Estimating Sustainable Long-Term Fluid Disposal Rates in the Alberta Basin

Abstract

Reliable regional-scale permeability data and minimum sustained injectivity rate estimates are key parameters required to mitigate economic risk in the site selection, design, and development of commercial-scale carbon sequestration projects, but are seldom available. We used extensive publicly available disposal well data from over 4000 disposal wells to assess and history-match regional permeability estimates and provide the frequency distribution for disposal well injection rates in each of 66 disposal formations in the Alberta Basin. We then used core data and laboratory analyses from over 3000 cores to construct 3D geological, geomechanical and petrophysical models for 22 of these disposal formations. We subsequently used these models and the history-matched regional permeability estimates to conduct coupled geomechanical and reservoir simulation modeling (using the ResFrac™, Palo Alto, CA, USA, numerical simulator) to assess: (i) well performance in each formation when injecting carbon dioxide for a 20-year period; (ii) carbon dioxide saturation and reservoir response at the end of the 20-year injection period; (iii) reliability of our simulated rates compared to an actual commercial sequestration project. We found that: (i) the injection rate from our simulations closely matched actual performance of the commercial case; (ii) only 7 of the 22 disposal formations analyzed appeared capable of supporting carbon dioxide injectors operating at greater than 200,000 tons per year/well; (iii) three of these formations could support injectors operating at rates comparable to the successful commercial-scale case; (iv) carbon dioxide presence and a formation pressure increase of at least 25% above pre-injection pressure can be expected at the boundaries of the (12 km × 12 km) model domain at the end of 20 years of injection.

1. Introduction

Evolution of energy supply policies globally has the potential to fundamentally change subsurface formation pore space utilization in hydrocarbon producing basins across the globe. In the Alberta Basin, large-scale hydrocarbon fluid extraction over the last 60 years has resulted in extensive pressure depletion in most of the higher permeability (i.e., “conventional”) oil and gas formations [1]. However, emerging energy policies, such as net-zero emissions goals, can progressively increase subsurface fluid pressures in such formations, by gradually decreasing (future) fluid extraction rates while increasing (future) fluid injection rates. Additionally, reconciliation of (i) the anticipated volume of fluids to be injected into the subsurface with sustainable subsurface formation capacity, and (ii) injection rates required to support industrial needs with sustainable injection rates supported by the targeted formations are required to support optimal utilization of the existing basin subsurface pore space capacity and minimize risks to project proponents and clients as well as surface (e.g., induced seismicity) and subsurface users [1]. Extensive work has been conducted on developing methods to assess theoretical subsurface capacity for injected fluids such as carbon dioxide (CO₂) at various scales, based mainly on sparse data, probabilistic matrix porosity and generic permeability estimates (e.g., [2,3,4,5,6]). Recent methods have emerged that enable refinement of capacity estimates to account for additional subsurface risk factors, such as subsurface pressure sensitivities (e.g., [1,7]).

Assessment of the sustainability of predicted long-term injection rates and the anticipated effect on the receiving formations are also key requirements for the development of long-range plans for future net-zero projects. Reasonable certainty is required that predicted injection rates are realistic, can be sustained for decades, and the effects on the receiving formations are negligible or at least predictable. Estimation of such rates is fundamental for engineering and economic purposes, such as determination of the number of injectors, pipelines, and compression infrastructure requirements, and consequently the commercial service rates for future common-use CO₂ sequestration hubs. However, so far there has been limited focus on developing methods to estimate sustainable formation-scale long-term injection rates [8], despite the importance of this parameter to the successful long-term operation of high-volume fluid injection projects. Additionally, greater rigor may be required in evaluating sustainable injection rates, because long-term formation (e.g., thermal) impacts may be more important than short-term (e.g., pressure) impacts in such projects. For instance, initial results using fully-coupled models have suggested that thermal effects can be more important than fluid pressure effects over the long term in CO₂ injection, and significant errors could be expected when isothermal models are used [8]. Injection of increasingly larger volumes of non-traditional fluids such as CO₂ is anticipated to occur in the Alberta Basin and other sedimentary basins globally as net-zero emission energy policies are implemented over the next decade.

Our study aimed to estimate the most likely long-term fluid injection rates for key disposal formations in the Alberta Basin under current regulatory constraints (maximum injection pressure up to 90% of the minimum horizontal stress—σ_hmin), using actual laboratory (formation core-sample porosity, fluid saturation, uniaxial and triaxial test data), actual field data (formation thickness, pore pressure, minimum and vertical in situ stress, formation temperature) and a fully-coupled compositional three-dimensional integrated hydraulic fracturing and reservoir simulator. We used a compositional physics based simulator (ResFrac™, Palo Alto, CA, USA) which allowed for fully-coupled field-scale simulation of the geomechanical effects of sustained fluid injection on a target formation over the entire lifecycle of an injection well [9,10]. While the objective of this work was primarily to help benchmark realistic (sustainable) long-term Alberta Basin formation injection rates, the approach presented can also contribute to the development of a consistent approach (i.e., a standard protocol or workflow) for such assessments. Such a workflow has been identified as an important tool that can help to mitigate both economic and subsurface risks in the underground storage industry [11,12]. Our analysis suggests that, in a geoscience data-rich environment such as the Alberta Basin, the use of such a model and approach can provide reliable indicators of the long-term injection performance of injector wells and formation geomechanical response.

2. Materials and Methods

Alberta has one of the most extensive collections of publicly available geoscience and operational data, originating from its long history of hydrocarbon exploitation and open data policies. This includes operational data such as fluid production and injection volumes, formation pressures and well logs, as well as petrophysical, geological, geomechanical, chemical and other types of laboratory analyses conducted on formation cores collected from subsurface projects developed over the last 60 years across the province. Operational data (including monthly fluid injection volumes and pump run times for each injection well) were available in the geoSCOUT^TM database (available from geoLOGIC Systems https://www.geologic.com/geoscout/ [accessed on 9 January 2023]). Data from laboratory tests conducted on almost all core samples collected in the Alberta Basin were available on request from the Alberta Energy Regulator (https://static.aer.ca/prd/documents/sts/GOS-REPS.xlsb [accessed on 9 January 2023]).

2.1. Identification of Key Injection Formations in the Alberta Basin

Samaroo et al. (2022) used operational data to identify the geologic formations that had received the largest proportions of injected fluid in the Alberta Basin over the period January 1962–December 2021 [13]. This information shows that approximately 87% of the water injected into the Alberta Basin over this period was injected into only 12 geologic formations, comprising the Paleogene sands, the Lower Cretaceous sandstones, the Jurassic sandstones, the Devonian carbonates, and the Devonian sandstones (

Table 1. Key recipient formations for all types of fluid injection in the Alberta Basin over the period January 1962 to December 2021. After Samaroo et al. (2022) [13] and used with permission. Detailed production/injection and net fluid balance data for these formations are contained in Samaroo et al. (2022) [13].

Lithotype and Geologic Formation	Water Injected 1	Gas Injected 1	Steam Injected 1
Paleogene sands (Swan Hills)	17.1%	14.0%	0%
Lower Cretaceous sandstones (McMurray, Clearwater, Cardium, Viking, Nikanassin)	30.6%	9.8%	97.9%
Jurassic sandstones (Sawtooth)	11.5%	15.1%	0%
Triassic carbonates (Charlie Lake, Halfway)	0.4%	1.0%	0%
Triassic siltstones (Montney, Doig)	0.3%	0.6%	0%
Permian sandstones (Belloy)	0.2%	0%	0%
Carboniferous carbonates (Banff, Debolt, Elkton, Livingston, Turner Valley)	0.7%	1.8%	0%
Devonian carbonates (Arcs, Grosmont, Keg River, Leduc, Muskeg, Nisku, Slave Point, Wabamun, Winterburn)	29.2%	43.3%	1.4%
Devonian sandstones (Granite Wash, Gilwood)	4.1%	0.6%	0%
Cambrian sandstones (Basal Sandstone Unit)	0.3%	0%	0.8%
Total volumes injected in above-listed formations	23.8 km3	596.7 km3	3.41 km3
Total fluid volumes injected into all formations in the Alberta Basin	25.2 km3	692.2 km3	3.41 km3

¹ Totals may not add to 100% due to rounding. Fluid volumes are reported in cubic kilometers (i.e., km³ = cubic kilometer = 10⁹ m³), measured at surface conditions. Steam volumes reported in cold-water equivalent volume.

