functional_2010; @uriarte_trait_2010; @ruger_functional_2012; @kunstler_competitive_2012; @lasky_trait-mediated_2014]. The traits were either extracted from the [TRY](http://www.try-db.org/) data base [@kattge_try_2011] for most temperate forest and local data based collected on site for most tropical forest and then missing species were filled with TRY (see Table S1). Several of the NFI data provided height measurement that allowed to compute the maximum height per species (as the 99% quantile) which are more reliable than the maximum height reported by the TRY data base. We estimated maximum height for each species for the most NFI data (France, US, Spain, Switzerland; for Sweden we used the estimate from the French data and for Canada we used the estimate from the US data). For the few missing data we extracted measurement from the TRY database. For the other data set we used estimate from the local traits data base or TRY. Because in our approach we need to account for the traits of all competitors species present in the neighbourhood, when species traits data was missing we used the genus mean and if no genus data was available the mean of the species present in the Country. To avoid to analysis data were the proportion of missing traits was too high, we restricted our analysis to plots were the percentage of basal area of tree with no species level trait data were less than 10% and 5% for tree with no genus level trait data.
+
+We used functional traits extracted from existing data base for each species. We focused on wood density, species specific leaf area ($SLA$) and maximum height because these traits have prevously been show to be related to tree competitive interaction [@wright_functional_2010; @uriarte_trait_2010; @ruger_functional_2012; @kunstler_competitive_2012; @lasky_trait-mediated_2014]. The traits were either extracted from the [TRY](http://www.try-db.org/) data base [@kattge_try_2011] for most temperate forest or local data based collected locally for most tropical forest and then missing species were filled with TRY (see Table S1). Several of the NFI data provided height measurement that allowed to compute the maximum height per species (as the 99% quantile) which are more reliable than the maximum height reported by the TRY data base. We estimated maximum height for each species for the most NFI data (France, US, Spain, Switzerland; for Sweden we used the estimate from the French data and for Canada we used the estimate from the US data). For the few missing data we extracted measurement from the TRY database. For the other data set we used estimate from the local traits data base or TRY. Because in our approach we need to account for the traits of all competitors species present in the neighbourhood, when species traits data was missing we used the genus mean and if no genus data was available the mean of the species present in the Country. To avoid to analysis data were the proportion of missing traits was too high, we restricted our analysis to plots were the percentage of basal area of tree with (i) no species level trait data were less than 10% and (ii) tree with no genus level data was less than 5%.
# Analysis
-The general approach . The approach follows the approach of neighbourhood modelling[@canham_neighborhood_2006] which estimate the effect of different species of neighbouring competitors $c$ on the growth of focal species $f$ by estimating a competition parameters $\alpha$ representing the effect of the local abundance (basal area) on the growth of the focal species. The estimation relies on analysing tree individual growth (in term of their basal area growth) in function of tree size and the abundance of the competitive species in its neighbourhood. After exploring different equation we selected the equation \label{G1}, which provides good fit and can be linearised in log-log.
+The approach follows the neighbourhood modelling framework[@canham_neighborhood_2006] which estimate the effect of different species of neighbouring competitors $c$ on the growth of focal species $f$ by estimating a competition parameters $\alpha$ representing the effect of the local abundance (basal area) on the growth of the focal species. The estimation relies on analysing tree individual growth (in term of their basal area growth) in function of tree size and the abundance of the competitive species in its neighbourhood. After exploring different equation we selected the equation 1, which provides good fit and can be linearised in log-log.
\begin{equation} \label{G1}
-G_{s,f,p,i,t} = G_{max} \times D_{i,t}^{\gamma} \times exp(\sum_{c=1}^{C_p} {\alpha_{c,f} \times B_{c}}).
+G_{s,f,p,i} = G_{max} \times D_{i,t}^{\gamma} \times exp(\sum_{c=1}^{C_p} {\alpha_{c,f} \times B_{c}}).
\end{equation}
+where:<!-- **need to decide if we use $- \alpha$ or $\alpha$ ???**-->
-**need to decide if we use $- \alpha$ or $\alpha$ ???**
-
-where:
-
-- $G_{s,f,p,i,t}$ is the basal area growth, of the individual $i$ on plot $p$ at time $t$ and species $f$ from set $s$,
-- $\log{G_{max}}$ is the basal area growth with no competition,
+- $G_{s,f,p,i}$ is the basal area growth, of the individual $i$ on plot $p$ and species $f$ from set $s$,
+- $G_{max}$ is the basal area growth with no competition,
- $\gamma$ represents the diameter $D$ power with a random focal species $f$ effect, and
-- $B_c$ is basal area of the competitor species $c$. Here if the competitive parameters $\alpha_{c,f}$ is negative this represents competition, if positive facilitation.
+- $B_c$ is basal area of the competitor species $c$. Here if the competitive parameters $\alpha_{c,f}$ is negative this represents competition, if positive facilitation. $C_p$ is the number of competitor species on plot $p$.
The model was fitted as:
@@ -50,46 +51,40 @@ We then included the effect of trait in $G_{max}$ and $\alpha$ as previously don
\end{equation}
where:
-- $m_0$ is mean Gmax, including a plot (20x20m quadrats were used in large $xy$ plots) $p$ random effect to account for the variation of the local abiotic conditions, a random focal species $f$ effect, and a random set effect $s$ to account for the difference between data set.
+- $m_0$ is mean Gmax, including a plot or quadrats for $xy$ data $p$ random effect (20x20m quadrats were used in large $xy$ plots) to account for the variation of the local abiotic conditions, a random focal species $f$ effect, and a random set effect $s$ to account for the difference between data set.
