Ecosystem impact of invasive species

Biological invasions not only affect biodiversity and ecosystem structure they also impact ecosystem functioning via changes in biochemical pools and nutrient cycling, threatening the provision of ecosystem services (Ehrenfeld et al. 2010). Observational and experimental studies have already shown the effect of plant invasions (Vitousek et al. 1987, Liao et al. 2008) and the challenge now is to understand how invasions affect the ecosystems from a large scale perspective and how this knowledge can be used to inform management and conservation decisions.

Now, using plant functional traits as key indicators of ecosystem functioning (Chapin 2003) in combination with newly developed remote sensing tools to measure and predict such traits (Homolova et al. 2013), we can finally evaluate the impact of biological invasions, scale it from individuals to landscape and, generate spatially explicit assessments of their risks and consequences.

In this tutorial we will explore the use of hyperspectral data in combination with ground data (describing a selected set of biochemical leaf traits) of native tree species to assess the impact of an invasive shrub (Prunus serotina) on Phosphorus and Nitrogen foliar concentrations of the native vegetation.

The Dataset
The method and dataset presented here are thoroughly explained in Ewald et al. (in prep), a publication of the DIARS project. The dataset consists of an artificial hyperspectral image (click HERE to learn more about the artificial image) composed of a series of 150 field plots and 460 randomly chosen background plots accompanied by a table with P and N values for each of the field plots and the percentage cover of the invasive shrub Prunus serotina. You can download the data HERE, and remember to save it in the same directory you will use for your analysis.

The method

Lets start by installing and loading the packages with the commands for image processing and (raster and rgeos) and modeling (autopls) and, set the working directory:

# Run this command to install packages
# install.packages('raster','rgeos','autopls')

library(raster)
library(rgeos)
library(autopls)

# Set working directory
setwd("/home/garzonc/Desktop/DIARS/")

Once the working directory is set we can import the hyperspectral image and apply a brightness normalization which reduces the effects of shadows in the image. After that process is finished, we must aggregate the resulting image from its original size 3x3 meters per pixel to the new 12x12 meters per pixel resolution.

# import hyperspectral image
r.hs <- stack("mosaicCo_datasetComp.tiff")
# apply brightness normalisation
r.hs.bn <- prepro(r.hs, method = "bn")
# aggregate raster (using a factor of 4) for the prediction maps later.
r.hs.bn.agg <- aggregate(r.hs.bn, fact = 4)

Let’s import the field data. The first table contains the invasive species cover, recorded for each forest layer (herb, shurb and tree) and the geographic location of the plots (grid_allData1.csv). The second table contains the biochemical foliar measurements which consist of community weighted mean values of leaf nitrogen and phosphorous content for each plot (leafNP.csv).

# read field data
plotdata <- read.csv("grid_allData1.csv")
# plot band 100 of the hyperspectral image
plot(r.hs, (100), legend = FALSE)
# add the plot locations to the figure
points(plotdata$xgrid, plotdata$ygrid, pch = 3, cex = 0.5)