). Water volumes injected into these formations included reinjection of produced water into active production reservoirs (e.g., as part of pressure maintenance and enhanced oil operations) as well as water injected into formations without associated hydrocarbon recovery (i.e., water disposal).

While the 12 key injection formations identified above have historically been the recipients of the highest total injection volumes, significant changes in fluid injection types and patterns are anticipated to occur over the next decade [1]. These changes are likely to result in a substantial increase in future fluid injection volumes, distributed over formations that were not previously considered major injection targets and without concurrent associated fluid (hydrocarbon) recovery (i.e., an increase in fluid disposal). Therefore, all formations listed in Table 1 above could be considered future injection targets and consequently formations of interest to this study.

2.2. Historical Water Disposal Injection Rates in Formations of Interest

The geoSCOUT^TM database also contains the monthly fluid injection, fluid license type and run-time records for all licensed wells in the Alberta Basin for the period 1960 to present. Well license regulatory requirements in Alberta [14] stipulate (among other conditions) that the purpose of the well (e.g., disposal, production, observation, etc.), current status and fluid type (e.g., water, CO₂, methane, etc.) must be specified, and the corresponding monthly volumes reported when operational. All injection wells with a (historical or current) status corresponding to water disposal, the monthly volumes injected (in m³), injection run-time (in hours), unique injection well identifier (UWID) and the injection formation were extracted from this database. Injection wells classified in the water disposal category are standalone injectors, and are not associated with hydrocarbon extraction operations, such as waterflooding, pressure maintenance or other enhanced oil recovery activities. Disposal wells are exempt from the pool-associated regulatory obligations for replacement of formation fluids and pressure maintenance in hydrocarbon reservoirs (i.e., voidage replacement ratios [15]) and consequently are considered representative of the type of operations associated with CO₂ sequestration injection.

Injection records with zero monthly volume or zero monthly hours were then removed, and the wells grouped by injection formation to create formation-specific databases consisting of approximately 4000 injection wells and 1,000,000 data records. The monthly water disposal injection rate for each well in m³/h was then calculated by dividing the injection volume in m³ by the corresponding monthly run time in hours. Characteristic injection rate statistical parameters required for determining the most likely injection rate (mean, mode, median, variance, standard deviation and max) and the injection rate frequency distribution were then calculated for each formation of interest. These injection rate statistical parameter results are presented in Section 3, and the injection-rate frequency distribution for each formation is provided in Appendix A.

2.3. Geological, Geomechanical and Petrophysical Data for Formations of Interest

Samaroo et al. (2022) compiled and published a database of geological, geomechanical and petrophysical properties for selected cores from the major confining and injection formations in the Alberta Basin [13]. This dataset contains geological, geomechanical and petrophysical laboratory tests conducted on approximately 3000 cores collected from 260 wells, as well as in situ minimum horizontal and vertical stress, reservoir pore pressure, and temperature data for the major injection and confining formations in the Alberta Basin. Sufficient information was contained in this dataset to construct 3D geological, geomechanical and petrophysical models for 26 of the injection formations shown in Table 1. The key parameters used to construct the respective geological, geomechanical and petrophysical ResFrac™ models for each of the 26 injection formations (14 Mesozoic and 12 Paleozoic) are listed in

Table 2. Stratigraphic zones and minimum–maximum ranges of geological, geomechanical and petrophysical parameters used to construct geomechanical models for the Mesozoic Era formations of interest. Upper and lower numbers represent the minimum and maximum ranges of values for each parameter across the core sample depth interval.

Formation	Core Depth Interval (m)	ϕ	Sw (%)	Po (MPa/km)	ν	σh (MPa)	E (MPa)	k (m2)	ρ (kg/m3)	σt (MPa)	t (°C)	σv (MPa/km)	σc (MPa)
Upper Cardium	1794	0.04	0.35	7.10	0.12	22	13,872	2.60 × 10−16	2149	6	54	22.7	55
	1810	0.11	0.35	7.10	0.31	23	54,174	2.77 × 10−15	2149	19	54	22.7	261
Lower Cardium	2156	0.07	0.14	7.10	0.10	27	18,621	9.87 × 10−18	2149		65	22.7
	2477	0.12	0.15	7.10	0.30	31	59,966	2.76 × 10−16	2149		74	22.7
Pouce Coupe	1751	0.03	0.07	11.14	0.19	22	27,540	3.95 × 10−19	2542		53	22.7	185
	1823	0.06	0.13	11.14	0.31	23	55,020	3.95 × 10−19	2639		55	22.7	250
Viking	2176	0.21	0.40	7.70	0.06	30	13,790	5.97 × 10−17	2290		65	22.2	144
	2182	0.23	0.40	7.70	0.19	30	57,290	4.59 × 10−16	2570		65	22.2	144
Falher	2928	0.06	0.40	7.10	0.07	42	3360	3.95 × 10−18	2550	6	108	23.7	21
	3064	0.12	0.40	7.10	0.41	44	47,150	4.05 × 10−17	2550	15	113	23.7	248
Notikewin	2876	0.08	0.35	7.84	0.21	34	17,950	3.95 × 10−17	2550	5	106	23.5	84
	2892	0.10	0.35	7.84	0.31	34	32,760	5.92 × 10−16	2640	9	107	23.5	101
Middle Mannville	2332	0.09	0.43	7.84	0.19	28	18,459	1.48 × 10−18			86	22.1	70
	2455	0.09	0.43	7.84	0.57	29	44,974	1.16 × 10−16			91	22.1	147
Lower Mannville	2714	0.07	0.57	7.84	0.19	32	1170	2.76 × 10−17	2554		100	22.1	105
	2778	0.07	0.57	7.84	0.31	33	45,814	1.16 × 10−16	2699		103	22.1	194
Nikanassin	3362	0.06	0.27	9.50	0.02	46	22,389	4.93 × 10−17	2615		134	24.8
	3385	0.08	0.78	9.50	0.11	46	63,612	9.87 × 10−16	2669		135	24.8
Upper Charlie Lake	1478	0.04	0.00	7.49	0.20	29	2687	2.08 × 10−18			72	25.2	32
	1587	0.23	0.00	7.49	0.27	31	32,530	1.35 × 10−13			78	25.2	60
Lower Charlie Lake	1980	0.10	0.11	7.49	0.08	39	7053	4.67 × 10−16	2326		97	25.2	37
	2241	0.25	0.81	7.49	0.36	44	53,967	4.93 × 10−13	2617		110	25.2	218
Upper Doig	2406	0.09	0.47	7.90	0.11	38	36,966	4.93 × 10−14	2590		118	21.2	52
	2451	0.09	0.47	7.90	0.32	39	60,414	4.93 × 10−14	2650		120	21.2	111
Lower Doig	2956	0.07	0.45	7.90	0.11	47	10,246	3.95 × 10−16	2523		145	21.2	36
	2990	0.07	0.45	7.90	0.34	48	37,521	3.40 × 10−14	2595		147	21.2	37
Belloy	2476	0.06	0.79	10.05	0.26	48	22,770	2.55 × 10−18	2476	6	128	22.8	108
	2672	0.13	0.84	10.05	0.33	48	74,976	1.18 × 10−14	3093	22	131	22.8	211

Table header key: ϕ: total porosity; S_w: water saturation; P_o: pore pressure gradient; ν: Poisson’s ratio; σ_h: minimum horizontal stress: E: Young’s modulus; k: permeability; ρ: bulk density; σ_t: tensile strength; t: reservoir temperature; σ_v: vertical stress gradient; σ_c: unconfined compressive strength.

and

Table 3. Stratigraphic zones and minimum–maximum ranges of geological, geomechanical and petrophysical parameters used to construct geomechanical models for the Paleozoic Era formations of interest. Upper and lower numbers represent the minimum and maximum ranges of values for each parameter across the depth interval.

