‹ Weekly Natural Gas Storage Report

Methodology for EIA Weekly Underground Natural Gas Storage Estimates

Latest Update: March 10, 2015

This report consists of the following sections:


The Energy Information Administration (EIA) provides weekly estimates of working gas volumes held in underground storage facilities at the national and regional levels. Weekly estimates of working gas in storage were first provided by the American Gas Association (AGA) in 1994. In October 2001, the AGA announced that it would discontinue its survey due to resource considerations. AGA's last report appeared May 1, 2002, containing data for the report week ending April 26, 2002. The EIA first released its estimates of underground storage for the week ending May 3, 2002, on May 9, 2002, in the Weekly Natural Gas Storage Report (WNGSR). The WNGSR is posted on the EIA web site on the internet at http://ir.eia.gov/ngs/ngs.html 10:30 a.m. each Thursday, except for certain weeks that include Federal holidays.

This report is a description of the methodology that has been used by the Energy Information Administration to prepare weekly estimates of working gas in storage, starting in August 2005. The first estimates from the revised methodology appeared in the August 4, 2005, edition of the WNGSR with stock estimates as of July 29. The overall approach relies on weekly survey data from a sample of operators of underground storage facilities. These data are used to prepare regional and national estimates for all underground storage.

Survey and Survey Processing

The goal of the weekly storage data program is to provide weekly estimates of the level of working gas in underground storage for the United States and three regions. The total volume of natural gas in underground storage reservoirs is classified as either base gas or working gas (see Notes and Defintions for definition of terms). Underground storage facilities may be reservoirs in depleted oil and gas fields, aquifers, or salt caverns (a more detailed discussion of underground storage concepts is available in the EIA report, The Basics of Underground Natural Gas Storage).

EIA designed Form EIA-912, "Weekly Underground Natural Gas Storage Report," to collect weekly survey data on working gas in storage. The currently approved form and instructions are presented as Exhibit 1 and Exhibit 2. The Form EIA-912 requests reports of the volumes of working gas in storage as of 9 a.m. Friday of the previous week from a sample of underground natural gas storage operators. Respondents are asked to provide data representing working gas in storage fields their company operates in each of three regions of the United States. For most respondents operating both salt and nonsalt fields in the Producing region, EIA internally classifies the entire inventory reported based on the field type of the preponderance of the volume. In unique situations where respondents in the Producing region operate both salt and nonsalt fields with working capacities in excess of 15 billion cubic feet for their salt facilities, they report the working gas inventory of the salt fields separately from the working gas inventory of the nonsalt fields in the Comments section of Form EIA-912. Estimates for the Producing region are calculated by adding the estimates for Producing Salt and Producing Nonsalt fields.

Respondents are also instructed to submit revisions to data for previous weeks if those revisions were greater than 500 million cubic feet and to include notes explaining any unusual activity. Examples of unusual activity might include reclassification of working and base gas, or changes in ownership or operation of storage fields. In practice, respondents do not often have need to provide notes on unusual activity.

EIA has provided copies of the survey to all respondents and has also provided electronic versions of the form on its web site that can be accessed as an Excel spreadsheet or a PDF file. In order to provide data on a timely basis, companies are required to report by fax or E-mail.

Respondents provide estimates for working gas in storage as of 9 a.m. Friday each week. The deadline for submitting reports to the EIA is 5 p.m. Eastern Time the following Monday, except when Monday is a Federal holiday. In that case, forms are due at 5 p.m. on Tuesday. The estimates are released on Thursday at 10:30 a.m. (Eastern Time) on EIA's web site, except for certain weeks that include Federal holidays. Notification of changes to this general schedule is maintained on the EIA web site at http://ir.eia.gov/ngs/schedule.html, which also is directly accessible by use of a hyperlink titled, "Schedule," located on the side of the WNGSR web-page.

EIA employs a number of editing processes to ensure that the data collected each week are accurate. For example, the current week's data are compared with data reports for recent and relevant historical weeks and to compilations of the monthly data reports for the company's fields on the EIA monthly storage form. EIA also employs secondary source information such as weather data and stock or net change predictions in evaluating the validity of reported data. Companies with responses that are outside the edit bounds or with notes about special issues are contacted by survey personnel for confirmation or correction. An explanation is obtained and accommodated in the estimation process, if necessary.


