Pattern-Informed Energetics: Inferring energy allocation and the emergence of life‑history strategies​

Cara A. Gallagher , Viktoriia Radchuk, Melanie Dammhahn, Florian Jeltsch

Energy allocation among survival, growth, and reproduction is central to population dynamics, yet remains difficult to quantify directly. Its consequences, however, are widely observed in growth, body condition, reproduction, survival, and population dynamics. We describe Pattern-Informed Energetics (PIE), a pattern-oriented modeling workflow that treats the parameters of condition-dependent allocation functions within a mechanistic energy budget as inference targets and constrains them against multiple empirical patterns. PIE does not infer the energy-budget structure or the functional form of allocation, which are specified; it estimates their parameters, quantifies uncertainty, and shows which parameters the available data can constrain. We applied PIE to six mammal species spanning a 1,000-fold mass range and to a spatially explicit population model of the bank vole (Myodes glareolus). In both, allocation was a sigmoid function of body-fat percentage, described by a midpoint (the body condition at which allocation reaches half its maximum) and a steepness (how abruptly allocation changes around that point). Across species, allocation midpoints scaled strongly with body mass, with larger mammals shifting energy to growth and reproduction at higher body conditions. Life-history patterns constrained these midpoints, but steepness was more difficult to infer, despite close fit to all calibration patterns. In the bank vole model, 16 more diverse patterns constrained all four midpoints but only two of four steepnesses. Re-analysing the simulations with subsets of these patterns showed that different types of data informed different parameters, and that adding patterns did not always improve identifiability. The retained parameter sets reproduced field and experimental patterns not used for calibration, including seasonal dynamics and the effects of litter manipulation on reproductive costs and maternal survival. PIE thus offers a transparent way to determine what can be inferred about allocation from available data and which data would most improve inference.