Evaluating North America Paleoclimate Simulations for 6 ka and 21 ka Using a Combination of Observed Paleovegetation Data and Process-Based Vegetation Model Simulations

Abstract:

Paleoclimate model simulations are often evaluated using observed paleovegetation data (e.g., pollen and plant macrofossils) that record vegetation responses to past climate changes. These observed vegetation data can be combined with mechanistic vegetation model simulations to develop process-based evaluations of paleoclimate model simulations. The use of mechanistic vegetation model simulations allows us to identify the particular spatial and temporal features of individual paleoclimate simulations that may be producing agreement or disagreement between the observed and simulated vegetation data. We used this approach to evaluate a set of eight PMIP3 (Paleoclimate Modelling Intercomparison Project phase 3) paleoclimate simulations for 6 ka and 21 ka from the CMIP5 (Coupled Model Intercomparison Project phase 5) database. Climate data were regridded onto a 10-km grid of North America using the PMIP3 vegetation simulation protocol. The regridded climate data were used as input to BIOME4, an equilibrium vegetation model, to simulate 6 ka and 21 ka biomes across the study area. The simulated biome data were compared with observed paleovegetation data from the BIOME 6000 (version 4.2) dataset. In general, agreement between simulated and observed biomes was greater for forest biomes than for non-forest biomes. We evaluated specific instances of disagreement between the simulated and observed biomes to determine whether the biome disagreement was produced by the climate model simulation (e.g., temperature bias), the vegetation model simulation (e.g., inability to simulate important disturbance regimes), the observed paleovegetation data (e.g., limits in the biomization method), or a combination of these factors. The results are summarized and we describe some of the strengths and limitations of this data-model comparison approach for evaluating paleoclimate simulations.