Oil sand, also known as bitumen-saturated sand, is a mixture of sand, clay, bitumen (or heavy oil), and water^[1]. The oil sand containing heavy oil, super-heavy oil, or bitumen is an important kind of unconventional resource with massive reserves widely distributed all over the world^[2]. Currently, conventional recovery technologies of oil sand include open-pit mining, cyclic strengthened steam injection (CSS), steam-assisted gravity drainage (SAGD), and cold production, etc^[3,4]. Among them, the SAGD and its derivative techniques have been widely applied in the development of oil sand and super- and extra-heavy oils. Specifically, the SAGD development must be achieved through pair horizontal wells (i.e. a steam injection well and a production well). The horizontal section of the production well is located at the bottom of the reservoir, and that of the steam injection well is located about 5m above that of the production well. High temperature (> 200 °C) steam is injected into the reservoir through the steam injection well. After being heated, the bitumen or heavy oil in the oil sand will decrease sharply in viscosity and finally turn into a flowable liquid. Due to the density difference between the liquid phase and gas phase, the steam will migrate to the upper zone of the reservoir and form a continuously expanding steam chamber. Meanwhile, the crude oil with lower viscosity will flow to the production well along the edge of the steam chamber under the effect of gravity^[5,6,7,8]. In actual production, the development range of the steam chamber is mainly affected by factors such as reservoir heterogeneity, interlayer, and formation water distribution, etc, so the steam chambers develop in various shapes and are difficult to predict^{[9,10,11,12,13,14,15]}.

Oil sands are usually buried shallow, with the characteristics of high porosity, high permeability, weak cementation, and strong attenuation^[16]. In the process of oil sand SAGD development, on the one hand, with the continuous injection of high-temperature steam, the heavy oil in the oil sand will change significantly in density, viscosity, and phase state, leading to changes in the coupling relationship between heavy oil and sand matrix and elastic parameters of the reservoir; on the other hand, the injected high-temperature steam also changes from gas phase to liquid phase. The changes in the temperature field of the steam chamber will lead to great changes in rock physical parameters and obvious differences in time-lapse seismic responses. In recent years, the time-lapse seismic technology has developed rapidly and improved in applicability continuously^{[17,18,19,20,21,22,23,24,25,26,27,28,29,30]}, but didn’t work well in dynamic monitoring of steam chamber development during SAGD development of oil sand. One of the main reasons is that the physical characteristics of oil sand sensitive to temperature have complex variation laws with the increase of temperature and are difficult to measure at high temperatures in the laboratory.

In the respect of experimental research on temperature-sensitive physical characteristics of oil sand, Han et al.^[31] first found that the phase state and properties of heavy oil were mainly affected by temperature; and with the increase of temperature, the heavy oil gradually changed from glassy state to quasi-solid and liquid state. According to the temperature points of glass state and liquid state, the heavy oil is divided into three stages: glass state, quasi-solid state, and liquid state. Glass state heavy oil basically appears as solid and strengthens the rock skeleton. Quasi-solid heavy oil with high viscosity can support the rock skeleton but shows strong dispersion. Liquid heavy oil decreases sharply in viscosity and can flow completely. Li et al.^[32] carried out experiments on physical properties sensitive to temperature by using glass bead skeleton and mineral skeleton after extracting heavy oil respectively. They found that these two oil sand samples decreased in P- and S-wave velocities gradually with the increase of temperature, but as the mineral skeleton after heavy oil was extracted still had certain support, its overall velocities are higher than those of the glass bead sample. Yuan et al.^[33] made ultrasonic tests on oil sand samples taken in situ and after steam flooding in the temperature range of 10-129 °C, and found that P- and S-wave velocities of the samples gradually decreased with the increase of temperature, but the change rates of P-wave and S-wave velocities were various in different temperature sections.

Based on the existing rock physical experimental methods and understandings on oil sand, firstly, in light of the low impedance characteristic of oil sand, we changed the material of the probe base to improve the energy and signal-noise ratio of the transmission wave. Also, the maximum temperature in the experiments of this study was increased to 175 °C. Based on the ultrasonic measurements of typical oil sand samples from Kinosis, Canada and calculations with the theoretical model, combined with logging data, the variation laws of parameters such as P-wave velocity, S-wave velocity, and density, and frequency dispersion characteristic of the oil sand with temperature and pressure have been analyzed. Then the quantitative relationship between P-wave impedance and temperature of the oil sand has been constructed for the time-lapse seismic monitoring of the steam chamber in the SAGD development of oil sand in the study area.

