Hurricanes and production: Do more disruptions loom?

Aug. 14, 2006
The Atlantic hurricane season of 2005 was the most active season since accurate record-keeping began more than 60 years ago.

The Atlantic hurricane season of 2005 was the most active season since accurate record-keeping began more than 60 years ago. There were 27 named storms, including 15 hurricanes, seven of which were classified as Category 3 or higher. The paths of five major hurricanes passing through the Gulf of Mexico disrupted crude oil and natural gas production.

Hurricanes Katrina and Rita, in particular, crossed straight through the heart of the gulf’s producing region, resulting in widespread production shut-in, some of which remains in effect.

As of June 17, total oil production in the gulf had been reduced by more than 166 million bbl and gas production reduced by 804 bcf since the first of the hurricanes struck nearly a year ago. This amounts to about 30% of a normal year’s oil production and 22% of gas production from federal offshore fields.

However, hurricanes and tropical storms in the Gulf of Mexico typically cause only temporary disruptions to oil and gas operations. This article examines the incidence of severe weather in the Gulf of Mexico over the last 55 years and the impact on the production of oil and gas from the Outer Continental Shelf (OCS).

An econometric model relating the level of hurricane activity and shut-in production was developed by the Energy Information Administration (EIA) and used to simulate the potential impacts during the upcoming hurricane season.1 This simulation is based on the seasonal hurricane outlook published by the National Oceanic and Atmospheric Administration (NOAA) in May. EIA emphasizes the inherent uncertainty of any forecast of shut-in production because of the difficulty of predicting both the location of future gulf storms and the impact of any storms on OCS production. However, given NOAA’s current forecast, the likelihood of repeating last year’s oil and gas production disruptions is relatively small.

Classifying storms

Severe tropical weather is classified by the intensity of the weather system’s wind speeds. Tropical depressions have a maximum sustained wind speed of 38 mph, tropical storms have a maximum sustained wind of 39-73 mph, and hurricanes have winds exceeding 73 mph.

In the US, hurricanes are categorized on the Saffir-Simpson scale by maximum observed wind speed: Category 1 (74-95 mph), Category 2 (96-110 mph), Category 3 (111-130 mph), Category 4 (131-155 mph), and Category 5 (156+ mph).

Hurricane season in the Atlantic Basin officially runs from June 1 through Nov. 30, when ocean waters are warm enough to spawn tropical weather systems. However, about 75% of major hurricanes (Category 3 or greater) occur in August and September.

Click here to enlarge image

NOAA describes the intensity of tropical storms and hurricanes using the Accumulated Cyclone Energy (ACE) index, which measures the collective intensity or strength of the wind and the duration of all tropical storms and hurricanes in the Atlantic Ocean. Specifically, it is defined as “the sum of the squared maximum sustained wind speeds (in knots) for all 6-hr intervals during which the weather system is classified as either a tropical storm or hurricane.”

The ACE index can be used to describe the intensity of particular storms or to describe the seasonal intensity (as the sum of ACE indices for all storms). Detailed 6-hr observations of wind speed, atmospheric pressure, and location for all named storms since 1851 are available in NOAA’s Atlantic Basin hurricane database (http://www.aoml.noaa.gov/hrd/hurdat/Data_Storm.html). EIA used this database to construct seasonal ACE indices for the entire Atlantic Basin and for the Gulf of Mexico, which is defined as the region bounded by 18º N-31º N latitude and 81º W-98º W longitude.

Click here to enlarge image

Fig. 1 shows the annual Atlantic Basin ACE index for 1950-2005 along with the number of tropical storms and hurricanes. The seasonal intensity of tropical weather systems runs in what is termed multidecadal cycles. During the 1950s and 1960s, hurricane activity was above normal, while the 1970s, 1980s, and early 1990s exhibited below-normal activity. Since 1995, the Atlantic has again been experiencing above-average activity, with an ACE index almost twice the median value for the entire period.

Click here to enlarge image

Table 1 shows a summary of the seasonal average levels of hurricane activity for 1950-2005 and for the most recent above-average period. Since 1950, an average of 10 named storms have occurred in the Atlantic Basin, with 20% of those eventually becoming major hurricanes. These statistics highlight the cyclical increase in hurricane activity since 1995. The ACE index for the entire Atlantic over the period 1995-2005 is nearly double the median value for 1950-2005 and about 40% higher in the Gulf of Mexico. Much of this higher average is due to the hyperactive 2005 hurricane season.

Atlantic tropical storms and hurricanes usually follow one of three general paths: along the Eastern seaboard, across Florida and striking the Gulf Coast, or through the Caribbean Sea and moving north to strike the Gulf Coast. About one third of Atlantic storms eventually pass through the Gulf of Mexico. The map tracks major hurricanes that passed through the gulf during 1995-2005 and highlights the intense weather experienced by the Texas, Louisiana, Mississippi, and Alabama coasts over the last decade.

