ERROR TYPES
All measurements we make (except for counting discreet units) are analogue at some point, even if the final output appears digitally. All analogue measurements have errors. More specifically, all analogue measurements contain random and systematic errors. Some may have gross errors, also known as blunders.
To look at examples of each type of error, let us look at conventional (total station) surveying, which once dominated land seismic surveying. It is a good system to use as an example, because with these methods, almost all errors were field observational errors.
- An example of random errors is how a human observes the same target repeatedly but sets the cross-hairs slightly differently each time due to the resolution of the instrument and variations in their perception.
- A systematic error might be the collimation error in a theodolite where the vertical angle has a constant bias on one face due to its state of calibration.
- A gross error might be turning an angle from a back-sight to a fore-sight but sighting a wrong back-sight target due to obstruction in the line of sight. Logging errors, i.e., writing down the wrong measurement, once very common, could also be considered a gross error.
While conventional surveying is still used in jungles, most modern seismic positioning utilises GPS (now GNSS) measurements in one form or another. This removes many of the traditional sources of random and systematic errors, but not all of them.
Whereas most errors used to be observational, now many come from, recording, reporting, processing and office functions. Examples of processing errors are filtering errors, smoothing errors, interpolation errors and logic errors. Other office errors include reporting errors, quality control errors, editing errors and formatting errors.
Processing errors are most often found in processing software of Integrated Navigation Systems (INS), and predominantly with airgun source positioning, but also with acoustic processing. The next four examples below relate to INS and are errors that we continue to find today.
- Filtering errors: These systems read and process thousands of measurements from various sensors. Many of these sensors are subject to occasional spurious measurements or the cables and connectors that transmit the measurements might have intermittent faults. The processing software must decide the difference between a good measurement and a bad one, and they don’t always get it right.
- Smoothing errors: They also generally use smoothing, such as a Kalman filter to blend out spurious measurements. Once again, they must decide what is real and what is not. They sometimes smooth too much and remove spikes that really should stay there.
- Logic errors: Sometimes they have logic problems such as wrongly calculating a time of arrival at a shot point, for example on an offset on an angled detour. These problems arise when the software designer does not anticipate a certain combination of circumstances and the software therefore doesn’t account for the situation correctly.
- Interpolation errors: In many situations in source processing, regularly timed measurements have to be interpolated into events at discrete times, or with acoustic positioning, intermediate receiver positions have to be interpolated between those with pingers. This can be done with or without a bias applied from some other factor. Sometimes the processor can have these settings wrong, and the effect can be significant.
Other common office errors are described below:
- Reporting errors occur when the events in the field are not correctly conveyed to the processors. For example, the drillers might have to offset several shot points and then forget to pass on the information to the QCs to request a resurvey.
- Quality control errors can occur when quality indicators generated by the measurement system are not passed on through intermediate processing or they are not viewed and acted upon. They can also occur when QC procedures are not designed or followed correctly.
- Editing errors are very common in complicated surveys, especially transition zone or urban surveys, where different types of measurements are collated into aggregate datasets and repositioning and point moves have to be edited in.
- Formatting errors are mostly a problem in final data where header information is incorrect, data are omitted from columns, data are misaligned, or files don’t cross reference properly.
All these errors occur in modern seismic surveys. Some are detected. Too many are not because of either a lack of a dedicated QC professional, or a lack of the right software tools.
Please click on Error Causes to read on.