Electromagnetics

Long Term Seismic Monitoring

Introduction

In the mid-1960s seismic networks made for micro-earthquake recording began widespread operation. The distances between stations were a few kilometers to a couple of hundred kilometers and the signals were transmitted a central recording station where all data was recorded with central timing. The common time base enabled very accurate relative timing between stations and therefore allowed accurate locations of local earthquakes. Recording was initially analog and has now evolved to be almost all digital. Better communication has allowed these networks to have a global reach.

In addition to the general description of local or global networks, this discussion will also outline the basic steps in design and implementation of a new seismic network.

Seismic Network Purpose

General seismicity assists the long-term mitigation of seismic risk as well as providing resolution of seismo-tectonic features. Seismic hazard maps of the region are produced and the development and implementation of proper building codes is placed on a scientific foundation. Seismic monitoring related to seismic risk caused by human activity is of special political concern.

Induced seismicity in the vicinity of large dams or around large mines is often the object of network investigation. Monitoring of seismicity in a volcanic region to assist in prediction of eruptions and the monitoring of underground nuclear explosions are seismic network functions of substantial governmental attention.

Local, regional and global research into the earth's interior is a primary goal of seismology. Seismic networks are the only tool enabling the study of the detailed structure and physical properties of deep portions of our planet. Closer to the surface, the tectonic state of visible geologic structures remains a vibrant research topic.

The purpose of a new seismic network defines the design considerations and the technical solutions. If a variety of goals is postulated, the design may be unsatisfactory for some of the objectives. However, modern networks can deal with several goals better than older networks, which were restricted by technical limitations.

Within this discussion 'global' will refer to data, events, or investigations directed at studies of the earth that are regional or planetwide. An example would be studies of the subduction zone earthquakes that are clearly below the Mohorovich discontinuity. The word 'local' will refer to subregional or structure-specific studies which generally record local microearthquakes (unfelt events) in support of the understanding of tectonic-geologic processes. An example would be a network operated to monitor reservoir induced seismicity and separate it from unrelated tectonic seismicity.

Seismic Sensors

The number of components per seismic station is another consideration. The cost penalty for multiple components is so low that three components are the rule rather than the exception. In fact, to balance the strong-motion requirement with the local network requirements a tack is to have one six component station (two types of sensors) with the rest having three orthogonal sensors each.

The sensitivity and dynamic range of the sensor must be established. Again, the purpose of the network has maximum impact on these parameters. Recording to capture the ground-motion of a major earthquake will differ from recording to separate unfelt tectonic events from manmade seismic events. Modern instruments usually have more dynamic range in the A/D conversion and communication systems than can be utilized except at the quietest sites.

The sensor's frequency range is another consideration.The following table illustrates typical seismological applications and their approximate frequency range of interest. Application Frequency range (in Hz) Period range (in seconds) Seismic events associated with mining process 5 - 2000 0.2-0.0005 Very local and small earthquakes, reservoir induced seismicity 1 - 100 1-0.01 Local seismology 0.2 - 80 5-0.0125 Strong motion applications 0.0 - 100 ??-0.01 General regional seismology 0.05 - 20 20-0.05 Frequency dependence of seismic wave absorption 0.02 - 30 50-0.033 Energy calculations of distant earthquakes 0.01 - 10 100-0.1 Scattering and diffraction of seismic waves on core boundary 0.02 - 2 50-0.5 Studies of dynamic process in earthquake foci 0.005 - 100 200-.01 Studies of crustal properties 0.02 - 1 50-1 Dispersion of surface waves 0.003 - 0.2 300-5 Free oscillations of the Earth, silent earthquakes 0.0005 - 0.01 2000- 100.

A conclusion might be that local networks can use a sensor that is flat above the microseism belt at 3-5 seconds (and up to 1-200 Hz) but that global studies require a broader band sensor with good fidelity from roughly 50 seconds to 50 Hz. Previously proprietary networks consisted of a set of stations that were hardwired and softwired together. What is meant by softwired is that the data were used for the purposes of the network owner and seldom, due to formatting or proprietary considerations, were shared or even offered to others. As the internet has developed as the premier communication tool for seismic networks and as formats and protocols have converged into only a few 'standards' the sharing and common use of almost all seismic data is the norm.