Production impacts

In 2004, oil production from federally administered Gulf of Mexico fields accounted for nearly 27% of total US production, and marketed production of gulf OCS gas was about 20% of the US total.2 3 Texas, Louisiana, Alabama, and Mississippi also contribute onshore and state-administered offshore production.

When severe weather threatens producing areas, however, platform operators must shut in production to protect workers and facilities. Hurricanes Katrina and Rita passed directly over the OCS producing region, and at one point just prior to the landfall of Katrina, 79% of gulf platforms were evacuated. This caused 1.4 million b/d of oil (95% of normal gulf OCS daily production) and 8.8 bcfd of gas (88% of normal daily production) to be shut in.4

Katrina destroyed 44 platforms as it passed over the OCS producing region, including some deepwater projects that are still under construction.5 As Rita passed over the producing region, 93% of the platforms were evacuated, forcing operators to shut in 100% of normal oil production and 81% of normal gas production. Hurricane-force winds destroyed 69 platforms.

In addition to the upstream impacts on gulf production, the hurricanes damaged 457 underwater pipelines, and the Louisiana Offshore Oil Port had to temporarily stop accepting shipments during both hurricanes.6 Some onshore refineries and gas processing facilities suffered heavy damage. After Katrina hit Louisiana, nearly 2 million b/d of refinery capacity was shut down due to direct damage or interruption of power supplies. A total of more than 4.9 million b/d of refinery capacity was shut down after Hurricane Rita hit the Texas coast.

Click here to enlarge image

Although Hurricanes Katrina and Rita caused long-lasting effects on oil and gas operations in the gulf, severe weather in the region historically has had only a relatively minor impact. Fig. 2 shows oil and gas production in the gulf OCS for 1995-2005 with the effects of various tropical storms and hurricanes highlighted. The disruptions from severe weather can often be identified by temporary dips in production from normal levels. There have been six major hurricanes during the past decade that caused significant disruption in oil and gas production: Opal (1995), Georges (1998), Lili (2002), Ivan (2004) and Katrina and Rita (2005).

However, with the exception of Ivan, which shut in about 25% of monthly production, and Katrina-Rita, which shut in about 70% of monthly production, most disruptions have been temporary with near-normal production returning the following month. In fact, most weather systems in the gulf shut production in for only a few days. For example, in 1997 Hurricane Danny passed within 50 miles of the center of OCS production, yet it registered a barely perceptible drop in daily production rates, shutting in about 2% of that month’s oil and gas production. Hurricane Bret (1999) with 125 mph winds slightly impacted oil production but had almost no effect on the trend in gas production.

Annual shut-in outlook

Seasonal hurricane-related disruptions to oil and gas production are difficult to forecast, primarily due to the uncertainty involved in predicting the location and intensity of severe weather. However, an analysis of historical impacts can provide some insight into the range of potential effects given a seasonal hurricane forecast. EIA simulates annual shut-in oil and gas production, using information about the NOAA seasonal hurricane outlook. This simulation is based on an estimated econometric model that relates seasonal hurricane activity as measured by the ACE index and shut-in gulf OCS production.

In May of each year, NOAA publishes an outlook for hurricane activity in the Atlantic Basin, including the Caribbean Sea and the Gulf of Mexico, during the upcoming hurricane season. One of the primary components of this outlook is NOAA’s forecast of the seasonal ACE index measuring storm intensity and length. For 2006, NOAA currently expects the seasonal Atlantic ACE index to fall between 118 and 179 (135-205% of its normal level) and forecasts an 80% chance of an above-normal, very active hurricane season but one considerably lower than the Atlantic activity observed last year, which had an ACE index 280% of the normal level.

The Minerals Management Service (MMS) collects shut-in information directly from oil and gas producers in the gulf. However, detailed shut-in production statistics reports are only available for a few recent hurricanes, so some other method of estimating historical shut-in amounts is required. One recent paper analyzing the vulnerability of gulf production modeled shut-in volumes as the difference in production between hurricane season months and months outside the season.7

EIA’s analysis takes a different approach and estimates shut-in volumes based on short-term trends in gulf OCS oil and gas production data reported by MMS (http://www.gomr.mms.gov/homepg/pubinfo/pdfindex.html). The methodology in the EIA analysis is based on the observation that hurricanes and tropical storms often cause those noticeable but temporary dips below production trends. Hurricane Betsy in 1965, for example, caused a 36% production shut-in because actual oil production dropped to 254,842 b/d from what would have been a normal production of 400,240 b/d on a production curve joining production figures before the lost production and those following resumption of normal production.

