## Aix

where (a*) denotes the best linear predictor of xt given its past history; analogously (a,P) denote the best linear predictors of xt given its past and the past of yt_v and e„ utrepresent the error terms. Obviously, the restricted equation does not use information of the dataset Y, while unrestricted equation includes information of both datasets X and Y.

(2) Granger causality tests a full model against a null model with no possible causality, where the parameters of interest are set to zero. Thus, the null hypothesis of noncausality will be H0: = = •■■ = fip = 0. We conduct an F-test of the null hypothesis by comparison of respective sums of squared residuals from the restricted and the unrestricted models:

If the test statistics

is greater than the specified critical value, then the null hypothesis that Y is not the Granger cause of X, can be rejected. This procedure should be repeated for multiple values of lag with each p-value being tested independently of others.

It should be mentioned that with lagged dependent variables, such as Granger causality regressions, the test is valid only asymptotically. An asymptotically equivalent S1 value is given by

Another caveat of the GCTs is that they are sensitive to the choice of lag value and to the methods employed in dealing with the possible non-stationarity of the time series. If the Granger causality is applied, e.g., to the daily time series, one may expect that most of the residual variance will be concentrated at short timescales. In other words the Granger causality statistics will be dominated by the effect of daily prediction errors (Mosedale et al., 2006). The non-stationarity is a common problem of all kinds of the time series analysis. If time series are non-stationary, the model can be applied to temporally differenced quantities, Axt = xt —xt-1, rather than to the original data. An effective approach for the testing of causality that avoids the problems arising from ignoring possible non-sta-tionarity and/or co-integration between series has been proposed by Toda and Yamamoto (1995). A certain lack also represents the general practice of the researchers who investigate a simple story, e.g., Y is Granger cause of X, and do not look the other way. In the real examples, more complicated situations can occur such as neither time series Granger causes the other or that each of them causes the other.

Since the 90s of the last century, the GCT has been extensively used in climate research. Among other things it was applied in the concept of coupled climatic processes to investigate the ocean feedback on the North Atlantic Oscillation (NAO) (see Mosedale et al., 2006, and the references therein). It was found that the sea surface temperatures (SST) represent a Granger causal for the daily values of the NAO. Kaufman and Stern (1997) have examined the Northern and Southern Hemisphere instrumental temperature records from 1869 to 1994. Using the Granger causality analysis, they have found that there is a statistical relationship in which the Northern Hemisphere temperature depends on temperature in the Southern Hemisphere and have concluded that the detected causality is the result of anthropogenic climate influence. Almost immediately these attempts attracted benign attention and/or obtained trenchant criticism. The effect of this discussion can be found in the works by Triacca (2001, 2005; http://www.isi-eh.usc.es/ trabajos/122_41_fullpaper.pdf) or on numerous web sites such as "Still waiting for Greenhouse" (www.john-daly.com/granger.htm) and the Union of Concerned Scientists (www.ucsusa.org). The reaction was quite different, somewhere little bit trenchant, e.g., that the GCT is a creature of economic models, and should perhaps have remained so. Most of the researchers have laid stress on the necessity of the proper allowance for the non-stationarity of the data and utilizing more wide climatic databases, e.g., data simulated by global circulation models (Triacca, 2001).

Despite the criticism that has been directed against the use of GCT, its merits in the time series forecasting are indisputable, and it is extensively employed in climatology. Probably, it was above discussion that stimulated the interest to this kind of analysis and further application of the Granger causality test in different research fields. In the recent years, the Granger causality analysis has been extensively used to test the hypothesis that GHG are responsible for at least a part of the observed twentieth century global warming. Sun and Wang (1996) have suggested that global CO2 emission represents the Granger causal for the global surface warming. Tol (1994) and Tol and de Vos (1993,1998) have tested the hypothesis that the global mean SAT has increased due to an increase in the atmospheric concentrations of GHG. Obtained results were reliable enough to conclude that at least part of the recent warming with high probability can be attributed to the increase of the atmospheric concentration of CO2. Kaufman and Stern (2002) have found that there is statistically significant relationship between the SAT and the changes in the radiative forcing caused by natural variability and human activity. Combined with other similar investigations, the results mentioned above provide further evidence of the statement discussed in the previous sections that observed over the past 100 years warming trend in global mean temperature is unlikely to be entirely natural in origin.

Below we present an example of the application of GCT to demonstrate its pervasiveness in the borehole climatology research. Borehole temperature measurements contain direct information on the GST history. The GSTs represent important climatic variable and thus, in principle, do not need the calibration with the independent data. On the other hand, it is the air column temperatures, including the most important SATs taken at the screen height (1.5 m above the ground surface), that are typically of interest in the discussions of climate variability. The SAT responds to the convective heat transfer in an atmospheric boundary layer, while the GST represents continuously integrated ground temperature variations in the vicinity of the borehole that occur mainly by the heat conduction process. In addition, the GST is significantly influenced by the land surface and soil properties (vegetation, snow cover, etc.; for details see Sections 2.5-2.8, Chapter 2). The problem of coupling of the GST and SAT has arisen from the very beginning of the borehole climatology. The vast volume of repeated studies suggests that on a large scale mean annual GST corresponds linearly to the mean annual surface temperature; thus, it seems reasonable to view borehole temperatures as filtered versions of the SAT. This statement can be confirmed by the GCT.

For the analysis below we have used reconstructions of the annual global surface temperature (SAT) over the last five centuries (1500-1980), based on the multivariate calibration of high-resolution proxy climate indicators (tree rings, ice cores, corals, historical documents) combined with the long-term instrumental records by Mann et al. (1998) (Figure 39, Chapter 2) and similarly long GST reconstruction based exclusively on the terrestrial borehole data (Mann et al., 2003; Figure 105, Chapter 3; see also discussion in Section 3.3 of this chapter). The goal is to test the hypothesis that above SAT series (dataset Y) represents the Granger cause of the GST (dataset X), in other words, that the GST changes occur due to the changes in the SAT.