Seismic Data Acquisition

Buyers of digital seismic networks sometimes ask for additional paper drum recorders in order to continuously monitor incoming signals or in order to use the recorders as an educational tool. One must weigh the problems with paper drum recorders in digital systems. "Hardware wise" they are incompatible with digital systems and require additional digital to analog converters. Being mechanical devices, they are and will continue to be expensive and require specialized maintenance and consumables. On the other hand, nearly all modern observatory seismic software packages allow continuous observation of the incoming signals in (near) real time and can even simulate the appearance of paper seismograms. Once the user becomes familiar with the digital system paper drum recorders soon prove to be of little use and prove to be a purely political investment. There is not much benefit in purchasing drum recorders with digital systems. Triggered seismic networks.

Continuous digitally acquired seismic signals by their very nature yield a huge amount of data. A small local digital seismic network operating in continuous mode will produce a huge volume of data. Yet, only a small portion of that data is in fact useful information.

The storage and analysis problem has frequently led seismic network users to operate their systems on a "triggered" basis. Triggered systems still do continuous, real-time acquisition and processing of seismic signals, but for trigger purpose only. They analyze (or bring to the attention of the operator) signals only if the system's trigger algorithm recognizes a seismic event. A decision between continuous or triggered operation usually means a decision between higher network event detectability and reduced detectability. The difference is significant. And it easily becomes very significant if man-made seismic noise at the remote station sites is high or if the trigger parameters are not adjusted optimally. In modern high-capacity systems, the decision is less important since these systems often provide large temporary storage.

Note that the continuous seismic signal recording provides the most complete data, but storing and processing of all that data can become expensive. Systems in triggered mode will lose some weak events and produce a certain number of false triggers. The completeness of the data set is impaired because the efficiency of trigger algorithms is inferior to pattern recognition by a trained seismologist.

Trigger Algorithm Types

Every triggered seismic system must have an adjustable band-pass filter in front of the trigger algorithm. The adjustable pass-band filter allows the trigger algorithm to be sensitive to the frequency band of interest.

In seismic networks with standalone stations, each remote station has its own independent trigger. In such networks data is usually transferred to the central recording site on request-only or it is collected in person. These seismic networks have the lowest effectiveness of triggering and consequently often become very insensitive. This type of triggering functions well where high sensitivity is not desired - for example in strong motion networks.

Seismic networks which use a coincidence trigger algorithm offer much better detectability and completeness of acquired data. In these systems, all the data is transmitted from the remote stations to the central recording site where a trigger algorithm discriminates between seismic events and seismic noise. The coincidence algorithm considers not only signal amplitudes but also signal correlation - the number of active stations within a given time window. Fewer false triggers results. Signals of all stations in a network are recorded for every trigger, which greatly improves completeness of the recorded data.

An even better solution is provided by systems which temporally store continuous signals in memory for a given period of time ranging from several hours to several days. After the specified time, these systems remove the old data and replace it with the new incoming data. While this method requires prompt analysis of seismic signals, excellent completeness of data and detectability is obtained. In addition during aftershock sequences or earthquake swarms, the data can be stored in a continuous manner.

Such systems can still have an automatic trigger algorithm operating simultaneously, which enables automatic processing.

Seismic Data Transmission

Key technical parameters for the data-transmission links are:

The required channel bandwidth.

The distances over which data must be transmitted (becomes unimportant with internet based seismic networks).

The desired reliability

The physical network used, i.e. the access point to the Internet.

A remote seismic station, which can be reached by the Internet or a dial up phone line presents the most general purpose and flexible system available. Getting seismic data from a station using the Internet via a local computer is simple. The user uses the phone net to login to the station. Once logged in, he can check available seismic data and use a FTP file transfer protocol to copy the data to the local computer. The process is easy to automate.

All digital data acquisition and transmission systems exert time delays. The size of the time delay depends on the digitizer, the digital protocol used for transmission and the computer at the central station. For this reason, most digital field stations time stamp the data at the remote station. GPS clock prices are now a small fraction of total digitizer costs.