Formation	Depth (m)	ϕ	Sw (%)	Po (MPa/km)	ν	σh (MPa)	E (MPa)	k (m2)	ρ (kg/m3)	σt (MPa)	t (°C)	σv (MPa/km)	σc (MPa)
Wabamun	2238	0.03	0.01	6.39	0.13	30.4	12,828	5.76 × 10−20	2584	4	67	23.15	50
	2380	0.07	0.73	6.39	0.45	32.4	115,172	3.33 × 10−14	2931	11	71	23.15	192
Leduc	1618	0.02	0.15	7.87	0.08	22.4	17,655	9.87 × 10−18	2803	3	49	22.75	18
	1855	0.10	0.58	7.87	0.74	25.6	119,241	4.42 × 10−13	2842	9	56	22.75	29
Upper Slave Point	595	0.04	0.44	5.40	0.14	7.3	23,520	6.95 × 10−13	2490	3	18	23.71	96
	795	0.15	0.58	5.40	0.28	9.7	75,430	2.06 × 10−12	2620	10	24	23.71	169
Lower Slave Point	1323	0.03	0.44	5.40	0.08	16.1	15,240	5.00 × 10−16	2610	5	50	23.71	86
	1366	0.07	0.58	5.40	0.22	16.7	57,690	2.38 × 10−12	2680	16	52	23.71	86
Upper Muskeg	739	0.10	0.76	7.90	0.13	11.3	24,320	7.00 × 10−17	2580	5	22	21.05	63
	744	0.25	0.86	7.90	0.24	11.3	47,520	2.17 × 10−13	2820	8	22	21.05	63
Lower Muskeg	1510	0.09	0.75	7.90	0.18	23.0	47,200	7.00 × 10−17	2744	5	59	21.05	52
	1523	0.21	0.88	7.90	0.49	23.2	83,200	2.17 × 10−13	2865	9	59	21.05	160
Upper Keg River	1014	0.09	0.15	9.01	0.19	16.3	2120	4.18 × 10−13	2620	7	43	22.88	70
	1046	0.38	0.24	9.01	0.27	16.8	53,830	2.22 × 10−12	2670	10	44	22.88	140
Middle Keg River	1510	0.15	0.18	9.01	0.18	24.3	47,200	1.86 × 10−13	2620	7	63	22.88	52
	1522	0.36	0.25	9.01	0.49	24.5	83,200	1.96 × 10−12	2667	11	64	22.88	52
Lower Keg River	1739	0.09	0.15	9.01	0.06	28.0	38,370	1.59 × 10−13	2621	7	73	22.88	58
	1778	0.38	0.25	9.01	0.36	28.6	92,500	2.21 × 10−12	2667	11	75	22.88	177
Red Beds	1494	0.02	1.00	9.79	0.10	23.5	14,280	3.56 × 10−22	2546	3	67	23.32	41
	1625	0.05	1.00	9.79	0.31	25.5	64,800	1.59 × 10−21	2598	8	73	23.32	92
Basal Sandstone Unit	2050	0.05	1.00	9.50	0.11	41.2	9900	6.91 × 10−16	2620	2	82	23.43	36
	2265	0.19	1.00	9.50	0.36	45.5	29,500	1.48 × 10−14	2700	5	91	23.43	98

Table header key: ϕ: total porosity; S_w: water saturation; P_o: pore pressure gradient; ν: Poisson’s ratio; σ_h: minimum horizontal stress: E: Young’s modulus; k: permeability; ρ: bulk density; σ_t: tensile strength; t: reservoir temperature; σ_v: vertical stress gradient; σ_c: unconfined compressive strength.

below.

The data shown in Table 2 and Table 3 represent a summary of the minimum and maximum ranges of each of the parameters required to construct a representative geomechanical model for each formation. Each formation depth interval contains multiple core samples, with higher resolution data corresponding to the stratigraphic variations within each respective core depth interval. These detailed stratigraphic data (i.e., the core testing results from the 3000 laboratory core tests summarized in [13]) were imported into the ResFrac™ software and used to construct the corresponding 3D stratigraphic, geomechanical and petrophysical models for each formation of interest.

While extensive geomechanical data were available for the specified depth intervals listed in Table 2 and Table 3 above, in some cases there were insufficient core laboratory test data over the core (run) interval for petrophysical parameters such as permeability, porosity, and water saturation. In such cases, a random function generator was used to interpolate between the minimum and maximum values from adjacent (upper and lower) depth intervals and thereby complete the petrophysical model for the corresponding stratigraphic zone. The relative locations of disposal and sample core wells across Alberta are shown in Figure 1 below.

2.4. Common Injection Parameters Used in All Simulations

The injection well configuration used in all ResFrac™ models was based on the actual design of the Radway CO₂ injectors used at the commercial-scale Quest Carbon Capture and Storage (CCS) Project in Alberta, and the corresponding injector well completion report is available in [16]. The CO₂ injector was assumed to be vertical for all formation models, with a perforated zone length of approximately one-third of the formation thickness (to a maximum perforated zone length of 35 m). Other common injection parameters used in all simulations, along with the corresponding source/references, are provided in

Table 4. Common injection parameters used in all simulations.

Parameter	Value
Well geometry	Vertical [16]
Maximum horizontal stress (σHMAX) azimuth	45.5° [17]
Injection well casing internal diameter	0.15037 m [16]
Perforated zone length	1/3 of formation thickness, to a maximum of 35 m [16]
Perforation shot density	17 shots per meter [16]
Perforation hole diameter	0.0106 m [16]
Matrix (x and y) dimensions	12 km × 12 km
No. of matrix elements in x direction	100
No. of matrix elements in y direction	100
Matrix region z dimension	Thickness of cored region of formation
Matrix element max height (z)	1 m
Matrix boundary conditions	Closed
Vertical permeability	0.1 × horizontal permeability [18]
Fluid model	Compositional
Bottomhole pressure (BHP) injection constraint	90% of σhmin [19]
Injection simulation run time	20 years [20]
CO2 density	761 kg/m3 [20]
Brine density	1193 kg/m3 [20]
Brine salinity	320,000 ppm [20]
Relative permeability model	Brooks-Corey [21]
Injection fluid temperature	15 °C [22]
Biot coefficient (Mesozoic Formations)	0.7 [23]
Biot coefficient (Paleozoic Formations)	0.6 [24]
Geothermal gradient	Formation specific in situ measurements

below.

While limited injection fluid temperature records were available, the average injection fluid temperature of 15 °C was selected to represent the lower temperature limit observed [22,25] in other CO₂ injection operations in Alberta. The maximum bottomhole injection pressure constraint selected (i.e., 90% of the formation σ_hmin) represents the regulatory criteria [19] applicable to disposal wells operating in Alberta. While the constraint of maximum injection pressure equals to 90% of σ_hmin used ensures regulatory compliance, it does not account for the potential for triggering slip on critically stressed faults that may exist in the proximity of the injection well. In cases where such critically stressed faults are present, an injection pressure threshold lower than 90% of σ_hmin may be required to mitigate the likelihood of triggering fault slip within the reservoir and/or adjacent formations [26]. Future enhancements to this workflow could include investigation of this geomechanical risk factor and the potential impacts to estimated injection rates. Closed (i.e., no flow) boundary conditions were used in all simulations to replicate conditions in which there may be multiple independent CO₂ injection projects operating adjacent to each other, with (regulatory) pressure buildup constraints at common boundaries, as well as to evaluate most-likely worst-case pressure-buildup conditions.

