In allanecology/BetaDivMultifun: Testing effects of multitrophic beta-diversity and land-use intensity (LUI) on ecosystem multifunctionality
Aim : assessing landscape variables and find appropriate variables to be included in analysis.
Requirements
library(corrplot)
add to analysis_nonpublic.R
landscape <- data.table::fread(paste(pathtodata, "/data_assembly/raw_data/31018_5_Dataset/31018_5_data.csv", sep=""))
landscape <- data.table(BEplotZeros(landscape, column = "EP_PlotID", plotnam = "EP"))
setnames(landscape, old = c("EP_PlotID", "EP"), new = c("oldEP", "Plotn"))
landscape2 <- data.table::fread(paste(pathtodata, "/data_assembly/raw_data/LandUse-31319_31369/LandUse/LandscapeMetrics_acrossYears_acrossScales_Long_UpdatedOct22.csv", sep=""))
landscape2 <- data.table(BEplotZeros(landscape2, column = "plot_id", plotnam = "EP"))
landscape2[, plot_id := NULL]
setnames(landscape2, old = c("EP"), new = c("Plotn"))
landscape2_long <- landscape2
landscape2[, (c("Land.te", "Land.pr")) := lapply(.SD, FUN = as.numeric), .SDcols = c("Land.te", "Land.pr")]
landscape2 <- melt(landscape2, id.vars = c("Plotn", "Year", "scale"))
landscape2 <- dcast(landscape2, Plotn ~ variable + Year + scale, value.var = "value", fill = NA)
Include new variables
- Landscape LUI :
- temporal LUI : delta LUI
- local LUI : LUI
- landscape LUI : to be done
- historic LUI : permanency
1 km radius whenever possible. Sensitivity with other radii.
landscape LUI
Various measures possible. Use PCA to determine.
- landscape LUI (2008 measures) :
- variables included in landscape-LUI index from BEF-Up ID 31018 and https://doi.org/10.1038/s41467-021-23931-1
- land-cover diversity
- forest cover and forest cover diversity (exclucde, as we do not use forest data)
maybe still include, because main source of colonisers?
- grassland cover
- grassland cover diversity
- annual crop cover [%]
- landscape composition (compositional diversity) : shannon diversity "Land.shdi"
- landscape connectivity : e.g. edge density "Land.te", the mean-euclidian nearest neighbor distance (which might work better than cohesion), or patchiness. --> use edge density because it's a measure for connectivity and includese some configuration aspect.
Years correlate?
Question 1 : Do the years correlate? Use Landscape configuration measures if yes.
If yes : we can use earlier years.
# landscape2_long[, .(Plotn, Year, scale, Land.shdi, Land.te)]
ylandscape2 <- landscape2[, .(
Land.shdi_2012_1000, Land.shdi_2015_1000, Land.shdi_2018_1000,
Land.te_2012_1000, Land.te_2015_1000, Land.te_2018_1000)]
m <- cor(ylandscape2, method = "spearman")
par(mfrow = c(1, 2))
corrplot(m, diag = F, order = "hclust", type = "lower")
corrplot(m, diag = F, order = "hclust", type = "lower", method = "number")
dev.off()
test <- merge(landscape2[, .(Plotn, Land.shdi_2012_1000)],
landscape[, .(Plotn, Shannon.1000)])
plot(test$Land.shdi_2012_1000, test$Shannon.1000) # correlation of 2008 and 2012 measure
rm(test)
Years strongly correlate. Decide for a year. The earlier the better, because some response variables were measured more recently. LUI is included from 2006 - 2016.
Use from 2012.
Radii correlate?
ylandscape2 <- landscape2[, .(
Land.shdi_2012_500, Land.shdi_2012_1000, Land.shdi_2012_1500, Land.shdi_2012_2000,
Land.te_2012_500, Land.te_2012_1000, Land.te_2012_1500, Land.te_2018_2000)]
m <- cor(ylandscape2, method = "spearman")
par(mfrow = c(1, 2))
corrplot(m, diag = F, order = "hclust", type = "lower")
corrplot(m, diag = F, order = "hclust", type = "lower", method = "number")
dev.off()
All positively correlate, between 0.45 and 0.95. larger radii cluster. 500m radius clusters rather across measures than across radii. Use 1000m based on ecological thinking rather than based on mathematical necessity. Will run sensitivity analysis.