Seasonal shut-in production is the sum of monthly shut-in levels for a given year. Gulf tropical storms and hurricanes typically cause small disruptions when measured on a seasonal basis. The average seasonal shut-in production (as a percentage of normal annual gulf OCS production) during 1960-2005 was 1.4% for oil and 1.3% for gas.

However, these averages are skewed upwards by the 19% of oil production and 18% of gas production shut in during 2005. Total cumulative shut-in production through June 19, 2006, totaled 30% of normal annual oil production and 22% of normal annual gas production, according to MMS. The median seasonal shut-in production has been only 0.6% of annual gulf OCS oil production and 0.5% of annual gas production.

Click here to enlarge image

The econometric model represented in Table 2 consists of two linear regression equations, using annual data for the sample period 1960-2005. The first equation relates the NOAA forecast of the Atlantic ACE to a Gulf of Mexico ACE. The second equation relates the Gulf of Mexico ACE to the percentage of oil or gas production that is shut in during the given year. Neither regression equation includes a constant term because a season with no hurricane activity implies no hurricane-related disruptions. In order to isolate the sources of forecast uncertainty, this dual-stage regression approach was adopted instead of a reduced form regression directly relating Atlantic hurricane activity to shut-in production.

The estimated coefficient for the first equation in the table shows that, on average, the annual gulf ACE index is about 17% of the total Atlantic ACE. Although this value is statistically significant, the low coefficient of variation (R2 = 0.48) indicates the relatively poor fit of this relationship. The Atlantic Basin covers a much larger area than the gulf; so greater hurricane activity within the Atlantic may not necessarily translate into greater activity within the gulf, depending on the path of individual hurricanes during a given season.

The second regression equation relates the square of the Gulf of Mexico ACE index to the percentage of normal annual gulf OCS oil or gas production shut in as a result of hurricanes or tropical storms. The explanatory variable is squared to reflect the observation that hurricane-related damage rises more than proportionally to hurricane intensity.8 The dependent variable is represented by the percentage rather than volume of shut-in production in order to control for the significant increase in offshore production over the last 45 years.

The results shown in Table 2 indicate that the fit of the shut-in oil (2a) and gas (2b) production equations is relatively robust, with 88% and 91%, respectively, of the variation in shut-in production explained by variation in the gulf ACE. However, the robust fit is primarily due to the 2005 observation. Estimating the equation over the period 1960-2004 causes the R2 value to fall to 32%, although the estimated coefficient does not significantly change. Thus, this second relationship also contains quite a bit of uncertainty.

EIA can use the estimated parameters from the two regression equations to produce a point estimate of the expected gulf ACE and a point estimate of the expected impact on oil and gas production, given the average of the NOAA Atlantic ACE forecast. But it is perhaps more important to describe the uncertainty of the shut-in production prediction. To examine this uncertainty, EIA performs a Monte Carlo simulation of the relevant random variables in order to create a sampling distribution for the amount of oil and gas production that may be shut in during the 2006 hurricane season. A Monte Carlo methodology encompasses a technique of statistical sampling used to approximate solutions to quantitative problems.

This particular simulation is a three-step process that begins with an assumption about the probability distribution of the NOAA Atlantic ACE value. A histogram of past values indicates that a log normal distribution fits the data well. EIA assumes that the Atlantic ACE forecast is based on this type of distribution, with a mean equal to the mean of NOAA’s forecast range (148.5). The standard deviation of the forecast distribution is equal to the standard deviation of ACE data from past years with above-average hurricane activity, defined by NOAA as an Atlantic ACE greater than 106.

The second step involves simulating an Atlantic ACE value from its sampling distribution and applying this value to the estimated parameters from the first regression. The resulting prediction for gulf ACE and its associated standard forecast error generate a normal sampling distribution from which a value for the gulf ACE index can be simulated. Any simulated negative values are converted to a value of zero. The final step is to apply the simulated gulf ACE value to the results from the second regression to calculate a predicted value and standard error for the percentage of annual production shut-in expected during 2006. A normal sampling distribution for shut-in production can be created from this information.

This sampling distribution provides a detailed description of both the expected level and uncertainty of possible lost production, given the initial NOAA 2006 Atlantic ACE forecast. This Monte Carlo simulation consisted of 10,000 random samples from the shut-in sampling distribution. EIA multiplied the simulated shut-in percentages by its 2006 annual production forecasts for federal Gulf of Mexico oil (530 million bbl) and gas (3.28 tcf) in order to calculate expected volumes of shut-in production.9 These simulated production values can be summarized with a cumulative probability distribution curve that plots the percentiles for various levels of shut-in production, i.e., the probability of no more than X million bbl of oil or X bcf of gas being shut in during 2006 (Fig 3).