This experiment mainly measured the density and P-wave velocity parameters of heavy oil in oil sand under varying temperatures, as well as the porosity and P- and S-wave velocities of the oil sand. The measurement of oil sand porosity is based on Archimedes' principle. The specific process is as follows: firstly, the sample volume was obtained based on Archimedes principle; then, combined with the known sample mass, the sample density was calculated; finally, based on the density of known mineral and pore fluid, the in-situ porosity of the oil sand was calculated. The ultrasonic pulse transmission method was used to measure the P- and S-wave velocities of the heavy oil and oil sand samples. The ultrasonic source is generally a piezoelectric crystal, which can not only send out sound waves but also receive sound waves, convert the received sound waves into a corresponding voltage, amplify and display the waveform through the acoustic instrument receiver. From the first arrival picked, the propagation time of sound waves in the medium can be obtained and the wave velocity can be calculated.

The rock ultrasonic measurement system used in this experiment (Fig. 1) mainly included four units: pressure control, sample clamping, temperature control, and signal recording. To heat the oil sand sample, the pressure kettle was wrapped by a layer of heating cotton. The pressure control system used two digital pressure pumps to control the confining pressure and pore pressure of the sample respectively. The pore pressure was usually smaller than the confining pressure to ensure the tightness of the sample. By using the sample clamping system, the base of the probe can be closely coupled with both ends of the sample, and then the sample and the probe base can be wrapped and sealed with a retractable rubber sleeve to prevent pore pressure oil from invading the sample. The system can measure the temperature of 10-175 °C and the pressure of 0-100 MPa.

As oil sand has weak cementation, easy deformation, and low impedance (the P-wave impedance of oil sand in the study area is about 5×10⁶ kg/(m²·s)), if the traditional ultrasonic probe base made of titanium alloy (with the P-wave impedance of about 27×10⁶ kg/(m²·s)) is used in ultrasonic measurement, there would be the following problems: (1) As big difference of wave impedance exist between common titanium alloy base and the oil sand sample, and the transmission wave signals would be weak. (2) As heavy oil in oil sand has a high viscosity, ultrasonic signals would have strong attenuation, resulting in unclear first arrival. Therefore, according to the principle of wave reflection and transmission, in this work, the PEEK resin with low impedance (p-wave impedance of about 4.5×10⁶ kg/(m²·s)) and high-temperature resistance (thermal conductivity changes linearly with the temperature below 250 °C) was used to make the base of the ultrasonic probe so that more transmission waves energy could pass through the sample and first arrival could be obtained clearly. This improved the signal-noise ratios of P-wave and S-wave waveform signals and the reliability of first arrival pickup to some extent.

According to the mineral composition and physical properties of the reservoir rock in the study area, we first selected rock samples with 10 different oil saturations (0-90%). Secondly, the mineral composition, bulk densities, porosities, etc of the samples were measured, and the micro-pore structures of the samples were scanned by the CT. Thirdly, the typical oil sand samples were washed to obtain heavy oil samples, densities, and ultrasonic (0.8 MHz) P-wave velocities of the heavy oil at different temperature points were measured under atmospheric pressure. Then the S-wave velocities of the heavy oil at the temperature points were calculated using the FLAG software. Finally, the ultrasonic P- and S-wave velocities of the oil sand samples under different temperatures and pressures were measured. According to the actual temperature and pressure ranges of oil sand in the study area (with the maximum temperature of 220 °C and the maximum pressure of nearly 4 MPa), the temperature variation range measured in this study was 10-175 °C, the pressure variation range was 0.5-3.5 MPa, and the temperature and pressure measurement intervals were 10 °C and 0.5 MPa, respectively.

According to the actual production situation in the study area, the experiments in this work were designed in a narrow pressure range to test the temperature-sensitive properties of the heavy oil and oil sand samples. Specifically, the experiments aimed to test the variation trend of density and elastic wave velocity parameters with temperature, dispersion effect, and phase state transition temperature points of the heavy oil; variation patterns of elastic wave velocity with the temperature of the oil sand samples with different oil saturations. Based on the results of oil sand temperature sensitivity experiments, logging data, and the Cole-Cole attenuation model, the quantitative relationships between P-wave impedance and temperature of oil sand samples with different oil saturations have been constructed to guide the time-lapse seismic inversion of the temperature field of the steam chamber during the steam-assisted gravity drainage process.

According to the experimental results, the temperature-sensitive characteristics of four typical oil samples were analyzed. All four samples have quartz as the main component (Table 1). Among them, the LLD-1 sample had the lowest quartz content (80.2%) and the highest clay content (12.2%). The samples had a porosity range from 36.9% to 40.1%, and an oil saturation range from 31.2% to 81.4% (Table 2). According to the classification standard of oil saturation, LLA-2 sample is oil sand with high oil saturation (greater than or equal to 70%), LLA-1 and LLD-1 samples are oil sand with medium oil saturation (greater than or equal to 50% and less than 70%), and LLB-2 sample is oil sand with low oil saturation (greater than or equal to 20% and less than 50%).

Based on previous experimental results, heavy oil would change significantly in-phase and physical properties with the increase of temperature, which is the most important factor affecting the temperature-sensitive properties of the oil sand. Therefore, the variations of heavy oil physical properties with temperature were analyzed first.

Fig. 2 shows the tested results of density, P-wave velocity, and S-wave velocity of the heavy oil with different temperature. Fig. 2a shows that the density of heavy oil decreases linearly with the increase of temperature. The main reason is that the heavy oil will expand in volume with the rise of temperature. Fig. 2b and Fig. 2c show that the P- and S-wave velocities of heavy oil gradually decrease with the increase of temperature, and velocity change rates are various at different temperature sections. The variation trends of heavy oil density, P- and S-wave velocity with temperature reflect phase state variation characteristics of the heavy oil.

Fig. 3 shows the FLAG simulation results of P- wave and S-wave velocities of the heavy oil. It can be seen that the P-wave and S-wave velocities of the heavy oil decrease sharply with the increase of temperature and show strong frequency dispersion between the glassy temperature point (about -34.4 °C) and the liquid temperature point (about 49.0 °C). For example, at about 10 °C, when the frequency increases from seismic frequency band (50 Hz) to the ultrasonic frequency band (0.8 MHz), the P-wave velocity gradually increases and shows a frequency dispersion of about 25%; the S-wave velocity has more significant frequency dispersion and can increase by 5 times. But when the temperature rises gradually and exceeds the liquid temperature point of the heavy oil, the velocity frequency dispersion of heavy oil basically disappears, and the heavy oil completely exists in the form of pore fluid.

Fig. 4 shows that for the high oil saturation sample LLA-2, the P-wave velocity gradually decreases with the increase of temperature, this is mainly because the heavy oil gradually changes from glass state to liquid state with the increase of temperature. When the temperature is 10°C, the P-wave velocity is at the highest of about 2500 m/s; when the temperature is 175 °C, the P-wave velocity is about 1250 m/s. In the measurement temperature range (10-175 °C), the decreased amplitude of P-wave velocity decreases gradually with the increase of temperature. The S-wave velocity has an overall variation trend similar to P-wave velocity. In addition, as the pressure increases from 0.5 to 3.5 MPa, the P-wave and S-wave velocities gradually increase in the low-temperature range (< 120 °C), but hardly change in the high-temperature section (> 120 °C).

The sample LLA-1 with medium oil saturation and the sample LLA-2 with high oil saturation have similar burial depth (about 307 m) and mineral composition, and thus similar variation trend of elastic wave velocity with temperature. But at 10 °C, the P-wave velocity of LLA-1 is higher, at about 2800 m/s (Fig. 5). From the analysis of CT scanning results, the reason is that the sample LLA-1 has good grain contacts.

Sample LLD-1 is also a sample with medium oil saturation. Compared with sample LLA-1, it has higher P-wave and S-wave velocities generally (Fig. 6). This is because that this sample has lower oil saturation (60.1%), slightly higher density (2.14 g/cm³), higher clay content (12.2%), and tight cementation between the grains from CT scanning.

With lower oil saturation (31.2%), the sample LLB-2 has smaller variation amplitudes of P- and S-wave velocities with temperature (Fig. 7). When the temperature rises from 10 to 175 °C, its P- and S-wave velocities reduce by about 800 m/s and 410 m/s, which is about 30% and 31% respectively. In contrast, the P- and S-wave velocities of the other three samples decrease by about 50% and 40%, respectively.

Based on the tested densities, porosities, and P-wave velocities of the oil sand samples and heavy oil, as well as logging data of the study area, the relationships between density, velocity, and temperature in the whole temperature range (0-250 °C) were predicted by Cole-Cole attenuation model, and the frequency dispersion was corrected to construct the quantitative relationship template between P-wave impedance and temperature of the oil sand samples in the seismic frequency band (Fig. 8). This template shows the variation trends and ranges of P-wave impedance with the temperature of oil sand samples with different oil saturations. Based on this template and oil saturation data of the target layer from logging, the data volume of time-lapse seismic P-wave impedance can be converted into temperature field to monitor the development of the steam chamber in SAGD development of oil sand.