PCA of landscape measures
Idea : Find 2-3 measures to include, representative of most important axes. Hypothesis : use grassland permanency, as well as a measure of "landscape LUI".
plandscape2 <- landscape2[, .(Plotn, Land.te_2012_1000)]
plandscape <- landscape[, .(Plotn, Shannon.1000, Arable.1000, Forest.1000, Grassland.perm.1000,
Grassland.1000)]
plandscape <- merge(plandscape2, plandscape, by = "Plotn")
rm(plandscape2)
# Plot information is taken as rownames of a data.frame (instead of data.table).
plandscape_rownames <- plandscape$Plotn
plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE])
rownames(plandscape) <- plandscape_rownames
rm(plandscape_rownames)
# run the pca
pca_landscape <- prcomp(plandscape, scale = T)
plot_grid(plotlist = list(ggplot(data.frame(PCs = paste0("PC", seq(1, 6)), var = (pca_landscape$sdev)^2),
aes(x = PCs, y = var)) + geom_bar(stat = "identity"),
autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T),
autoplot(pca_landscape, x = 2, y = 3, loadings = T, loadings.label = T),
autoplot(pca_landscape, x = 3, y = 4, loadings = T, loadings.label = T)
), nrow = 2)
# saved as "include_new_variables_landscape_pca_1000m.pdf"
grassland cover vs. grassland cover diversity
Sensitivity on other radius: 2 km
the choice of measures should not depend (too much) on the chosen radius.
# # # # #
# 2 km
# # # # #
plandscape2 <- landscape2[, .(Plotn, Land.te_2012_2000)]
plandscape <- landscape[, .(Plotn, Shannon.2000, Arable.2000, Forest.2000, Grassland.perm.2000,
Grassland.2000)]
plandscape <- merge(plandscape2, plandscape, by = "Plotn")
rm(plandscape2)
# Plot information is taken as rownames of a data.frame (instead of data.table).
plandscape_rownames <- plandscape$Plotn
plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE])
rownames(plandscape) <- plandscape_rownames
rm(plandscape_rownames)
# run the pca
pca_landscape <- prcomp(plandscape, scale = T)
plot_grid(plotlist = list(ggplot(data.frame(PCs = paste0("PC", seq(1, 6)), var = (pca_landscape$sdev)^2),
aes(x = PCs, y = var)) + geom_bar(stat = "identity"),
autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T),
autoplot(pca_landscape, x = 2, y = 3, loadings = T, loadings.label = T),
autoplot(pca_landscape, x = 3, y = 4, loadings = T, loadings.label = T)
), nrow = 2)
# saved as : include_new_variables_landscape_pca_2000m.pdf
# # # # #
# 500 m
# # # # #
plandscape2 <- landscape2[, .(Plotn, Land.te_2012_500)]
plandscape <- landscape[, .(Plotn, Shannon.500, Arable.500, Forest.500, Grassland.perm.500,
Grassland.500)]
plandscape <- merge(plandscape2, plandscape, by = "Plotn")
rm(plandscape2)
# Plot information is taken as rownames of a data.frame (instead of data.table).
plandscape_rownames <- plandscape$Plotn
plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE])
rownames(plandscape) <- plandscape_rownames
rm(plandscape_rownames)
# run the pca
pca_landscape <- prcomp(plandscape, scale = T)
plot_grid(plotlist = list(ggplot(data.frame(PCs = paste0("PC", seq(1, 6)), var = (pca_landscape$sdev)^2),
aes(x = PCs, y = var)) + geom_bar(stat = "identity"),
autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T),
autoplot(pca_landscape, x = 2, y = 3, loadings = T, loadings.label = T),
autoplot(pca_landscape, x = 3, y = 4, loadings = T, loadings.label = T)
), nrow = 2)
# saved as : include_new_variables_landscape_pca_500m.pdf
# # # # #
# all radii
# # # # #
plandscape2 <- landscape2[, .(Plotn, Land.te_2012_500, Land.te_2012_1000, Land.te_2012_2000)]
plandscape <- landscape[, .(Plotn, Shannon.500, Shannon.1000, Shannon.2000, Arable.500, Arable.1000,
Arable.2000, Forest.500, Forest.1000, Forest.2000,
Grassland.500, Grassland.1000, Grassland.2000)]
plandscape <- merge(plandscape2, plandscape, by = "Plotn")
rm(plandscape2)