Both datasets can be found on the web site www.ncdc.noaa.gov/paleo. Calculated under assumption of the lag length p = 2 years coefficients of the restricted linear equation for GST series are: a* = 1.968 ± 0.013, a*2, = - 0.968 ± 0.013, ct = 0.000191 ± 0.000075. Similar coefficients for the unrestricted regression equal: a1 = 1.969 ± 0.013, a2 = -0.969 ± 0.013, h = 0.000124 ± 0.000087, p2 = -0.000232 ± 0.000087, and c1 = 0.000182 ± 0.000071. The F-statistic and the significance F were used to test the validity of the regression. Because calculated F-value in both cases is much higher than the significance F (confidence level equals to 95%), we would reject the null hypothesis (H0: a* = a = 0 and/or a1 = a2 = P1 = & = 0) and conclude that at least one independent variable in both restricted and unrestricted equations is correlated to the dependent variable. Calculated sums of squared residuals are 0.0000151 and 0.0000148, respectively, that gives the value S1 = 4.794 (Eq. (51)). Because this value is greater than the specified critical value (Fp,T_2p_1 confidence level is 95%), we can reject the null hypothesis that the used SAT series is not a Granger cause of the examined GST.

Although simple statistics presented above does not constitute a total proof of the hypothesis about the GST-SAT causality, and further more accurate application of the Granger causality to this problem (e.g., using multiple lag values) is necessary to obtain unambiguous conclusion, even such simple test was robust enough to support strong long-scale GST-SAT coupling.

3.5 Deep Continental Drilling and Signature of Remote Climate Changes

As demonstrated in Chapter 2, under favorable conditions all Holocene climate can be evaluated if the precise borehole temperature log is available to the depth of 1-2km. On the other hand, Beltrami and Burlon (2004) have shown that under the restrictions that are essential to obtain robust spatial averages on hemispheric or global scale, the merging of the individual GST reconstruction results cannot retrieve reliable information on the climate variations at times before 1500. Information about more remote climate changes can be gained only from individual inversions of temperature logs measured in deep holes. Thus, borehole climatology can provide the reconstruction of remote climatic events only from separate locations where such holes were drilled. On the other hand, an importance of availability of temperature measurements from deep boreholes was recognized by geothermal community long ago. Such studies not only are performed for the knowledge of the composition, structure, and evolution of the Earth's crust but also can serve many economical purposes, like various deep mining and geothermal energy projects or even the next-generation nuclear waste repository. In the recent decades the deep drilling programs have represented an essential part of the geophysical research. A global overview of major completed as well as on-going deep drilling programs can be found, e.g., on the website of the International Continental Scientific Drilling Programme (the ICDP; www.icdp-online.de). The use of the temperature logs measured in deep holes for the reconstruction of the remote climate change began from the very beginning of borehole climatology.

3.5.1 Late Quaternary GST changes inferred from the deep hole measurements

In traditional paleoclimatology the reconstruction of remote climatic changes is based on a variety of proxy records. Because climatic variables are only indirectly reflected in these data and their evaluation requires an interpretation of physical, chemical, and/or biologic phenomena, inferred results may contain systematic biases and errors. To compile the most meaningful and complete climatic history, it is necessary to consider the information of many independent records, and the measurements of underground temperatures in deep boreholes performed worldwide at the recent decades represent a valuable source of paleoclimatic information. In contrast to proxy data that are indirect inferences of the climate change, the subsurface temperatures measured in boreholes

Fig. 110. Left: The reference Pleistocene climate history used as a forcing function for calculation of subsurface temperature-depth profiles. The ice-base temperature adopted as —1.5° corresponding to glacier thickness of 2-3km; interglacials are generally warmer by 1° than the reference temperature. Right: transient component of the subsurface temperature calculated in response (1) to the last 0.1Ma GST variations and (2) to the last 10000 years temperature change.

Fig. 110. Left: The reference Pleistocene climate history used as a forcing function for calculation of subsurface temperature-depth profiles. The ice-base temperature adopted as —1.5° corresponding to glacier thickness of 2-3km; interglacials are generally warmer by 1° than the reference temperature. Right: transient component of the subsurface temperature calculated in response (1) to the last 0.1Ma GST variations and (2) to the last 10000 years temperature change.

directly archive the past GSTs. Because the low thermal diffusivity of rocks, the GST changes propagate slowly downwards and remain recorded as transient perturbations to the otherwise steady subsurface temperature field. For typical rock diffusivities of about 10_6m2/s temperature changes that occurred 1000 years ago penetrate to approximately 250 m depth into the Earth, while temperature excursions that date back to 30000 years (the last glacial period) produce anomalies at 1300-1400m depth. Thus, 1.5-2 km deep holes may already yield the GST history up to the last glacial period and subsequent variations of the Holocene climate. Because the depth of the ordinary deep drillholes rarely exceeds 1-2 km the most often time period for the remote GST reconstruction is one to several tens of thousand years.

Figure 110 shows the synthetic transient T-z profiles resulting from the long-term GST variations. Profile (1) is based on the last 0.1Ma changes including the last glacial and major postglacial excursions, while curve (2) represents the underground response to the last 10000 years (the Holocene). The reference climate history (Figure 110, left) includes two Wisconsin10 glacial stages that lasted to 9500 B.P. and prolonged period of warm conditions with its maximum at the Atlantic Optimum when the warming of approximately 6K according to the glacial conditions was longed for at least 2000 years. Generally warm interglacial conditions were interrupted by provisional returns of cold, of which the main were the Cochrane re-advance between 7250 and 6000 B.P. and the Sub-Atlantic period. While the effect of the post-Wisconsin warming on the simulated T-z profile is quite strong, the signature of the more recent postglacial excursions is comparatively weak. Maximum disturbance reaches only 0.05 K; thus, the possibility for its reconstruction is not so definite for the warming which terminated the last glacial.

10The final glacial advance of the Pleistocene in North America, 115000-10000 B.P., corresponding to the Scandinavian Weichsel and Alpine Würm.

As mentioned above (Chapter 2, Section 2.4.3), the reconstructed GST history represents a weighted average. In other words, the further we go into the past the less detail can be resolved and the smoother trend of the real temperature conditions on the Earth's surface can be obtained. A 100-year long event that occurred 300-500 years ago can be resolved with the relative variance of 10-15%. For as early as 2000-3000 years ago, it is only possible to resolve a 500-year interval with the same reliability, and the corresponding duration of event is 1000 years if it occurred 7000-9000 years ago. Because the cold climate of the Wisconsin glacial has spread approximately over the period of 90000 years, there is a real chance to reveal it in the GST reconstructions. Together with the average postglacial GST variations obtained GST histories will thus yield an estimate of the glacial/interglacial temperature differences. Similar conclusion was made by Safanda and Rajver (2001) on the basis of synthetic calculations corresponding to the Weichselian ice age in Central Europe.