- $m_1$ is the slope for the link between $G_{max}$ and the trait $t_f$.
\begin{equation} \label{alpha}
-\alpha_{c,f}= c_0 + c_r \times t_f + c_e \times t_c + c_l \times \vert t_c-t_f \vert
+\alpha_{c,f}= \alpha_0 + \alpha_r \times t_f + \alpha_e \times t_c + \alpha_s \times \vert t_c-t_f \vert
\end{equation}
-
where:
-- $c_0$ is the mean competition effect, and including a random focal $f$ species effect,
-- $c_r$ is the slope for the link between competition and the trait $t_f$ of the focal species, this thus the competitive response part in the framework of the workshop,
-- $c_e$ is the slope for the link between competition and the trait $t_c$ the trait of the competitor trees, this thus the competitive effect part in the framework of the workshop,
-- $c_l$ is the slope for the link between competition and absolute distance of the trait $\vert{t_c-t_f}\vert$ , this thus the absolute distance/limiting similarity part in the framework of the workshop.
+
+- $\alpha_0$ is the mean competition effect, including a random focal $f$ species effect,
+- $\alpha_r$ is the slope for the link between competition and the trait $t_f$ of the focal species, this thus the competitive response in Fig. 1 and include a random focal $f$ species effect,
+- $\alpha_e$ is the slope for the link between competition and the trait $t_c$ the trait of the competitor trees, this the competitive effect in Fig. 1,
+- $\alpha_s$ is the slope for the link between competition and absolute distance of the trait $\vert{t_c-t_f}\vert$, this the trait similarity in Fig. 1.
When the equation \label{alpha} is developed in the competition index of equation \label{G1} the parameters are directly related to community weighted means of the different traits variables.
\begin{equation} \label{alpha}
-\sum_{c=1}^{C_p} {\alpha_{c,f} \times B_{c}} = c_0 \times B_{tot} + c_r \times t_f \times B_{tot} + c_e \times B_{t_c} +c_l \times B_{\vert t_c - t_f \vert}
+\sum_{c=1}^{C_p} {\alpha_{c,f} \times B_{c}} = \alpha_0 \times B_{tot} + \alpha_r \times t_f \times B_{tot} + \alpha_e \times B_{t_c} +\alpha_l \times B_{\vert t_c - t_f \vert}
\end{equation}
-
Where:
+
- $B_{tot} = \sum_{c=1}^{C_p} {B_{c}}$,
- $B_{t_c} = \sum_{c=1}^{C_p} {t_c \times B_{c}}$,
- and $B_{\vert t_c - t_f \vert} = \sum_{c=1}^{C_p} {\vert t_c - t_f \vert \times B_{c}}$.
-All explicative variables and the response variable were standardized by Std. Dev. so the parameters are effect size (in addition the trait, the basal area growth and the diameter were centred to facilitate convergence). The model were fitted using $lmer$ in [lme4](http://cran.r-project.org/web/packages/lme4/index.html).
-
-
-**NEED TO EXPLAIN THAT WE HAVE TWO MODELS ONE WITH FIXED BIOMES AND ONE WITH RANDOM ECOREGION?**
-
-fit with random kopper zones
-or
-random kopper zones and biomes fixed need to try
+All explicative variables and the response variable were standardized by Std. Dev. so the parameters are effect size (in addition the trait, the basal area growth and the diameter were centred to facilitate convergence). The model were fitted using $lmer$ in [lme4](http://cran.r-project.org/web/packages/lme4/index.html). We fitted two version of this model. In the first version all parameters $m_{0 to 1}$ and $\alpha_{0,r,e,s}$ were estimated with and ecoregion random effect to account for the variation between data set and different abiotic region. In the second version, we repeated the same analysis but including different fixed estimates of these parameters per biomes to explore variation between biomes[^options].
-explain multi traits as well
+[^options]: There is different options here, that can still changes slightly, I will run a R-inla estimation (Bayesian faster than stan or jags to decide that).
+**\color{red}TO BE DONE** Because traits similarity may be more strongly related to multi-traits distance than a single trait distance we also explored a model with all three traits and a traits distance based on the euclidean distance of the three traits (standardized). This model expand the equation 1 with a parameter $m_1$, $\alpha_e$ and $\alpha_r$ per trait and an effect of multi-traits similarity (with the parameter $\alpha_s$).
# Table and Figures
-% latex table generated in R 3.1.2 by xtable 1.7-4 package
-% Wed Dec 17 17:33:43 2014
+% latex table generated in R 3.1.1 by xtable 1.7-4 package
+% Sat Dec 20 22:42:03 2014
\begin{table}[ht]
\centering
\scalebox{0.3}{
@@ -114,6 +109,7 @@ explain multi traits as well
\hline
\end{tabular}
}
+\caption{Table M1. Description of all data set used in the analysis.}
\end{table}
## Table References
@@ -147,4 +143,97 @@ a tropical forest. Ecological Applications 12:1344–1363.
20. Wiser, S.K., Bellingham, P.J. & Burrows, L.E. (2001) Managing biodiversity information: development of New Zealand’s National Vegetation Survey databank. New Zealand Journal of Ecology, 25, 1–17.
