In the second of our series of blogs to accompany the recently-published Special Feature, Toward prediction in the restoration of biodiversity, Katharine Stuble describes her work in Every restoration is unique: testing year effects and site effects as drivers of initial restoration trajectories.

Restoration practitioners are acutely aware that the outcomes of their efforts are strongly contingent on a slew of factors.  Some are well understood, while others affect restoration success in far less well-understood (or not even identified) ways. For example, species selection, control of exotic species, site preparation, and planting method can all play important roles in driving restoration success and ultimately community composition, and do so in ways that are at least partially understood. On the other hand, it is known in the restoration world that identical restoration efforts can yield radically different results when implemented in different years and/or sites. Such year and site effects can make it difficult for restoration practitioners to predict the outcomes of their efforts and thereby plan on when and where to best expend resources to maximize outcomes. However, ecological research studies are notorious for establishing experiments at a single site and/or year, creating a disconnect between scientific knowledge and restoration reality.

In our recent study, we explore contingencies in restoration outcomes in the context of strong site and year effects. California grasslands are of great conservation interest, as a highly threatened ecosystem under the dual threats of land conversion and invasion by exotic annual grasses. Working in these California grassland communities, we established native grasses and forbs in plots at three sites across northern California in four consecutive years. Planting practices and seed mixes were identical across sites and years, allowing us to examine differences in restoration outcomes based solely on site and year effects.

2013-01-13 08.08.35

Results indicate that restoration outcomes are likely to be highly variable across space and time, with community composition across our restoration plots differing substantially across sites and planting years. For example, grass species that dominated in some sites and years were completely absent in others. And, while California poppy was often the dominant forb, occasionally it was failed to establish and yarrow dominated the forb component of the plots. Even the balance of forbs to grasses was strongly contingent on when and where plots were established. In half of the site-year combinations, forbs dominated the plots (achieved greater cover than grasses) while in the other half of the site-year combinations grasses dominated the plots. While one establishment year yielded forb-dominated communities across all sites and another year yielded grass-dominated communities across sites, two of the years were forb-dominated at some sites and grass-dominated at others. We were able to relate about nearly 40% of the variation we observed in plant community composition to temperature and rainfall. While this suggests that the abiotic conditions during establishment strongly influence the composition of restored communities, there are clearly drivers of restoration success that we don’t yet have a handle on.

While conditions during restoration seem to play an important role in early community trajectories, over the four years this experiment ran, we found evidence that these differences in community composition based on conditions during establishment are likely to persist. After some re-shuffling of each community over the first two years of growth, community compositions were relatively stable for the remainder of the study. As such, targeting places and times with abiotic conditions that might yield desired plant communities could be a restoration strategy with long-lasting implications for community composition.


When ecologists study restoration practices and outcomes, it is imperative that we consider possible contingencies associated with our outcomes. Results of a restoration experiment are almost certainly highly contingent on when and where the experiment was conducted. Ideally, ecologists could replicate our experiments far more to better understand the generalizability of our conclusions, and the sources of variation among outcomes. At a minimum, results of restoration studies (and perhaps most ecological studies) should be interpreted as a result of strong (and often not explicitly recognized) contingencies associated with the conditions during experimentation. As additional data on contingencies in restoration become available, we can hopefully push toward improving the predictability of restoration outcomes.

The full article, Every restoration is unique: testing year effects and site effects as drivers of initial restoration trajectories is a part of the Special Feature, Toward prediction in the restoration of biodiversity and available in read in the Journal of Applied Ecology.

Read other blogs from the series: