Characterizing the response of weeds to canopy shade is important for improved understanding of crop–weed competition and weed population dynamics. In 2000 and 2001, field studies were conducted in central New York state to examine the influence of three neighbor types (none, broccoli, or broccoli plus winter rye) and two locations (between or within rows of broccoli) on the morphology, phenology, and seed germination characteristics of Powell amaranth. Reductions in light availability and in the ratio of red-to-far red light were associated with increases in (1) partitioning of dry weight to stem tissue, (2) stem elongation, and (3) specific leaf area. Canopy shade also resulted in fewer main leaves at flowering and a reduced rate of leaf appearance but had no effect on the number of days to flowering. The relationship between Powell amaranth fecundity and aboveground dry weight was allometric, with both parameters declining significantly under competition. The weight of seeds produced did not vary significantly according to the competitive environment experienced by the maternal parent. However, the germination percentage of viable seeds was 40 to 50% lower for seeds maturing on plants grown under competition than without competition. Reductions in the number of main leaves at flowering and increased seed dormancy may be adaptive responses to canopy shade. Both mechanistic crop–weed competition models and population dynamic models would benefit from incorporation of data on the phenotypic plasticity of morphology, phenology, and seed germination characteristics of weeds.

Summary Estimates of weed fecundity and its variability are critical for the development of population dynamic models and for evaluating the long‐term consequences of weed management practices. The purpose of this research was to estimate Amaranthus powellii fecundity across years, seasons and competitive environments using a mechanistic model. Existing models were modified to account for weed responses to shade, and to dynamically simulate seed production among subthreshold densities of A. powellii. The model was parameterized and tested using five sets of field data in which A. powellii was grown either alone or with transplanted broccoli. The model overestimated A. powellii height, but predicted both dry weight and fecundity well. Mean simulated fecundity for A. powellii ranged from 0 to 268 000 seeds per plant depending on year, crop maturity date, relative time of emergence and location. Year‐to‐year variability in simulated fecundity was large with coefficients of variation ranging from 77 to 128 under low and high competitive environments respectively. Our results suggest that estimates of weed fecundity based on 1 or 2 years of empirical data may result in significant errors in population dynamic models, and that the use of economic optimum thresholds based on weed density alone entails considerable risks.