21. https://nvs.landcareresearch.co.nz/
+\newpage
+
+
+
+```
+##
+## Attaching package: 'pander'
+##
+## The following object is masked from 'package:knitr':
+##
+## pandoc
+```
+
+
+------------------------------------------------------------
+ Var Sweden NVS US Canada NSW
+-------------------- -------- ------ ------- -------- ------
+ # of trees 202605 53775 1371107 495008 906
+
+ # of species 26 117 493 75 101
+
+# of plots/quadrats 23487 1415 59840 15995 63
+
+ % of angiosperm 0.2702 0.9401 0.6331 0.3438 0.9989
+
+ % of evergreen 0.7295 0.991 0.3727 0.6485 0.9238
+
+ % cover Leaf N 0.9971 0.9911 0.9231 0.9949 0
+
+ % cover Seed mass 0.9971 0.9995 0.9918 0.9958 1
+
+ % cover SLA 0.9971 0.9986 0.9134 0.9945 0
+
+% cover Wood density 0.9967 0.9959 0.9443 0.9943 0.9923
+
+ % cover Max height 0.9791 0.9999 1 0.9986 1
+------------------------------------------------------------
+
+Table: Table M2. Data description (continued below)
+
+
+----------------------------------------------------------
+ France Swiss Spain BCI Paracou Japan Fushan
+-------- ------- ------- ------ --------- ------- --------
+ 184316 28404 418805 27123 46404 4630 14701
+
+ 127 62 122 239 712 140 72
+
+ 17611 2597 36462 2033 2157 318 623
+
+ 0.7414 0.3631 0.3469 0.9974 1 0.7352 0.9204
+
+ 0.285 0.5526 0.8163 0.7778 0.8345 0.7037 0.7527
+
+ 0.9886 0.9923 0.9936 0.9283 0.7252 0 0.8034
+
+ 0.9916 0.9924 0.9901 0.9352 0.5577 0.9702 0.6441
+
+ 0.9921 0.9674 0.9731 0.9302 0.733 0.997 0.9997
+
+ 0.989 0.9503 0.9895 0.9307 0.735 0.997 0.9935
+
+ 0.9999 0.9973 1 0.9542 0.635 0.9985 0.9585
+----------------------------------------------------------
+
+Table: Table continues below
+
+
+-------------------
+ Luquillo Mbaiki
+---------- --------
+ 14011 17600
+
+ 82 204
+
+ 399 989
+
+ 1 0.9948
+
+ 0.9904 0.7257
+
+ 0.9928 0.4019
+
+ 0.9928 0.01403
+
+ 0.9928 0.4019
+
+ 0.9928 0.4702
+
+ 0.9928 0
+-------------------
+
+
# References
diff --git a/docs/paper/image/fig1d.svg b/docs/paper/image/fig1d.svg
index 7bf410c8b2ad19c9decb953217b7b5b785065bb5..b41187ffa1b6629f5a254272691cd615cfcb600e 100644
--- a/docs/paper/image/fig1d.svg
+++ b/docs/paper/image/fig1d.svg
@@ -293,21 +293,21 @@
transform="matrix(1,0,0,-1,407.1,287.9)"
id="text338"><tspan
style="font-size:16px;font-variant:normal;font-weight:normal;writing-mode:lr-tb;fill:#4daf4a;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial Black;-inkscape-font-specification:ArialMT"
- x="0 8.7040005 16.608 20.191999 29.087999 34.383999 43.279999 52.175999 60.175999 69.071999 73.472 77.952003 86.751999"
- y="0"
+ x="0"
sodipodi:role="line"
- id="tspan340"><tspan
- style="-inkscape-font-specification:Arial;font-family:Arial"
- id="tspan4330">Tolerance to </tspan></tspan></text>
+ id="tspan340"
+ y="0"><tspan
+ style="font-family:Arial;-inkscape-font-specification:Arial"
+ id="tspan4330">Competitive </tspan></tspan></text>
</g><g
id="g342"><text
transform="matrix(1,0,0,-1,373.4,269.9)"
id="text344"><tspan
style="font-size:16px;font-variant:normal;font-weight:normal;writing-mode:lr-tb;fill:#4daf4a;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial;-inkscape-font-specification:Arial"
- x="0 8.0799999 16.976 30.271999 39.167999 47.967999 52.448002 56.032001 60.431999 64.015999 72.912003"
- y="0"
sodipodi:role="line"
- id="tspan346">competition</tspan></text>
+ id="tspan346"
+ x="0"
+ y="0">response</tspan></text>
</g><g
id="g348"><text
transform="scale(1,-1)"
@@ -332,18 +332,18 @@
transform="matrix(1,0,0,-1,390.1,251.9)"
id="text362"><tspan
style="font-size:16px;font-variant:normal;font-weight:bold;writing-mode:lr-tb;fill:#4daf4a;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial;-inkscape-font-specification:Arial Bold"
- x="0 8.0799999 12.48"
+ x="0 8.0799999"
y="0"
sodipodi:role="line"
- id="tspan364">x c</tspan></text>
+ id="tspan364">x α</tspan></text>
</g><g
id="g366"><text
transform="matrix(1,0,0,-1,410.7,245.1)"
id="text368"><tspan
- style="font-size:9.30000019000000044px;font-variant:normal;font-weight:bold;writing-mode:lr-tb;fill:#4daf4a;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial;-inkscape-font-specification:Arial Bold"
+ style="font-size:9.30000019px;font-variant:normal;font-weight:bold;writing-mode:lr-tb;fill:#4daf4a;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial;-inkscape-font-specification:Arial Bold"
x="0"
y="0"
- id="tspan370">1</tspan></text>
+ id="tspan370">r</tspan></text>
</g><g
id="g378"><path
d="m 255.6,339.3 0,-39.9"
@@ -420,18 +420,18 @@
transform="matrix(1,0,0,-1,434.28,447.5)"
id="text422"><tspan
style="font-size:16px;font-variant:normal;font-weight:bold;writing-mode:lr-tb;fill:#984ea3;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial;-inkscape-font-specification:Arial Bold"
- x="0 8 12.48"
+ x="0 8"
y="0"
sodipodi:role="line"
- id="tspan424">x c</tspan></text>
+ id="tspan424">x α</tspan></text>
</g><g
id="g426"><text
transform="matrix(1,0,0,-1,454.8,440.8)"
id="text428"><tspan
- style="font-size:9.30000019000000044px;font-variant:normal;font-weight:bold;writing-mode:lr-tb;fill:#984ea3;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial;-inkscape-font-specification:Arial Bold"
+ style="font-size:9.30000019px;font-variant:normal;font-weight:bold;writing-mode:lr-tb;fill:#984ea3;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial;-inkscape-font-specification:Arial Bold"
x="0"
y="0"
- id="tspan430">2</tspan></text>
+ id="tspan430">e</tspan></text>
</g><g
id="g438"><text
transform="matrix(1,0,0,-1,373.8,369.1)"
@@ -458,15 +458,16 @@
id="tspan4326">Trait </tspan></tspan></text>
</g><g
id="g450"><text
- transform="matrix(1,0,0,-1,373.8,351.1)"
- id="text452"><tspan