These common injection parameters and relationships were used to complete the dataset required to construct the corresponding ResFrac™ 3D models for each of the 26 formations of interest, and to conduct the reservoir simulations described in the sections below.

2.5. Reservoir Simulations

The ResFrac™ simulator is a three-dimensional physics-based, integrated hydraulic fracturing and reservoir simulation numerical model consisting of a geologic model containing the properties of the formation, a compositional fluid flow model containing the phase behavior and fluid flow characteristics of the formation, a wellbore geometry model containing the wellbore architecture used for injection, and a fluid injection operations model containing the fluid injection operations over the lifecycle of the well [9]. The ResFrac™ simulator was designed as a practical engineering tool that can be used to create a reasonable representation of the physical processes occurring during field-scale fluid injection (and hydraulic fracturing), in order to quantitatively and qualitatively understand, predict and optimize fluid injection operations [27].

The ResFrac™ simulator fully couples (implicitly) subsurface injection and (multi-phase) fluid flow processes with injection-induced thermal and mechanical stresses in the matrix in three dimensions (x, y and z) to evaluate well injection and production rates (while considering the effects of fluid-phase composition on fluid flow), as well as the dimensions and location of any subsurface fractures created during injection [10]. The compositional model optimizes computational efficiency by assuming that (i) fluid flow (wellbore to elements and between elements) is governed by a combination of Darcy and non-Darcy (Forchheimer) flow, (ii) the stress state of each element can be approximated by the boundary element method, which assumes small-strain conditions within each element, that each element is elastically homogenous and isotropic and deformation of each element is linear-elastic, (iii) fractures generated are planar at field-scale, and can initiate and propagate arbitrarily, and (iv) porothermoelastic stress changes in the model domain do not affect the total volumetric strain [9,10,27]. Additionally, all model domains were constructed under the (simplified) assumption that stratigraphic layers corresponding to core samples exhibited vertical transverse isotropy across the entire model domain. Model outputs include quantification of daily injection rates for fluids, fluid saturation and formation pressure across the model domain, as well as the dimensions of any large-scale planar fractures (i.e., larger than the dimension of the smallest mesh element) generated over the simulation period. A detailed technical description/evaluation of the initial/boundary conditions and the governing/constitutive equations used in the Resfrac^TM model to simulate matrix (hydraulic and thermal) fracture initiation and propagation, multiphase fluid flow, thermal transport, stress shadowing, and porothermoelastic responses from temperature/pressure change in the matrix are outside of the scope of this work, but can be found in the Resfrac^TM Technical Writeup document [27].

Models were constructed with variable mesh spacing, with denser 45 m × 45 m × 1 m (x-y-z) elements close to the injection well and gradually increasing to kilometer-scale (x-y) spacing towards the edges of the model domain. Mesh elements in the z-direction were maintained at a maximum of 1 m thickness, to account for vertical stratigraphic data resolution and variations. A mesh refinement strategy of increasing resolution in the near-wellbore region is considered an effective method for generating reliable results in large-domain simulation models, while maintaining geomechanical model computational efficiency [28]. Figure 2 below shows a diagram of the mesh strategy used to optimize model run times while maintaining near-wellbore simulation reliability.

Two identical 3D models were constructed for each of the 26 formations listed in Table 2 and Table 3, with one model containing brine as an injection fluid and the other CO₂. Formations located at a depth of less than 1 km were removed from this analysis, as Alberta’s regulations stipulate that such formations cannot be used for the purposes of CO₂ sequestration [29]. The brine model for each formation was then run for an injection period of 20 years at a maximum bottomhole pressure constraint of 90% of σ_hmin. The output from each brine model run was evaluated and the model (i.e., the global permeability adjustment parameter) was adjusted iteratively to match the historically observed injectivity. The settings of the CO₂ injection fluid model were then adjusted to mirror those of the history-matched model. The resulting CO₂ injection model was then run for a 20-year period, and the injection rate, dimensions of any subsurface fractures created, CO₂ saturation, and pressure increase at the domain boundary noted. Model run durations varied between one hour and up to 96 hours, depending on the number of elements (i.e., formation thickness) and formation permeability, with a total of approximately 11,000 hours of simulator time used.

2.6. History-Matching of Modeled and Historical Injectivity Rates and Model Calibration

Laboratory permeability measurements obtained from core (and/or core-plug) samples generally do not account for the impact of sampling bias and geological heterogeneities such as fractures, planes of weakness and other preferential flow paths and consequently require correction (i.e., upscaling) before they can be used at the formation or regional-scale [30,31,32,33]. Core-plug derived permeability measurements have been noted to underestimate regional-scale permeability measurements by up to six orders of magnitude in regions of Alberta [32]. Previous work has noted that in one specific case in Alberta (the Viking Formation), the upper range (i.e., the maximum) of the core-plug permeability data was likely representative of actual field-scale fluid flow conditions [33]. However, limited regional-scale permeability data are currently available, and assessment of this parameter can be challenging, especially in dual-porosity systems and fractured rock formations. History-matching of modeled and actual flow rates in regional well networks is considered a reliable method of assessing the kilometer-scale impact of geological heterogeneities, such as fracture networks [34]. The ResFrac™ simulator contains a global permeability multiplier variable, which allows for an efficient adjustment of all permeability values across the entire geologic model domain by a specified scalar constant. The use of this variable facilitates rapid adjustment of core-scale permeabilities across the entire domain and iterative (trial and error) model re-runs to obtain a match to field-scale permeability, when required.

Each formation model was initially run with water as the injection fluid and a global permeability multiplier of 1 for an injection period of 20 years, and the stabilized water injection rates obtained compared with the median historical water (injection) disposal rate for the corresponding formation. The global permeability multiplier variable was then adjusted (using trial and error) to increase or decrease the model’s stabilized water injection rate to approximate the corresponding median historical injection rate for that formation. This process was repeated iteratively until the model’s 20-year stabilized water injection rate closely matched (+/− 10%) the historical median rate for the corresponding formation, with an average of five model runs required to obtain a history-match. The global permeability multiplier and the history-matched regional-scale permeability values for each formation are shown in

Table 5. History-matched regional-scale permeability values for formations of interest.

Formation	Depth Interval (from–to; m)	Average Core-Scale Permeability (m2)	Global Permeability Multiplier Required for History-Match	History-Matched Regional-Scale Permeability (m2)
Upper Cardium 1	1794–1810	1.23 × 10−15	100	1.23 × 10−13
Lower Cardium 1	2156–2477	2.50 × 10−16	10	2.50 × 10−15
Pouce Coupe (Dunvegan) 1	1751–1823	3.95 × 10−19	100,000	3.95 × 10−14
Viking 1	2176–2220	2.31 × 10−16	75	1,73 × 10−14
Falher 1	2928–3064	1.69 × 10−17	1000	1.69 × 10−14
Notikewin 1	2876–2892	1.44 × 10−16	300	4.31 × 10−14
Middle Mannville 1	2332–2455	4.03 × 10−17	3000	1.21 × 10−13
Lower Mannville 1	2714–2778	6.52 × 10−17	10,000	6.52 × 10−13
Nikanassin 1	3362–3385	5.70 × 10−16	5000	2.85 × 10−12
Upper Charlie Lake 1	1478–1587	1.18 × 10−14	0.05	5.89 × 10−16
Lower Charlie Lake 1	1980–2241	1.91 × 10−13	0.01	1.91 × 10−15
Upper Doig 1	2406–2451	4.93 × 10−14	1	4.93 × 10−14
Lower Doig 1	2957–2990	1.05 × 10−14	30	3.15 × 10−13
Belloy 1	2476–2672	4.63 × 10−15	300	1.39 × 10−12
Wabamun 2	2238–2380	2.26 × 10−15	100,000	2.26 × 10−10
Leduc 2	1618–1855	2.74 × 10−14	20	5.48 × 10−13
Lower Slave Point 2	1323–1366	5.71 × 10−13	0.5	2.86 × 10−13
Lower Muskeg 2	1510–1523	7.02 × 10−14	1	7.02 × 10−14
Upper Keg River 2	1014–1046	1.26 × 10−12	1	5.61 × 10−13
Middle Keg River 2	1510–1522	1.34 × 10−12	0.3	4.81 × 10−13
Lower Keg River 2	1739–1778	1.13 × 10−12	0.1	1.32 × 10−13
Basal Sandstone Unit 2	2050–2265	3.55 × 10−15	30	1.11 × 10−13

¹ Mesozoic formations. ² Paleozoic formations.

below. It is assumed that these disposal wells were not creating/propagating subsurface fractures during operation, because regulatory criteria in Alberta stipulate that fluid disposal wells must operate at all times below the fracture gradient [19].