# Plot information is taken as rownames of a data.frame (instead of data.table).
plandscape_rownames <- plandscape$Plotn
plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE])
rownames(plandscape) <- plandscape_rownames
rm(plandscape_rownames)
# run the pca
pca_landscape <- prcomp(plandscape, scale = T)
plot_grid(plotlist = list(
ggplot(data.frame(PCs = factor(paste0("PC", seq(1, length((pca_landscape$sdev)))), levels = paste0("PC", seq(1, length((pca_landscape$sdev))))),
var = (pca_landscape$sdev)^2),
aes(x = PCs, y = var)) + geom_bar(stat = "identity"),
autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T),
autoplot(pca_landscape, x = 2, y = 3, loadings = T, loadings.label = T),
autoplot(pca_landscape, x = 3, y = 4, loadings = T, loadings.label = T)
), nrow = 2)
# saved as : include_new_variables_landscape_pca_500_to_2000m.pdf
Sensitivity : not perfectly the same, but main directions. Arable vs. Land.te/ Shannon.
Lanscape covariate
Shannon diversity and edge density. Landscape diversity and landscape configuration. Measures are correlated.
plandscape2 <- landscape2[, .(Plotn, Land.te_2012_1000)]
plandscape <- landscape[, .(Plotn, Shannon.1000)]
plandscape <- merge(plandscape2, plandscape, by = "Plotn")
rm(plandscape2)
plot(plandscape$Land.te_2012_1000, plandscape$Shannon.1000)
# Plot information is taken as rownames of a data.frame (instead of data.table).
plandscape_rownames <- plandscape$Plotn
plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE])
rownames(plandscape) <- plandscape_rownames
rm(plandscape_rownames)
# run the pca
pca_landscape <- prcomp(plandscape, scale = T)
ggplot(data.frame(PCs = paste0("PC", seq(1, 2)), var = (pca_landscape$sdev)^2),
aes(x = PCs, y = var)) + geom_bar(stat = "identity")
autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T)
Higher distance means that the landscape around the plots is very differently structured, in terms of how many different land uses are there, and how large the patches are. E.g. high diversity would indicate many different land uses. High total edges would indicate that there are many small patches (because many small patches have more edges than few large patches).
Including variables
Historic LUI: Grassland perm 1000
will be included. PCA corroborated idea that historic land-use explains other dimension than present land-use.
histLUI <- landscape[, .(Plotn, Grassland.perm.1000)]
Landscape LUI : % Arable around plot
We know from study https://doi.org/10.1038/s41467-021-23931-1 that % arable land is correlated with landscape LUI. The measure is treated as "lanscape LUI" there.
Lansdcape covariates
- Shannon diversity
- total edges
Diversity of landscape use and total edge are correlated. Makes sense because the more diverse land-uses are present in a landscape, the more edges we can observe. Is not completely the same tough: total edge measures landscape fragmentation, the diversity measures the number and abundance of land-uses at the landscape scale.
If we take them separately, we can interpret separately. Higher edges means smaller patches. But actually some of the functions (e.g. Groundwater recharge) need more space --> better if fragementation is not too high. And some functions might profit from a diverse surrounding in terms of habitat, or not.
We take as separate covariates, because we don't only want to correct for them, but even might interpret their effects.
Check local correlations
plandscape2 <- landscape2[, .(Plotn, Land.te_2012_1000)]
plandscape <- landscape[, .(Plotn, Grassland.perm.1000, Arable.1000, Shannon.1000)]
plandscape <- merge(plandscape2, plandscape, by = "Plotn")
rm(plandscape2)
m <- cor(plandscape[, -1], method = "spearman")
corrplot(m, diag = F, order = "hclust", type = "lower")
corrplot(m, diag = F, order = "hclust", type = "lower", method = "number")
OK to include them in a model together (from a local perspective). Will be checked again at distance level.
allanecology/BetaDivMultifun documentation built on Nov. 9, 2023, 8:47 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Aim : assessing landscape variables and find appropriate variables to be included in analysis.
