Informing management decisions for ecological networks, using dynamic models calibrated to noisy time‐series data

Abstract

Well‐intentioned environmental management can backfire, causing unforeseen damage. To avoid this, managers and ecologists seek accurate predictions of the ecosystem‐wide impacts of interventions, given small and imprecise datasets, which is an incredibly difficult task. We generated and analysed thousands of ecosystem population time series to investigate whether fitted models can aid decision‐makers to select interventions. Using these time‐series data (sparse and noisy datasets drawn from deterministic Lotka‐Volterra systems with two to nine species, of known network structure), dynamic model forecasts of whether a species’ future population will be positively or negatively affected by rapid eradication of another species were correct > 70% of the time. Although 70% correct classifications is only slightly better than an uninformative prediction (50%), this classification accuracy can be feasibly improved by increasing monitoring accuracy and frequency. Our findings suggest that models may not need to produce well‐constrained predictions before they can inform decisions that improve environmental outcomes.

Structural forecasting of species persistence under changing environments

“…This field is providing the opportunity to anticipate rather than simply explain biodiversity changes in ecological communities contingent on explicit scenarios for climate change, land‐use and species re‐distributions (Dietze, 2017). Importantly, biodiversity forecasting spans and integrates many model‐driven (parametric) and data‐driven (nonparametric) methodologies, such as uncertainty propagation, statistics, informatics, Bayesian approaches, machine learning, Markov chain approaches, empirical dynamic modelling (Sugihara et al ., 2012; Harfoot et al ., 2014; Cazelles et al ., 2016; Dietze, 2017; Cenci and Saavedra, 2019; Adams et al ., 2020; Maynard et al ., 2020), as well as parameterising complex mechanistic models using either demographic, eco‐physiological or allometric information (Preston, 1962; Pacala et al ., 1996; Dietze, 2017). However, the majority of these methodologies demands extensive amounts of data, their explanatory power has been contested, and their generalisation has not always been validated with experimental work (Dietze, 2017; Clark et al ., 2020).…”

Predicting seagrass decline due to cumulative stressors

“…The algorithms that we used for SMC sampling and DSVI are described in Appendices D and E respectively. Briefly, our SMC sampling algorithm (Adams et al, 2020) is adapted, although not substantially different, from ideas presented in and . For DSVI, we followed the algorithm of Titsias & Lázaro-Gredilla (2014), and within this algorithm we used ADADELTA to calculate the learning rate .…”

“…To implement Bayesian inference we used two posterior-computation methods, Sequential Monte Carlo sampling (Doucet et al, 2000) and doubly stochastic variational inference ; the former involves computations which are embarrassingly parallelisable, and the latter scales well to higher dimensions because it transforms the simulation problem into an optimisation problem. Despite these advantages, both methods have been used in only a few biological and ecological applications (Quiroz et al, 2020;Adams et al, 2020). Increasing the uptake of these advanced methods of implementing Bayesian inference will improve our ability to provide ecological forecasts, a topic of growing interest within ecology (Petchey et al, 2015;Pennekamp et al, 2017;Dietze et al, 2018) that has thus far been hindered due to concerns about uncertainty quantification and biological complexity.…”

“…Below we describe our specific SMC sampler (Adams et al, 2020) which is adapted, although not substantially different, from ideas presented in and .…”

Recent advances of quantitative modeling to support invasive species eradication on islands

“…including ecosystem ensemble modelling (Baker et al 2017a(Baker et al , 2019aAdams et al 2020), fuzzy cognitive mapping (Dexter et al 2012;Baker et al 2018b) and qualitative modelling (Dambacher et al 2003;Dambacher & Ramos-Jiliberto 2007;Raymond et al 2011). Despite differences in mathematical approaches, each of these share the same core: a network of species interactions, and a large degree of uncertainty about the direct and indirect consequences of ecosystem interventions.…”