The seismic enclosure at seismic stations assure a good mechanical contact between seismic sensors and non-weathered, solid bedrock and protect equipment from temperature and humidity impact, dust and dirt, lightning, vandalism, and from small animals. The shelter should also provide a good, low resistance electric ground for sensitive electronic equipment and lightning protection system grounding, and easy and safe access to maintain and service the equipment.

Two vital issues, leading to potentially fatal consequences if neglected, are lightning protection and grounding system.

Note that lightning is the most frequent cause of seismic equipment failures. The lightning threat varies dramatically with station latitude, topography, and local climate and an investment in the best protection available is money well spent. A good, low impedance grounding system keeps instrument noise low and allows proper grounding and shielding of equipment and cables. In some areas a light fence around the station will minimize man and animal-made seismic noise and help protect stations against vandalism.

Inadequate site selection and poor seismometer placement can easily negate the benefits of expensive, high-sensitivity, high-dynamic-range seismic equipment. For example, a station which sits on unconsolidated alluvial deposits instead of bedrock can, due to thermal and wind effects, make broadband recording useless. Money invested in expensive seismic equipment can have its benefits wasted because of improper site conditions.

Shallow boreholes with a depth from a few to 15 m are sometimes used instead of surface vaults. A 15 m deep station in a difficult terrain may cost more than a shallow borehole of the same depth. Seismic noise improvement in such shallow boreholes is small. In terms of network cost, it might be cheaper to increase seismic station density rather than install fewer stations with expensive boreholes.

How To Get Started

The range of interest in ranked by distance: local seismology (epicentral distances < 150 km), regional seismology (epicentral distances between 150 and 2.000 km), and/or global seismology.
The main purpose of setting up a network: usually either to monitor a region's general seismicity or to perform special studies of specific seismotectonic features, important civil engineering structures, of civil and/or nuclear explosions or man-induced seismicity.
Relative importance of the project's functions: the network's alarm function for civil defense purposes, the seismological research aimed at the mitigation of seismic risk, the separation of tectonic from man-made seismicity, or research into the Earth's deep interior. Questions to keep in mind are:
- How big is the region to be monitored?
- What is the level of seismicity in the region?
- What are the institional resources available to administer seismological problems?
- What is available in terms of communication infrastructure?
- What resources are available to operate and maintain the system, and to support research work using the system's data?
- How much money is available to install the system?

There are always trade-offs between desires and reality.

Often a new seismic network is established without the knowledge about how to allocate the finances such that a seismic network has optimal scientific and political effectiveness. Too often a majority of the network funds are spent on buying equipment (boxes), even though the conditions for proper operation of this complex equipment are equally important. A proper budget must include money for:

A feasibility study that examines potential layouts, sites, and seismic systems.
Preparation of remote stations and housing for the central recording site.
Purchase of the network equipment.
Manufacturer's services such as installation, training, and repair.
Scientific and technical personnel and training.
Network operation costs (personnel, travel, data-transmission costs, processing, consumables, and spare parts).
Reporting on results, which requires personnel, processing software, computers, printing and backup storage.
Network servicing and maintenance cost.

Once the goals are clear and the funds are properly allocated, one has to clarify the entire project's interrelated seismological and technological aspects. One should pay attention to:

The size and the layout of the proposed seismic network.
The seismicity level to be monitored - in other words, the amount of data one will record/deal with.
How accurate the network's event determination should be.
How wide a dynamic range and resolution are desired for the data acquired from the network.
The importance of the new system having alarm capabilities for civil defense purposes and the desired response time.
The amount of technical reliability one expects from the system.
The desired robustness of the system in terms of functioning through a damaging earthquake.

The combination of the uniqueness of each situation and the consideration of the above questions will lead to a coarse system-engineering approach. Eventually the stakeholder will have to agree to certain tradeoffs. The number of stations in the network, the capabilities of any given network, and the cost will interact.

With respect to the number of stations for determination of a location the theoretical minimum is three stations which record both P-wave and S-wave arrivals. However, due to their uncertainty, such results usually have little value. Six stations acquiring records of an event provide scientifically credible evidence of an event's location and ten or more stations acquiring good quality records of an event should provide an acceptable basis for more sophisticated studies of an earthquake. A waveform analysis of digital, high-dynamic range, three-component records leads to good results with fewer stations. In principle, one three-component station can determine the magnitude, epicenter and the origin time, however this requires a very well known model of the ground.