Requirements
library(corrplot)
add to analysis_nonpublic.R
landscape <- data.table::fread(paste(pathtodata, "/data_assembly/raw_data/31018_5_Dataset/31018_5_data.csv", sep="")) landscape <- data.table(BEplotZeros(landscape, column = "EP_PlotID", plotnam = "EP")) setnames(landscape, old = c("EP_PlotID", "EP"), new = c("oldEP", "Plotn")) landscape2 <- data.table::fread(paste(pathtodata, "/data_assembly/raw_data/LandUse-31319_31369/LandUse/LandscapeMetrics_acrossYears_acrossScales_Long_UpdatedOct22.csv", sep="")) landscape2 <- data.table(BEplotZeros(landscape2, column = "plot_id", plotnam = "EP")) landscape2[, plot_id := NULL] setnames(landscape2, old = c("EP"), new = c("Plotn")) landscape2_long <- landscape2 landscape2[, (c("Land.te", "Land.pr")) := lapply(.SD, FUN = as.numeric), .SDcols = c("Land.te", "Land.pr")] landscape2 <- melt(landscape2, id.vars = c("Plotn", "Year", "scale")) landscape2 <- dcast(landscape2, Plotn ~ variable + Year + scale, value.var = "value", fill = NA)
Include new variables
- Landscape LUI :
- temporal LUI : delta LUI
- local LUI : LUI
- landscape LUI : to be done
- historic LUI : permanency
1 km radius whenever possible. Sensitivity with other radii.
landscape LUI
Various measures possible. Use PCA to determine.
- landscape LUI (2008 measures) :
- variables included in landscape-LUI index from BEF-Up ID 31018 and https://doi.org/10.1038/s41467-021-23931-1
- land-cover diversity
- forest cover and forest cover diversity (exclucde, as we do not use forest data) maybe still include, because main source of colonisers?
- grassland cover
- grassland cover diversity
- annual crop cover [%]
- landscape composition (compositional diversity) : shannon diversity "Land.shdi"
- landscape connectivity : e.g. edge density "Land.te", the mean-euclidian nearest neighbor distance (which might work better than cohesion), or patchiness. --> use edge density because it's a measure for connectivity and includese some configuration aspect.
Years correlate?
Question 1 : Do the years correlate? Use Landscape configuration measures if yes.
If yes : we can use earlier years.
# landscape2_long[, .(Plotn, Year, scale, Land.shdi, Land.te)] ylandscape2 <- landscape2[, .( Land.shdi_2012_1000, Land.shdi_2015_1000, Land.shdi_2018_1000, Land.te_2012_1000, Land.te_2015_1000, Land.te_2018_1000)] m <- cor(ylandscape2, method = "spearman") par(mfrow = c(1, 2)) corrplot(m, diag = F, order = "hclust", type = "lower") corrplot(m, diag = F, order = "hclust", type = "lower", method = "number") dev.off() test <- merge(landscape2[, .(Plotn, Land.shdi_2012_1000)], landscape[, .(Plotn, Shannon.1000)]) plot(test$Land.shdi_2012_1000, test$Shannon.1000) # correlation of 2008 and 2012 measure rm(test)
Years strongly correlate. Decide for a year. The earlier the better, because some response variables were measured more recently. LUI is included from 2006 - 2016.
Use from 2012.
Radii correlate?
ylandscape2 <- landscape2[, .( Land.shdi_2012_500, Land.shdi_2012_1000, Land.shdi_2012_1500, Land.shdi_2012_2000, Land.te_2012_500, Land.te_2012_1000, Land.te_2012_1500, Land.te_2018_2000)] m <- cor(ylandscape2, method = "spearman") par(mfrow = c(1, 2)) corrplot(m, diag = F, order = "hclust", type = "lower") corrplot(m, diag = F, order = "hclust", type = "lower", method = "number") dev.off()
All positively correlate, between 0.45 and 0.95. larger radii cluster. 500m radius clusters rather across measures than across radii. Use 1000m based on ecological thinking rather than based on mathematical necessity. Will run sensitivity analysis.
