OCN_to_AEM: Construct asymmetric eigenvector maps (AEM) from an OCN

Description

Function that computes asymmetric eigenvector maps from an OCN. These can be used as spatial variables to assess spatial gradients in environmental or ecological data.

Usage

OCN_to_AEM(OCN, level = "AG", weight = NULL, resistance = "length", moranI = FALSE)

Value

A list as produced by a call to aem. If moranI = TRUE, a krandtest resulting from the call to moran.randtest is appended to the output list.

Arguments

OCN: A river object as produced by aggregate_OCN.
level: Aggregation level at which AEMs are to be calculated. It must be equal to either "RN" or "AG".
weight: Determines how and if weights should be used to compute the AEMs. Defaults to NULL, which applies equal weights to all links of the OCN. It can be one of "gravity", "exponential", "linear", "parabolic" or a user-specified function. See details.
resistance: Identifies how resisitance (i.e., the variable negatively related to the link weight) is calculated. Defaults to "length" (i.e., resistance equal to link length). Alternatively, resistance = "time" uses water travel time to weigh the links. In such a case, water velocities must be contained in the OCN (e.g. via rivergeometry_OCN).
moranI: Logical. Should Moran's I statistics be computed and random tests be performed via moran.randtest?

Details

Possible character strings for weight are:

"gravity": \(w(r) = r_{max}/r\)
"exponential": \(w(r) = \exp(-r/r_{max})\)
"linear": \(w(r) = 1 - r/r_{max}\)
"parabolic": \(w(r) = 1 - (r/r_{max})^2\)

Examples

Run this code

OCN <- aggregate_OCN(landscape_OCN(OCN_20), thrA = 5)

res <- OCN_to_AEM(OCN) # unweighted AEMs
res$values # eigenvectors associates with the AEMs
plot(OCN, res$vectors[,1], drawNodes = TRUE, 
colLevels = c(-max(abs(res$vectors[,1])), max(abs(res$vectors[,1])), 100), 
	colPalette = hcl.colors(100,"Blue-Red 2")) # plot first eigenvector

res_g <- OCN_to_AEM(OCN, weight = "gravity") # weighted AEMs based on gravity model

fn <- function(r) {1 - r^0.5} 
res_f <- OCN_to_AEM(OCN, weight = fn) # weighted AEMs based on user-specified weight function

# compute Moran's I and perform permutation test to assess which eigenfunctions should be retained
# \donttest{
res_g <- OCN_to_AEM(OCN, weight = "gravity", moranI = TRUE)
selectedAEM <- which(res_g$moranI$pvalue < 0.05) 
# selected eigenfunctions are those with significantly positive spatial autocorrelation

# plot selected eigenfunctions 
# (these could be e.g. used as spatial covariates in a species distribution model)
par(mfrow=c(3,4))
for (i in selectedAEM){
plot(OCN, res$vectors[,i], drawNodes = TRUE, 
	colLevels = c(-max(abs(res$vectors[,i])), max(abs(res$vectors[,i])), 100), 
	colPalette = hcl.colors(100,"Blue-Red 2"))
	title(paste0("AEM",i))
	}
# }