The spatial distribution of the stations in a seismic network is very important for the network's capabilities of event location. The design of a network is most often done by pitching pennies on a map. After a geometrically uniform pattern is morphed into a logistically feasible network, the purpose and financial considerations are reviewed. If depth is a primary objective, there is a rule of thumb for a reliable depth determination. First, the depth range of concern in a region is estimated. The average distance between stations in the seismic network should be less than the midpoint of that range. This requirement is a tough requirement, especially in the large regions and in the regions where the events are typically shallow.

Realistic expectations about the system's earthquake epicenter determinations is required. For events outside the seismic network, large errors in determining epicenters will be present. Generally, reliable determination of event locations occurs if the "seismic gap" (the largest of all angles among the lines connecting a potential epicenter with all the stations in the network, which recorded the event) is less than 200 degrees. To increase the accuracy of epicenter determinations, especially for the events outside a seismic network, one needs to include data in the analysis from seismic stations in neighboring regions. Acquiring this wider database is usually necessary for determining reliable event locations outside a seismic network.

The matter of seismic site selection is often not given recognition in spite of the fact that a local seismic network can only have good detectability if the sites have low noise levels. If seismic noise at the sites is high the benefits of a modern, high dynamic range equipment are lost. If an excessive bursts, spikes or other man-made seismic noise is present, high trigger thresholds and poor network detectability will result. If stations are situated on soft ground, the noise levels will be up to 12 db higher that those stations with seismometers firmly attached to bedrock. If the network geometry is inappropriate, the location of events will be inaccurate. A systematic site selection procedure is therefore essential for success of any local seismic network.

Only the basic steps of the site selection procedure will be presented here. Map studies are relatively inexpensive and are therefore the first to be performed. The first step is defining the geographical region of interest. The next step is to examine existing geological faults, seismotectonic features, and other available information about seismicity in the area. Then prepare a simplified map of regional geological conditions showing the quality of bedrock. The rule is: the higher the acoustic impedance of the bedrock, the higher the maximum possible gain of a seismic station. Next, study the topographical aspects of the possible locations. Moderately changing topography is desired. To study man-made and natural noise sources in the region, one should evaluate road traffic, railway traffic, heavy industry, mining and quarry activities, agricultural development of the region, topographic barriers to wind or weather systems, along with the natural sources like oceans and lakes, rivers, and waterfalls.

Topographic profiling of RF paths based on topographical maps is performed for radio linked networks. The availability of phone lines for data transmission, if postulated, has to be checked. The availability of power lines or the decision for solar panels must be made. It is also important to research land ownership and land use plans for the sites. Property ownership or future development can a site unsuitable for seismic stations. The climate at the sites also influences site selection: temperatures, wind, precipitation, insolation data, lightning threat should be considered.

The field studies are the next step in the site selection process. Expect to make several visits to each site. A seismologist familiar with seismic noise measurements, a seismo-geologist, and a communication expert should all visit the sites. The ease of seasonal access to the site, local man-made seismic noise sources are evaluated. The site group can perform seismic noise measurements, study the local seismo-geological conditions at the site, investigate the local RF data transmission conditions and verify power and phone line availability(if applicable) To the extent that is possible, uniform local geology is preferred for seismic stations. An ideal approach is to make shallow seismic profiling of the sites.

After all these studies one ends up with two or three potential sets of the best suitable seismic stations. The resulting network layouts are then studied for the best network performance by computer modeling. By comparing the results, one will be able to make an informed decision about the final seismic network layout. The cost of a professional RF survey represents generally a few percent of the total investment. We believe that the combined benefits of an RF survey are well worth the investment. It is a major step toward the reliable operation of the seismic network.

