The mapped 0–9 k temperature anomalies reveal considerable spatial variability, with marked different responses over land and ocean surfaces (Fig.  4 a–e).

Over the continents, the Holocene temperature decrease is strongest in summer (June–July–August) in line with the insolation forcing, with a maximum cooling of 7°C in the Canadian Arctic (Fig.  4 c).

The western part of the Arctic Ocean experiences a maximum cooling of up to 7°C during autumn (September–October–November, Fig.  4 d), whereas in the Barents Sea the greatest temperature depression occurs in winter (December–January–February, Fig.  4 a).

In most regions, the season with the minimum change in temperature is spring (March–April–May, Fig.  4 b).

However, on an annual basis, all areas north of 60°N experience a decrease in temperature in the model (Fig.  4 e)..

The latter anomaly is strongly influenced by the strong winter cooling over the Nordic Seas (Greenland-Icelandic-Norwegian Seas, see Fig.  4 a), of which a part is included in the calculation of the average depicted in Fig.  5 a.

The relatively strong response during January reflects the impact of the Arctic Ocean, where a maximum temperature difference was noted in winter (see Fig.  4 a)..

In addition, the simulated cooling is stronger in northern Greenland than in the south (Fig.  4 c), which is in agreement with the difference between NorthGRIP and GRIP.

Our model produced a long-term cooling of ~1.5°C (Figs.  4 c, 8 b) and a precipitation reduction (Fig.  10 d) that agrees well with this paleoclimatic evidence.

In our experiment, the overall Holocene summer cooling in the Nordic Seas is of the same order (1–2°C, Fig.  4 c) as suggested by the proxy data.

The simulated climatic evolution of the Barents Sea region gives a reasonable match with the proxy data, showing a long-term cooling trend of 1°C during summer (Fig.  4 ) and a marked expansion of sea ice (Fig.  13 )..

In our experiment, these two feedbacks caused an amplification of the orbitally forced cooling, especially in the Canadian Arctic, as can be deduced from a comparison of the increase in surface albedo (up to +30%, Fig.  13 b) and the pattern of the 9–0 kyr BP temperature anomalies (locally −7°C, Fig.  4 d).

Considering that the duration of this lag is a few months, both effects caused a clear maximum temperature difference over the Arctic Ocean in the winter half year instead of summer (Fig.  4 ).

A substantial part of the temperature reduction in these areas is probably due to this biogeophysical feedback, especially in spring (compare Fig.  4 b, c with Fig.  13 d).