+ transform="scale(1,-1)"
+ id="text452"
+ x="357.79996"
+ y="-351.10001"><tspan
style="font-size:16px;font-variant:normal;font-weight:normal;writing-mode:lr-tb;fill:#ff7f00;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial Black;-inkscape-font-specification:ArialMT"
- x="0 8.8959999 12.48 20.559999 28.559999 32.144001 45.439999 49.023998 52.608002 61.504002 66.800003 70.384003 74.783997 82.783997"
- y="0"
+ x="378.35995 386.35995 389.94394 403.23993 406.82394 410.40796 419.30396 424.59995 428.18396 432.58395 440.58395"
sodipodi:role="line"
id="tspan454"><tspan
- style="-inkscape-font-specification:Arial;font-family:Arial"
- id="tspan4328">dissimilarity </tspan></tspan></text>
+ style="font-family:Arial;-inkscape-font-specification:Arial"
+ id="tspan4328">similarity </tspan></tspan></text>
</g><g
id="g462"><text
transform="matrix(1,0,0,-1,379.1,333.1)"
@@ -513,20 +514,20 @@
transform="matrix(1,0,0,-1,409.1,333.1)"
id="text494"><tspan
style="font-size:16px;font-variant:normal;font-weight:normal;writing-mode:lr-tb;fill:#ff7f00;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial Black;-inkscape-font-specification:ArialMT"
- x="0 4.1919999 8.6719999 16.672001 21.072001"
+ x="0 4.1919999 8.6719999 16.672001"
y="0"
sodipodi:role="line"
id="tspan496">| <tspan
- style="-inkscape-font-specification:Arial Bold;font-family:Arial;font-weight:bold"
- id="tspan3236">x c</tspan></tspan></text>
+ style="font-weight:bold;font-family:Arial;-inkscape-font-specification:Arial Bold"
+ id="tspan3236">x α</tspan></tspan></text>
</g><g
id="g498"><text
transform="matrix(1,0,0,-1,438.3,326.3)"
id="text500"><tspan
- style="font-size:9.30000019000000044px;font-variant:normal;font-weight:bold;writing-mode:lr-tb;fill:#ff7f00;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial;-inkscape-font-specification:Arial Bold"
+ style="font-size:9.30000019px;font-variant:normal;font-weight:bold;writing-mode:lr-tb;fill:#ff7f00;fill-opacity:1;fill-rule:nonzero;stroke:none;font-family:Arial;-inkscape-font-specification:Arial Bold"
x="0"
y="0"
- id="tspan502">3</tspan></text>
+ id="tspan502">s</tspan></text>
</g><g
id="g510"><path
d="m 215.7,380 0,-79"
diff --git a/docs/paper/paper.md b/docs/paper/paper.md
index 7f20d78ca651e85fe4d2892638e8110420f2f377..24ca3cd9b0cd4efa08afb534260b6f58d7fc9720 100644
--- a/docs/paper/paper.md
+++ b/docs/paper/paper.md
@@ -1,5 +1,11 @@
%Title : Functional traits have globally consistent effects on plant competition
-%Authors: Georges Kunstler and ...
+%Authors: Georges Kunstler; David A Coomes; Daniel Falster; Francis Hui;
+ Rob Kooyman; Daniel Laughlin Lourens Poorter; Mark Vanderwel;
+ Ghislain Vieilledent; Joe Wright; Masahiro Aiba; John Caspersen;
+ Sylvie Gourlet-Fleury; Marc Hanewinkel; Bruno Herault; Hiroko Kurokawa;
+ Yusuke Onoda; Maria Uriarte; Sarah Richardson; Paloma Ruiz;
+ I-Fang Sun; Goran Ståhl; Nathan Swenson; Jill Thompson; Miguel Zavala;
+ Hongcheng Zeng; Jess Zimmerman; Niklaus E Zimmermann; and Mark Westoby.
# Summary paragraph outline (max 200 words)
@@ -8,15 +14,15 @@ Competition is a very important type of ecological interaction, especially in te
# Main text
-**(MAX 1500 words till the end of Methods = 1499)**
+**(MAX 1500 words till the end of Methods = 1494)**
Competition is a fundamental type of interaction in ecological communities. Each individual modify their neighbouring environment and thus influences the performance of neighbouring individuals [@keddy_competition_2001]. Competition influences species composition and its changes over time. Maybe competition is especially important for vegetation on land because most vegetation types have high enough cover that shading and water and nutrient depletion are conspicuous. There have thus been a very large number of studies on competition among plants [@goldberg_patterns_1992], but firm generalizations have yet to be established about its outcomes.
-When competition is described as interactions between pairs of species (as it traditionally has been), the number of different interactions to be measured grows explosively with the number of species ($N^2$), and becomes quickly intractable. Also this species-pair approach does not lead naturally to generalization across forests on different continents with different composition. Here we quantify competition between trees (in the sense of influence of neighbours on growth of a focal tree) within a framework, which is novel in two important ways: (i) competition is modelled as a function of traits rather than of species and (ii) we partition how traits drive the outcome of competition in four different key processes (Fig. \ref{ilustr}). Competition can select trait values that are the most competitive. This competitive advantage of trait values can arise because (1) there are correlated with higher potential growth (in absence of competition) [@wright_functional_2010], (2) they are correlated with a higher tolerance to competition [@goldberg_competitive_1996], or (3) they are correlated with higher competitive effect [@gaudet_comparative_1988]. In contrast, competition can promotes the coexistence of a mixture of traits values, if (4) competition decrease with increasing dissimilarity of the traits of the competitors and the focal tree [@macarthur_limiting_1967]. These four processes are likely to be connected to the key traits used to describe plant strategies[@hillerislambers_rethinking_2012; @lasky_trait-mediated_2014], however there is no agremenent on their relative contributions to each of these processes and whether the magnitude and direction of these effects are conserved across large scale.
