Variability of Data Sources

The differences in how individual wells can be identified was discussed in this report’s “Introduction” section. Another example of variation can be found in well production data. Production data for a well listed as a single surface location may come from multiple underground formations or reservoirs at different depths, which is often referred to as comingled production. The level of detail at which this production is reported can vary between data sources. In some data sources, hydrocarbon production is reported as a single volume from an individual well; this is known as well-level production. Another way hydrocarbon production can be reported is as a single volume from multiple wells, which is known as lease-level production. The methodology developed in this study to classify wells relies on S&P Global’s U.S. oil and gas Petroleum Information Data Model (PIDM) database (S&P Global, 2023; hereafter S&P Global’s database), which contains lease- and well-level data through February 2023. PIDM is a database structure used by S&P Global’s U.S. oil and gas database and is based upon the Public Petroleum Data Model association standards; the PIDM is no longer active.

Limitations of Recorded Data

Hydrocarbon production data records in the United States are highly variable because each State defines and regulates its own hydrocarbon exploration and extraction operations. Data collection may include regular reporting. Therefore, consistency and timeliness are the main limitations of state-level databases. For example, Illinois does not require hydrocarbon volume reporting (General Assembly of Illinois, 2023) at the well level, whereas Colorado collected hydrocarbon volumes at lease levels before 1999 and has collected hydrocarbon volumes at the well level since 1999 (Colorado Oil and Gas Conservation Commission, 2002). Texas collects oil and gas volumes at the lease level but also collects enough well data such that S&P Global can allocate the oil and gas production down to the well level using a proprietary approach (Dick Catto and Bruce Smith, IHS Inc. [now S&P Global], written commun., 2014; Railroad Commission of Texas, 2024). Simply put, allocation allows S&P Global to take a single production volume from multiple wells and use associated lease and field-specific information to allocate the production volumes to the individual wells that composed the leave-level production volume.

Another limitation of hydrocarbon data records is consistency. State-specific reporting requirements create challenges when gathering all hydrocarbon production data across the United States. Some States require reporting of all hydrocarbons produced along with those injected, and other States only require a single hydrocarbon constituent (for example, oil but not natural gas) to be reported. Timing is also an issue because each State defines and manages its own hydrocarbon data reporting requirement, which can range from six months to as many as 2 years. Given that each State also typically has its own format and schedule for releasing available hydrocarbon production data, compiling data from all States into a single database can be challenging. Companies like S&P Global and Enverus provide comprehensive U.S. hydrocarbon production databases. Because these databases include publicly sourced and proprietary data and calculations, there are likely to be variations between these commercial data sources as well. Given these persistent limitations of the data, data sources, and the wide range of reporting requirements, this report takes a conservative approach to defining abandonment. For example, because the maximum time for an operator to report to the State is 2 years, our definition conservatively uses 5 years as the cutoff to avoid misidentifying wells.

United States Production Data

The methodology used in this study accounts for the variability of data sources and limitations in reported data while utilizing the proprietary data provided by S&P Global for U.S. well records and production data. At the time of this analysis (2024), S&P Global’s database contained more than five million wells and almost three million production entities. The S&P Global database organizes well information using a 14-digit UWI, whereas their production information uses the producing “entity” and “entity type” as additional unique identifiers specifically for reporting volumes. The entity identifier is a combination of 2 to 23 alphanumeric characters that uniquely identify a specific record containing production volumes recorded through time. The entity type identifier indicates whether the production is recorded at the level of a lease, a well, or has been allocated (as previously explained). Because wells may produce hydrocarbons from multiple formations and because some States report their production by lease, there often exists a many-to-many relationship between well and production data.

S&P Global’s database includes not only production but also injection volumes, which may contain data for fluids other than oil and gas that may be withdrawn or injected as part of the hydrocarbon extraction process. The methodology in this report only considers produced hydrocarbons when determining whether a well is active or abandoned. This precludes considerations for injection, water, and stratigraphic wells (exploration wells used to gather geologic information). There are research opportunities to modify the methodology to consider these wells when making the active or abandoned determination.

Analysis of Production Gaps

From a data-analysis perspective, the most common issue with using production data to classify a well as abandoned is the possibility of a short-term production pause (Seely, 2016). Hydrocarbon production is subject to market changes in supply and demand (Seely, 2016). Also, hydrocarbon production wells are mechanical operations and require periodic maintenance. Most, if not all, well maintenance requires the production operation to be halted while the maintenance is performed (Seely, 2016). Therefore, it is not uncommon for a well to pause production for economic or mechanical reasons.

For this study, it was essential to quantify these production pauses and their durations. To accomplish the goal of quantifying production pauses for millions of wells, a Java program was written that used the S&P Global PIDM production database to count production pauses and their durations. The program output specifying the produced fluid and the duration of the production pause allowed for analysis by duration and fluid. For example, a pause of one month for oil has a frequency of 1,832,874 occurrences in the database. Analysis was performed on individual hydrocarbon fluids (oil, gas, gas condensate, and casinghead gas) separately and all fluids together. Even though there is some variation in frequency between these analyses, the percentages of frequency all follow the same decline trend. The trend shows at least 90 percent of the production gaps lasting 2 or fewer years, 96–98 percent of the gaps lasting 5 or fewer years, and 98–99 percent of the gaps lasting 10 or fewer years (fig. 1).