Early attempts of GST inversions of the temperature-depth profiles measured in the deep-drilled holes were made in the works by Bodri and Cermak (1997b), Rajver et al. (1998), Safanda and Rajver (2001). Figure 111 illustrates results of one of these reconstructions. Almost 6-km deep borehole Jablunka (JAB-1) was drilled in the eastern Carpathian flysch zone. This area was outside of the last Eurasian ice sheet (Grosswald, 1980). Temperature logging was performed in 1982 to the depth of about 6000 m with a sampling interval of 50 m (Figure 111). The quality of the temperature log permitted to penetrate into the past up to 30000 B.C. The reconstruction by Bodri and Cermak (1997b; Figure 111) detected climatic trends that are consistent with other interpretations of the past climatic history. Crowley (1983) has collected the evidence on the climate change in the Holocene. His data indicate two main climaxes during the last glaciation (Late Weichsel): the first 23000-25000 years ago, and the second about 15000-20000 years

Fig. 111. Left: measured temperature log (solid line) and thermal conductivity (squares) for borehole JAB-1. Right: reconstructed GST history for borehole JAB-1 (solid line). Dashed line represents average height changes of the upper tree line in the Alps and other temperate mountain regions over the last 15000 years. (Data by Lamb (1977).)

Fig. 111. Left: measured temperature log (solid line) and thermal conductivity (squares) for borehole JAB-1. Right: reconstructed GST history for borehole JAB-1 (solid line). Dashed line represents average height changes of the upper tree line in the Alps and other temperate mountain regions over the last 15000 years. (Data by Lamb (1977).)

ago. In the GST history reconstructed for borehole JAB-1 (Figure 111) these intervals have been resolved as a single event. The general background situation at the period of maximum cooling of the last glaciation (about 20000-18000 B.P.) seems to have been a drop of the annual mean air temperatures in Central Europe by about 8-14 K compared to the present (Frenzel et al., 1992), which is in good accordance with the values of 9-11K obtained for the JAB-1 GST history. The latter values were calculated as the temperature differences between mean values for -20 B.C., -15 B.C., and 1900-1982 A.D. periods. The last datum corresponds to the time of borehole JAB-1 temperature logging. The Holocene (the last 0.01 Ma) represents an interglacial epoch. Generally warm conditions have prevailed since the last glacial maximum. The glacial period was transformed into the interglacial in two relatively rapid steps of warming. The first step has begun 13 000 years B.P. and the latter occurred between 10000 and 7000 years B.P. (Figure 2, Chapter 1). Between these periods there was a brief return to near full glacial climate. The resolving power of the SVD inversion method used for the GST reconstruction presented in Figure 111 appeared too low to detect the sudden climatic cooling that was documented between 11000 and 10000 years ago. As shown in Figure 111, except for some less pronounced oscillations, the GST history reconstructed for borehole JAB-1 quantitatively reproduces the generalized climate trend derived from the data of height changes of the upper tree line in the Alps and other temperate mountain regions over the last 15 000 years. The later variations are predominantly due to differences of summer temperature and duration of the season for growth (effective temperature control).

Similar information on the climate conditions since the Würm Ice Age (the last 45 000 years) that represents the latest central European and/or Alpine glaciation was inferred from temperature log measured in the 2 km deep borehole Ljutomer, Slovenia (Rajver et al., 1998). The GST history was formally reconstructed for the last 90000 years; however, because of the progressive decrease in the resolution only climatic episodes of the recent 20 000-10 000 years were clearly outlined. Inversions were performed with both the ramp and FSI method for different values of the standard deviations of a priori conductivity model and measured temperature. The trial runs with different standard deviation values were necessary because of the poor knowledge of the thermal conductivity of rocks at the borehole site. The set of various a priori models was chosen to investigate the quality of estimated GST histories and select the more informative variant of the reconstruction. Figure 112 presents results of inversions obtained by Rajver et al. (1998). Because of the differences in a priori assumptions, estimated GST histories appear more or less smoothed preserving, however, general course of the past climate. The reconstruction revealed glacial maximum 13-14 years ago. The certainty of this event was supported by the results of independent reconstruction with simpler ramp approach. The postglacial warming occurred about 2000-3000 years B.P. and was followed by the long-term "wavy" warming trend with superimposed provisional returns of cold conditions.

Attempts to reconstruct climate variations for the past 100000 years were continued in the work by Safanda and Rajver (2001). Authors have used temperature logs measured in the 1.5-2.5km deep boreholes from the territory of the Czech Republic and Slovenia. Figure 113 shows GST histories estimated by the joint inversion of three Czech and two Slovenian T-z profiles, respectively. As in the previous figure, different GST histories were calculated using three sets of values of a priori standard deviations for the thermal conductivity and the temperature data. Calculated glacial/interglacial temperature

100000 10000 1000 100 10 TIME, year B.P.

100000 10000 1000 100 10 TIME, year B.P.

Fig. 112. GST histories reconstructed for the Ljutomer deep temperature-depth profile, Slovenia (data by Rajver et al., 1998). GST changes were inverted by ramp/step method as well as by FSI inversion with different standard deviations of a priori conductivity model and measured temperature.

variations for both borehole groups do not exceed 8 K. Inversions performed with highest values of a priori standard deviation for the thermal conductivity gave quite smooth GST histories with variations oscillating in the range of 2-3 K. Preferred values of a priori standard deviation for the thermal conductivity and the temperature data are 0.5 W/mK and 0.2 K, respectively. Inversions exhibit very similar times for the occurrence of the glacial maximum between 19000 and 10000 B.P. and rapid warming since then. The range of temperature excursions was significantly lower in the last two millennia. These results agree well with information extracted earlier from the German KTB super deep hole (see below) and borehole JAB-1, where inversion of the T-z profile gave approximately 10 K temperature increase from the last glacial to the present. Above reconstructions also coincide with the most detailed climatic history of the last 100000 years reconstructed by Zoth and Haenel (1988) on the basis of proxy data series. The GST histories described above probably reflect the Holocene climatic trends typical for the region of Central Europe.