PCA of landscape measures
Idea : Find 2-3 measures to include, representative of most important axes. Hypothesis : use grassland permanency, as well as a measure of "landscape LUI".
plandscape2 <- landscape2[, .(Plotn, Land.te_2012_1000)] plandscape <- landscape[, .(Plotn, Shannon.1000, Arable.1000, Forest.1000, Grassland.perm.1000, Grassland.1000)] plandscape <- merge(plandscape2, plandscape, by = "Plotn") rm(plandscape2) # Plot information is taken as rownames of a data.frame (instead of data.table). plandscape_rownames <- plandscape$Plotn plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE]) rownames(plandscape) <- plandscape_rownames rm(plandscape_rownames) # run the pca pca_landscape <- prcomp(plandscape, scale = T) plot_grid(plotlist = list(ggplot(data.frame(PCs = paste0("PC", seq(1, 6)), var = (pca_landscape$sdev)^2), aes(x = PCs, y = var)) + geom_bar(stat = "identity"), autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T), autoplot(pca_landscape, x = 2, y = 3, loadings = T, loadings.label = T), autoplot(pca_landscape, x = 3, y = 4, loadings = T, loadings.label = T) ), nrow = 2) # saved as "include_new_variables_landscape_pca_1000m.pdf"
grassland cover vs. grassland cover diversity
Sensitivity on other radius: 2 km
the choice of measures should not depend (too much) on the chosen radius.
# # # # # # 2 km # # # # # plandscape2 <- landscape2[, .(Plotn, Land.te_2012_2000)] plandscape <- landscape[, .(Plotn, Shannon.2000, Arable.2000, Forest.2000, Grassland.perm.2000, Grassland.2000)] plandscape <- merge(plandscape2, plandscape, by = "Plotn") rm(plandscape2) # Plot information is taken as rownames of a data.frame (instead of data.table). plandscape_rownames <- plandscape$Plotn plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE]) rownames(plandscape) <- plandscape_rownames rm(plandscape_rownames) # run the pca pca_landscape <- prcomp(plandscape, scale = T) plot_grid(plotlist = list(ggplot(data.frame(PCs = paste0("PC", seq(1, 6)), var = (pca_landscape$sdev)^2), aes(x = PCs, y = var)) + geom_bar(stat = "identity"), autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T), autoplot(pca_landscape, x = 2, y = 3, loadings = T, loadings.label = T), autoplot(pca_landscape, x = 3, y = 4, loadings = T, loadings.label = T) ), nrow = 2) # saved as : include_new_variables_landscape_pca_2000m.pdf # # # # # # 500 m # # # # # plandscape2 <- landscape2[, .(Plotn, Land.te_2012_500)] plandscape <- landscape[, .(Plotn, Shannon.500, Arable.500, Forest.500, Grassland.perm.500, Grassland.500)] plandscape <- merge(plandscape2, plandscape, by = "Plotn") rm(plandscape2) # Plot information is taken as rownames of a data.frame (instead of data.table). plandscape_rownames <- plandscape$Plotn plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE]) rownames(plandscape) <- plandscape_rownames rm(plandscape_rownames) # run the pca pca_landscape <- prcomp(plandscape, scale = T) plot_grid(plotlist = list(ggplot(data.frame(PCs = paste0("PC", seq(1, 6)), var = (pca_landscape$sdev)^2), aes(x = PCs, y = var)) + geom_bar(stat = "identity"), autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T), autoplot(pca_landscape, x = 2, y = 3, loadings = T, loadings.label = T), autoplot(pca_landscape, x = 3, y = 4, loadings = T, loadings.label = T) ), nrow = 2) # saved as : include_new_variables_landscape_pca_500m.pdf # # # # # # all radii # # # # # plandscape2 <- landscape2[, .(Plotn, Land.te_2012_500, Land.te_2012_1000, Land.te_2012_2000)] plandscape <- landscape[, .(Plotn, Shannon.500, Shannon.1000, Shannon.2000, Arable.500, Arable.1000, Arable.2000, Forest.500, Forest.1000, Forest.2000, Grassland.500, Grassland.1000, Grassland.2000)] plandscape <- merge(plandscape2, plandscape, by = "Plotn") rm(plandscape2) # Plot information is taken as rownames of a data.frame (instead of data.table). plandscape_rownames <- plandscape$Plotn plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE]) rownames(plandscape) <- plandscape_rownames rm(plandscape_rownames) # run the pca pca_landscape <- prcomp(plandscape, scale = T) plot_grid(plotlist = list( ggplot(data.frame(PCs = factor(paste0("PC", seq(1, length((pca_landscape$sdev)))), levels = paste0("PC", seq(1, length((pca_landscape$sdev))))), var = (pca_landscape$sdev)^2), aes(x = PCs, y = var)) + geom_bar(stat = "identity"), autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T), autoplot(pca_landscape, x = 2, y = 3, loadings = T, loadings.label = T), autoplot(pca_landscape, x = 3, y = 4, loadings = T, loadings.label = T) ), nrow = 2) # saved as : include_new_variables_landscape_pca_500_to_2000m.pdf
Sensitivity : not perfectly the same, but main directions. Arable vs. Land.te/ Shannon.