A sample for the EIA-912 was drawn from the list of respondents to Form EIA-191, "Monthly Underground Natural Gas Storage Report," for the purpose of preparing weekly estimates of natural gas in underground storage facilities in each of three regions. The Form EIA-191 is completed by all operators of underground natural gas storage fields in the United States. Data are provided for storage volumes in each storage field and reservoir. Each month EIA presents these data in the Natural Gas Monthly aggregated by State and storage region. Approximately 127 underground natural gas storage operators report for their operations at approximately 399 storage fields. For the sampling and estimation for the WNGSR, the data from the EIA-191 are converted to the sums of field volumes by company and region.

The procedure used a stratified sample design with probability proportionate to size (PPS) sample selection, with a measure of size equal to the average between selected October and March monthly volumes. There are four strata that are based on geographic region and field type (salt or nonsalt): East, West, Producing-Nonsalt, and Producing-Salt.

Reported volumes of working gas in storage as reported on the Form EIA-191 were aggregated by storage operator and region to prepare the sampling frame for the EIA-912 for each region. For each stratum, the data were divided into two groups. The first, the certainty group, consisted of the larger operators in the region as well as all operators with storage fields in more than one region. All operators in the certainty group were selected for the sample. The second, the noncertainty group, consisted of all other operators, who were selected for the sample with a probability proportional to their size.

A stratified sample of companies was selected from the list of operators to achieve a target standard error of the estimate that was no greater than 3 percent for total working gas in storage in each region. Reports of operators in the certainty group contribute nothing to sampling error. This process ultimately resulted in a set of roughly 70 respondents to the EIA-912 survey that are used for estimation.

Ensuring the quality and accuracy of the weekly estimations depends upon maintenance of the EIA-191 frame. Because the respondents to the EIA-191 are subject to change as a result of births, deaths, and evolving industry structure, ensuring the correspondence between the sample of storage operators and the universe of storage operators from which the sample is drawn necessitates updating the sample periodically with a full sample reselection, usually every 1 to 2 years. During the period in between full sample re-selections, changes in the sampling frame are conducted as indicated in the estimation methodology discussed below.


The method of estimation uses both a 12-month moving average of working gas volumes (based on the latest published EIA-191 monthly data), and the latest EIA-912 weekly data. The trends exhibited by the data from the weekly sample group are used to estimate the latest weekly values for the storage operators not reporting on the EIA-912. Each individual company outside the weekly sample is considered based on its own past history in the monthly data from the EIA-191 survey and the weekly behavior of sampled companies within the same stratum. These company-level weekly estimates of working gas stocks for the reporting and non-reporting operators are summed to form the weekly estimated totals for each of the four noncertainty strata.

More specifically, for each company in the EIA-191 frame, 12-month equal-weighted moving averages and corresponding standard deviations and coefficients of seasonal variation are computed as described below. For each weekly reporting company i, we calculate a company-level ratio, ri,t , of the weekly stock volume reported in the EIA-912 for week t (denoted wi,t ) to the most recent available 12-month moving average of reported monthly stock volumes:

    formula graphic

    formula graphic

    and T represents the reference month for the most recent available monthly stock volume, mi,T , reported by company i . (Usually month T is one or two months prior to the month that includes week t.)

In addition to the mean, the standard deviation across months,formula graphic, and a coefficient of seasonal variation, ci,T , are calculated for formula graphic :

    formula graphic

For each sampling stratum k, let Ck,T denote the median of the ci,T for all weekly-reporting companies i in stratum k. Similarly, let Rk,t represent the median value of ri,t for all weekly-reporting companies i in the stratum.

Let j denote a company in stratum k that is not in the weekly-reporting sample, and let

    formula graphic

We compute formula graphic and cj,T as described above for the weekly-reporting units. The volume of natural gas stored by company j in week t is estimated as

    formula graphic

    where p(j)=cj,T /Ck,T , a factor representing the seasonal pattern of unit j relative to that of the weekly-reporting units in stratum k.