Vast deep drilling efforts have been performed in the territory of Russia in the recent one to two decades, and deep temperature logs are available from boreholes of the East European Platform, Middle Asia, West Siberia, Kamchatka, and other regions at up to 5-6 km depths. These drillings were mainly of industrial interest due to the fact that the search for hydrocarbons in the recent years has moved into greater depth. Most of the temperature measurements were performed since 1990s in a short time after drilling, thus, in boreholes that probably did not achieve thermal equilibrium. A successful attempt to reconstruct remote GST changes even on the basis of such less precise

Fig. 113. GST histories obtained by joint inversion of the three temperature logs from the Czech Republic (top) and two temperature logs from Slovenia (bottom). The GST histories were inverted by FSI technique. Curves are marked by the value of a priori standard deviation of the thermal conductivity and the temperature data.

Fig. 113. GST histories obtained by joint inversion of the three temperature logs from the Czech Republic (top) and two temperature logs from Slovenia (bottom). The GST histories were inverted by FSI technique. Curves are marked by the value of a priori standard deviation of the thermal conductivity and the temperature data.

temperature logs was demonstrated in the work by Demezhko and Shchapov (2001). These authors have inferred 80000-year long GST history from temperature-depth profile measured in 5.4 km deep hole located in the Middle Urals (Russia). Because only about eight months passed between the logging and the date when the drilling was ceased, the authors have applied correction for drilling disturbances to the measured data (see Section 2.1, Chapter 2). Similar to the above-described reconstructions, inferred GST history included the end of the Würm glaciation, rapid warming about 10000 years ago (beginning of the Holocene), the Holocene Climatic Optimum (6000-4000 B.P.), and the later major climatic oscillations, such as the Medieval Warm Period that culminated near 1000 A.D. and the Little Ice Age between 1500 and 1800 A.D. Estimated amplitude of the postglacial warming appeared somewhat higher than in the above-cited works and reached 12-13K. Authors ascribe this inconsistency to the effect of the long-term snow thickness variations (Section 2.6.2). New estimations of the amplitude of the postglacial warming have been made for the South Urals from the temperature-depth profiles measured in 4500 m deep borehole Leuza-1 (the Cis-Ural Trough) and in 2000 m deep Ilmenskaya-1 borehole (55.00°N, 60.17°E, 340ma.s.l.). Estimations of the postglacial warming reached approximately 11 and 8.5 K, respectively (Golovanova and Valiyeva, 2006; www.ig.cas.cz/activities/abstrakty_total1.htm).

The longest (100ka) GST histories were inferred from four deep (1600-2900m) temperature-depth profiles measured in boreholes situated within the area 50.2-55.5°N, 66.6-102°W in Canada (Rolandone et al., 2003). During the last glaciation this area belonged to the Laurentide11 ice sheet. Ice thickness at borehole locations varied from 1 to 3km at 21ka B.P. (Peltier, 2002). Inversion of the measured T-z profiles has yielded the temperature history at the base of the ice sheet and the surface temperature course after the glacial retreat. Calculated GST histories are generally consistent and contain information on the minimum temperature around 20-10ka B.P. Relatively rapid warming since then has culminated at 3-4ka, and can likely be attributed to the ice retreat. Further climatic events were cooling with minimum at 1700-1800 A.D. and subsequent warming. Obtained results also indicate slight spatial differences in the basal temperature history across the Laurentide ice sheet. Temperature at the base of the glacier depends on the accumulation and flow of ice as well as on the geothermal heat flow from the Earth's interior (for details see Section 2.9, Chapter 2). During the last glacial maximum basal temperatures were lower at the southeastern edge of the glacier than southwest of the glacier center. Terrestrial heat flow increases in the SE-SW direction from 34 to 51 mW/m2. According to the studies by Rolandone et al. (2003), the more rapid flow of ice at the southeastern edge of the ice sheet combined with the lower heat flow from below has resulted in colder conditions. The minimum temperatures were reached at different times from 8-12 to 20-30ka B.P. At all sites basal temperatures were above the melting point throughout 100ka reconstructed period. The GST values during the last glacial minimum range from —1.5 to +0.4°C. This fact could explain unstable nature of the ice sheet that has been derived from proxy sources (Dyke and Prest, 1982; Licciardi et al., 1998). For comparison, present bedrock temperatures measured beneath the central part of the Greenland ice sheet range between —8 and — 13°C (Dahl-Jensen et al., 1998). Some attempts of modeling physical processes and glaciological reconstruction of the Laurentide ice sheet (Liccardi et al., 1998; Marshall et al., 2000; see also the references therein) were made. The latter work has emphasized significant inconsistency in the simulated basal temperatures between models that likely occur due to poorly constrained parameters and physical processes. Results of the GST reconstructions can thus provide additional useful constraint for the glaciological reconstruction of the Laurentide ice sheet history.

All described long-term GST histories represent the local-scale reconstructions. As mentioned above, Beltrami and Burlon (2004) have shown that under the restrictions that are essential to obtain robust spatial averages on hemispheric or global scale, the merging of the individual GST reconstruction results generally cannot retrieve reliable information on the climate variations at times before 1500. The reason is a well-known decrease in the signal strengths and resolution of the "borehole" method into the past, smoothing of the resulting GST history obtained from the joint inversion or from the merging of individual GST histories calculated using temperature logs with different noise levels. Disturbances caused by remote climatic events can also be removed by the procedure of steady-state thermal gradient correction below 1 km depth. An attempt of the continental-scale reconstruction of the late Quaternary climate changes was made in uThe Laurentide ice sheet was a massive ice sheet that covered most of Canada and significant part of the northern US between approximately 90 and 18k years B.P.

the work by Huang et al. (1997). Authors have suggested the procedure of the GST reconstruction that was believed to avoid above restrictions. Their technique is based on the merging heat flow determinations at different depths in different boreholes and thus requires the availability of the heat flow determinations in deep boreholes. The fact is that two-thirds of all terrestrial heat flow measurements have been performed in less than 1 km deep holes. Situation in North America is even more critical. There 87% of the heat flow determinations originate from less than 500 m deep holes. This puts some limitations on the application of suggested method.

