At a time when the combination of state of the art observing networks and modeling tools only provide weak constraints on carbon fluxes at continental (~10,000 km) scales, pressing scientific questions and societal concerns demand hard constraints on fluxes at regional (~1000 km) scales. It is clear that to move forward we need an expanded observing network and new modeling tools, but it is unclear how to best proceed on both fronts. One major question associated with planning new CO₂ observations is, given fixed resources, would a network of relatively few high-accuracy instruments or many relatively low-accuracy instruments provide better constraints on regional scale CO₂ fluxes? Recent modeling work has suggested that considerable flux information can be extracted from the large high-frequency variations in CO₂ over and near continents. However, the annual-mean flux signals are still very small and we can not be sure that future data assimilation schemes will be insensitive to small systematic biases in the measurements. Other important tradeoffs to consider in designing future networks include the location of measurement sites in high signal, high variability versus low signal, low variability regions, and the need for airborne measurements versus higher density surface measurements. The resolution of all of these issues will depend critically on the capabilities of future modeling and data assimilation tools. As a first step, we have developed a network design tool that allows us to quantitatively assess various instrumentation and platform combinations using one data assimilation scheme. We will report on results from assimilation runs for a variety of hypothetical future networks, focusing on their ability to resolve regional scale CO₂ fluxes over North America.