The company-level estimation equation captures the seasonal variation of a specific company relative to that of the weekly-reporting companies in its stratum. The generality of the estimation equation makes it quite flexible and capable of representing a wide variety of possibilities. For example, if a given company has no variation over the previous 12 months for which estimates are available, so that cj,T = 0 , the estimation equation implies that the latest weekly volume is equal to the 12-month moving average, formula graphic . If the company's variation matches the median pattern of its stratum, then cj,T = Ck,T , and the estimation equation implies that the latest weekly volume is equal to the strict moving average estimator, formula graphic .

EIA estimates current working gas inventory for a region as the sum of two total components: the total volume of the reporting companies and the total of the estimated volumes for companies not reporting on the weekly (the Producing-Salt and Producing-Nonsalt strata are combined to obtain the Producing region total). For each stratum k, the estimated volume of working gas stored in week t is computed as:

    formula graphic

    Nk is the number of companies in stratum k,
    nk is the number of weekly-reporting companies.

Estimates for the Producing region are computed by adding the estimates from the Producing Salt and Producing Nonsalt strata. The estimate for the Lower 48 States is the sum of the estimates for the three regions.

The estimation technique is not model-based. However, specific 12-month moving averages and their associated coefficients of seasonal variation are updated with the most recent month's data as the data become available. Additional adjustments to the working gas data will be incorporated on an ongoing basis as necessary. Some examples are discussed below:

Reclassification of base and working gas by a sample member: Occasionally operators of storage fields change the classification of some gas in the storage field from base gas to working gas or from working gas to base gas. (See Glossary for definition of terms.) Because the EIA-912 is a survey of working gas, this reclassification leads to an apparent change in the volume of available gas. A reclassification of significant size is handled by including the reclassified working gas volume in the data used for the calculations of the 12-month moving average and the coefficient of variation. When EIA observes such a change in classification, EIA notifies report users of this change in inventory levels with a note about the reclassification if the effect of the reclassification and other changes are 7 billion cubic feet or more.

Purchase or sale of a storage field: The adjustments to the estimation system will vary depending on whether the company is in the sample company or non-sample company groups.

Startup or shutdown of a field: Field shutdowns are handled by simply eliminating the company in the case of a single-field company or by removing the associated volume from the 12-month moving average for the appropriate company. In the case of a field startup, the moving average is based on the available information from the EIA-191 survey, so a complete 12-month average is not formed until a year's worth of data is reported on the EIA-191. Field startups exceeding a certain threshold level will become part of the EIA-912 sample.

Negative values: Although not a common occurrence, from time to time a company may report a negative working gas value, which may be caused by removal of base gas from storage in anticipation of sale or abandonment of a storage field or withdrawal from base gas at the end of the withdrawal season. When confirmed negative values are received for a field expected to be shut down, a value of zero is entered into the database. If the report is caused by temporary withdrawals from base gas, such as occurred toward the end of the winter 2002-2003, EIA announces the occurrence and describes the adopted changes in estimation procedures at that time.

Occurrences of non-response or data quality issues of individual companies are addressed through implicit or explicit imputation. Imputation of company data values for any stratum occurs either by treating the nonrespondent as a company not reporting on the weekly or by using company-specific linear extrapolation methods based on the most recent data.

Computing the 5-year Averages, Maxima, Minima, and Year-Ago Values for the Weekly Natural Gas Storage Report

1.) Current Method

Beginning with the Weekly Natural Gas Storage Report for the week ended January 7, 2005, EIA implemented a daily interpolated average approach to calculating the 5-year summary statistics and year-ago values reported in the WNGSR. The daily interpolated average approach facilitates simple yearly comparisons of the data, unencumbered by possible confusion associated with ordering weekly data.

The calculations are based on daily linear interpolations of the working gas inventory values between each week reported in the Historical Weekly Storage Estimates Database. Using the historical data, daily data are derived by (1) computing an average daily rate of stock change for each week based on the published weekly data; (2) using this daily rate and the published weekly stock levels to estimate daily stock levels for each day in between the report days. The interpolated daily values are derived as follows:

    formula graphic

    It = EIA-estimated interpolated daily inventory level for day t,
    Es = EIA-estimated weekly inventory level for day s,
    t = {0,1,2,..., T}
    sformula graphic t = {0,7,14,..., T} corresponds to the weekly report day, i.e., Friday.
    0 ≤ t - s ≤ 6 is the number days between t and the most recent report day s.