For their research Huang et al. (1997) have used more than 6000 continental heat flow measurements compiled by the International Heat Flow Commission (IHFC; www.geo.lsa.umich.edu/IHFC). Most of the data are concentrated in the mid-latitudes of North America and Eurasia; thus recovered from these data GST trend mainly characterizes climate excursions that occurred in the Northern Hemisphere. Available heat flow data were arranged according to the depth range over which the heat flow was determined, and then averaged over 50 m intervals (Figure 114). Calculated standard errors generally reflect the regional variability of the data. Observed slight increase of the standard deviation with depth reflects the diminishing of the heat flow determinations in

48 52 56 60 64 68

HEAT FLOW, mW/m2

Fig. 114. Global representative of the heat flow distribution with depth calculated from ~6000 heat flow data in 50m intervals. Shading means ± 1 standard deviation (data by Huang et al., 1997). Dashed line represents steady-state component of the heat flow.

48 52 56 60 64 68

HEAT FLOW, mW/m2

Fig. 114. Global representative of the heat flow distribution with depth calculated from ~6000 heat flow data in 50m intervals. Shading means ± 1 standard deviation (data by Huang et al., 1997). Dashed line represents steady-state component of the heat flow.

TIME, 103 years B.P.

Fig. 115. The 20000-year long GST histories reconstructed from the global heat flow distribution. The null hypothesis means that there was no climate change. Other three curves correspond to the progressive weighting of the data that allows more and more significant deviation from the null hypothesis. Arrows indicate the long-term average surface temperature before the reconstructed 20000-year period. (Redrawn from Huang et al. (1997).)

TIME, 103 years B.P.

Fig. 115. The 20000-year long GST histories reconstructed from the global heat flow distribution. The null hypothesis means that there was no climate change. Other three curves correspond to the progressive weighting of the data that allows more and more significant deviation from the null hypothesis. Arrows indicate the long-term average surface temperature before the reconstructed 20000-year period. (Redrawn from Huang et al. (1997).)

the deeper holes. The distribution of globally averaged heat flow with depth exhibits weak decreasing trend, however, with sizeable variations that do not show any clear pattern. Authors ascribed the detected heat flow variations to the climate change.

To reveal the corresponding global GST history, they converted heat flow-depth data into temperature-depth profile using expression similar to Eq. (5) (Chapter 2). The GST inversion was performed by means of FSI technique. Although temperature variations in this type of reconstruction are expected to be highly smoothed, an analysis of heat flow measurements as a function of depth yielded a broad enough set of climatic excursions over the last 20000 years (Figure 115). The null hypothesis means the complete absence of the past climatic change. To investigate the likely range of the GST variations, authors have demonstrated three examples of the possible GST histories that progressively enlarge the amplitudes of the inferred GST change by a greater weighting given to the data. The most reliable range for the GST changes is restricted by the curves "A" and "C". Of course, it is possible to obtain variations smoother than the curve "A", but they will require stronger forcing of the null hypothesis and less comprehensive inclusion of the information from observations. More pronounced amplitudes than curve "C" are unrealistic. Inferred long-scale climate course (Figure 115) is generally coherent with the climatic history presented in Figure 110 (left). The calculations indicated low long-term temperatures. The powerful recovery from the previous cold conditions began at least 20000 years ago. The early to mid-Holocene time appears as a relatively long warm period with its maximum around 8000-6000 years B.P. These warm conditions were changed by the strong cooling that culminated around 2000 B.P. and subsequent fingerprints of the Medieval Warm Period, the Little Ice Age, and the recent warming. Although being relatively smoothed, obtained GST history is clearly coherent with above-described local remote GST reconstructions as well as with the broad outlines of the late Quaternary climate changes revealed by proxy sources.

Recent investigations by Gosnold (2006) have shown that the amplitude of the postglacial warming in the Northern Hemisphere is of the order of 10-15 K. The author has also brought into attention of the borehole climatologists that the amplitudes of warming may significantly differ between both hemispheres. This study was based on a similar technique of the heat flow determinations in different depths at different boreholes as the previous work. The heat flow-depth variations were calculated from more than 1500 deep European boreholes. In North America, the number of deep holes is significantly lower, and similar depth distribution of the heat flow was obtained from 759 determinations of the IHFC Global Heat Flow Database (www.geo.lsa.umich.edu/IHFC). The author has shown that while average heat flow in Southern Hemisphere shields is approximately 61mW/m2 (this value falls close to the global estimate presented in Figure 114), this value in Northern Hemisphere shields of similar ages is equal to only 37mW/m2. According to Gosnold (2006), post-glacial warming may be among the main possible explanations of such a discrepancy. Other possible reasons may be of physical and/or chemical nature. For example, there is some evidence that crustal radioactivity is greater at sites of heat flow measurements in the Southern Hemisphere continents and thus could provide greater contribution to the heat flow. Further thorough investigations need to separate different effects and support the conclusion on the differences in the amount of the climate warming after the last glacial epoch in both hemispheres.

Recently, Demezhko et al. (2006) have performed an analysis of the spatial pattern of the Pleistocene-Holocene warming in Northern Eurasia based on the 48 GST reconstructions. This study has found that the amplitude of warming has increased from 8 to 23 K in the SE to NW direction from Greenland to the Urals. Maximum amplitudes were confined to the North Atlantic region. Authors have concluded that detected warming pattern may be a result of the formation of the present system of the warm surface currents in the North Atlantic (the Gulf Stream, North Atlantic, and Norwegian currents). The recent study of ocean circulation in the North Atlantic has found a 30% reduction in the warm currents that carry water north from the Gulf Stream. If the ocean current that gives western Europe its relatively soft climate is becoming weak, it is feared that it might shut down entirely and plunge the continent into a mini ice age12 (Marsh et al., 2005).