The global permeability multiplier values above were then entered into the corresponding ResFrac™ CO₂ injection model for each formation, thereby adjusting (i.e., calibrating) permeabilities in all stratigraphic layers across the entire model domain. This calibrated permeability accounts for sample bias and the impact of formation-scale heterogeneities and discontinuities often under-represented in laboratory core and core plug samples [35], and is considered a reliable method of accounting for the kilometer-scale impact of geological heterogeneities, such as fracture networks [34].

2.7. Estimation of Carbon Dioxide Injection Rates and Formation Geomechanical Impacts

Each CO₂ injection model, containing the calibrated (regional-scale) permeability values, was then run with the bottomhole injection pressure constraint of 90% of σ_hmin for an injection period of 20 years. The ResFrac™ model outputs allow the user to continuously track the mass fraction rate of CO₂ injected over time, as well as the surface area of new fractures generated over the lifetime of the injection sequence. An example of the output of the model is shown in Figure 3 below.

In order to verify that the simulator was capable of detecting the occurrence of fracturing during injection simulations, the injection pressure was increased to 10% above the fracture gradient for a subsample of formations. These simulations were then run for 20 years and the fracture surface area, gas injection rate and fracture aperture outputs of the simulator examined to ensure that the occurrence of fracturing was reflected in the fracture surface area output parameter of the model. The results of one of these simulations is shown in Figure 4 below.

These simulations showed that fracturing within the simulation period can be detected by both instability in the model’s (CO₂) gas injection rate parameter output and the generation of non-zero values in the fracture surface area parameter output (Figure 4). Therefore, both model output parameters were examined carefully for evidence of the occurrence of fracturing for all simulations conducted. The stabilized (i.e., minimum) CO₂ injection rates over the 20-year injection simulation period and the fracture surface area for each formation were then tabulated and the results are presented in Section 3 below.

3. Results

The sections below present the results of our analyses of the benchmark water injection rates and the CO₂ injection rates obtained from the calibrated models for each of the formations of interest in the Alberta Basin.

3.1. Historical Water Injection Rates in Formations of Interest

Table 6. Historical water injection rates and key statistical parameters for water disposal wells operating in each of the formations of interest in the Alberta Basin during the period January 1962–December 2022.

Geologic Era	Formation	Historical Water Injection Rates (m3/h)						No. of Injectors
		Mean	Mode	Median	Variance	Std. Dev.	Max
Upper Cretaceous 2	Belly River	2.2	0.6	1.0	10	3.1	60	49
Upper Cretaceous 2	Cardium	5.9	0.8	1.1	92	9.6	64	28
Upper Cretaceous 2	Dunvegan 1	5.3		4.0	72	8.5	70	2
Lower Cretaceous 2	Viking Sandstone	4.9	3.0	2.6	43	6.6	65	89
Lower Cretaceous 2	Paddy	11.2	3.0	9.5	117	10.8	206	20
Lower Cretaceous 2	Cadotte	6.6	13.5	3.3	78	8.9	63	56
Lower Cretaceous 2	Peace River	8.5	4.2	6.7	65	8.1	35	10
Lower Cretaceous 2	Colony	3.8	5.3	2.3	16	4.0	48	11
Lower Cretaceous 2	McLaren Sandstone	19.5	9.9	13.1	276	16.6	71	12
Lower Cretaceous 2	Notikewin	2.8	0.6	1.9	15	3.9	28	33
Lower Cretaceous 2	Grand Rapids	7.7	2.7	2.4	387	19.7	287	49
Lower Cretaceous 2	Clearwater Sandstone	8.2	0.4	3.2	154	12.4	132	44
Lower Cretaceous 2	Sparky	4.7	0.4	2.5	35	6.0	38	77
Lower Cretaceous 2	Falher	8.3	14.5	9.0	65	8.0	37	2
Lower Cretaceous 2	Rex	9.5	20.7	2.6	230	15.2	115	6
Lower Cretaceous 2	Lloydminster Sandstone	12.4	0.1	9.9	166	12.9	477	123
Lower Cretaceous 2	Glauconitic Sandstone	26.3	1.0	18.1	789	28.1	1178	270
Lower Cretaceous 2	Ostracod	11.3	0.3	5.1	148	12.2	48	5
Lower Cretaceous 2	Cummings	39.8	0.0	20.1	3354	57.9	1718	103
Lower Cretaceous 2	Dina	36.6	0.0	23.6	2072	45.5	2290	358
Lower Cretaceous 2	Detrital 1	17.9		19.4	43	6.5	32	3
Lower Cretaceous 2	Spirit River	5.4	1.3	1.8	29	5.4	21	8
Lower Cretaceous 2	Bluesky	4.4	1.0	1.6	82	9.1	84	39
Lower Cretaceous 2	Wabiskaw Sandstone	17.8	0.1	1.8	809	28.5	120	15
Lower Cretaceous 2	McMurray Sandstone	19.9	0.0	8.9	1006	31.7	1102	233
Lower Cretaceous 2	Cadomin	14.0	11.6	10.2	165	12.9	49	18
Lower Cretaceous 2	Gething	4.3	0.3	1.0	94	9.7	68	16
Lower Cretaceous 2	Sunburst	17.7	0.0	14.3	241	15.5	126	122
Lower Cretaceous 2	Ellerslie	31.4	0.1	17.0	1460	38.2	282	205
Lower Cretaceous 2	Taber	11.2	0.8	8.5	111	10.6	73	97
Lower Cretaceous 2	Nikanassin	9.5	10.8	9.7	24	4.9	25	3
Jurassic 2	Sawtooth	22.2	0.1	17.8	453	21.3	655	200
Jurassic 2	Nordegg	7.0	13.7	4.0	51	7.1	45	19
Triassic 2	Baldonnell	5.0	15.5	2.6	26	5.1	21	5
Triassic 2	Boundary Lake	0.9	0.0	0.5	5	2.3	18	4
Triassic 2	Charlie Lake	3.1	1.6	1.7	20	4.5	49	14
Triassic 2	Halfway	5.7	0.3	2.5	49	7.0	45	7
Triassic 2	Doig	4.1	5.1	4.8	2	1.5	6	2
Triassic 2	Montney	5.6	0.0	3.2	69	8.3	90	27
Permian 2	Belloy	10.8	1.0	8.1	125	11.2	245	100
Mid-Carboniferous 2	Debolt	9.6	0.4	4.8	181	13.5	115	48
Mid-Carboniferous 2	Livingstone	12.5	6.2	9.2	143	12.0	69	24
Mid-Carboniferous 2	Elkton	14.5	1.3	4.6	811	28.5	391	29
Mid-Carboniferous 2	Turner Valley	3.2	4.0	2.1	19	4.4	50	15
Lower Carboniferous 2	Shunda	3.3	2.0	1.8	20	4.5	54	7
Lower Carboniferous 2	Pekisko	5.9	2.0	3.6	43	6.5	102	57
Lower Carboniferous 2	Banff	9.5	0.8	6.4	101	10.1	67	56
Upper Devonian 3	Blueridge	16.5	17.6	17.7	121	11.0	56	4
Upper Devonian 3	Wabamun	13.9	0.7	4.3	866	29.4	321	229
Upper Devonian 3	Nisku	19.1	0.0	9.4	1870	43.2	933	265
Upper Devonian 3	Arcs	10.6	12.2	7.5	253	15.9	367	26
Upper Devonian 3	Grosmont	6.8	0.5	2.1	151	12.3	280	51
Upper Devonian 3	Peechee 1	6.3		7.1	6	2.4	10	2
Upper Devonian 3	Camrose	3.6	7.0	2.7	13	3.6	42	5
Upper Devonian 3	Leduc	53.9	0.1	21.3	5232	72.3	1353	338
Upper Devonian 3	Duvernay 1	7.3		6.8	39	6.2	47	2
Upper Devonian 3	Cooking Lake	18.1	0.1	15.5	281	16.7	169	19
Upper Devonian 3	Cairn	10.0	16.5	8.0	56	7.5	47	10
Middle Devonian 3	Slave Point	16.5	0.0	11.7	513	22.7	636	74
Middle Devonian 3	Gilwood	12.2	13.3	10.8	96	9.8	109	13
Middle Devonian 3	Keg River	19.6	0.6	7.0	3854	62.1	873	135
Middle Devonian 3	Muskeg	5.9	3.4	4.7	47	6.8	111	7
Middle Devonian 3	Contact Rapids 1	5.9		5.3	9	3.0	26	1
Lower Devonian 3	Lotsberg *	294.7 *		287.7 *	14,129 *	118.9 *	614 *	3 *
Lower Devonian 3	Granite Wash	10.9	0.0	8.8	116	10.8	213	22
Cambrian 3	Basal Sandstone Unit	49.5	5.4	39.7	1798	42.4	348	45