Click here to enlarge image

Fig. 3a shows the cumulative distribution curve for the Monte Carlo simulation of shut-in annual gulf oil production given NOAA’s May outlook for the Atlantic ACE index of hurricane activity. The expected level (mean) of shut-in oil production during 2006 is 11 million bbl or 2.1% of forecast annual production, and the median level (50th percentile) is 8.7 million bbl. The distribution is skewed to the right, with most simulated shut-in values being relatively low and only a small proportion of large shut-in amounts.

The y-axis intercept of the curve indicates that there is a 15% probability of no oil production being shut in during 2006. Although there is a 90% probability of less than 25 million bbl being shut in (i.e., a 10% probability of more than 25 million bbl being shut in)-similar to the impact from Hurricane Ivan in 2004-the likelihood that more than 50 million bbl will be shut in is less than 1%. In fact, only 3 of the 10,000 simulated values matched or exceeded the cumulative levels of shut-in production greater than 100 million bbl that was seen after Katrina and Rita last year.

Results for lost gas production are similar (Fig. 3b). The expected level of shut-in gulf production during 2006 is 64.3 bcf, or 1.96% of annual forecast production, and the median level is 50.5 bcf-1.5% of annual forecast. The likelihood of zero shut-in production is 14%, and there is a 10% probability of more than 145 bcf of shut-in production.

None of the simulated shut-in gas values approached the 2005 levels.

Forecast caveats

The sizable range in possible shut-in gulf production for both oil and gas highlights the extreme uncertainty implicit in attempting to project possible impacts from hurricanes and tropical storms so early in the season, yet EIA can conclude that the likelihood of witnessing a repeat of 2005 is relatively remote, given NOAA’s current hurricane outlook. The shut-in production values simulated in this report are conditional on the NOAA forecast for Atlantic ACE. However, the May Hurricane Outlook indicates that there is a 20% probability that hurricane activity in the Atlantic Basin could be near-normal (ACE 66-103) or even below-normal.

Furthermore, NOAA emphasizes that its May outlook is based on climatological conditions that are still evolving. An updated hurricane outlook will be issued this month when conditions favorable for hurricanes are more predictable. It is quite possible that NOAA could substantially revise its projections for seasonal hurricane activity, as it did in 2005 when the May outlook that projected a seasonal Atlantic ACE range slightly lower than the current 2006 projection was revised substantially upward in August prior to the formation of Hurricane Katrina. If a similar situation occurs this month, estimates of shut-in oil and gas production could be significantly higher.

And, of course, the gulf’s producing region is particularly vulnerable to hurricane-related production disruptions. The level of shut-in production depends on a variety of factors, including the intensity and location of storms and the ability of platform operators to respond to the threat of damaging hurricanes.

Oil and gas producers along the OCS have responded to this vulnerability in numerous ways, including stocking up on replacement parts to speed recovery, upgrading communication systems, and developing stronger, more-stable platform infrastructure. As technology improves over time, the threat of shut-in production likely also will be reduced.

Acknowledgment

The authors thank Less Goudarzi, Mark Hutson, and Dave Costello for their help in developing this article.

References

  1. Energy Information Administration (EIA), “The Impact of Tropical Cyclones on Gulf of Mexico Crude Oil and Natural Gas Production,” May 2006.
  2. EIA, Petroleum Supply Annual 2004, Vol. 2.
  3. EIA, Natural Gas Monthly, May 2006.
  4. Minerals Management Service (MMS), Hurricanes Katrina-Rita evacuation and production shut-in statistics reports, various dates, 2005.
  5. MMS, “MMS Updates Hurricanes Katrina and Rita Damage,” MMS press release, May 1, 2006.
  6. US Department of Energy, Hurricane situation reports, various dates, 2005.
  7. Kaiser, M.J., Pulsipher, A.G., “Empirical models help gauge production lost due to storms,” OGJ, May 8, 2006, p. 36.
  8. Pielke, R.A. Jr., Landsea, C.W., “Normalized Atlantic Hurricane Damage 1925-1995,” Weather and Forecasting, Vol. 13, pp. 621-31, 1998.
  9. EIA, “Short Term Energy Outlook,” June 2006.

The authors

Tyler Hodge ([email protected]) is an industry economist with the Energy Information Administration. He is a PhD candidate at the Colorado School of Mines and has a BS and MS from the University of Colorado. Hodge is a member of the International Association for Energy Economics.

Tancred Lidderdale is a senior economist with the Energy Information Administration. He has been with EIA for 15 years and previously worked for 13 years in the petrochemical and refining industries. He has a PhD in economics from George Mason University and a BS in chemical engineering from Georgia Tech.