A significant part of the local remote GST reconstructions was obtained from permafrost areas. Numerous investigations have proved that temperature logs measured in boreholes drilled in permafrost represent a very useful tool for the reconstruction of remote climate changes. Because heat transfer within thick permafrost occurs almost exclusively by conduction, permafrost is affected primarily by the long-term temperature changes. Under proper choice of the mathematical description of the freezing/thawing process, an inversion of borehole temperatures from permafrost areas that nowadays occupy about 25% of the land in the Northern Hemisphere can provide reliable GST histories for the past few ten 10000 associated with the last glaciation. It was shown that at least 100ka long GST histories can be reconstructed from the 1.5 to 2km deep holes (see, e.g., Mottaghy and Rath, 2006; and the references therein). An ensemble of results reported from the borehole temperature measurements in vast regions of the Alaskan, Russian Arctic, Scandinavia, and European mountains has revealed ~ 17-18 K

temperature increase since the last glacial. The GST reconstructions have also proved that an impact of the glaciation and fingerprints of the past permafrost can be detected from borehole temperature logs in many permafrost free regions (for details see Section 2.8, Chapter 2). Similar to the permafrost boreholes, past ice surface temperature reconstructions from the temperature logs measured in the ice boreholes provide valuable estimates of remote temperature changes in polar environments that are complementary to paleo-climatic records obtained from ice core oxygen isotopes. Ice borehole information could spatially complete information provided by land boreholes. Boreholes drilled through Greenland and Antarctica ice-caps provide the most detailed picture of the timing of climate change over the last 700ka (for details see Section 2.9, Chapter 2).

### 3.5.2 Climate signature in superdeep boreholes

The success in inferring the remote climatic changes from the temperature data measured in deep boreholes has inspired borehole climatologists for further achievements. Superdeep boreholes belonging to the ICDP (www.icdp-online.de) attracted special attention of the "borehole" community dealing with remote climate changes. Scientific deep drilling is a valuable tool for understanding the Earth's structure and the ongoing processes. Information from the deep holes provides direct insight into the Earth's interior and can be used for critical testing of the geophysical/geological models. This information also represents an essential component for a responsible management strategy for the Earth's natural resources and environment. An interest in the superdeep drilling was also enhanced by the recent proposal to use deep holes (>5km) as the next-generation nuclear waste repositories.

The ICDP is a multinational program to further and fund geosciences in the field of the continental scientific drilling. Currently Austria, Canada, China, Czech Republic, Finland, Germany, Iceland, Japan, Mexico, Norway, Poland, South Africa, and the USA are its members through their National Funding Organizations and/or major research institutions. In addition, UNESCO and some international companies are the associated members. From the very beginning, geothermal and paleoclimatic investigations have appeared among the most important directions of the ICDP scientific research. Numerous processes occurring in the continental crust are temperature dependent. Measurements of subsurface temperature distribution and associated quantities (thermal conductivity, heat production, heat flow) are of vital importance to the understanding of these processes. Paleoclimatic directions include the following research fields: (1) the manner in which Earth's climate has changed in the recent as well as in the remote past and the reasons for these changes, and (2) the subsurface effects of major impacts on climate and mass extinctions.

The German KTB continental deep drilling program represents one of the primary and celebrated attempts of climate investigation in the superdeep holes. The drilling site is located at Oberpfalz area in NE Bavaria (Germany). This region is quite suitable for the study of deep-seated crustal processes. The drill site is located at the boundary between two major tectonic units of the Hercynian fold belt in Central Europe: the Saxothuringian and Moldanubian. The region represents a suture zone formed by the closure of an oceanic basin 320 million years ago. This process gave way to a continent-continent collision and the formation of the huge mountain chain comparable to the present extension of the Himalayas. Now the high mountain relief is eroded and previously deeply buried rocks are exposed at the surface (Burkhardt et al., 1989).

During the KTB project two deep boreholes were drilled: so-called pilot hole (VB, 4 km) and ultra deep main hole (HB, 9.1km). Both drill holes are located at a distance of only 200 m from each other. It is the unique constellation of two deep boreholes that are very close at one site. Both holes are drilled in the crystalline metamorphic rocks of the Hercynian continental collision zone, where the dominate rock types are paragneisses and metabasites. First studies for the KTB project began in 1978, and official inauguration of the KTB pilot hole occurred in 1987. The project has included collection, compilation, analysis, and interpretion of a high-quality dataset. Geothermal investigations represented significant part of the KTB scientific program. That time this deep-drilling project has produced one of the world's best collections of the geothermal data and provided a unique opportunity for the study of heat transfer processes in the deep continental crust.

Priliminary research on the GST changes on the 0.01Ma scale using the KTB-hole information has examined the temperature log measured in the 4 km deep pilot hole. The drilling ceased in 1989, and perturbation to the subsurface temperature field caused by drilling had almost entirely dissipated to the moment of the last temperature logging in 1996 (Huenges and Zoth, 1991). Figure 116 shows examples of the temperature logs measured in the KTB-VB hole. Full set of the temperature logs measured in both the KTB-VB and the KTB-HB drillholes can be found on the web site of the ICDP (www.icdp-online.de/sites/ktb). Except for the small-scale temperature oscillations, the most striking feature of the measured T-z profiles is their distinct non-linearity: the curve is concave. Temperature deficit relative to a linear T-z profile is especially pronounced in the depth range 0.5-3.5km. It was very enticing to attribute observed curvature to the remote climate change, and a number of forward models were simulated to interpret the curvature of the T-z profile from the KTB drilling site in this context (Rybach, 1992; Jobmann and Clauser, 1994; Kohl and Rybach, 1996). Numerical modeling has demonstrated that the subsurface temperature at the VB-hole bears a clear signature of the pale-oclimatic temperature change and quantitatively agrees with the reference climatic series of the last 0.1 Ma for Germany reconstructed by Zoth and Haenel (1988) on the basis of the proxy records. Clauser et al. (1997) have inverted temperature log measured in the 4km deep KTB pilot hole where temperatures probably were close to the original pre-drilling conditions (Figure 116). Even under simplified approach of the half-space with homogeneous thermal properties, the authors obtained reasonable well timing of the post-glacial warming with amplitude of nearly 10 K. It was also shown that concave shape of the KTB-VB could in principle be explained by the paleoclimatic effect alone.

Further investigations have revealed several factors contributing to the thermal field. Probably, one of the major findings of the KTB program was the discovery of the presence of free fluids at significant depths. The KTB researchers expected bone-dry deep crystalline rocks. To their surprise, fluid inflow occurred at several depths from open fractures. Numerous experiments/tests were performed to ascertain the properties of the hydrogeological system at the borehole site. Because the thermal regime in the KTB hole is possibly affected by the groundwater flow, several authors (e.g., Kohl and Rybach, 1996; Clauser et al., 1997) have investigated the thermo-hydraulic field near the KTB in 2- and 3-D approaches. All models were based on detailed knowledge of the geological structure in the drilling site and have taken into account vertical contrasts of rocks with

TEMPERATURE, deg.C

Fig. 116. Two temperature logs measured in the KTB pilot hole. (Data from the public database www.icdp-online.de/sites/ktb).