+When competition is described as interactions between pairs of species (as it traditionally has been), the number of different interactions to be measured grows explosively with the number of species ($N^2$), and becomes quickly intractable. Also this species-pair approach does not lead naturally to generalization across forests on different continents with different composition. Here we quantify competition between trees (in the sense of influence of neighbours on growth of a focal tree) within a framework, which is novel in two important ways: (i) competition is modelled as a function of traits rather than of species and (ii) we partition how traits drive the outcome of competition in four different key processes (Fig. \ref{ilustr}). Competition can select trait values that are the most competitive. This competitive advantage of trait values can arise because (1) there are correlated with higher potential growth (in absence of competition) [@wright_functional_2010], (2) they are correlated with a higher tolerance to competition [@goldberg_competitive_1996], or (3) they are correlated with higher competitive effect [@gaudet_comparative_1988]. In contrast, competition can promotes the coexistence of a mixture of traits values, if (4) competition decrease with increasing dissimilarity of the traits of the competitors and the focal tree [@macarthur_limiting_1967]. These four processes are likely to be connected to the key traits used to describe plant strategies[@hillerislambers_rethinking_2012; @lasky_trait-mediated_2014], however there is no agreement on their relative contributions to each of these processes and whether the magnitude and direction of these effects are conserved across large scale.
Here we dissect how three key traits[@westoby_plant_2002; @chave_towards_2009] (maximum height, wood density and specific leaf area - *SLA*) affect these four processes involved in competition between trees using neighbouring modeling approach[@uriarte_neighborhood_2004]. We compiled data of growth along side local abundance of their competitor for more than 7 million trees representing more than 2500 species covering all the major biomes of the earth (Fig. \ref{res2}b). We analysed how the potential growth of each individual tree was reduced by the local abundance of its competitors. Our analysis accounts for the trait of both the focal tree and its competitors estimating the trait effect for each of the processes presented in Fig. \ref{ilustr}.
-Across all biomes we found that strongest drivers of individual growth was first-ranked the local abundance of competitors; and second ranked the direct influence of the focal plant’s traits on its growth (Fig. \ref{res1} Extended data Table 1). We detected only negative effect of the abundance of competitors showing that competition was predominant. Among the three traits wood density had the strongest direct effect, followed by maximum height whereas SLA had no detectable effect (Fig. \ref{res1}). Then our results show the influence of neighbour traits on their competitive impact, and of focal species traits on tolerance of competition (Fig. \ref{res1}). Taken together these two effects are in the range of half or quarter as big as the direct trait effect (Extend data Table 2), down to zero influence depending on the trait (Fig \ref{res1}). Finally, there is a small but consistent effect whereby the wider is the absolute trait separation between focal and neighbour species, the weaker is competitive suppression of growth (Fig. \ref{res1}). This may arise because of negative density dependence arising for species with similar trait because, for instance, a higher load of herbivores or pathogens [@bagchi_pathogens_2014] or less efficient use of resources (such a less efficient packing of crown in the canopy[^Jucker] or the root in the soil). An analysis using a multiple-traits distance rather than a single trait distance didn't show different pattern (supplementary Figure **\color{red}?[^todo]**). Analyses allowing for different effect between biomes did no show strong evidence for any particular biome behaving consistently differently from the others (Fig. \ref{res3}). The exception is the temperate biomes where SLA showed much stronger effect, probably ought to the dominance of deciduous species in this biome (Fig. \ref{res2}).
+Across all biomes we found that strongest drivers of individual growth was first-ranked the local abundance of competitors; and second ranked the direct influence of the focal plant’s traits on its growth (Fig. \ref{res1} Extended data Table D1). We detected only negative effect of the abundance of competitors showing that competition was predominant. Among the three traits wood density had the strongest direct effect, followed by maximum height whereas SLA had no detectable effect (Fig. \ref{res1}). Then our results show the influence of neighbour traits on their competitive impact, and of focal species traits on tolerance of competition (Fig. \ref{res1}). Taken together these two effects are in the range of half or quarter as big as the direct trait effect (Extend data Table D1), down to zero influence depending on the trait (Fig \ref{res1}). Finally, there is a small but consistent effect whereby the wider is the absolute trait separation between focal and neighbour species, the weaker is competitive suppression of growth (Fig. \ref{res1}). This may arise because of negative density dependence arising for species with similar trait because, for instance, a higher load of herbivores or pathogens [@bagchi_pathogens_2014] or less efficient use of resources (such a less efficient packing of crown in the canopy[^Jucker] or the root in the soil). An analysis using a multiple-traits distance rather than a single trait distance didn't show different pattern (extended data Figure **\color{red}?[^todo]**). Analyses allowing for different effect between biomes did no show strong evidence for any particular biome behaving consistently differently from the others (Fig. \ref{res3}). The exception is the temperate biomes where SLA showed much stronger effect, probably ought to the dominance of deciduous species in this biome (Fig. \ref{res2}).
[^Jucker]:Refs from T. Jucker PhD. to come.
@@ -29,7 +35,7 @@ Across all biomes we found that strongest drivers of individual growth was first
species having high wood density[@wright_functional_2010]. High wood density also resulted in a
higher competitive effect, that may be related to deeper crown
[@poorter_architecture_2006; @aiba_architectural_2009]. The lack of direct effect of *SLA* on maximum growth (but with a positive tendency)
- agree well with the weak correlation prevously reported for adult
+ agree well with the weak correlation previously reported for adult
trees [@wright_functional_2010]. Increasing *SLA* was also weakly
related to decreased competitive effect and no effect or possible
weakly decreased competitive tolerance, which agree with previous
@@ -38,15 +44,15 @@ Across all biomes we found that strongest drivers of individual growth was first
positively related to maximum growth as previously
reported[@wright_functional_2010]. Species with small maximum height
were also much more tolerant to competition than taller species, in
- line with the idea that subcanopy tree are more shade-tolerant. For
+ line with the idea that sub-canopy tree are more shade-tolerant. For
wood density and *SLA* the link with competitive effect and
- competitive response was oppposed, in agrement with a coordinated
- selection under which trait value that confere high competitive
+ competitive response was opposed, in agreement with a coordinated
+ selection under which trait value that confer high competitive
effect also confer high competitive tolerance (a competitive
hierachy[@kunstler_competitive_2012; @mayfield_opposing_2010]). This
was not the case for maximum height because short species were more
tolerant to competition but had a lower competitive effect. This
- match well the observation of the persistance of subcanopy species
+ match well the observation of the persistence of sub-canopy species
under a close cover of tall tree and the stratification theory of
species coexistence[@kohyama_stratification_2009].