With the resulting daily working gas time series, 5-year averages are computed for each day of the year, using the preceding 5 years of daily estimates. For example, the 5-year average used as reference for 01-07-2005 includes the data for 01-07-2000, 01-07-2001, 01-07-2002, 01-07-2003, and 01-07-2004.

The Lower-48 States minimum and maximum values that are used in the storage graph included in the WNGSR are based on a similar treatment of the weekly data. However, regional working gas maxima or minima are calculated independently and so do not necessarily sum to the maxima or minima determined for the Lower 48 States.

The year-ago values in the weekly report are those estimates for the same day of the prior year as the current release. As such, they do not match published data for the prior year. For example, the year-ago values associated with the January 7, 2005, data release are the estimates calculated for January 7, 2004, not those for the report weeks in 2004, which would be for either January 2, 2004, or January 9, 2004.

The 5-year (2000 -2004) averages, maxima, minima, and year-ago values for each report week in 2005 are available for download at: http://ir.eia.gov/ngs/ngsstats.xls

2.) Method Used for Estimates from May 2002 Through December 2004

The 5-year average volume was computed as a simple arithmetic mean of recorded volumes in corresponding weeks during the 5-year period. Typically each Friday in a given year was assigned a number from 1 to 52. The first Friday of January was considered the first week of the year, so it was assigned the number 1. The following Friday was identified as week 2, and the procedure continued for all remaining weeks in the year. This procedure orders the weeks, allowing calculation of the 5-year average results presented in the Weekly Natural Gas Storage Report (WNGSR). The historical data for the weeks from January 1994 to the present are updated each week and are located on the EIA website at: http://ir.eia.gov/ngs/ngshistory.xls. There also is a link located on the WNGSR that directs users to this site.

As an example, the following table contains the data for the first Friday of each year from 1999 to 2003. These weekly values were the basis for the calculated 5-year (1999-2003) averages, which were used as the reference values in the first week of 2004.

Table 1. Sample Calculation for 1st Week of 2004, Using First Weeks in 1999-2003
Date East West Producing Total Week #
1/1/1999 1,615 330 762 2,707 1
1/7/2000 1,435 297 686 2,418 1
1/5/2001 1,035 224 352 1,611 1
1/4/2002 1,600 354 835 2,790 1
1/3/2003 1,333 342 656 2,331 1
5-year average 1,404 309 658 2,371  

Given that a year consists of 52 weeks plus one or two days, some years contain 53 Fridays. This circumstance occurs only once every few years, so generally there was not a sufficient number of data observations in a given 5-year period to average only observations for the 53rd week. For example, 1999 and 2004 both contain 53 Fridays. The values of the 5-year averages used in releases of the WNGSR during 2004 were based on data for 1999-2003. For the release of storage data for December 31, 2004, which was the 53rd and final week, the use of average data for only the 53rd weeks in the period would be limited to only the 1999 observation.

As an approach to establish the average value when a 53rd Friday occurs, which occurs on either December 30 or 31, the 5-year statistics were based on estimated values for corresponding dates during the 5-year reference period. The calculations were based on a linear interpolation of the working gas inventory values between the last week of each year and the following week. The interpolated daily values were derived using the equation in the preceding section.

As an example, the following table contains the interpolated estimates for December 31 of each year from 1999 to 2003. These weekly values were the basis for the calculated 5-year (1999-2003) averages, which were used as the reference values for December 31, 2004.

Table 2. Sample Calculation for 53rd Week of 2004, Using Final Weeks in 1999-2003
Date East West Producing Total
12/31/1999 1,482 321 720 2,523
12/31/2000 1,104 230 387 1,721
12/31/2001 1,674 359 869 2,904
12/31/2002 1,362 347 659 2,368
12/31/2003 1,505 324 753 2,582
5-year average 1,425 316 678 2,419

The minimum and maximum values that are used in the storage graph included in the WNGSR were based on the same treatment of the weekly data.

Derivation of the Weekly Historical Estimates Database

The Energy Information Administration (EIA) began collection and estimation of weekly volumes of working gas in U.S. underground natural gas storage facilities in the Spring of 2002. The first survey responses to the EIA-912, "Weekly Underground Natural Gas Storage Report," were due Monday, March 18, 2002 with data for the week ended Friday, March 15, 2002. The initiation of this new series of survey estimates raised interest in a comparable set of weekly estimates for prior periods to provide a historical context for the new values. The method for developing the weekly data varied over the historical period.

