To evaluate the spatial variability of Arctic climate change during the pre
sent interglacial, CAFE Project Members compiled well-dated terrestrial, ma
rine, and ice-core paleoenvironmental records spanning the past 10-12 thous
and years (ka). Six tundra biomes of increasing summer temperature requirem
ents were defined based on regionally coherent pollen assemblages. Using a
rule-based approach, pollen spectra were converted to tundra, forrst/tundra
, or Forest biomes ranked by their average growing season requirements. Mar
ine sea-surface reconstructions were based on proxy data following a simila
r rule-based approach. From these data-based reconstructions. departures in
summer temperatures from modern normals were calculated in 1 ka time slice
s through the Holocene. To test predictive models, data-based summer temper
ature reconstructions were compared with general circulation model (GCM) si
mulations for 10 ka and 6 ka ago. Paleodata and model results both show tha
t warming occurred earlier across Beringia and Asia relative to lands adjac
ent to the North Atlantic, and that Late Holocene cooling was most apparent
in the North Atlantic region. However, the GCM over-predicts the magnitude
of Mid-Holocene warming over northern Asia and underestimates the intensif
ication of the North Atlantic drift in the early Holocene. Strong spatial v
ariability in environmental response during the Holocene, despite symmetric
(insolation) forcing, suggests that any future changes, whether caused by
anthropogenic or natural factors, are unlikely to result in a uniform chang
e across the Arctic, adding additional complexity to forecasts of global im
pacts. (C) 2001 Elsevier Science Ltd. All rights reserved.