Very little is known about the demography in size-structured seaweed populations, and this is especially true for populations in variable environments. Thus, the demography of the brown alga Ascophyllum nodosum was analyzed with a matrix population model. This was built on a 3-yr study in two populations on the Swedish west coast, where > 1100 individuals were marked and followed twice each year. The environmental variability is due to the presence of ice during some of the winters. The frequency of ice years in the study area is known, and is higher at one site compared to the other. During the study there were 2 yr with ice and one without at both sites, and these temporal changes in the environment resulted in large variations in the vital rates of A. nodosum. The individuals were divided into five size classes and the population dynamics at years without ice were characterized by low mortality rates and high transition probabilities for growth to larger sizes, while years with ice had high mortality rates and high transition probabilities for breakage to smaller sizes. Of the 25 possible transitions in the life cycle graph all except 2 had nonzero entries, which means that the adult life of A. nodosum individuals can be described as plastic growth between all five size classes. A crude estimate of the recruitment showed that both populations will increase in numbers in ice-free years, but will decrease in years with ice. The size-dependent fertility rate is probably subject to errors, and thus the survival matrices were scaled with a fertility function built on total reproductive biomass per individual, allowing analyses of the demography at different levels of the asymptotic growth rate lambda(1). One of the main differences between the populations was that the stable size distribution for survival matrices in ice-free years was dominated by size class 4 at one population but class 5 at the other. Thus, to achieve the same value of lambda(1) higher fertility rates were needed at the population dominated by class 4. Elasticity analysis of the transition matrices showed the same trend, although the main result from that analysis was that growth to larger sizes or remaining in the same class contributed more to lambda(1) than reproduction. This was valid for all levels of lambda(1) investigated.