  • Data for December 31, 1993, are based on published data from the EIA monthly survey, EIA-191, 'Monthly Underground Gas Storage Report.'
  • The data for the weeks from January 7, 1994, through February 22, 2002, are estimates derived from a computation process that used both EIA monthly survey data and AGA weekly survey data. (The derivation of these data is described in the remainder of this section.)
  • Data for March 1, 2002, are the end-of-February published EIA-191 data plus one-seventh of the net inventory changes reported by AGA for the week ended March 1, 2002.
  • The data cell values for March 8, 2002, are the midpoint values for March 1 and March 15, 2007.
  • The data for the weeks beginning with March 15, 2002, to the present are the product of the EIA weekly survey, EIA-912, 'Weekly Underground Natural Gas Storage Report.'

Derivation of data from January 7, 1994, through February 22, 2002. The EIA generated a set of weekly estimates for the period from 1994 to March 2002 based on the EIA monthly data series and the weekly AGA series. The derived series of weekly estimates consists of the AGA weekly series calibrated to the EIA monthly series and is expected to provide a consistent basis for comparison with the EIA weekly series based on the EIA-912 survey.

The estimation method adjusts the AGA weekly values by linearly interpolated values reflecting the ratios of monthly EIA storage volumes and monthly AGA volumes. The steps used to implement the methodology were as follows.

1) Estimates for end-of-month inventory levels based on the AGA series first were calculated for the majority of months covered by the AGA data series, because the ends of those months did not coincide with the end of an AGA "storage week" (Friday). End-of-month estimates for the AGA series were computed by linear interpolation of the corresponding weekly AGA estimates based on the number of days of the storage week contained in the adjacent months.

2) The ratio of EIA end-of-month working gas in storage to AGA estimated end-of-month working gas in storage was calculated for each month. Weekly equivalents of these values were calculated by linear interpolation as described in steps 3 and 4.

3) A step value was calculated for each month by subtracting the end-of-month ratio for the prior month from the ratio for the month of interest and dividing the result by the number of days in the month of interest.

4) A "weekly adjustment factor" for each Friday (i.e., each storage week) in the month of interest was derived by multiplying the step value for that month by the number of days elapsed in the month (in practice, the number of 'elapsed' days is equal to the date of the Friday marking the end of the storage week) and adding that value to the EIA/AGA ratio for the prior month. For example, if the first Friday of the month occurred on the third day of the month, the weekly adjustment factor for the storage week ending on that Friday would be computed by multiplying the month's step value by 3 and adding that increment to the EIA/AGA ratio for the prior month.

5) Finally, EIA weekly volume estimates were derived by multiplying AGA weekly volumes by the weekly adjustment factors, thus yielding a weekly series calibrated to the EIA monthly data and reflecting weekly variation similar to that exhibited by the AGA weekly series.

The full implementation of steps 1-5 can be represented in equation form. Specifically, the stock level in any given week in a region was derived as follows:

    formula graphic


    Ew = EIA-estimated weekly inventory level for Friday of week w in month m,
    Em inventory level for the end of month m, estimated using data from the EIA-191,
    Aw = weekly inventory level for Friday of week w, as estimated by the AGA,
    Am = inventory level for the end of month m, estimated by linear interpolation from weekly AGA estimates,
    Dm = number of days in month m, and
    dw,m = number of days between the beginning of month m and Friday of week w in month m.

In creating this derived weekly series, EIA's primary focus was to produce a weekly series that was fully consistent with the EIA monthly data. (As one can see from the methodology formula, when the end of a month occurs on a Friday, so that it coincides with the end of a reporting week (i.e., D(m)=d(m)), the equation reduces to the multiplication of AGA end-of-month volumes by the EIA/AGA end-of-month ratio, which yields the EIA end-of-month volumes.) The adopted approach produces a series of weekly estimates calibrated to the EIA monthly data and that reflects the variation in AGA weekly volumes.

