NOAA/WDS Paleoclimatology - Guiot et al. 1996 European Land Surface Conditions 6K Reconstruction

Biome models are good tools for comparing the results of climate model simulations in an integrated way and such comparisons become more imformative if the simulated biomes can be compared directly with paleodata. It has been achieved in a recent paper (Prentice at al. 1996) where a method is proposed to assign to pollen spectra a biome compatible with those of the biome model of Prentice et al. (1992). This approach enables one to compare model results to data in a point by point mode, which is indispensable. But a precise evaluation of the boundaries between biomes needs a space interpolation. We propose here a stepwise method based on the relationship between biomes and climate to interpret these biomes from sparse data and to map them on a geographical grid. For that, neural network techniques are particularly adapted to take into account the non linearities of the system studied. Another and maybe more important application is the estimate of land surface boundary condition for palaeoclimate model simulations. The method is illustrated with pollen data at 6000 yr B.P. in Europe.