Operators of oil and gas pipelines face a big challenge in meeting federal and state safety regulations. A new subscription service built on DigitalGlobe’s geospatial big data platform, GBDX, enables operators to identify areas of concern, receive frequent updates, validate the data, and prioritize their risk mitigation activities.Pipeline operators are required by federal and state regulations to classify each segment of pipeline by the risk it poses to nearby population. Often, operators have only a very crude estimate of this risk. What’s more, there’s the greater challenge of monitoring changes on and around each segment. How can an operator monitor, say, 50,000 miles of pipe on a regular basis?
In some cases, operators employ aerial photography, visualizing about a third of their assets each year. If they cannot afford that, they rely on Google Maps or buy piecemeal satellite imagery. However, even if they could afford to buy satellite imagery around every mile of pipe once every year, that would still leave them blind to day-to-day changes and activities present in every new satellite image.
In the United States, there are 2,200,000 miles of oil pipelines and 300,000 miles of gas pipelines, operated by hundreds of companies. [Figure 1] Pipelines typically traverse varied environments, mostly on land not owned by their operators. In some areas they are very isolated, while in other areas, such as the Gulf region, they cross very dense population networks. Some pipelines are used for transmission, such as shipping oil from wells to refineries, while others are used for distribution, such as providing natural gas to individual homes.
Federal law requires pipeline operators to
- categorize each segment of pipeline depending on its proximity to population
- assess the risk of impact on that population in case of leak or explosion, and
- perform certain types of maintenance and inspection activities depending on that risk.
The higher the potential impact of a disaster on nearby population, the more frequent and stringent the requirements for inspections, maintenance, and upgrades. For example, a segment that runs through Dallas, within hundreds of feet of homes, offices, and schools, requires much more monitoring than one along the Rockies, hundreds of miles from the nearest town. High consequence areas (HCAs) include those locations where commercial business occurs (such as shopping malls), where people with limited mobility gather (such as schools, prisons, and hospitals), and where people gather outdoors (such as sports stadiums).
In order to ensure the safe operation of pipelines, the Office of Pipeline Safety within the Pipeline and Hazardous Materials Safety Administration of the U.S. Department of Transportation establishes the criteria for safe pipeline operation. It uses manufacturer’s specifications to regulate operating conditions and establishes repeatable and verifiable procedures, assisted in implementation and compliance by state regulatory agencies. The potential impact radius (PIR) of a release due to a pipeline breach, which depends on the pipe’s diameter and the pressure of the product flowing through it, determines the width of the required buffer around it. For instance, if the edge of a building is detected inside that buffer, that area is of higher consequence than one that does not include any buildings.
If pipeline operators could receive regular updates, what would they want to know? All of the following would be relevant:
- new buildings
- new roads
- changes in soil or land use
- any other indications of widespread human activity
A lone hiker passing through is not a concern, but a new campground in close proximity to a pipeline is. Permanent changes, such as new construction, pose an even greater concern. Perhaps what was once a farm field is now a school or a residential neighborhood. New road construction can serve as an early warning of such developments.
Most pipeline operators are not equipped to conduct this kind of monitoring independently. The sheer magnitude of potentially relevant data vastly exceeds the analytical capabilities of most operators’ monitoring teams and IT systems. For example, they would have to sift through a large percentage of the 3.3 billion OpenStreetMap (OSM) features and 2.6 trillion tweets posted worldwide every year to identify the tiny percentage directly relevant to their operations, and then integrate that information into their databases. Even if operators could set up such a system, they would also want to analyze satellite data as frequently as it was available, in order to identify new buildings and roads. DigitalGlobe alone collects 70 terabytes of satellite data every day to add to its 90 petabyte archive. A collection of that magnitude would be a huge technical undertaking for any one organization to do this on its own, let alone for every operator to do it. Pipeline operators need more than the occasional aerial or satellite snapshot. They need a monitoring service that combines data from Earth observation satellites, open geo data vector sources, and social media to alert them promptly to any relevant activities or changes near their pipes. Additionally, they need to be able to analyze this data and extract the features, measurements, and statistics required to plan and act to prevent accidents.
DigitalGlobe recently released a cloud-based subscription service built on our geospatial big data platform, GBDX, for the oil and gas industry. It
- provides frequent updates
- alerts operators to changes [Figure 2] and
- enables operators to analyze raster data (such as aerial and satellite imagery) and vector data (such as roads and borders) at scale and on demand
Because the service focuses on identifying changes as well as present and past activities, an organization can use it to forecast changes based on current trends, or employ it after a failure to analyze what caused it.