TEMPERATURE, deg.C

Fig. 116. Two temperature logs measured in the KTB pilot hole. (Data from the public database www.icdp-online.de/sites/ktb).

significantly different thermal conductivities as well as information on hydraulic properties which was necessary for an interpretation of advective heat transfer process. Forward numerical experiments by Kohl (1998), who used a complex 3-D transient model of the KTB site accounting for advection, topography, lithologic heterogeneities, and paleocli-matic GST variations, confirmed conclusion made by the earlier research that, together with the lithologic effects, the Pleistocene temperature changes induced by the last glaciation represent the most dominant influence on the temperature field in the KTB. The effect of the thermal advection by subsurface fluid movement is traceable but of minor importance. The author has demonstrated that even in the strongly advection-dominated systems that at certain depth ranges can significantly perturb conductive temperature distribution the paleoclimatic signal cannot be completely "washed out". On the other hand, the estimation of the paleoclimate fingerprints from advectively disturbed environments cannot be performed using pure conductive approach. For certain recovery of the paleo-climate signal the use of improved advection inversion techniques is indispensable (see Section 2.7, Chapter 2). However, in most of the field situations the construction of both realistic forward models and inversion parametrization schemes represents an extremely complex task because of the lack of sufficient field (especially hydrological) information. At present time, geothermal data from advectively disturbed boreholes can be used rather for the investigation of the temperature-dependent processes occurring in the continental crust than for paleoclimate reconstructions.

The Kola superdeep project represents similarly well-known deep-drilling effort. The Kola (SG-3) borehole site is located on the northern rim of the Fennoscandian (Baltic) shield near the Norwegian border at about the same latitude as Prudhoe Bay, whose GST

reconstruction results were described in Section 3.1.2 (Table 7). It was a Russian-funded project to drill deep into the Earth's crust. As in the case of the KTB, the Kola project planned wide-ranged geophysical/geological studies. The research areas were: (1) the deep structure of the Baltic Shield, the physical and chemical composition of the deep crust, and the hypothetical transition from granite to basalt; (2) lithosphere geophysics; (3) seismic discontinuities; and (4) the thermal regime in the Earth's crust.

The drilling of the main Kola hole began in 1970, and a number of boreholes were made from a central branch. The deepest of them (SG-3) reached its final depth of 12262m in 1994. It is currently the deepest borehole in the world and penetrates about a third through the Baltic continental crust. Extensive geophysical studies have been performed at the project site. Geophysical loggings and other measurements in this hole began almost immediately after the drilling was ceased. Undisturbed temperature-depth profile was measured there in 1998 after four years of continuous shut-in time of borehole (Popov et al., 1999). The Kola drill hole exhibits a considerable variation in the vertical component of heat flow density (Kukkonen and Clauser, 1994; Mottaghy et al., 2005). Measurements revealed significant growth of the vertical heat flow across the borehole. It is about 30mW/m2 in the uppermost 1 km and equals approximately 70mW/m2 at 4-5km depth. Observed variation in the vertical component of the heat flow cannot be attributed to the technical disturbances caused by the drilling procedures, but reflects the complex impact of three main natural processes.

The SG-3 hole is located at slightly elevated terrain (150-300ma.s.l.). Similar to the German KTB holes, the presence of free fluids was indicated in the Kola site down to a depth at least of some kilometers (Huenges et al., 1997). Forward 2-D numerical models by Kukkonen and Clauser (1994) were simulated using the vast available data on lithol-ogy, hydrogeology, topography, and the thermophysical structure in the area. Modeling results indicated that contrary to the KTB situation that archives significant paleoclimatic information, the main factors affecting the heat flow at Kola site are advective heat transport (especially in upper 2-3 km) and the complicated crustal structure. The area was covered by the Weichselian glaciation. Kukkonen and Clauser (1994) calculated the pale-oclimatic influence using the reference late Pleistocene and Holocene climate history as the forcing function. Their simulations have shown that paleoclimate influence appears to be considerably smaller than the advective and structural effects. Paleoclimatic disturbances to the heat flow decrease rapidly with depth from approximately16mW/m2 to less than 2mW/m2 at 1.5 km.

This conclusion was supported by the recent investigations of 36 shallow boreholes situated in the vicinity of the Kola SG-3 hole (Mottaghy et al., 2005). Except for the temperature logging, detailed studies of thermal conductivity as well as other important geophysical variables (density, specific heat capacity, radioactive heat generation rate, porosity, and permeability) on numerous samples were performed. Obtained data appear to be in good agreement with the corresponding quantities early measured for the superdeep SG-3 borehole. Detected heat flow values fall in the range of 31-45mW/m2. Moreover almost all boreholes exhibit significant increase of the heat flow with depth similar to that observed early in the SG-3 hole. For simulation and/or interpretation of observed regularities the authors constructed a realistic 3-D numerical model that incorporated wide amount of available data on the crustal structure at the SG-3 surrounding. Results of numerical trial runs have shown that at least in the upper 4 km advection heat transport is the main reason for heat flow growth and the crustal heterogeneity is of only secondary importance. On the other hand, the latter study has shown that at the deeper levels at least half of measured temperature disturbances occur due to the paleoclimate influence. The authors have concluded that the advection of heat by groundwater flow and the paleoclimate play significant role in the downward increase of the heat flow. As in the case of the KTB for the reliable reconstruction of the past climate history both effects should be interpreted together.

The above-described results of the geothermal investigations in two superdeep holes probably appear somewhat disappointing concerning the possibilities to infer remote climate change. It is clear as to why numerous measurements/investigations did not provide expected results. Borehole sites were chosen for mainly geotectonic reasons. Due to the lack of previous attempts, both the technical experience of drilling to a great depth and the knowledge of the deep crustal conditions were insufficient in the beginning. For example, numerical ingenuities were applied during the Kola drilling experiment. The main innovation was that, instead of turning the drill bit by rotating the stem, in the Kola well the bit alone was turned by the flow of drilling mud. Thus, it became possible to eliminate rotation of the entire drill string above. The researchers expected to find highly compact rocks at the deep crustal levels. On the contrary, deep rocks were strongly fractured and saturated with water. Initially it was planned that the Kola superdeep hole would be 15000m deep. However, mainly because of the higher temperatures that reached 180°C instead of expected 100°C, the final depth did not approach even 13000 m (after 24-year drilling). Further penetration down to 15000 m would have meant working at approximately 300°C, and the drill bit could no longer work at such conditions.

