Evaluation of ecological network analysis: Validation of output

Ecological network analysis (ENA) is a modeling approach increasingly being used to evaluate food webs and provide an ecosystem-based approach to resource management. Unfortunately, validation of ENA output is rarely performed. This study represents part of a larger effort to critically evaluate ENA. Here we validate ENA output using stable isotope analysis (SIA), and where validation is not met, determine the effects of modifying trophic networks to reflect validation.Quantitative trophic networks representing four salt marsh ponds during late summer 2002 were constructed from an extensive field sampling program augmented by literature values. Ponds were used because they contain relatively simple food webs, have well defined boundaries, and allow for relatively complete sampling. SIA data (δ13C, δ15N, δ34S) were used to validate effective trophic levels calculated by Ecopath for four separate networks, and carbon source estimates from NETWRK's total dependency matrix for four compartments in a single network.Effective trophic levels calculated by Ecopath matched those from δ15N data for three of the four networks. Mean differences of trophic levels between methods ranged from 0.12 (95% CI = 0.27) to 0.34 (95% CI = 0.35) in three networks, and 0.53 (95% CI = 0.28) in the fourth. Agreement in carbon source estimates was determined graphically using ternary plots. Percent carbon from three sources (meiofauna, epiphytic algae, phytoplankton/POC) given in the total dependency matrix (NETWRK output) did not fall within the range given by stable isotope mixing models for three of four compartments. Modifications to reflect validation of carbon source estimates did not result in significant differences in ENA output. Lack of validation was often due to inherent differences in methods and assumptions imbedded in the models.Our study highlights the need for feedback between model construction, analysis and validation in improving trophic networks. Moreover, validation allows the effect of uncertainty in trophic networks to be evaluated by quantifying the sensitivity of ENA output to modifications in the models.