Given some of the limitations of the data, like variations among States and differences in the amount and type of data collected, we chose to be conservative in our definition and defined an abandoned well as a well that has not produced for 5 or more years. Most production pauses are 2 or fewer years, but given the limitations of the data, it was determined that 2 years could fall within the minimum time frame required by some States for data reporting. Additionally, using a gap of 10 years would capture almost all occurrences of inactivity but would miss many wells due to the lag time to reach 10 years for inclusion in a cutoff (for example, a well that had not produced for 6 years would be considered active because it had not been inactive for 10 years). Setting the threshold at 5 years accommodates long periods of inactivity while excluding records with short production gaps due to reporting, economic, or maintenance issues.

Definition of Well Abandonment

The methodology in this report does not use well attributes reported in the S&P Global database, like well status or production status, when defining well abandonment. Because data sources vary in how they define statuses, this methodology relies on production data as a starting place for defining hydrocarbon well abandonment. With quantified production gap frequency data on hand, the definition of well abandonment can be tailored using well data as a guide. For example, if a user has enough well data to quantify production gaps, then they could use their data to justify using production gaps greater than 2 years as indicating well abandonment instead of our use of greater than 5 years. This report uses the structure of the PIDM database to quantify production gaps because the database organizes production by entity, entity type, fluid, and year and includes production volumes for all months within each year recorded. This structure also allows well production to be classified as either well-level or lease-level production. Analyzing production gaps on a national scale permits a cutoff to be determined between production pauses and abandonment using the data as a guide. We define an abandoned well as a well with a spud date (the date when drilling operations begin for a new well) before January 1, 2021, that either has no associated production or has no production reported in the last 5 years.

This methodology only considers wells started 2 years before the most recent vintage of the hydrocarbon data because State reporting requirements vary across the United States, and not all States require 2 years for their data to become available. This results in two ways a well can be considered abandoned. First, if a well completed before January 1, 2021, has no associated recorded hydrocarbon production, it is considered abandoned. Second, if a well completed before January 1, 2021, has associated recorded hydrocarbon production (at the well or lease level) but has no reported production in the last 5 years, the well is considered abandoned.

Method for Identifying Abandoned Wells from Oil and Gas Data

In developing this methodology, a Java (Oracle, 2023) program was created to systematically query the proprietary S&P Global U.S. PIDM oil and gas database (which is no longer active) to identify abandoned and active wells using the abandonment definition described in the section of this report titled “Definition of Well Abandonment.” Given the proprietary nature of the underlying database and the small potential userbase, this report does not include the data or the Java program code but describes the general steps taken by the program to identify active and abandoned wells. The S&P Global U.S. PIDM contains proprietary data restricting the release of raw data, which limits the usefulness of the Java code because only users with access to PIDM would be able to use the Java code. Also, S&P Global offers many different products related to U.S. oil and gas data, with PIDM being an older product. Only a small userbase would be able to utilize the Java code. However, the program is logically straightforward and easily recreated from figure 2, which is a flowchart of the steps in the methodology as described in the rest of this section.

The program starts by compiling, from the main well table, all UWIs not including the UWIs that have a spud date 2 years or less from the analysis date (February 2023). This initial step creates a virtual list of all available unique wells stored within the larger database. Once all UWIs of interest have been compiled, the code then loops through all the UWIs and looks at each UWI individually. For each UWI, all associated producing entities are compiled. Because the production is organized separately from the well information, this step links the UWI to the associated production entities and determines if well or lease-level production is analyzed. If any of the production entities associated with the UWI has an entity type of “lease,” lease production will be analyzed; otherwise, well production will be used. Once the type of production has been determined, the process continues by looping through the compiled entities. This loop allows the program to consider each producing entity associated with each UWI.

For each entity, the database is queried to determine which hydrocarbon fluids have been produced during the entity’s lifetime. For each fluid the entity has produced, the production volumes for the previous 5 years are compiled. This step allows the program to compile all hydrocarbon volumes for the previous 5 years on a per fluid and per entity basis for each UWI. Each fluid is evaluated, and if all produced volumes for the past 5 years are empty or zero, the program marks the fluid as “not present” and progresses to analyze the next fluid. If any of the production volumes are greater than zero for any of the previous 5 years, the program marks the fluid as “present” and progresses to the next fluid. This step permits all entities and fluids to be checked individually and a decision made regarding whether production is present or not.

After all hydrocarbon fluids have been analyzed, if all fluids are marked as “not present,” the entity is marked as “abandoned.” If any of the hydrocarbon fluids are marked as “present,” the entity is marked as “active.” This programming step decides whether a unique entity associated with a unique UWI is considered abandoned or active by analyzing all hydrocarbon fluids associated with the unique entity. Once all the entities associated with a UWI have been analyzed, if all entities are marked as “abandoned,” the UWI is considered abandoned. If any of the entities are marked as “active,” the UWI is considered active. This is the final determination for each UWI whether the well is considered active or abandoned. Once all the millions of UWIs have been classified as either active or abandoned, the program is almost complete. The final step is to write the data to two separate files, one file containing the abandoned well information, and the other containing the active well information. As of the date of analysis (February 2023), one file contained more than three million abandoned records, and the other file contained more than one million active records.