* Injection wells in this formation are associated with salt cavern disposal operations and injection rates are representative of operational requirements and not geological constraints. ¹ Insufficient number of wells to calculate mode. ² Mesozoic Era. ³ Paleozoic Era.

presents the analysis of the historical water injection rates of all licensed disposal wells that have operated in each of 66 disposal formations in the Alberta Basin over the last 60 years (January 1962–December 2022). The injection rate frequency distribution for wells in each formation is included in Appendix A.

While the data presented in Table 6 above constitutes a useful reference of the water injection rate activity in each of the listed formations, visualization of key trends in the data is simpler when this information is plotted as a stacked bar chart, as presented in Figure 5.

Table 6 shows that several cases of extreme (outlier) maximum injection rate values exist in the database, which can affect the representativeness of the calculated mean. Examination of the injection rate frequency distributions for all formations of interest (Appendix A) show that these are predominantly positively skewed, with the presence of (extreme in some cases) outlier values. In such cases of skewness, the median is the preferred measure of central tendency as it is a resistant parameter and less likely to be affected by outliers in the dataset. Therefore, the median value was selected as the most likely injection rate and hence a more appropriate benchmark for comparing injection rates across formations in this study, as well as for subsequent calibration of the geological model and to history-match the water injection reservoir simulation for each formation of interest.

Figure 5 shows that only 20% (13) of the (66) listed disposal formations recorded median water disposal injection rates exceeding 10 m³/h (approximately 1500 barrels per day [bpd]). Wells located in the Lower Cretaceous (Glauconitic, Cummings, Dina, Detrital, Ellerslie) and Jurassic (Sawtooth) sandstones and three Devonian carbonate (Blueridge, Leduc and Cooking Lake) formations showed higher than average median injection rates (around 20 m³/h; 3000 bpd) and the highest concentrations of water disposal injectors. The highest median water disposal injection rates (around 40 m³/h; 6000 bpd) in the Alberta Basin were recorded by wells located in the Basal Cambrian (Sandstone) Unit formation. However, geological heterogeneities that can significantly enhance secondary porosity and local-scale permeability are common in many of the formations listed above [30,32,36] and are likely contributors to the presence of injection rate outliers noted in Table 6.

3.2. Estimated Regional-Scale Permeability, per-Well CO₂ Injection Rates and Formation Geomechanical Impacts

Table 7. Modeled annual CO₂ injection rates for a single injector well operating for 20 years at 90% of σ_hmin in each of the formations of interest in the Alberta Basin.

Formation	Depth Interval (from–to; m)	History-Matched Regional-Scale Permeability Estimates (m2)	Average Well-Scale Permeability Reported in Previous Studies (m2)	Modeled Sustained 20-Year CO2 Injection Rates per Injector Well (tpa *)	Fracture Surface Area Generated (m2)
Upper Cardium	1794–1810	1.23 × 10−13	8.33 × 10−14 [30]	90,212	0
Lower Cardium	2156–2477	2.50 × 10−15	2.16 × 10−16 [37]	22,439	0
Pouce Coupe/Dunvegan	1751–1823	3.95 × 10−14	9.87 × 10−14 [38]	134,202	0
Viking	2176–2220	1.73 × 10−14	1.83 × 10−14 [39,40]	48,826	0
Falher	2928–3064	1.69 × 10−14	2.96 × 10−14 [41]	166,552	0
Notikewin	2876–2892	4.31 × 10−14	2.96 × 10−14 [41]	33,838	0
Middle Mannville	2332–2455	1.21 × 10−13	4.93 × 10−13 [42]	184,184	0
Lower Mannville	2714–2778	6.52 × 10−13	4.93 × 10−13 [42]	484,058	0
Nikanassin	3362–3385	2.85 × 10−12	1.97 × 10−16 [43]	323,565	0
Upper Charlie Lake	1478–1587	5.89 × 10−16	5.09 × 10−15 [44]	37,102	0
Lower Charlie Lake	1980–2241	1.91 × 10−15	1.48 × 10−15 [44]	32,782	0
Upper Doig	2406–2451	4.93 × 10−14	1.38 × 10−14 [44]	156,845	0
Lower Doig	2957–2990	3.15 × 10−13	1.38 × 10−14 [44]	118,147	0
Belloy	2476–2672	1.39 × 10−12	5.33 × 10−14 [45]	192,471	0
Wabamun	2238–2380	2.25 × 10−12	2.40 × 10−14 [40,46]	130,794	0
Leduc	1618–1855	5.48 × 10−13	5.0 × 10−13 [36]	327,794	0
Lower Slave Point	1323–1366	2.86 × 10−13	2.93 × 10−13 [36]	202,459	0
Lower Muskeg	1510–1523	7.02 × 10−14	9.87 × 10−16 [47]	50,462	0
Upper Keg River	1014–1046	1.31 × 10−12	8.98 × 10−14 [40]	297,922	0
Middle Keg River	1510–1522	4.81 × 10−13	8.98 × 10−14 [40]	236,957	0
Lower Keg River	1739–1778	1.32 × 10−13	8.98 × 10−14 [40]	187,215	0
Basal Sandstone Unit	2050–2265	1.23 × 10−13	9.87 × 10−13 [20]	423,821	0

* tpa: tons per annum.

presents the regional-scale permeability required to history-match the median per-well water disposal rates in each of the 22 formations for which sufficient data were available to both establish the median individual-well water disposal rate and to construct 3D geomechanical and petrophysical models. This table shows that the history-matched (upscaled) regional-scale permeability values derived in this study were in general consistent with those—derived from drill stem tests—contained in the previous studies listed in Table 7. However, while the permeability values presented below are considered regional scale, they may be subject to sample bias and are likely only representative of the areas within Alberta for which both core and water disposal data were collected (shown in Figure 1). Within such areas, the permeability estimates provided below can be used as a reasonable approximation of the regional-scale fluid flow and injectivity characteristics (at 90% of σ_hmin) of the 22 formations contained in Table 7. This information may be useful in the planning, site selection, and design stages of future large-scale fluid disposal projects in the Alberta Basin and can help to mitigate economic risks associated with formation-scale injectivity uncertainty.

