Gaining insight into food webs reconstructed by the inverse method

Carbon and nitrogen cycling on intertidal mudflats of a temperate Australian estuary. IV. Inverse model analysis and synthesis

“…There are, however, also downsides to this principle: since the parsimony principle is not based on ecological theory, flux values are typically positioned at the lower or upper boundaries and this solution typically recovers zeros for some of fluxes in the model (Steele 2009). The search for a selection criterion that is better rooted in ecological theory is still ongoing (Véz- Although this method is also not based on ecological theory, it draws the solution away from the boundaries and avoids zeroing out fluxes in the model (Kones et al 2006). Moreover, a direct estimate is obtained on the uncertainty with which the flux can be recovered.…”

“…See Table 2 for acronyms cal solution and ensures standardization in the solution method of inverse models. However, this optimization criterion is based on a minimization of the sum of squared flows, and it is presently unknown whether the selected solution really resembles the 'true' solution (Kones et al 2006). Moreover, most inverse studies have not given an account on the uncertainty that is associated with the inferred flux values, although sensitivity analysis on the input data generally show that the inferred fluxes are not very sensitive to changes in the input data of the model (Jackson & Eldridge 1992, Niquil et al 1998, Vézina & Savenkoff 1999.…”

“…Moreover, most inverse studies have not given an account on the uncertainty that is associated with the inferred flux values, although sensitivity analysis on the input data generally show that the inferred fluxes are not very sensitive to changes in the input data of the model (Jackson & Eldridge 1992, Niquil et al 1998, Vézina & Savenkoff 1999. A novel method was suggested by Kones et al (2006), which is based on sampling the infinite set of solutions. In brief, the data in the matrix Eqs.…”

“…After sampling a sufficient number of solutions (10 000 for the present model), the whole solution space is covered by the sampling. The mean of this set of sampled solutions is taken as an average model solution and the SD is taken as an uncertainty measure (see for details), which has been found to represent a good measure of uncertainty (Kones et al 2006). Flux estimates that are well defined by the data will have low standard deviations, whereas fluxes that are poorly defined will have a high SD (Van Oevelen et al in press).…”

Carbon, nitrogen and phosphorus resource supply and utilisation for coastal planktonic heterotrophic bacteria in a gradient of nutrient loading

“…Inverse modelling has previously been applied with success in several studies of planktonic systems (e.g. Jackson & Eldridge 1992, Stone et al 1993, Olsen et al 2006, 2011, Kones et al 2006.…”