In this study, equations are developed that predict for synthetic sedimentary rocks (clastics, carbonates and evapourates) thermal properties comprising thermal conductivity, specific heat capacity and thermal diffusivity. The rock groups are composed of mineral assemblages with variable contents of 15 major rock-forming minerals and porosities of 0–30 per cent. Petrophysical properties and their well-logging-tool-characteristic readings were assigned to these rock-forming minerals and to pore-filling fluids. Relationships are explored between each thermal property and other petrophysical properties (density, sonic interval transit time, hydrogen index, volume fraction of shale and photoelectric absorption index) using multivariate statistics. The application of these relations allows computing continuous borehole profiles for each rock thermal property. The uncertainties in the prediction of each property vary depending on the selected well-log combination. Best prediction is in the range of 2–8 per cent for the specific heat capacity, of 5–10 per cent for the thermal conductivity, and of 8–15 for the thermal diffusivity, respectively. Well-log derived thermal conductivity is validated by laboratory data measured on cores from deep boreholes of the Danish Basin, the North German Basin, and the Molasse Basin. Additional validation of thermal conductivity was performed by comparing predicted and measured temperature logs. The maximum deviation between these logs is <3 °C. The thermal-conductivity calculation allowed an evaluation of the depth range in which the palaeoclimatic effect on the subsurface temperature field can be observed in the North German Basin. This effect reduces the surface heat-flow density by 25 mW m⁻².

Data on rock thermal conductivity (TC) are important for the quantification of the subsurface temperature regime and for the determination of heat flow. If drill core is not retrieved from boreholes and thus no laboratory measurement of TC can be made, other methods are desired to determine TC. One of these methods is the prediction of TC from well logs. We have examined the relationships between TC and standard well-log data (gamma ray, density, sonic interval transit time, hydrogen index and photoelectric factor) by a theoretical analysis and by using real subsurface data from four boreholes of the North German Basin. The theoretical approach comprised the calculation of TC from well-log response values for artificial sets of mineral assemblages consisting of variable contents of 15 rock-forming minerals typical for sedimentary rocks. The analysis shows different correlation trends between TC and the theoretical well-log response in dependence on the mineral content, affecting the rock matrix TC, and on porosity. The analysis suggests the development of empirical equations for the prediction of matrix TC separately for different groups of sedimentary rocks. The most valuable input parameters are the volume fraction of shale, the matrix hydrogen index and the matrix density. The error of matrix TC prediction is on the order of 4.2 ± 3.2 per cent (carbonates), 7.0 ± 5.6 per cent (evaporites), and 11.4 ± 9.1 per cent (clastic rocks). From the subsurface data, comprising measured TC values (n = 1755) and well-log data, four prediction equations for bulk TC were developed resembling different lithological compositions. The most valuable input parameters for these predictions are the volume fraction of shale, the hydrogen index and the sonic interval transit time. The equations predict TC with an average error between 5.5 ± 4.1 per cent (clean sandstones of low porosity; Middle Buntsandstein), 8.9 ± 5.4 per cent (interbedding of sandstone, silt- and claystones; Wealden), and 9.4 ± 11 per cent (shaly sandstones; Stuttgart Fm.). An equation including all clastic rock data yields an average error of 11 ± 10 per cent. The subsurface data set also was used to validate the prediction equation for matrix TC established for clastic rocks. Comparison of bulk TC, computed from the matrix TC values and well-log porosity according to the geometric-mean model, to measured bulk TC results in an accuracy <15 per cent. A validation of the TC prediction at borehole scale by comparison of measured temperature logs and modeled temperature logs (based on the site-specific surface heat flow and the predicted TC) shows an excellent agreement in temperature. Interval temperature gradients vary on average by <3 K km^–1 and predicted compared to measured absolute temperature fitted with an accuracy <5 per cent. Compared to previously published TC prediction approaches, the developed matrix and bulk TC prediction equations show significantly higher prediction accuracy. Bulk TC ranging from 1.5 to 5.5 W (m K)^–1 is always predicted with an average error <10 per cent relative to average errors between 15 and 35 per cent resulting from the application to our data set of the most suitable methods from literature.

We compile and analyze all available geothermal heat flow measurements collected in and around Greenland into a new database of 419 sites and generate an accompanying spatial map. This database includes 290 sites previously reported by the International Heat Flow Commission (IHFC), for which we now standardize measurement and metadata quality. This database also includes 129 new sites, which have not been previously reported by the IHFC. These new sites consist of 88 offshore measurements and 41 onshore measurements, of which 24 are subglacial. We employ machine learning to synthesize these in situ measurements into a gridded geothermal heat flow model that is consistent across both continental and marine areas in and around Greenland. This model has a native horizontal resolution of 55 km. In comparison to five existing Greenland geothermal heat flow models, our model has the lowest mean geothermal heat flow for Greenland onshore areas. Our modeled heat flow in central North Greenland is highly sensitive to whether the NGRIP (North GReenland Ice core Project) elevated heat flow anomaly is included in the training dataset. Our model's most distinctive spatial feature is pronounced low geothermal heat flow (< 40 mW m⁻²) across the North Atlantic Craton of southern Greenland. Crucially, our model does not show an area of elevated heat flow that might be interpreted as remnant from the Icelandic plume track. Finally, we discuss the substantial influence of paleoclimatic and other corrections on geothermal heat flow measurements in Greenland. The in situ measurement database and gridded heat flow model, as well as other supporting materials, are freely available from the GEUS Dataverse (https://doi.org/10.22008/FK2/F9P03L; Colgan and Wansing, 2021).