The δ18O record from ice cores serves as a proxy paleoclimatic temperature record, through the association of isotopic ratio to air temperatures at time of precipitation. Climatic change may be preserved also as a signal in ground temperatures, not as a proxy indicator of past climate but as a direct consequence of the effect of past air temperature variations and associated physical processes at the ground surface. In the Canadian Arctic Archipelago, δ18O records are available from the Devon and Agassiz ice caps, and precision ground temperatures to depths of up to 1000 m are available from 40 petroleum exploration wells, about one third of which are suitable for paleoenvironmental reconstruction. There is an opportunity to compare these two methods of looking at the paleoenvironment, and to show how complementary they are to each other.

Geothermal analysis is predicated on the fundamental hypothesis that the terrestrial heat flow, which arises largely from the decay of radioactive elements within the crust, does not vary measurably in the upper few km. But at many wells, the heat flow, calculated as the product of the measured temperature gradient and rock thermal conductivity, does vary systematically with depth in the well. While more random variations may be attributed to measurement errors, and corrections may be made for such known effects as local topography, the residual coherent “long wavelength” variation may be ascribed to effects arising from climate change.

Can we, then, determine whether a particular temperature history is consistent with the geothermal record, or ideally, invert the geothermal data to reveal a record of past surface temperatures? Attempts with varying success at paleoclimatic reconstruction from ground temperatures have been reported in the literature (e.g. Lane, 1923; Hotchkiss and Ingersoll, 1934; Birch, 1948; Cermak, 1971; Vasseur and others, 1983; Lachenbruch and others, 1986) and from temperature profiles in ice sheets (e.g. Paterson, 1968; Weertman, 1968; Budd and Young, 1982).

In this study, standard techniques in geothermics (e.g. Jaeger, 1965) have been used (1) to show the effect of any hypothesized surface paleotemperature model upon subsurface temperatures, or (2) on the hypothesis that the variation in heat flow is attributed to paleoclimatic effects, to derive a surface temperature model at each well that minimizes the variation in a statistical sense. The resolution of the method and limitations in our measured temperature and rock thermal conductivity data restrict the application of the second method to the past few hundred to one thousand years. The paper considers the first approach for the period 1 ka-10 kaB.p. at about a dozen wells and gives an example of the second approach at a well west of the Agassiz Ice Cap.

Aproach (1). In studying the Devon Island ice core, Fisher and Koerner (1979) present a detailed record of the mean annual air temperature at the site throughout the Holocene, based on the δ18O record. A simplified time-temperature model of this record is applied to the ground temperature data set for the period 1 ka-10 ka B.P. Although the effect on the ground temperatures is only subtly perceptible, the model has the effect of reducing the apparent climatically-related curvature in the data, as reflected in an improvement in the standard deviation in the calculated heat flow profile by 5% to 30%. Hence, the geothermal record provides quantitative support for Holocene climatic information derived from the ice core record.

Approach (2). This inversion technique is analogous to Paterson’s (1968) reconstruction of the surface temperature during the past century from a temperature profile taken in the small Meighen Ice Cap, Arctic Canada. A unique model is not obtained; rather, a small set of possible surface temperature variations consistent with the deeper subsurface temperatures is produced. Such modelling suggests that subsurface temperatures at a well 180 km west of the Agassiz Ice Cap are consistent with ground surface temperatures some 4–6 Κ lower at the well during the Little Ice Age; this is considerably more severe than the mean annual air temperatures projected from the δ18O record at Agassiz. It is possible that the large increase in ground surface temperature at the wellsite since the Little Ice Age may be attributed to some climatically-related phenomena such as increased incidence of snow cover coherent with the changing climate. A well on Devon Island is not deep enough for a comparison to that ice cap.

The oxygen isotope data provide a valuable estimate of Holocene climate with which to correct ground temperature data for terrestrial heat flow, or other studies. However, examination of the signal of more recent events suggests that ground temperatures may be considerably modified by associated transient phenomena such as snow cover, vegetation, etc. Hence, one would expect that such a Holo¬cene correction might either understate or overstate the actual experience of the ground surface at a site.