The use of these history-matched regional-scale permeability data and the 3D geomechanical and petrophysical (ResFrac™) models enabled estimation of the injectivity performance of a single CO₂ injector (operating at an injection pressure of 90% of σ_hmin) located in the center of a 12 km × 12 km block of each of the 22 formations listed below. The modeled estimates of the sustained (minimum) annual CO₂ injection rates for a single injection well operating in each of the formations of interest for a period of 20 years are also shown in Table 7 below.

In all simulations conducted, zero fracture surface area was generated, which indicates that the sustained injection of CO₂ (over a 20-year period at 90% of σ_hmin and a temperature of 15 °C) into the 22 formations listed in Table 7 is unlikely to result in the creation of large-scale thermally-induced planar fractures. However, given the relatively large mesh element size in the near wellbore region (45 m × 45 m × 1 m; x-y-z) it is possible that smaller (i.e., below mesh element resolution) thermally-induced fractures could have occurred during the simulation but not have been detected by the model because of the (relatively low) mesh element resolution. Zero fracture surface area was consistently recorded in all simulations conducted, which indicates that conditions for the propagation of such fractures beyond the mesh element size (45 m × 45 m × 1 m) are unlikely to have existed during the simulation.

Table 7 shows that there are several disposal formations that may be capable of supporting the operation of high-rate CO₂ injectors for 20-year periods, at rates comparable to those experienced at the only commercial-scale CO₂ sequestration project operating in Alberta currently (i.e., the Quest project). These include regionally extensive Mesozoic sandstones such as the Middle and Lower Mannville, Nikanassin, Pouce Coupe/Dunvegan, Falher, Doig and Belloy, which are relatively ductile [13] depleted formations likely to be hydraulically isolated from the Precambrian basement [1]. Formation ductility, pressure depletion and hydraulic isolation from the Precambrian basement are major factors that reduce the potential for generation of large-magnitude induced seismicity during sustained industrial-scale fluid disposal activities [1,13]. Additionally, Table 7 shows that the regionally extensive (depleted) Devonian carbonate formations also located distant from the Precambrian basement (such as the Wabamun, Leduc, Slave Point and Keg River) could sustain CO₂ injection rates comparable to those demonstrated at the Quest project. The generation of large magnitude induced seismicity and the loss of (lateral and/or vertical) containment are considered critical failure modes that can permanently and materially impact the chances of success of a large scale CO₂ injection project [48].

In all simulations presented in Table 7, a single CO₂ injector operating for 20 years resulted in the presence of CO₂ distributed throughout the 12 km × 12 km model domain at varying levels of saturation in each stratigraphic horizon. Figure 6 shows the example of the simulation output for the Middle Mannville formation.

Additionally, in all simulations it was observed that the reservoir pressure at the matrix boundary (located 12 km from the injection well) had increased by a minimum of 25% by the end of the 20-year simulation period. The gradual increase in reservoir pressure at the (no-flow) matrix boundary resulted in a corresponding decrease in modeled injectivity over the simulation history, as shown in Figure 6. Under the closed (no-flow) boundary conditions used in these simulations, a single CO₂ injector operating for 20 years is likely to result in a simulated reservoir pressure increase of at least 25% at a radius of at least 12 km (i.e., size of matrix region) from the injector well location for all the formations modeled. However, this is likely representative of a worst-case scenario because actual boundary conditions are not likely to be closed and therefore actual reservoir pressure increase is likely to be less than the modeled reservoir pressure increase.

A reliable assessment of the credibility of a simulation requires assessment of the congruence between model predictions, actual observed external data not used in the development of the model, and prospective/predictive outcomes (i.e., model validation) [49]. In the geoscience field, a history-match of a reservoir model to past reservoir performance data is considered a good indicator of model reliability/credibility [50]. History-matching requires the use of observed reservoir behavior to estimate reservoir properties that resulted in the behavior [50]. In this study, we employed history-matching to estimate regional-scale permeability for the formations of interest by using the historical water disposal rates in each formation, and then used this information to calibrate our CO₂ injection models.

Validation of the CO₂ injection rates predicted by our simulations requires a comparison of predicted rates to actual CO₂ injection rates in the formations assessed. The monthly CO₂ injection rates for the Quest CCS project are available from the date of commissioning of the project in 2015, and are published in the project’s Knowledge Sharing Program Records submitted to the Alberta Department of Energy [51]. Figure 7 shows a comparison of the actual monthly CO₂ injection rates for all (three) injectors at this project for the period August 2015 (i.e., Year 0)–December 2021 (i.e., Year 7) to the ResFrac™ model’s predicted CO₂ injection rate over a 20-year period for the Basal Sandstone Unit formation (also referred to as the Basal Cambrian Sandstone or BCS).

Figure 7 shows that the per-well CO₂ injection rates predicted by the ResFrac™ model closely matched the actual injection rates of the three CO₂ injectors operating at the Quest project for years 1–4. The decline in Quest injectivity which occurred at Year 4 is associated with the precipitation of halite in the near wellbore environment at these wells, and injectivity can be restored to original rates by conducting hot water workovers (flushes) [52]. The fluctuation in Quest well injectivity post-Year 4 shown in Figure 7 is therefore considered an operational issue, and not a fundamental geologic limitation. Figure 7 therefore validates that our workflow and simulation methodology appear capable of predicting, with a reasonable degree of accuracy, the actual individual CO₂ well injectivity rates (at least) in the Basal Sandstone Unit formation.

4. Discussion

Extensive guidance criteria and methods have been developed for identification and assessment of geologic containment (such as caprock integrity) risks in CO₂ injection projects (e.g., [53,54,55,56]) and for estimation of gross capacity [57,58], which are critical factors that determine the suitability of a site for geologic CO₂ storage. Recently, some progress has been made on the estimation of sustainable capacity, which consists of the ability of the storage container to safely maintain the injected CO₂ in the target reservoir for geologic time, while minimizing impacts to other current and future surface and subsurface users [1]. However, the regional economic feasibility of commercial-scale CO₂ injection and storage is largely controlled by both the available (sustainable) formation storage capacity in the region and the ability to consistently inject (into the subsurface) the desired CO₂ volumes required to meet project goals at rates that minimize the number of injector wells over the lifetime of the project [12]. The importance of these parameters is reflected in the current selection criteria for CO₂ sequestration sites, which provide guidelines for both capacity (i.e., reservoir thickness and porosity) and injectivity (i.e., permeability, reservoir pore pressure and fracture pressure) [58]. However, in many cases there is high confidence in permeability estimates at the local-scale but limited confidence in regional-scale permeability estimates, which limits the ability to assess the probability of successful well performance over the multi-decade operating life of large volume fluid disposal (such as commercial-scale CO₂ sequestration) projects [59]. In addition to injection well performance, other key sources of uncertainty specific to CO₂ sequestration projects include the geomechanical response of target disposal formations and the expected (CO₂ and pressure) plume behavior over several decades of injection [60]. Disposal well injectivity is therefore a key risk in commercial-scale CO₂ sequestration projects, with high confidence required in assessments of both long-term storage capacity, the minimum per-well injection rates, and the maximum number of wells before such projects can obtain corporate financial approvals to proceed and secure commercial sequestration service contracts with industrial emitters (clients).