Finally our study show that trait values that favour tolerance to
@@ -66,12 +72,12 @@ The globally consistent link between competition effect on growth and traits tha
To examine the link between competition and traits we first compiled
forest plot data from both national forest inventories and long-term
permanent plots from 14 countries covering all the major biomes of the
-world (Fig. 1b, see extanded Table 1). We restricted our analysis to
+world (Fig. 1b, see extended methods Table M1 \& M2). We restricted our analysis to
tree with trunk diameter $>= 10cm$ to have a common minimum size
across all data set. Second, we extracted traits mean per species (not
accounting for intra-specific variability) from either a global
([TRY](http://www.try-db.org/) data base[@kattge_try_2011] or local
-data base (extanded Table 1), for three key traits: wood density, SLA,
+data base (extended Table 1), for three key traits: wood density, SLA,
and Maximum height. Third, we computed the local abundance (measured
as the basal area - $m^2/ha$) of competitors per species in the
neighborhood of each tree. The neighborhood was defined as a 15m
@@ -98,7 +104,7 @@ we included biomes as a random effect in each parameters to estimate
the overall effect across all biomes and (ii) we included biomes as a
fixed effect to analyse the difference between biomes.
-**Supplementary Information line** is available in the online version of the paper.
+**Supplementary Information** is available in the online version of the paper.
**Acknowledgements**
We are thankful that people whose long term commitment allowed the established and
diff --git a/docs/paper/paper.txt b/docs/paper/paper.txt
index 97f93774d21bfac55be03474ba07f48962e92a61..862aea487d6b39f95950939a45f653bb7563d7ef 100644
--- a/docs/paper/paper.txt
+++ b/docs/paper/paper.txt
@@ -1,9 +1,10 @@
Competition is a fundamental type of interaction in ecological communities. Each individual modify their
-neighbouring environment and thus influences the performance of neighbouring individuals1 . Competition thus influences species composition and its changes over time. Maybe competition is especially
-important for vegetation on land because most vegetation types have high enough cover that shading and
-water and nutrient depletion are conspicuous. There have thus been a very large number of studies on
-competition among plants2 , but firm generalizations have yet to be established about its outcomes.
+neighbouring environment and thus influences the performance of neighbouring individuals1 . Competition influences species composition and its changes over time. Maybe competition is especially important
+for vegetation on land because most vegetation types have high enough cover that shading and water and
+nutrient depletion are conspicuous. There have thus been a very large number of studies on competition
+among plants2 , but firm generalizations have yet to be established about its outcomes.
+
When competition is described as interactions between pairs of species (as it traditionally has been), the
number of different interactions to be measured grows explosively with the number of species (N 2 ), and
becomes quickly intractable. Also this species-pair approach does not lead naturally to generalization
@@ -12,16 +13,13 @@ trees (in the sense of influence of neighbours on growth of a focal tree) within
novel in two important ways: (i) competition is modelled as a function of traits rather than of species
and (ii) we partition how traits drive the outcome of competition in four different key processes (Fig.
1). Competition can select trait values that are the most competitive. This competitive advantage of trait
-1
-
-
values can arise because (1) there are correlated with higher potential growth (in absence of competition)3 ,
(2) they are correlated with a higher tolerance to competition4 , or (3) they are correlated with higher
competitive effect5 . In contrast, competition can promotes the coexistence of a mixture of traits values,
-if (4) competition decrease with increasing dissimilarity of the traits of the competitor and the focal tree6 .
-These four processes are likely to be connected to the key traits used to describe plant strategies7,8 , however
-there is no agremenent on their relative contributions to each of these processes and whether the magnitude
-and direction of these effects are conserved across large scale.
+if (4) competition decrease with increasing dissimilarity of the traits of the competitors and the focal
+tree6 . These four processes are likely to be connected to the key traits used to describe plant strategies7,8 ,
+however there is no agremenent on their relative contributions to each of these processes and whether the
+magnitude and direction of these effects are conserved across large scale.
Here we dissect how three key traits9,10 (maximum height, wood density and specific leaf area - SLA) affect
these four processes involved in competition between trees using neighbouring modeling approach11 . We
compiled data of growth along side local abundance of their competitor for more than 7 million trees
@@ -32,64 +30,62 @@ each of the processes presented in Fig. 1.
Across all biomes we found that strongest drivers of individual growth was first-ranked the local abundance of competitors; and second ranked the direct influence of the focal plant’s traits on its growth (Fig.
2 Extended data Table 1). We detected only negative effect of the abundance of competitors showing
that competition was predominant. Among the three traits wood density had the strongest direct effect,
-followed by maximum height whereas SLA had no detectable effect (Fig. 2). Then our results show
-the influence of neighbour traits on their competitive impact, and of focal species traits on tolerance of
-competition (Fig. 2). Taken together these two effects are in the range of half or quarter as big as the
-direct trait effect (Extend data Table 2), down to zero influence depending on the trait (Fig 2). Finally,
-there is a small but consistent effect whereby the wider is the absolute trait separation between focal and
-neighbour species, the weaker is competitive suppression of growth (Fig. 2). This may arise because of
-negative density dependence arising for species with similar trait because of a higher load of herbivores
-or pathogens12 or lower sharing of ressources (such as a lower packing of crown in the canopy1 ). An
-analysis using a multiple-traits distance rather than a single trait distance didn’t show different pattern
-(supplementary Figure ?2 ) . An analysis allowing for different effect between biomes did no show strong
-evidence for any particular biome behaving consistently differently from the others (Fig. 4). The exception is the temperate biomes where SLA showed much stronger effect, probably ought to the dominance
-of deciduous species in this biome (Fig. 3).
