Ecological theories of community assembly have been used to understand and predict community structural changes in response to agricultural practices. This is a fundamental issue to design sustainable cropping system. Weed infestations are potentially an important threat to crop production. Factors affecting weed community composition and diversity have traditionally been studied at the within-field level, since it is generally assumed that the main drivers of weed dynamics are farming practices at this spatial scale. However, spatial dispersion and landscape characteristics may strongly impact weed diversity both at local and regional scales. Moreover, weeds can persist for many years in the soil seed bank, which thus trace the lagged response of weeds along the crop sequence. Since the scale (field and year) of management of agricultural land is relatively invariant (at field and farm scales), we may assume a potential mismatch between the scale of management and those of population processes. We will present a synthesis of several studies that provide empirical evidence of the strong impact of spatio-temporal factors on weed diversity and will discuss the consequences of these results for designing sustainable cropping system. Analyses were conducted on a weed community data set consisting in over 1 000 arable fields sampled in the LTER ‘Plaine et Val de Sèvre’ (450 km², central western France), in which agricultural use is known in the 19 000 parcels over the last 20 years. We specifically asked two questions. First, is weed community assembly shaped by filtering and/or stochastic processes? Second, what is the relative contribution of local (field-scale) and regional (landscape-scale) factors on weed diversity? We combined a functional approach with a null model analysis to discriminate filtering by crop type and agricultural practices, from stochastic processes. We show that the weed functional diversity is predominantly shaped by stochastic processes. Non random distribution was observed for the timing of germination emphasizing its crucial role for the maintenance of weeds. In a second analysis, we aimed at disentangling the confounding effects of local management and landscape complexity with a sampling design which dissociated farming systems from landscape properties. We show that local management effects in the cropping season, i.e. intensity of herbicide uses, or in the rotation, i.e. crop sequence diversity, are overweighed by the effect of the percentage of organic fields in the surrounding landscape which positively affects weed richness. To sum up, even though crop type and management practices have been previously identified as the main drivers of weed diversity within arable fields, others processes, most probably stochastic ones such as spatial dispersion, also affect weed diversity suggesting that landscape-scale might be the appropriate scale for weed management.