1861 Ecological Applications, 12(6), 2002, pp. 1861–1871  2002 by the Ecological Society of America MEASURING THE SUCCESS OF WILDLIFE COMMUNITY RESTORATION EARL D. MCCOY 1 AND HENRY R. MUSHINSKY Department of Biology and Center for Urban Ecology, University of South Florida, Tampa, Florida 33620-5150 USA Abstract. It is difficult to measure the success of wildlife community restoration when the restoration goal concerns a relatively broad geographic area, rather than a particular piece of land, since many restored sites need to be compared to many reference sites simultaneously. A review of the methods used to measure success in previous restoration efforts indicated the potential value of multimetric methods to make the comparison. We designed a new method that retains some of the advantages of multimetric methods but also removes some of the associated problems. The new method was applied to data from 30 restored sites (phosphate-mined land) and 30 reference sites in central Florida (USA), and it showed the difference in wildlife composition between restored and reference sites to be large, relative to the maximum possible difference. Alternative methods were applied to the same data. Multivariate methods were unable to measure success adequately. Percentage similarity analysis yielded results seemingly comparable to those derived with the new method, but the new method possesses advantages over percentage similarity analysis. The advantages derive largely from the first step of the new method, which is ranking species according to the magnitude of the difference between distributions at restored and reference sites. Species ranking clearly reveals which species contribute most to the distance between the restored system and the target; enables well- founded decision-making about which species can be excluded, if necessary, without se- riously compromising the restoration goal; promotes rigor in the subsequent steps of the model; and overcomes the problem that a single species may inordinately influence the result. Key words: cluster analysis; communities; focal species; multimetric methods; multivariate meth- ods; ordination analysis; percentage similarity analysis; restoration success; vertebrates; wildlife community restoration. INTRODUCTION As restoration of populations of organisms and their habitats becomes an increasingly important human con- cern, measuring the success of restoration is vital (Na- tional Research Council 1992, Lubke and Avis 1998, Smyth and Dearden 1998, Hackney 2000, McCoy et al. 2000). We use ‘‘success’’ to mean an acceptably small difference between the structure and/or function of a restored system and that of a reference system. Similar definitions are employed for studying theoret- ical concepts, such as ecological stability (Pimm 1991, McCoy and Shrader-Frechette 1992, Shrader-Frechette and McCoy 1993, Rapport and Whitford 1999), as well as for accomplishing practical tasks, such as biological monitoring and evaluation (Spellerberg 1992, Karr and Chu 1999). Under this definition, measuring the suc- cess of restoration requires a four-step procedure: (1) develop a goal, (2) set a target based on the goal, (3) determine the difference between the restoration and the target, and (4) decide if the difference is acceptably small. Despite the apparent simplicity of this proce- dure, measuring the success of restoration is not a straightforward exercise. For example, a clear goal 1 E-mail: mccoy@chuma.cas.usf.edu must be developed in each case (Jorgensen et al. 2000, Kentula 2000), which often is difficult, because the goal must be reasonable to both the persons doing the res- toration and the persons—usually agents of the pub- lic—for whom the restoration is being done (Kelly and Harwell 1990, Wilson and Mitsch 1996, Mitsch et al. 1998, Pouyat 1999, Cairns 2000). Perhaps the principal difficulty in measuring the suc- cess of restoration is selecting a method that will de- termine the difference between the restored system and the target reliably and also assess the significance of the difference rigorously. Although many candidate methods can be found in the literature, a great deal of uncertainty exists concerning their suitability (Suter 1993, Haila and Kouki 1994, Simon and Lyons 1994, Hengeveld 1996, Karr 1996, Rapport et al. 1998, Wil- kins 1999). This situation is true even in the relatively sophisticated arena of wetlands restoration (e.g., Zedler 1995, Mitsch and Wilson 1996, Mitsch et al. 1998, Zedler and Callaway 2000). In fact, none of the avail- able methods may be completely suitable in every case. It is important to keep in mind that a particular method should not be employed simply because of its avail- ability or popularity—the target should dictate the method that is employed, and not the other way round. A contemporary example of the difficulty of method selection comes from the restoration of phosphate-