Skip to content

When you choose to publish with PLOS, your research makes an impact. Make your work accessible to all, without restrictions, and accelerate scientific discovery with options like preprints and published peer review that make your work more Open.

PLOS BLOGS The Official PLOS Blog

Average Distance to the Nearest Forest is Increasing with Forest Loss in the US

During the 1990s, over 90,000 square kilometers of forest cover in the continental United States were lost—an approximate decline in coverage of nearly 3% or an area roughly equivalent to the size of Maine. Primarily this lost in forest cover was due to the disappearance of key connecting forest patches. According to a new study published in PLOS ONE by Sheng Yang and Giogros Mountrakis from the State University of New York College of Environmental Science and Forestry, the average forest distance increased by over 500 m during this time.


Forest attrition is the removal of clustered or isolated forest patches and results in habitat loss, a decline in animal and plant populations, a decrease in the number of species in a given region, and changes in regional environmental conditions. Forest loss can have localized effects on weather and climate—altering both seasonal and annual precipitation as well as affecting soil and air temperatures.


Forest canopies not only have a direct cooling effect from shading, but also through transpiration—the movement of water from the soil, through plants, to the atmosphere. As air moves over forested areas, it picks up moisture and cools. When forests are removed, this process is severely altered, resulting in drier air, lower rainfall, and typically higher temperatures. Over 30% of the continental United States is covered by forest and declines in forest cover would have profound ecological and economic effects.


While many landscape studies of forest change focus on fragmentation, Yang and Mountrakis have created a new metric to quantify geographic forest change—forest attrition distance, or FAD. FAD is calculated as the average distance from non-forest and forest areas to the nearest forests in a landscape. National Land Cover Database maps of binary forest cover were used to calculate FAD for the US.


Higher FAD values indicate larger patches of non-forested areas. Yang and Mountrakis claim that FAD is superior to previous metrics based on similar distance calculations (e.g. GISfrag and mean distance to forest edge). The argument for FAD is that its derivation includes distances between forested and non-forested areas instead of just measuring forest to forest. FAD would therefore allow for a more “robust” measure of the evolution of the entire landscape. FAD is also more easily interpreted by a non-technical audience. FAD increased by over 500 m during the 1990s—or more plainly, on average, to get to a forest you would have to walk 500 m longer at the end of the decade as you had to at the beginning. That’s a pretty straight-forward visual that drives home the magnitude of the change.


One thing that I like about this paper is its explicit consideration of what forest change looks like. What do I mean by that? The first figure, shown below, offers seven different ways that forest cover can change. We have example A, our control, which gives the initial forest state. Example B and D show “aggregated attrition” or the loss of complete intact sections—say, you have a large pizza and someone grabbed a giant slice out of it. Examples C and E show equal decreases in area as B and D, but are more random—the term used here is “dispersed attrition.” Think more like one of those thin crust pizzas where you can grab pieces out of the middle.

Figure One, from Yang and Mountrakis (2017).
Figure One, from Yang and Mountrakis (2017). Case (a) shows the initial forest state followed by seven forest change examples (b-h).Forest area is depicted in green pixels. Red squares with doted lines indicate forest loss. Values of forest attrition distance and other relative metrics can be found in Table One (Yang and Mountrakis, 2017).

Upon first consideration, it might be assumed that a loss of area is a loss of area. However, there can be stark differences given the geographic or geometric pattern of forest loss. Aggregated attrition, as in examples B and D where contiguous sections of the forest are lost create “less fragmented” forests than do examples C and E, dispersed attrition. Forest fragmentation is classically measured using perimeter to area ratios. An increase in perimeter when the area of the forest remains the same, results in a higher perimeter to area ratio, and therefore more fragmentation. However, while the examples of aggregated versus dispersed attrition show the same losses in total forest cover, FAD increases substantially more with aggregated attrition, though forest fragmentation as measured by perimeter to area would be higher in in the dispersed attrition examples. The argument here would be that fragmentation may not necessarily tell the entire story and that FAD offers different information of potential use.


More interestingly, Yang and Mountrakis also consider more complex forest change regimes–shrinkage, subdivision, and perforation, shown in examples F, G, and H respectively. Figure one shows how these more complex forest cover loss regimes can play out spatially. In this examples, shrinkage, subdivision and perforation all have the same forest cover loss as examples D and E, but exhibit much higher edge density, lower average patch size, and higher perimeter to area ratios. But, reduced FAD as compared to aggregate attrition. Yang and Mountrakis make the claim that FAD is the only landscape metric that can accurately and quantitatively differentiate between all the considered scenarios, making it of potentially novel and significant importance in land use and land cover change studies.

And if you want to stick with that ill-advised pizza metaphor, imagine someone who only eats crusts or just takes bites out of random pieces and puts them back. Hopefully no one eats pizza like that, though. Ugh.


Yang and Mountrakis delve into the connection between forest cover and forest attrition. They find a non-linear relationship where initially, as forest cover decreases, forest attrition increases slowly. As forest cover decreases past an apparent threshold of 20-30% total cover, forest attrition increases exponentially.


Rural areas, rather than urban areas, are suffering higher rates of forest attrition and greater increases in FAD. Of particular note are the high rates of forest attrition on federal and state lands, highlighting a need for more focused management and conservation strategies that consider forest attrition and adequately take counter-steps.


Forest attrition losses in the western US are driven by losses in gap areas, or more isolated patches of forest. In the eastern US, forest losses are primarily in the forest interior or near the edge of the forest, resulting in lower attrition. Yang and Mountrakis attribute this difference in part to lower tree densities and higher elevation and terrain variance in the west. There are also different disturbance regimes and stressors on western forests than their eastern cousins. Western forests are more likely affected by fires or insect outbreaks. While eastern forests may be more heavily and intensively managed. Not explicitly considered here is the role of moderate disturbance. Western forests can experience stand-clearing fires. Fires in eastern forests still occur, but can be incredibly localized—often involving arson or human malfeasance—with results that are rarely stand-clearing. Eastern forests are also typically closer to habitation and this most certainly plays a role. Eastern forests also experience pressures from pest and insects including gypsy moth and eastern ash borer. Again, these disturbances are much more moderate in scope than those facing western forests–typically.

Yang and Mountrakis (2017).

Yang and Mountrakis also claim that forest attrition results in irreversible carbon losses compared to other geographic patterns of forest loss. I personally find myself divided on this claim, though it does point to an interesting research question—how does carbon cycling at the ecosystem or landscape scale respond to changes in FAD? Forest cover loss results in more than just the loss of trees. Soils are exposed to direct sunlight, resulting in lower soil moisture and higher soil temperatures that in turn affect decomposition and soil respiration—the production and emission of carbon dioxide from soils by microbes and roots. There would be complex effects on forest microclimates given the varying geographic and geometric shape of forest change. The magnitude of differences may be an open question, though from a conservation or climate mitigation standpoint, it is an important one.


Yang and Mountrakis produce an interesting new metric here that has potential for conservation and land use change work, and also for applied and basic research. But as with any novel technique, time is the ultimate judge—or in some cases, data availability. Application of this metric at more broad scales may be possible with high enough resolution global forest cover maps. Many sensitive areas around the globe, from Indonesia to the Amazon, are experiencing high rates of deforestation. If adequate data are available, FAD may be useful tool. Hopefully in the future we won’t have to walk so far to reach the forest.


Back to top