Extensive efforts have been focused on estimating regional and basin-scale CO₂ sequestration capacity both globally and within Alberta (e.g., [61,62,63,64,65]), but there has been significantly less attention directed to the need to improve the reliability of injectivity rates [8], despite the critical importance of this parameter [66]. The inability of CO₂ injector wells to achieve the minimum design injection rate compromises project economics and can ultimately result in project economic failure. Therefore, the ability to minimize long-term injection rate uncertainty (i.e., estimate realistic long-term injection rates) can contribute significantly to reducing economic risk for CO₂ sequestration projects. Design injection rates vary depending on a number of site-specific factors including well construction and operation cost, with the minimum individual well design injection rate in the Quest project in Alberta established at 394,000 tons-per-annum/well (tpa/well) in 2012 [67], while injection rates below 25,000 tpa/well were considered uneconomic in the case of projects contemplated in Europe [68]. However, carbon penalties in Canada have since increased substantially from approximately $15/ton in 2012 [69] (year of the final investment decision of the Quest project) to $65/ton in 2023, which is likely to significantly reduce the minimum injectivity rate threshold per injector well required for positive return rates in CO₂ sequestration projects.

Consequently, if an arbitrary injection rate threshold of 100,000 tpa/well is assumed for the Alberta Basin, then 15 of the 22 disposal formations analyzed appear to be capable of supporting CO₂ sequestration wells operating at this rate for 20 years without incurring large-scale (thermally induced) fracturing. If this threshold is increased to 200,000 tpa/well then only 7 of the 22 formations appear to be capable of supporting sustained CO₂ injection at this rate for 20-year periods. These 7 formations include previously overlooked geographically extensive Mesozoic sandstones and carbonates that appear to be capable of supporting injectors operating at rates equivalent to those at the only commercial-scale CO₂ sequestration (Quest) project in Alberta. Additionally, these Mesozoic formations are depleted (a result of over 6 decades of hydrocarbon extraction), are more likely to be hydraulically isolated from the Precambrian basement [1], and are more ductile [13]. Geologic sealing capacity (i.e., containment) uncertainty is also lower in these formations, because these formations are predominantly legacy oil and gas reservoirs which have successfully contained fluid under pressure over geologic timescales, and extensive geologic and petrophysical characterization datasets are publicly available. The combination of these factors can reduce the likelihood of sustained industrial-scale fluid injection into these formations generating induced seismicity of concern as well as reduce geologic uncertainty, when compared to injection into (previously unexploited and underexplored) saline aquifers overlying the Precambrian basement. The generation of induced seismicity and loss of containment are considered two critical failure modes for large-scale CO₂ sequestration projects [48].

Deep saline aquifers have been the main targets for large-scale CO₂ sequestration interest, based on their lack of legacy wellbore penetrations, extensive geographic extent, depth, thickness and potential pore volume capacity [70]. However, in the Alberta Basin there is very limited geologic data available on the Basal Sandstone Unit formation (the deep saline aquifer of interest) despite its extensive geographic coverage, while its stratigraphic location on top of the Precambrian basement increases its likelihood of a hydraulic connection with high seismogenic hazard Precambrian fault systems. The lack of geologic data increases uncertainty while its proximity to the Precambrian basement and virgin reservoir pressure increase the potential of generating induced seismicity of concern, especially if scenarios of multiple CO₂ sequestration projects operating simultaneously in the same formation over multi-decade periods are contemplated [1]. Therefore, the ability of the geographically extensive and ultrathick intermediate-depth Mesozoic sandstone and carbonate formations in Alberta to support long-term CO₂ injection rates similar to those of the deep saline aquifers may be of interest to project developers and regulators in the province. The history-matched permeability and the simulated CO₂ injection rate estimates presented in this study may also be useful to project developers to help manage injectivity and economic risk in CO₂ sequestration projects proposed within the Alberta Basin.

However, it should be noted that the presence of legacy oil and gas wellbores in these Mesozoic formations can increase containment risk and project operating (monitoring, measurement, and verification) costs. Such risks could be managed by a combination of upgrading wellbore abandonment to meet modern standards and maintaining pressure in the sequestration horizon below the original (virgin) formation pressure throughout the project lifecycle and post-closure [71]. Additionally, while only vertical wells and injection pressure below the fracture gradient were considered in our simulation study, individual well injectivity rates can be increased (nominally) by using horizontal injector wells and (significantly) by hydraulic stimulation of the injection formations [12].

Our CO₂ injection pressure strategy (i.e., injection pressure limit = 90% of σ_hmin) and consequently our injection estimates presented are based on the requirements for regulatory compliance in the Alberta Basin. However, lower injection pressures may be required in cases in which there are critically stressed faults present in or in proximity to the disposal formations, even when such formations are depleted, to avoid triggering fault slip [26]. Substantial injection pressure reduction is likely to result in substantial injection rate reduction, and such an investigation represents an opportunity for a future refinement to the results presented in this study.

While we have focused on estimation of long-term CO₂ injection rates in the Alberta Basin in this study, the workflow we have used can be applied in many hydrocarbon producing basins in which there is an extensive amount of geoscience data. The use of such a method to estimate regional-scale CO₂ injection rates can help to reduce project design injectivity uncertainty and economic risk, as the history-matching of modeled and actual flow rates in regional well networks is a reliable method of accounting for kilometer-scale impact of geological heterogeneities [34]. A summary of our workflow is provided in Figure 8.

The geomechanical, geological, petrophysical and water disposal data, and the reservoir simulator used in this study (and represented in the workflow in Figure 8), are commonly required for hydrocarbon operational activities and are therefore generally available in hydrocarbon producing basins. The method described above can therefore be of interest to CO₂ injection project proponents, regulators, and policy makers considering the development of industrial-scale sequestration projects in such locations.

5. Conclusions

Our analysis shows that there is large variability in the range of factors required to upscale laboratory measured (core-scale) permeability to history-matched regional-scale permeability across the Alberta Basin, with formation-specific correction factors ranging over six orders of magnitude (from 0.1 to 1 × 10⁵). This observation is consistent with the formation-scale estimates of upscaling factors obtained from well-scale drill stem testing conducted in two select areas of the Alberta Basin (the Peace River [32] and the Pembina Cardium [30]).

Our analysis also indicates that, when populated with formation-specific geology/geomechanics/petrophysics and history-matched permeability parameters, physics-based compositional 3D geomechanical models (such as ResFrac™) appear to be capable of providing useful indicators of likely well and reservoir performance for commercial CO₂ sequestration projects. Our simulations also show that, when injection pressures are constrained to 90% of σ_hmin, only 7 of the 22 disposal formations assessed in the Alberta Basin are capable of sustaining CO₂ injection at commercial-scale rates (i.e., greater that around 200,000 tpa/well) over a 20-year period. Sustained injection of CO₂ into these formations appears unlikely to result in large-scale thermally-induced fracturing over a 20-year injection period. Four of these formations (the Lower Mannville, Leduc, Nikanassin, and Upper Keg River) appear to also be capable of sustaining commercial-scale long-term CO₂ injection rates comparable to those of the principal CO₂ sequestration target formation in Alberta (the Basal Sandstone Unit formation). Commercial-scale CO₂ sequestration into such formations may present a lower induced seismicity hazard, because of the greater likelihood of hydraulic isolation from the high seismogenic hazard Precambrian basement.

Our analysis also indicates that, under the conditions used in this study, (i) the presence of CO₂ at varying levels of saturation can be expected throughout the 12 km × 12 km model domains for all formations analyzed, and (ii) a minimum pressure increase of at least 25% over initial formation pressure can be expected (under closed boundary conditions) at the boundaries of the model domain (radius of 12 km from the injector well) at the end of the 20-year injection period. The methods and workflow utilized in this study can be used for the estimation of long-term CO₂ injectivity rates and geomechanical effects in any geoscience-data-rich sedimentary basin, and consequently may be of value to project developers, regulators, and policy makers globally.