Global Abiotic Edge Effects

The edges of forests are places of great dynamism. As the intersection of a forest and another habitat type, they may support species familiar to either habitat type and often attract highly adaptable ‘generalist’ species. Edges have been researched heavily as a site of increased stress and extinction for species which thrive in a forest’s interior, protected space due to increased heat, light, and dryness.

Despite the assertions that edges are hotter, drier, and brighter that have persisted over decades, the data has not been collated to determine how far, or where, this holds true. Below is an excerpt of the thesis project conducted by Elan Simon Parsons which combines data points from 30+ years of ecological edge research to characterize so-called “edge effects” in forests and forest fragments on abiotic (non-living) features of the environment.

Fully documented code may be found on GitHub, and all raw data has been stored at Open Science Framework.

To catch all studies potentially including environmental data of edge effects per se or from fragmentation, I conducted a search in March 2017 of Web of Science with the below keyphrase:

(“forest fragment*" or forest “edge eff“) AND (”soil (temperatur or moisture)” or “light intensit*" or “(air or ambient) temperatur" or humidity or “wind speed” or precip)

returning 205 results whose abstracts (and, if necessary, texts) were scanned to determine whether the study actually fell into these parameters. The cited and cited-by lists of all 205 were similarly scanned and collected from; only studies which explicitly measured at least one environmental variable (light, temperature, moisture, or air), rather than relying on weather station data or historical average values, were retained.

After this filtering, 71 studies remained. To examine the scientific methodology, results, and reach of each experiment, I gathered information from each paper concerning the geographic location of the study, its topical focus, the layout of sampling techniques, the environmental variables studied, the equipment used to take environmental data, the temporal dimension of sampling, kinds of statisical analyses done (if any), general results, and how many times the paper had been cited on Web of Science.

In total, 18 unique variables were found across studies. To have appropriate replication to find trends in the numerical data, I chose to focus on variables which were recorded by >10 studies:

air temperature (AT, n = 65)

relative humidity (RH, n = 53)

vapor pressure deficit (VPD, n = 21)

soil temperature (ST, n = 19)

soil moisture (SM, n = 18)

photosynthetically active radiation (PAR, n = 11)

wind speed (WS, n = 11)

To merge the 71 data files, use bash command:

cat ./Directory/*.csv > mergeddata.csv

And utilize OpenRefine to remove additional headers, fix minor errors, and remove blank spaces. Results in file mergedrefined7.csv.


Edge effects have not been studied in most countries, and studies have been concentrated in Brazil, the U.S., Australia, and Mexico.


Only about half of the 71 studies had data (in the format of an abiotic measurement per a distance from an edge/interior) that could be extracted from tables or using Web Plot Digitizer. Very little of the total research concentrates on boreal forests; because of this, we have no true idea of what edge effects look like in these areas. However, combined data from all three biomes show clear patterns:


LOESS curves, above as the grey line in each plot, indicate that there is at least some change in most environmental measurements between the edge (x=0) and more interior forest space. Interior conditions are shown at y=0, so where the curve passes near this point, that area is likely indistinguishable from a forest interior. Light levels experience the most change, followed by wind speed, vapor pressure deficit (VPD), air temperature, soil temperature, and relative humidity. Soil moisture shows an undulating pattern that only seems to stabilize past 200m into a forest.

Here, represented as heat maps:


Based roughly on the method of Laurance et al. in the work “Effects of forest fragmentation on recruitment patterns in Amazonian tree communities”, I ran linear mixed models against a 95% confidence interval of interior measurements for each environmental variable. Anything outside of this uncertainty, this confidence interval of the interior, thus signals an edge effect.


All variables and distances were transformed with log1n (natural log + 1) to linearize.

Following the logic of the above method, air temperature, relative humidity, soil temperature, and soil moisture do not show significant edge effects globally. VPD was higher than the interior up to 25 meters, light levels were higher only up to 3 meters, and wind speed was increased only up to 7 meters.

Based on my data, abiotic edge effects are short-lived within a forest if they exist at all. This does not, however, imply that edge effects generally are not meaningful or potentially hazardous to plant and animal species – edge habitat is growing rapidly worldwide due to human settlement and extraction efforts, and many species cannot thrive in these spaces. My results simply provide an interesting dimension to edge effects research, where focus is primarily on impacts to species and communities.