In addition, four major themes cutting across substantive research areas also repeatedly arose in workshop discussions:
Need for a long and continuing historical record of human activities shaping the carbon cycle. A long observational record of key activities (e.g., land use/land cover transformations, fossil fuel use, environmental treaties and policies, and agricultural land management practices) could be built from historical sources and from archaeological data, supplemented by remotely sensed data for recent times. Such data are necessary to quantify the trajectory of carbon sources and sinks in terms of social as well as biophysical drivers, to account for their current state, and to project future effects of human activities on the carbon cycle. A good historical record would provide the observational base needed for research on the substantive themes just noted, as well as for other substantive research on human interactions with the carbon cycle. Some of the necessary data collection is being done in two major international research programs, on Land Use/Land Cover Change (LUCC) and Past Global Changes (PAGES). However, because these programs have their own independent research priorities, this work is often not explicitly linked to the carbon cycle.
Need to develop emissions scenarios independently of the intergovernmental process so that plausible but politically unpalatable scenarios can be given due consideration.
Value of regional analyses for integrating the social sciences and natural sciences. Regional analyses, possibly including focused studies of selected regions, can provide venues for better interdisciplinary integration. Regional work should include efforts to scale up to the region (e.g., by using household-level data or agent-based models), analyses across regions (e.g., by comparing multiple case studies), and investigation of interactions across scales.
Need for more attention to uncertainties in scenarios and data. More attention is needed to the quality of data used as input to models, especially when data were estimated by “backfilling” methods. Models could be used to identify and elaborate uncertainties, and there could be stronger efforts to estimate the likelihoods of scenarios.