However, the abundance of the potential sites for the superdeep borehole research and the increased experience in the deep hole drilling permanently support the interest in such studies. An international workshop on continental scientific drilling was held at the GeoForschungZentrum, Potsdam, Germany, from March 30 to April 1, 2005. The purposes of the workshop were: (1) to review and summarize the achievements of the last decade of the ICDP, and (2) to define the opportunities for the future drilling projects addressing a broad set of topics in the earth sciences. The potential sites included subduction zones at the Izu Peninsula (Japan) and/or at Crete, the greatest continental collision zone in the Nanga Parbat region of the Himalayas, etc. The "Climate Change and Global Environment" was declared among the most important scientific research priorities in future. Some of the superdeep drilling programs are currently operating. Thus, e.g., since the 1990s the crater Chicxulub on the Yucatan Peninsula, Mexico, represents the area of extended geophysical and geological research (Hildebrand et al., 1991; Steinich and Marin, 1997). It is assumed that this structure resulted from the impact on the Earth of a large (more than 10km in diameter) asteroid or comet (Dressler et al., 2004). The study of the impact structure with a diameter of 180-200km and a center at the port of Chicxulub involved drilling of eight cored UNAM (Universidad Nacional Autónoma de México) boreholes inside the crater and in its immediate vicinity (Urrutia-Fucugauchi et al., 1996) with a depth range between near 60 (UNAM 4) and 700 m (UNAM 7) and culminated by drilling the 1.5 km deep borehole Yaxcopoil-1 (YAX-1) of the Chicxulub Scientific Drilling Project that represents a part of the International Continental Deep Drilling Program (Urrutia-Fucugauchi et al., 2004).

The precise high-resolution temperature logging was repeated nine times in the period March 2002-February 2004 (Wilhelm et al., 2003, 2004; Popov et al., 2004).

The long-term GST reconstructions using the temperature-depth profiles measured in the deep-drilled boreholes in Fennoscandia are of special interest. Two-millennia long GST histories reconstructed in Finnish boreholes have been already presented in Section 3.1.1. The study confirmed certain incoherence of the climatic history in Finland with the results obtained for other parts of Europe. The early section of the reconstructed GST histories covers a cold period between approximately 400 and 1000 A.D., followed by a long gradual warming up to 1500-1700 A.D. and a cold period around 1800 A.D. followed by strong subsequent warming. The fifteenth to sixteenth century warming in Finland appears to be different from the Little Ice Age conditions reported, e.g., for central Europe. During the Weichselian period this area was covered by glaciers. The analysis of the depth dependence of the heat flow in the Fennoscandian Shield and the neighboring parts of the East European Platform has shown its systematic variations with depth (Kukkonen and Joeleht, 2003). The authors have attributed these variations to the long-term climate change during the late Pleistocene glaciation and the Holocene. Inversion of the temperature-depth profiles from a suite of boreholes have shown that the lowest temperatures occurred during the last glacial maximum (—20000 years B.P.) and were followed by the average warming of 8.0 ± 4.5 K approximately 10 000 years B.P. Kukkonen et al. (1994) have estimated the 10000-year long GST history using T-z data measured in approximately 1 km deep borehole in Lavia, SW Finland. Their reconstruction revealed three steps of the long-term GST history in the region: (1) rapid recovery from the previous cold conditions about 9000 years ago that can be attributed to the retreat of the Weichselian ice sheet, when the temperature increased by approximately 4K, (2) the warm period that continued from 8000 to 5000years B.P., and (3) approximately 1K further warming that occurred at the beginning of the twentieth century. This GST history coincides well with the climate course after the latest ice age obtained for southern Finland on the basis of the proxy data (Donner, 1974).

The Geological Survey of Finland (GTK) is currently running the Outokumpu Deep Drilling project (www.gsf.fi/projects/o_k_deepdrilling). Drilling at Outokumpu (eastern Finland) site began in April 2004 and was successfully completed in January 2005 at the final depth of —2500m. The site under investigation belongs to the Paleoproterozoic formation that is well known for its polymetallic massive sulfide ore deposits. It is also one of the oldest ophiolitic formations all over the world. The main reason that has motivated this deep drilling project was the investigation of the deep structure of a classical ore province in the stable Precambrian terrain. On the other hand, it is expected that the Outokumpu deep hole will provide a wide range of research possibilities in numerous scientific branches. The Outokumpu hole is expected to be a deep geolaboratory for various in situ experiments. Among other important activities the research program includes carrying out numerous down-hole temperature logging experiments. The main goal of these measurements is inferring the GST history during the Weichselian glaciation and the Holocene from the geot-hermal data. Except for the possible new paleoclimatic contributions, results of the planned measurements are expected to provide an improved understanding of heat transfer and fluid flow in the crystalline bedrock. It can be believed that geothermal measurements at the Outokumpu deep hole will turn this site into a reference example of the paleoclimate reconstruction and the heat transfer regimes in the Precambrian crystalline crust.

These two efforts are neither the first nor the last attempts at drilling superdeep boreholes. The potential of the deep/superdeep holes for the borehole climatology is likely not fully revealed. The selection of scientifically useful sites, drilling of superdeep holes, and their investigation are ongoing and may present with unexpected discoveries.

The results described above confirm the possibility to extend the GST history back to the last glacial period. Due to their significant depth extension, the deep holes can reveal information about remote climate changes. In spite of somewhat lower resolving power of the geothermal method compared to most of the proxy indicators and its significant reduction on the timescales of 10000 years and longer, it is evident that obtained during numerous research efforts GST histories contain clear fingerprints of the last Pleistocene glacial/interglacial transition. The latter event possibly represents the most dominant signal archived in the T-z profiles measured at European deep boreholes. The consistency between GST histories obtained from borehole temperature logs by various authors as well as their coherence with existing independent climatologic records gives strong support to the possibility of using the borehole thermometry database to extract information on remote climate changes.