3.1 Data Collection
GST history reconstructions require high-quality data obtained from boreholes. Abandoned, low yield boreholes without any mechanical attachments (pumps etc.) are readily available for temperature-depth measurements. However, only a few boreholes can provide high-quality data although they fit the description. The most common problem is the groundwater flow. Secondly, even the groundwater activity is negligible, intra-borehole vertical flow due to differences in pressure heads can disturb background thermal conditions in a borehole (Erkan, 2015). Many measurement sites had to be eliminated due to these effects.
Spatial distribution of the borehole sites used for GST modeling is shown in Fig. 1. Data collection was completed in three phases. The first phase took place in September 2020 and boreholes in Kastamonu Province were investigated. Kastamonu is the largest province in the region (13064 km2). Among the 11 boreholes investigated in field studies, three of them yielded suitable temperature-depth profiles that were suitable for climate reconstructions. The second phase took place in September 2021, and boreholes in Bartin, Bolu, Duzce, Karabuk and Zonguldak provinces were investigated. Field studies on eight boreholes resulted in one suitable borehole in Bartin for climate reconstruction. Finally, the third phase of the study was completed in August 2022. The third phase covered Bolu and Bartin provinces with eight borehole sites. One borehole in Bolu yielded suitable results for climate reconstruction. The borehole in Bartin which was logged in 2021 was also relogged in 2022.
As shown in Fig. 1, boreholes in Kastamonu are located in Dizdarli, Devrekani and Hasirli Districts. These districts are located in inland parts of Kastamonu province. Each location has its distinct topographic characteristics. Dizdarli is located on the western region of Kastamonu. The studied borehole is located near a village on a hill. Devrekani and Hasirli Districts are located about 20km north of Kastamonu city center, south of Kure mountains. The borehole in Hasirli is located near a village, with a vast open space around it with very light vegetation. The borehole in Devrekani is located on a small hill with no vegetation and it is 500 meters away from the farmyards. The borehole in Bartin province is located about 8 km south of the central district, in an area that is mostly grassland. It is inside an orchard that was planted in 2017. Finally, borehole in Bolu province is located near Akoren village. The village is located on an open area. The properties of the boreholes are shown in Table 1.
Table 1
Properties of the studied boreholes and sites. H: elevation, LD: logging date as month.year, BD: borehole depth, WT: depth of water table, D: borehole diameter, T0: undisturbed GST at time = 0,λ: thermal conductivity, qb: background heat flow
Location
|
Latitude
|
Longit
|
H
m
|
LD
|
BD
m
|
WT
m
|
D
cm
|
T0
°C
|
λ
W/(m·K)
|
qb
mW/m2
|
Lithology
|
Akoren
|
40.9699
|
32.0531
|
865
|
08.22
|
115
|
15
|
20
|
12.01
|
2–3
|
32-47.5
|
Limestone (0-115)
|
Bartin
|
41.5633
|
32.3932
|
42
|
08.21
|
210
|
5
|
20
|
13.83
|
2–3
|
57.3–85.8
|
Marl (0-210)
|
Devrekani
|
41.6074
|
33.8432
|
1087
|
09.20
|
170
|
80
|
20
|
10.15
|
2.76
1.90
|
38.5
|
Limestone (0–48)
Diorite (48–170)
|
Dizdarli
|
41.6733
|
32.9234
|
885
|
09.20
|
87.5
|
15
|
20
|
8.87
|
2–3
|
15.3–22.4
|
Limestone (0–95)
|
Hasirli
|
41.6367
|
33.9837
|
1185
|
09.20
|
95
|
60
|
20
|
10.27
|
2.89
|
20.79
|
Marble (0-110)
|
Temperature logging device includes a thermistor sensor that can obtain high-resolution temperature-depth data (Erkan et al., 2017). During the measurement process, temperature logging probe was released into the boreholes and temperature and depth data were recorded with 2.5–5 meter logging intervals. When the probe is in water, about 20 seconds was employed to reach equilibrium at each logging depth. On the other hand, measurements above static water level needed to be approached differently as sensitive temperature logging probe took longer to reach equilibrium in air. Therefore, a time-lapse measurement procedure was necessary for measurements above the water level. During these measurements, temperatures at each 30-sec intervals were recorded for five minutes at each depth (Fig. 2).