Lanscape covariate
Shannon diversity and edge density. Landscape diversity and landscape configuration. Measures are correlated.
plandscape2 <- landscape2[, .(Plotn, Land.te_2012_1000)] plandscape <- landscape[, .(Plotn, Shannon.1000)] plandscape <- merge(plandscape2, plandscape, by = "Plotn") rm(plandscape2) plot(plandscape$Land.te_2012_1000, plandscape$Shannon.1000) # Plot information is taken as rownames of a data.frame (instead of data.table). plandscape_rownames <- plandscape$Plotn plandscape <- as.data.frame(plandscape[, !"Plotn", with=FALSE]) rownames(plandscape) <- plandscape_rownames rm(plandscape_rownames) # run the pca pca_landscape <- prcomp(plandscape, scale = T) ggplot(data.frame(PCs = paste0("PC", seq(1, 2)), var = (pca_landscape$sdev)^2), aes(x = PCs, y = var)) + geom_bar(stat = "identity") autoplot(pca_landscape, x = 1, y = 2, loadings = T, loadings.label = T)
Higher distance means that the landscape around the plots is very differently structured, in terms of how many different land uses are there, and how large the patches are. E.g. high diversity would indicate many different land uses. High total edges would indicate that there are many small patches (because many small patches have more edges than few large patches).
Including variables
Historic LUI: Grassland perm 1000
will be included. PCA corroborated idea that historic land-use explains other dimension than present land-use.
histLUI <- landscape[, .(Plotn, Grassland.perm.1000)]
Landscape LUI : % Arable around plot
We know from study https://doi.org/10.1038/s41467-021-23931-1 that % arable land is correlated with landscape LUI. The measure is treated as "lanscape LUI" there.
Lansdcape covariates
- Shannon diversity
- total edges
Diversity of landscape use and total edge are correlated. Makes sense because the more diverse land-uses are present in a landscape, the more edges we can observe. Is not completely the same tough: total edge measures landscape fragmentation, the diversity measures the number and abundance of land-uses at the landscape scale.
If we take them separately, we can interpret separately. Higher edges means smaller patches. But actually some of the functions (e.g. Groundwater recharge) need more space --> better if fragementation is not too high. And some functions might profit from a diverse surrounding in terms of habitat, or not.
We take as separate covariates, because we don't only want to correct for them, but even might interpret their effects.
Check local correlations
plandscape2 <- landscape2[, .(Plotn, Land.te_2012_1000)] plandscape <- landscape[, .(Plotn, Grassland.perm.1000, Arable.1000, Shannon.1000)] plandscape <- merge(plandscape2, plandscape, by = "Plotn") rm(plandscape2) m <- cor(plandscape[, -1], method = "spearman") corrplot(m, diag = F, order = "hclust", type = "lower") corrplot(m, diag = F, order = "hclust", type = "lower", method = "number")
OK to include them in a model together (from a local perspective). Will be checked again at distance level.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.