+followed by maximum height whereas SLA had no detectable effect (Fig. 2). Then our results show the
+influence of neighbour traits on their competitive impact, and of focal species traits on tolerance of competition (Fig. 2). Taken together these two effects are in the range of half or quarter as big as the direct
+trait effect (Extend data Table 1), down to zero influence depending on the trait (Fig 2). Finally, there is a
+small but consistent effect whereby the wider is the absolute trait separation between focal and neighbour
+species, the weaker is competitive suppression of growth (Fig. 2). This may arise because of negative
+density dependence arising for species with similar trait because, for instance, a higher load of herbivores
+or pathogens12 or less efficient use of resources (such a less efficient packing of crown in the canopy1 or
+the root in the soil). An analysis using a multiple-traits distance rather than a single trait distance didn’t
+show different pattern (extended data Figure ?2 ). Analyses allowing for different effect between biomes
+did no show strong evidence for any particular biome behaving consistently differently from the others
+(Fig. 4). The exception is the temperate biomes where SLA showed much stronger effect, probably ought
+to the dominance of deciduous species in this biome (Fig. 3).
The direction of the traits effect agree well with the existing literature. High wood density was lined with
-slow potential growth rate but high tolerance to competition, in agreement with shade-tolerant species having high wood density3 . High wood density also had a tendency to a higher competitive effect, that may
-be related to deeper crown13,14 . There was a small postive direct effect of SLA on maximum growth, with
-wide confidence interval, which agree well with the weak correlation prevously reported for adult trees3 .
-Increasing SLA was aslo weakly related to decreased competitive effect and no effect or possible weakly
-decreased competitive tolerance, which agree with previous study reporting a weak negative correlation
-betwen SLA and shade tolerance3 . Finally, maximum height was positively related to maximum growth as
-previously reported3 . Species with small maximum height were also much more tolerant to competition
-than taller species, in line with the idea that subcanopy tree are more shade-tolerant. For wood density and
-SLA the link with competitive effect and competitive response was oppposed, in agrement with a coordinated selection under which trait value that confere high competitive effect also confer high competitive
-tolerance (a competitive hierachy15,16 ). This was not the case for maximum height because short species
-were more tolerant to competition but had a lower competitive effect. This match well the observation
-of the persistance of subcanopy species under a close cover of tall tree and the stratification theory of
-species coexistence17 . Finally our study show that trait values that favour tolerance to competition also
+slow potential growth rate but high tolerance to competition, in agreement with shade-tolerant species
+having high wood density3 . High wood density also resulted in a higher competitive effect, that may be
+related to deeper crown13,14 . The lack of direct effect of SLA on maximum growth (but with a positive
+tendency) agree well with the weak correlation prevously reported for adult trees3 . Increasing SLA was
+also weakly related to decreased competitive effect and no effect or possible weakly decreased competitive tolerance, which agree with previous study reporting a weak negative correlation betwen SLA and
+shade tolerance3 . Finally, maximum height was positively related to maximum growth as previously
-render species slow growing in absence of competition. Our results just demonstrate that the trait-based
-(particularly for wood density Fig. 4) trade-off between fast growth in the absence of competition and
-lower tolerance to competition is a global phenomenon common to all forested biomes. This important
-because this is one of the most classical process proposed to explain species coexistence in forest18 .
+reported3 . Species with small maximum height were also much more tolerant to competition than taller
+species, in line with the idea that subcanopy tree are more shade-tolerant. For wood density and SLA
+the link with competitive effect and competitive response was oppposed, in agrement with a coordinated
+selection under which trait value that confere high competitive effect also confer high competitive tolerance (a competitive hierachy15,16 ). This was not the case for maximum height because short species were
+more tolerant to competition but had a lower competitive effect. This match well the observation of the
+persistance of subcanopy species under a close cover of tall tree and the stratification theory of species
+coexistence17 . Finally our study show that trait values that favour tolerance to competition also render
+species slow growing in absence of competition. Our results just demonstrate that the trait-based tradeoff (particularly for wood density Fig. 4) between fast growth in the absence of competition and lower
+tolerance to competition is a global phenomenon common to all forested biomes. This important because
+this is one of the most classical process proposed to explain species coexistence in forest18 .
The globally consistent link between competition effect on growth and traits that we report is promising to
simplify the complex interaction governing forest communities. Our results also demonstrates that most
-assumptions about the link and traits competition that are use to tease out community assembly are too
-simplistic19 . Analysis for other fitness component (survival and recruitment) are now need to be able to
-scale up these short-term interactions to population dynamics impacts on traits composition.
+assumptions about the link between traits and competition that are use to tease out community assembly
+are too simplistic19 . Analysis for other fitness component (survival and recruitment) are now need to be
+able to scale up these short-term interactions to population dynamics impacts on traits composition.
Methods summary
To examine the link between competition and traits we first compiled forest plot data from both national
-forest inventory and long-term permanent plots from 14 countries covering all the major biomes of the
-world (Fig. 1b, see extanded Table 1); We restricted our analysis to tree with trunk diameter >= 10cm
-to have a common minimum size across all data set. Second, we extracted traits mean per species (not
-accounting for intra-specific variability, and computing the genus mean if species data were lacking) from
-either a global (TRY data base20 or local data base (extanded Table 1), for three key traits: wood density,
-SLA, and Maximum height. Third, we computed the local abundance (measured as the basal area - m2 /ha)
-of competitors per species in the neighborhood of each tree. The neighborhood was defined as a 15m radius
-around the focal tree in large plot with xy coordinates of all trees or as the whole plot for national forest
-inventory data that are based on small plot (the plot size correspond to a circular plot ranging from 10 to
-25m in radius). Fourth, for each traits, we computed four crowding index that where representing (i) the
-overall crowding irrespective of the species trait (as the sum of basal area of local competitors), (ii) the
-overall crowding times the trait of focal species, representing how traits affect tolerance to competition,
-(iii) the mean traits of the competitor weighted by their abundance (basal area), representing how traits
-affect the competitive effect, and (iv) the mean of the trait absolute distance between the focal tree and the
-competitor species weighted per their abundance (basal area), representing how the dissimilarity of trait
-affect competition. Finally, for each trait, we fitted a model estimating how individuals tree basal area
-growth was affected by the focal species traits (direct traits effect) and by these four crowding indexes,
-while accounting for size effect. To facilitate comparison between parameters and traits all traits and
-variable were standaridised to a SD of one. We report these standardized parameters for each traits. Two
-models were fitted, (1) we included biomes as a random effect in each parameters to estimate the overall
-effect across all biomes and (ii) we included biomes as a fixed effect to analyse the difference between
-biomes.