Because pipes are often buried, thus not visible from the ground or the air, operators need to define their areas of interest, typically by uploading a GIS definition such as a shapefile. The monitoring service does the rest, autonomously and automatically.
Using either an interactive Web tool called AnswerFactory or a plugin for Esri ArcMap, customers can define the geometry of their pipeline, their rights of way, the buffers they need to monitor, and other variables of interest to them, depending on their business requirements—such as building footprints, roads, human activity, or land use. They can run their selections daily, weekly, monthly or whenever new imagery or vector data is available. They can then easily see the location, direction, and magnitude of changes that concern them and sort them by relevance. For example, a new school would be of much greater concern to them than a new barn.
How It Works
Our fleet of satellites rains pixels down from the sky every day. With the GBDX platform, those pixels are available in the cloud, where they can be analyzed using machine learning, including convolutional neural networks and computer vision techniques, and DigitalGlobe Crowdsourcing analysis, by distributing the data to tens of thousands of people who validate the machine’s work. For example, the crowd might confirm “Yes, that is a road. No, that is not a building.”
To make it possible to deliver the answers directly into the Esri tools most familiar to many pipeline operators, GBDX includes a plugin for Esri ArcMap. However, operators can also use Answer Factory.
In one scenario, GBDX monitors OSM daily for vector updates, then checks whether those updates are within a buffer area along the pipeline. However, OSM updates, which are contributed by volunteers, often miss infrequently travelled areas, such as new subdivisions under construction on former farmland. Therefore, to provide some “ground truth” to the OSM updates, GBDX can also compare them with road features extracted from DigitalGlobe’s most recent imagery. [Figure 3]
Pipeline operators already have algorithms to calculate the borders of their HCAs. However, to meet regulatory requirements they have to manually trace the footprint of each building within their buffer for the entire length of each pipeline. Then, they still only know what was there when they traced the footprint. If, for example, Don’s Place Bar, which is inside a buffer, decided to build a sand volleyball court on its lot, the pipeline operator would not know about it until someone manually examined the next satellite image of the area, months or years later. [Figures 4 and 5]
It would be prohibitively expensive for pipeline operators to manually monitor such changes in land use in near real time. However, when an issue arises, they may spend tens of millions of dollars to address it. By analyzing georeferenced open source data from social media overlaid on satellite imagery, DigitalGlobe’s automated algorithms can autonomously detect changes in land use and alert pipeline operators. They can then target mitigation and remediation efforts toward only the affected areas, thereby also reducing future risks and costs. This system can easily improve operators’ change detection by an order of magnitude. In turn, this increases by an equal amount their ability to avoid issues and disasters by focusing their efforts around change detection.
A single query can return all the coverage from the DigitalGlobe catalog for the past 12 months of, say, the electrical distribution lines in Richland County, Wisconsin; [Figure 6] or the 1,500 Bakken wells in North Dakota. The query returns a table that shows how many satellite images are available for each month.
Next, pipeline operators need to analyze the imagery. Traditionally, this would require hundreds of staff hours just to look at the images. DigitalGlobe replaces that with automated techniques that
- extract features such as roads and buildings
- integrate open source vector and social data
- create change indices to help operators understand what activities are taking place near each segment of their pipelines.
For example, the system might generate a report listing the top 20 places with the highest percentage of change in their entire pipeline. [Figure 7] The operators can then focus their resources on those places. If they access the service via ArcMap, they can explore the data in even greater detail.
In addition to machine learning, crowd sourcing, and the synergy between the two, there is also a key role for experts to do quality control and notice things that neither machines nor crowds can. For example, once the system identifies areas in which a high amount of change has occurred, it is the job of experts to analyze the imagery and determine the exact nature of that change. Every morning, an expert could look at a list of the areas where the most change has occurred and select each one to display both the pixels and the vectors for that area. Instead of having to comb through millions of features for her entire pipeline, she could focus on just the ones that meet a certain threshold of change that she has selected. She could also compare two images for the same geographic extent taken weeks or months apart, and review the machine’s change report. She might then tag one or more of these areas and send a field service team to inspect them and to take action as needed.
This service allows operators to monitor pipelines at an unprecedented scale. It is also very useful for any private company or public agency that manages other linear features that can be represented as vector lines—such as golf course trails, electric power lines, or logging roads.
Later offerings might also detect slope changes due to subsidence. These are critical when pipelines go through or along hills because such movements could snap the pipe. Currently, pipeline operators have little practical ability to detect changes in land use from new satellite images or to identify and count, let alone analyze, social media posts or new OSM features around their pipelines. By mining its petabytes of data and employing sophisticated algorithms and crowd sourcing, DigitalGlobe is enabling operators to monitor changes that can affect them at a scale that has never been possible before.