We formulated a number of recommendations:
- The dependence of the response of many surface variables on the amount of deforestation that was prescribed in each model suggest that, if the community agrees on the amount of deforestation that occurred over specific time periods, a significant part of the dispersion among the models will be reduced. We wrote a protocol in March 2010 that we circulated to a number of global climate groups that participated to the CMIP5 exercise.
- There is a need to revisit the way land-surface models are evaluated before being used for global climate simulations. The main problem we are facing is that land-surface models are generally evaluated off-line, forced with prescribed atmospheric forcing. Our results suggest that this is probably insufficient. Evaluation should try and account for atmospheric feedbacks. Such evaluation should then be completed by series of analysis to see how well do models simulate the contrasting dynamic properties of various vegetation types, which are relevant for biosphere-atmosphere interactions (e.g. water-use efficiency, dynamics of evaporative fraction, effective temperature sensitivity of carbon balance).
- We suggest that detection/attribution studies should account for LULCC as well as they do account for other anthropogenic changes. Increased greenhouse gases in the atmosphere, and the consequent changes in sea-surface temperatures and sea-ice extent, are often used as the main drivers of terrestrial changes. Our results suggest that such an assumption may lead to erroneous conclusions regarding the land-surface impacts of climate change since LULCC has, on a number of variables, an impact of similar magnitude, but of opposite sign.
- Important conclusion from the LUCID-CMIP5 experiments is that the information flow from the Integrated Assessment Models to ESMs needs to be much improved. In particular, the harmonization of LULCC scenarios should provide not only changes in land use, but also changes in land cover. Coherent protocols for implementing LULCC in ESMs are needed to coordinate simulations and to increase the statistical significance of results, especially, when analyzing the regional-scale impacts of LULCC.