The main oil sand target layer in the study area is the lower Cretaceous McMurray formation, which is composed of meandering river deposits affected by tide under the background of the estuary. The layer is at the buried depth of 150-350 m and about 30 m thick on average. The reservoir rock is mainly quartz sand, in which bitumen and clay minerals fill between skeleton particles. According to the results of the core experiment and physical properties interpreted from logging data, the oil sand is characterized by high porosity, high permeability, low density, and low P-wave and S-wave velocities. Under the original formation conditions (temperature of 7 °C and pressure of 1 MPa), the oil sand layer has an average porosity of about 28.9%, water saturation of 32%, crude oil density of 1.025 g/cm³, oil sand density of 1.95-2.12 g/cm³, and P-wave velocity of 2100-2600 m/s.

In the study area, oil sand reservoirs vary greatly in physical properties and oil saturation laterally and have many interlayers^{[34,35,36,37]}, making the effect of SAGD highly uncertain. Therefore, the time-lapse seismic survey was designed in the early stage to monitor the development of the steam chamber. 3D seismic data was acquired as the reference data once before the development of the oil sand. After the SAGD development started, seismic data were collected twice at an interval of two years, as shown in Fig. 9. Compared with the base seismic section before the SAGD development (Fig. 9a), it can be seen that the two seismic sections of the oil sand target layer acquired after the beginning of SAGD development have significant changes in reflection characteristics (in the red frame line) and "pull-down" phenomenon of the bottom event axis (Fig. 9b, 9c) (indicating increase of seismic wave propagation time), and the maximum "pull-down" time difference of the second seismic monitoring section has a maximum pull-down time difference of nearly 20 ms (equivalent to the reduction of the P-wave velocity of about 40%). The reason is that during SAGD development, with the injection of high-temperature steam, the temperature of the oil sand target layer increases and the P-wave velocity decreases.

P-wave impedance data were obtained through time- lapse seismic inversion of the study area, and calibrated by the relationships of P-wave impedance with temperature of oil sand samples with different oil saturations, to predict the temperature changes of the steam chamber in the target layer (Fig. 10). It can be seen that the tempera-tures around the steam injection well increased significantly, but were different in different parts around the injection well in the two times of seismic surveys. Combined with the logging interpretation results, the reservoir zones with low shale contents have temperatures significantly higher than reservoir zones with high shale content. The prediction results tally with geological knowledge. The predicted temperatures from inversions of the seismic data acquired in the two surveys can reflect the temperature changes in the steam chamber well. Comparison of the measured temperatures along the way of 12 newly drilled infill wells in the study area with the predicted temperatures from time-lapse seismic (Fig. 11) shows that the monitoring points with a prediction error of less than 5% account for 91.3%. The above results indicate that the time-lapse seismic inversion can effectively monitor the development of the steam chamber and guide the optimization of production measures.

For oil sands with weak cementation and low impedance, in the process of ultrasonic rock physical property test, using an ultrasonic base made of PEEK resin instead of conventional titanium alloy can enhance signal energy and signal-noise ratio and obtain clearer first arrival.

The experimental results of heavy oil in the study area show that heavy oil decreases sharply in viscosity with the increase of temperature. The corresponding glass temperature point and liquid temperature point of the heavy oil are -34.4 °C and 49.0 °C, respectively. P-wave and S-wave velocities of the heavy oil also decrease sharply with the increase of temperature and have various change rates in different temperature sections. When the temperature exceeds the liquid temperature point, the heavy oil gradually turns from glass state to liquid state, and the S-wave velocity decreases sharply to zero. The quasi-solid heavy oil between glassy temperature point and liquid temperature point shows noticeable frequency dispersion.

The physical characteristics of oil sand in the study area are affected by many factors such as temperature and pressure and vary in complex patterns. When the temperature increases from 10 °C to 175 °C, the P- and S-wave velocities through the oil sand decrease significantly by 1000 m/s and 500 m/s, respectively. The elastic parameters of oil sand samples with different oil saturations have different main controlling factors and variation laws. The elastic properties of oil sand with high oil saturation are mainly dependent on the properties of heavy oil, but the elastic parameters of oil sand with low oil saturation are mainly dependent on the stiffness of skeleton particles. Compared with oil sands with medium and high oil saturations, the P- and S-wave velocities of oil sand with low oil saturation are more sensitive to pressure change at high temperatures.

As the SAGD development of the target layer in the study area features low pressure (the maximum pressure of nearly 4 MPa) and high temperature (up to 220 °C), in this work, the influence of temperature on the rock physical characteristics of oil sand was examined primarily. Based on the experimental data, the quantitative relationships between P-wave impedance and temperature of oil sand samples with different oil saturations were constructed, and the temperature changes of the steam chamber during the SAGD development of oil sand were predicted by time-lapse seismic inversion. The comparison of the measured temperatures of 12 newly drilled infill wells in the study area and predicted temperatures from time-lapse seismic inversion during the SAGD shows that the monitoring points with a prediction error of less than 5% account for 91.3%.