Determinants of successful groundcover restoration in forests of the Southeastern United States

Restoration is a time-consuming and expensive endeavour. As such, it is a field that could benefit immensely from the adoption of an adaptive management approach, where knowledge gained from previous experiences is incorporated into future planning efforts. However, meta-analyses that synthesise results obtained over large spatial scales are rarely conducted, hindering the ability of restoration practitioners to learn from mistakes and successes of others. We present a case study of groundcover restoration in the southeastern United States as an example of the wealth of information that can be obtained from a synthesis of existing data. This example is characteristic of many restoration endeavours in that a large number of decisions must be made during the restoration process. We used Classification and Regression Trees (CART) to identify those restoration activities that most often lead to successful groundcover establishment in forests of the Southeastern US. The most important factors in determining mean survivorship of plants established through outplanting included planting season, the presence of existing canopy cover, and the use of prescribed fire after planting. These three factors alone explained 28% of the variation in mean survivorship (pPRE = 0.28, model PRE = 0.32). In contrast, the single greatest predictive variable in the establishment density of seeds sown on restoration sites was the use of herbicide prior to planting which explained 30% of the variation in establishment rate (pPRE = 0.30, model PRE = 0.38). Based on these analyses, we describe simple, discrete recommendations that should improve the survival and establishment of native plant species in restoration sites across the southeastern coastal plain of the US. We also discuss the utility of CART in distilling large volumes of existing data into easily interpretable recommendations.