For extrapolation of the time-lapse temperatures to the final equilibrium temperature, the following equation was used (Costain, 1970):
$$\text{T}={\text{T}}_{\text{f}}\left(1-{\text{e}}^{-\frac{\text{t}}{{\tau }}}\right)+{\text{T}}_{0}{\text{e}}^{-\frac{\text{t}}{{\tau }}}$$
where t is the time of recording, T is the temperature at time t, Tf is the final temperature, T0 is the initial temperature and τ is the time constant of the probe (taken as 3.5 minutes for our probe). Equilibrium temperature value (Tf) was then obtained by applying generalized least squares (GLS) method (Chapra and Canale, 2011) between the data and the equation above. For the detection and elimination of the problematic measurements, we applied a two-step procedure. First, we accepted the results at a certain depth satisfying |Tf-T0| ≤ 0.02 as this upper bound is within the practical limit of temperature measurements in the field. For the measurements in the remaining depths, we enforced that the standard error (SE) of the GLS fit is ≤ 5%. These thresholds were selected in compliance with the error margin used in the paleoclimatic reconstruction model discussed below.
Figure 2 shows example temperature measurements from Devrekani that were taken above the water table level. For 30 meters (Fig. 2-a), |Tf-T0| value is 0.04 which is above the threshold but SE is found as 4% which is below 5% limit. Therefore, Tf value of 10.793°C that is calculated by our algorithm is acceptable. As another example, Fig. 2-b shows temperature measurements and curve fit data from 45-meter temperature-depth log data. |Tf-T0| value here is found as 0.017 which is below the threshold so calculated value of 10.883°C is acceptable for Tf.
Temperature-depth data used as the input for GST modeling are shown in Fig. 3. Except Deverakani and Hasırlı, all data were recorded below the water table level. In the subsurface, short-term annual variations in surface temperatures are generally effective down to depths of ~ 20m (Harris and Chapman, 1998; Erkan et al., 2019). Our repeated T-D measurements in Bartin that is one year apart also showed that seasonal variations can be effective down to around 25 meters Therefore, temperature logs influenced by short-term annual temperature variations were omitted from GST model inputs.
Variations of the topography can have significant static effect on the background geothermal gradients and hence the background heat flow. The boreholes used for GST modeling are located with insignificant topography except Dizdarlı. As shown in Fig. 5, during generating the transient data (data after the static conditions are removed) the effect of topography on the background conditions are automatically eliminated. Hence, no topographic correction was applied for Dizdarlı site
3.2 GST Reconstructions
Ground surface temperature reconstructions were performed by utilizing the Functional Space Inversion (FSI) algorithm (Shen and Beck, 1991). FSI algorithm is based on functional space analysis instead of conventional discrete formulation; the solution is provided by using the iterative gradient method. FSI model requires temperature-depth profile, thermal properties of subsurface layers, background heat flow and undisturbed surface temperature as input. In our study, temperature-depth profiles were determined by instrumental loggings as described in the data collection section. Thermal properties of subsurface layers were determined by either measuring collected rock samples or using literature values based on lithological descriptions (see Table 1). Background heat flow (qb) was calculated by taking the deepest section of the temperature-depth log where temperature-depth profile shows a linear trend and can be defined as the undisturbed zone. According to Fourier’s Law:
where d1 and d2 are two depth points in linear portions of the temperature-depth curve, T1 and T2 are temperatures measured at depths d1 and d2 and finally, λ is the thermal conductivity (W/m/K) of the medium. Thermal conductivity values for Hasirli and Devrekani were determined by instrumental measurements of collected rock samples (Table 1) and remaining thermal conductivity values were obtained from literature after lithological identification of the collected rock samples (Gul at al., 2006, Balkan et al., 2017). In Bartin, Akoren and Dizdarli boreholes thermal conductivity values were taken as range of values due to unavailability of thermal conductivity measurements. Undisturbed surface temperatures (“pre-observational mean GST”, Harris and Chapman, 1997) were determined by extending background T-D profiles (see Fig. 3) to the surface. FSI model processes these data as inputs and produces a posteriori temperature-depth and background heat flow values. These products are used to test the model accuracy by comparing them with the observed values. If a priori and a posteriori profiles are in agreement, reconstructed GST values are accepted as the confirmed model.