Consequences of parameterization and structure of applied demographic models: a comment on Pardini et al. (2009)

Authors: EVANS, J - Dartmouth College; Davis, Adam

Submitted to: Ecological Applications
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/3/2010
Publication Date: 3/1/2011
Citation: Evans, J., Davis, A.S. 2011. Consequences of parameterization and structure of applied demographic models: a comment on Pardini et al. (2009). Ecological Applications. 21(2):608-613.

Interpretive Summary: Population dynamics simulation models can help land managers decide which management tactics to implement to control invasive species. However, errors in model contruction or scope of inference may result in counterproductive management efforts. We found several errors of this type in a recently published model of garlic mustard demography, resulting in incorrect management recommendations (focus on controlling adult plants only, leaving rosettes alone). After revising the model, we found that controlling both rosettes and adult plants would be productive for land managers in many parts of garlic mustard's range. Because of the tremendous spatiotemporal variability in garlic mustard population dynamics and the species’ large geographical range, more general management recommendations will only arise from a larger set of demographic data that has greater coverage in space and time.

Technical Abstract: In a recently published study, Pardini et al. (2009, hereafter "PDCK" after the authors' initials) developed a demographic model of the invasive weed Alliaria petiolata (garlic mustard, Brassicaceae [M. Bieb] Cavara and Grande). This was then used to identify optimal stages in the plant's life history to target with management actions. Model predictions indicated that management of newly recruited rosettes or second year flowering adults would have to reduce survival by at least 95% or 85%, respectively, to be effective, and that rosette management “could actually increase the overall density of A. petiolata” if managers killed fewer than 95% of rosettes annually. Based on these results, PDCK “recommend managers avoid [management of] rosette plants and instead focus on management of adult plants”. Several critical problems in PDCK’s model indicate that its management recommendations are incorrect and inappropriate for widespread application. These include errors in the statistical estimation of parameters in two density-dependent functions, and incorrect implementation of density-dependent functions in the model. Additionally, because the model was parameterized using a single year of data from only one site supplemented with data from other published sources, applying management predictions from this analysis to other A. petiolata populations is inappropriate. This is cause for concern given the variability of this species’ demography documented in past studies, and in light of work illustrating how demographic variation among populations can drive variation in the success of individual management strategies. Because the published recommendations in PDCK are already being implemented by some natural areas managers to control A. petiolata, these problems must be corrected and the scope of its predictions be more carefully defined. A revised version of PDCK’s model makes different management predictions than the original model. This new model indicates that (1) management of rosettes that is less than 63% efficient could cause increased population density, (2) any management which reduces adult survival probability or decreases per capita fecundity will reduce population size. Moreoever, (3) the counterproductive effect of rosette management described above is very sensitive to the value of the seedling survival rate (s1) used in the model and is predicted only if s1 is less than about 0.2, which is frequently exceeded in multiple study populations (Davis et al. 2006). When s1 exceeds 0.2, any reduction in rosette survival is predicted to suppress population size. Thus, although these predictions and implications have been corrected, they may still be of limited utility because of the limited scope of the data used to generate them. Because of the tremendous spatiotemporal variability in A. petiolata demographic rates and the species’ large geographical range, more general management recommendations will only arise from a larger set of demographic data that has greater coverage in space and time.