Another major objective was for the methodology to be relatively simple and straightforward, and easy to understand. The reliance on linear relationships is a practical and useful approach, although it must be understood to be an approximation of the underlying data.

Summary of Base Gas Withdrawal Activity During the 2002-2003 Heating Season

(This Summary originally appeared in the Weekly Natural Gas Storage Report released on August 8, 2003. It is reproduced here as a reference document.)

Withdrawals from base gas volumes occurred toward the end of the 2002-2003 heating season. EIA first reported withdrawals from base gas in the Weekly Natural Gas Storage Report (WNGSR) for the week ended March 7, 2003. This section of the methodology documentation provides an overview with summary statistics regarding the magnitude of base gas withdrawals.

The cumulative net withdrawals of base gas were treated as negative working gas stocks and were reflected in the working gas inventories reported in the WNGSR. This approach allowed the timely release of information about the total net gas supply from storage for the week while addressing potential risks regarding release of confidential information given the small number of individual respondents.

The resulting respondent data (working gas net of cumulative base gas withdrawals) were used according to the estimation methodology for the WNGSR. Estimated cumulative net withdrawals from base gas grew from the first week the practice was reported (March 7, 2003), and peaked at approximately 19 billion cubic feet (Bcf) for the week ending April 11 (Table B-1). As of May 9, the difference between working gas with the base gas adjustment and working gas without the base gas adjustment was less than 7 Bcf. Because only changes greater than 7 Bcf are required under the EIA revision policy, EIA has determined that the explicit release of region-specific base gas data should be limited to the nine weeks (the weeks included in the table) when the national difference met the revision threshold.

For purposes of making comparisons of weekly data from winter 2002-2003, weekly data from previous winters, and its monthly data series, EIA is providing revised estimates of weekly working gas in storage without base gas withdrawal adjustments for the period from March into May (Table B-2). The estimates excluding any base gas adjustments will be incorporated in the WNGSR historical database to ensure its consistency with other EIA publications, definitions, and data series. These estimates will be available to all customers as part of the standard downloadable data file in Excel format available in the historical data section. The replacements will be used to compute historical averages, and maxima and minima of working gas in storage.

The weekly stock estimates previously published for this period, which reflected the full quantity of gas withdrawn weekly from storage (base and working gas), are provided below for reference (Table B-3). The footnote about the withdrawals from base gas is discontinued with the release of the WNGSR on August 14, 2003, containing estimates for August 8.

Table 3. Cumulative Base Gas Withdrawals (Bcf)
Week ending Consuming Region East Consuming Region West Producing Region Total Lower 48 Incremental Net Change in Base Gas Withdrawals
07-Mar-03 12 0 3 15 N/A
14-Mar-03 13 0 5 18 3
21-Mar-03 12 0 5 17 -1
28-Mar-03 12 0 4 16 -1
04-Apr-03 13 0 4 17 1
11-Apr-03 13 0 6 19 2
18-Apr-03 15 0 3 18 -1
25-Apr-03 11 0 2 13 -5
02-May-03 7 0 1 8 -5

Table 4. Revised Weekly Natural Gas Stock Estimates, Without Adjustment for Base Gas Withdrawals (Bcf)
Week ending Consuming Region East Consuming Region West Producing Region Total Lower 48
07-Mar-03 343 179 214 736
14-Mar-03 280 172 202 654
21-Mar-03 289 167 204 660
28-Mar-03 317 167 212 696
04-Apr-03 311 167 210 688
11-Apr-03 277 167 198 642
18-Apr-03 313 174 215 702
25-Apr-03 344 180 230 754
02-May-03 392 192 252 836
Note: these estimates are based on the methodology and data available as of August 8, 2003.

Table 5. Original Weekly Natural Gas Stock Estimates, With Adjustment for Base Gas Withdrawals (Bcf)
Week ending Consuming Region East Consuming Region West Producing Region Total Lower 48
07-Mar-03 331 179 211 721
14-Mar-03 267 172 197 636
21-Mar-03 277 167 199 643
28-Mar-03 305 167 208 680
04-Apr-03 298 167 206 671
11-Apr-03 264 167 192 623
18-Apr-03 298 174 212 684
25-Apr-03 333 180 228 741
02-May-03 385 192 251 828
Note: these estimates are based on the methodology and data available as of August 8, 2003.