+forest inventories and long-term permanent plots from 14 countries covering all the major biomes of the
+world (Fig. 1b, see extanded methods Table 1). We restricted our analysis to tree with trunk diameter >=
+10cm to have a common minimum size across all data set. Second, we extracted traits mean per species
+(not accounting for intra-specific variability) from either a global (TRY data base20 or local data base
+(extanded Table 1), for three key traits: wood density, SLA, and Maximum height. Third, we computed
+the local abundance (measured as the basal area - m2 /ha) of competitors per species in the neighborhood
+of each tree. The neighborhood was defined as a 15m radius around the focal tree in large plot with xy
+coordinates of all trees or as the whole plot for national forest inventory data that are based on small plot
+(the plot size correspond to a circular plot ranging from 10 to 25m in radius). Fourth, for each traits,
+we computed for each individual, four crowding indices that where representing (i) the overall crowding
+irrespective of the species trait (as the sum of basal area of local competitors), (ii) the overall crowding
+times the trait of focal species, representing how traits affect tolerance to competition, (iii) the mean traits
+of the competitor weighted by their abundance (basal area), representing how traits affect the competitive
+effect, and (iv) the mean of the trait absolute distance between the focal tree and the competitor species
+weighted per their abundance (basal area), representing how the dissimilarity of trait affect competition.
+Finally, for each trait, we fitted a model estimating how individuals tree basal area growth was affected by
+the focal species traits (direct traits effect) and by these four crowding indexes, while accounting for size
+effect. To facilitate comparison between parameters and traits all traits and variable were standaridised
+to a SD of one. We report these standardized parameters for each traits. Two models were fitted, (i) we
+included biomes as a random effect in each parameters to estimate the overall effect across all biomes and
+(ii) we included biomes as a fixed effect to analyse the difference between biomes.
diff --git a/docs/results_knitr/quick.report.fixed.effect.R b/docs/results_knitr/quick.report.fixed.effect.R
index b99da2f03f16cfb35257c829cda7cb046a63d45c..fc21dfc671fcb9664826b9a6f01f4e2359b8a4ad 100644
--- a/docs/results_knitr/quick.report.fixed.effect.R
+++ b/docs/results_knitr/quick.report.fixed.effect.R
@@ -118,6 +118,11 @@ list.all.results.ecocode.id <-
list.all.results <-
readRDS.root('output/list.lmer.out.all.NA.no.log.rds')
+## No resampling
+list.all.results.s <-
+ readRDS.root('output/list.lmer.out.all.NA.simple.rds')
+
+
##+ check.convergence, echo = FALSE, results = 'hide'
@@ -134,10 +139,14 @@ vec.rel.grad.biomes <- sapply(list.all.results.biomes.id,
function(list.t) { max(abs(list.t[['relgrad']]))})
vec.rel.grad.NA <- sapply(list.all.results,
function(list.t) { max(abs(list.t[['relgrad']]))})
+vec.rel.grad.NA.s <- sapply(list.all.results.s,
+ function(list.t) { max(abs(list.t[['relgrad']]))})
vec.rel.grad.species < 0.001
vec.rel.grad.ecocode < 0.001
vec.rel.grad.biomes < 0.001
vec.rel.grad.NA < 0.001
+vec.rel.grad.NA.s < 0.001
+
## only one model with random ecocode for seed mass seems problematic
@@ -176,7 +185,7 @@ names.biomes <- c('Subtrop', 'Temp grassland', 'Medit', 'Temp forest',
pdf('../../figs/figres1.pdf',
height = 7, width = 14)
-plot.param(list.all.results ,
+plot.param(list.all.results.s ,
model = 'lmer.LOGLIN.ER.AD.Tf.r.ecocode.species',
traits = c('Wood.density', 'SLA', 'Max.height'),
param.vec = c("Tf","sumBn", "sumTfBn",
@@ -189,24 +198,26 @@ plot.param(list.all.results ,
col.vec = fun.col.pch.biomes()$col.vec,
pch.vec = fun.col.pch.biomes()$pch.vec,
names.bio = names.biomes ,
- xlim = c(-0.26, 0.17))
+ xlim = c(-0.26, 0.17),
+ data.type = 'simple')
dev.off()
-pdf('../../figs/figres2.pdf',
+pdf('../../figs/figres2s.pdf',
height = 7, width = 18)
-plot.param.biomes.fixed.elli(list.all.results,
- model = 'lmer.LOGLIN.ER.AD.Tf.fixed.biomes.species',
+plot.param.biomes.fixed.elli(list.all.results.s,
+ model = 'lmer.LOGLIN.ER.AD.Tf.fixed.biomes.species',
param.vec = c("Tf","sumBn", "sumTfBn",
"sumTnBn", "sumTnTfBn.abs"),
- param.names = c("Direct trait effect ",
- 'mean competition',
- 'Competive response x trait',
- 'Competive effect x trait',
- 'Competition & trait similarity'),
+ param.names = c("Direct trait effect ",
+ 'mean competition',
+ 'Competive response x trait',
+ 'Competive effect x trait',
+ 'Competition & trait similarity'),
col.vec = fun.col.pch.biomes()$col.vec,
pch.vec = fun.col.pch.biomes()$pch.vec,
- xlim = c(-0.4, 0.45) ,
- clim = clim)
+ xlim = c(-0.7, 0.45) ,
+ clim = clim,
+ data.type = 'simple')
dev.off()