Life history trade-offs predict that life history strategies are fundamentally constrained, and that patterns of covariance between traits should provide a basis to classify life histories. Demographic data, such as matrix population models (MPMs), provides a potentially valuable resource to explore these covariance patterns. MPMs are versatile population modelling tools and published MPMs have been compiled into the COMPADRE Plant Matrix Database. A principle component analysis (PCA) of key life history metrics derived from plant MPMs and found two important axes of life history variation: fast-slow continuum and reproductive strategies. Position on these axes predicted rate of recovery from disturbances and population growth rate.
MPMs can have any number of discrete classes defined by any combination of size, age, and function. It has been shown that class definition affects demographic outputs. Additionally, MPMs are constructed from field data so are subject to sampling variation in λ (long-term growth rate). Our approach is to simulate sets of density independent, size-structured integral projection models (IPMs) under the constraint that λ is close to 1. The variance in the λ distribution acts as a surrogate for sampling variation. These IPMs were discretized to MPMs to mimic the discretization of MPMs in COMPADRE.
Results/Conclusions
Increasing the standard deviation of the λ distribution of simulated life histories altered the contributions of life history metrics in the first two principle components. These principle components were a composite of life history strategy and population performance because the life history metrics calculated from MPMs were not independent of population performance. Inconsistencies in MPM class definition had some effect on the first two principle components. Class definition of MPMs introduced uncertainty about a life history strategy’s positioning in the PCA space. However, by discretising simulated IPMs to MPMs with different numbers of life stages we understood how this uncertainty arose.
Comparative analysis of matrix population models does not identify true life history trade-offs. A significant component of the covariance among MPM-derived life histories is consistent with non-adaptive constraints on these patterns, arising from density dependence. Life history traits calculated from simulated life history strategies were not independent of population performance. When projecting performance metrics onto the life history PCA space, the resulting associations should be interpreted with care.