In this post Jenny McCune discusses her recent paper ‘Species distribution models predict rare species occurrences despite significant effects of landscape context‘
According to the recent State of the World’s Plants report, there are about 391,000 known species of plants, one in five are at risk of extinction, and approximately 2,000 new plant species are identified each year – mainly from remote areas in the tropics. In the highly populated, well-explored forests of temperate North America, one might think that all our plants have been located, counted, and extensively studied. But that is not the case! While it is less likely that new species of plants will be found in the forests of temperate North America, our knowledge of the true status of many of our rare forest plants is incomplete. We’re not even sure where all the populations are, not to mention how they are fluctuating over time, or whether they are viable in the long term.
Why is this the case? Southern Ontario, Canada, is a great example. The landscape is made up of fragments of forest habitat in a sea of agricultural and residential land use. More than 80% of forest land is privately owned, and most landowners aren’t familiar with rare plants. Over 700 rare plant species are considered at risk or potentially at risk, but there are not enough botanists – or money – for the extensive fieldwork needed to survey for new populations of rare plants.
Finding plants with computer models
One solution is to prioritize sites to survey using species distribution models (or SDMs) to indicate suitable habitat. These models use the locations of known populations to predict other sites with similar climate, topography, and soils. However, other factors can affect whether or not a rare plant is found at a site, even if the climate, topography, and soils are perfect. For example, if the forest is too small or disturbed, or if it is too far away from a seed source, the plant may be missing even though everything else is just right.
I tested whether SDMs could help find new populations of 8 rare plant species. They varied in their range (some range mainly in southern Ontario, while others are at their northern limit here), their habitat specificity (some specialize on a narrow range of habitats, while others are generalists), and rarity on the landscape. For each species, I used SDMs to predict suitable habitat based only on climate, topography, and soils. Then I tested the models by surveying some of the sites predicted suitable, and recording all the plant species present at each site. I used the plant community data to ask whether the models were at least predicting the right habitat type, even if the rare plant was not found. Then, I looked at the importance of the amount of forest on the surrounding landscape, and the distance to the nearest known population, for predicting whether the rare species would be present.
For 7 of the 8 species, SDMs were a useful way to prioritize sites to search. I discovered new populations of 4 of the 8 species. A new population of a rare plant was more likely to be found at a site that the SDM predicted to have suitable habitat. So these models provide a good first cut of where to look for new populations.
The data on other plants at each site highlighted important factors that were not incorporated into the model. For example, “suitable” sites with a high abundance of white-cedar in the canopy never contained the rare hart’s-tongue fern – suggesting that canopy type is an important factor for that species.
Most importantly, the condition of the landscape surrounding a site influenced the likelihood of finding a rare species. Sites predicted to have suitable habitat were more likely to contain the rare plant if there was more forest within 500m of the site. So – while using SDMs can narrow down the search area, success is even more likely in sites with more forest on the surrounding landscape. In addition, the majority of new rare plant populations were found less than 5km away from a known population of the same species. This fits with the hypothesis that many of these rare forest plants have limitations to their dispersal that prevents them from accessing all the suitable habitat on this landscape.
Although plants make up a large proportion of species at risk in North America, compared to other groups like birds and mammals we are far behind in our understanding of what limits rare plants and how to help them. I found that SDMs are a relatively simple and useful way to prioritize field sites to survey for new populations of rare plants – even where the habitat is highly fragmented. However, the state of the surrounding landscape also makes a difference for whether or not these elusive plants are likely to be found.
2 thoughts on “Finding rare plants in forest fragments – species distribution models help, and landscape context matters”
Reblogged this on isabelmarquesevolution and commented:
A very cool post by The Applied Ecologist’s blog