Request For Proposal

In the request for proposal one should include the basic technical information, so that the manufacturer can put together a technical solution. However, the request for proposal should not contain an over-detailed technical description of the desired system as new approaches are constantly coming on the market. With too many technical details one can end up limiting one's choices. It is recommended not to push manufacturers to design a new system or new functionality in an existing system specifically for your needs. In spite of the fact that the seismic equipment manufacturers are willing to design custom made systems, one should know that there is usually a high price for this commodity, not only in terms of dollars but in terms of performance.

Avoid buying brand new systems in the market unless you are really assured of excellent support from the manufacturer. Simply, try not to be a 'guinea pig'. Any brand new system obviously has more technical imperfections and 'kid's illnesses' than more tested systems. This requires a higher level of knowledge and a really good co-working relationship with the manufacturer while solving these problems.

Data sheets of seismic equipment alone seldom provide enough information. Also, it is not easy to compare the data sheets of various manufacturers because each one will use a different system of specifications, measurement units, and definitions of technical parameters.

Be cautious about mixing products from different manufacturers in one system. It is not simple to interface between different products in terms of dynamic range, signal to noise ratio, baud rates, the processing capabilities and the power supply wattages. Use one manufacturer if possible, or, when that is not feasible, arrange to have one manufacturer who will be responsible for interface problems and the functioning of the system as a whole.

Each technical system, or element in it, properly operates within a set "range". One should be familiar with these parameter and know where, within the range, the system will operate successfully. If the system is to operate at the extreme end of the operational range on a regular basis, the system or element is most probably not the right choice.

The results will often be disappointing if, for example, one plans on using the maximum possible number of channels in a FM radio-frequency link. In such case it is often better to find another system or component, whose midrange operation can accommodate one's needs. A safety margin is required and do not expect a system to operate continuously, efficiently, and reliably in the extreme ranges.

The global market for seismic systems is quite limited. With a few exceptions, instruments are produced in small numbers. Inevitably, there is a limit to the quantity and thoroughness of testing of new equipment. In general, the equipment arrives with a higher than average number of bugs and technical imperfections.

The majority of seismic network suppliers have relatively little seismological experience and as a general rule, do not sell adequate software. On the other hand, there are public domain software packages available, which can solve the tasks and these are often offered by the manufactures. However, little training is offered and a new network often has expensive equipment and very primitive processing tools. thus the processing software and training are an important parts of the planning of a new network.

Currently, most seismic equipment is not as user friendly as most of us would like and the technical documentation frequently falls short. Customers are rarely given comprehensive and easy to follow instructions on how to setup and use the system. The reputable manufacturers of seismic equipment compensate for this situation with effective customer support.

There are four ways to install a new seismic system:

The user installs the new system. Only instruments are purchased. In this option, the customer is responsible for the operation of the system as a whole and the manufacturer supplies functioning elements. This approach gives the user great flexibility and responsibility.

The manufacturer demonstrates installation on a subsystem (a few stations, a sub-network). The user installs the rest. In this case, there is a sharing of the responsibility for the system's operation. This approach is often successful. However, the customer is required to have experience with seismic, computer, and communication equipment for this method to work.

The manufacturer installs the whole system with assistance from the local staff that will be responsible for running, maintaining, and servicing the network. Responsibility for making sure the system is operational lies with the manufacturer. The main benefit of this approach is that the users learn an enormous amount during the hands-on installation and associated problem solving time.

A customer orders a turnkey installation without user participation. The manufacturer has the complete responsibility for seeing that the system operates properly. In this case, the network will be successfully installed, but local staff members will learn nothing about solving potential future problems.

Two technical details relating to system installation are mentioned here. In case the system buyer will install the system or its parts, do not accept the 'standard length' cables offered sometimes by some seismic system manufacturers. The 'standard' cables rarely work well in the field. Note that badly installed connectors are among the most frequent failure causes.

Note also that, in case of purchased installation, the seismic station sites must be completely prepared before the manufacturer arrives to install the system. Failure in this area generally leads to less reliable functioning of the system due to compromises made in the field. Note that engineering services are based on time spent , and that an efficient use of this time is customer's direct benefit.

Starting Up

The region's geology, the seismic network and layout, the seismicity of the region, the amount of industrial blasting, seismic noise levels, seismic signal attenuation in the region and local earth structure at each site, are components of these adjustments. One will not be able to correctly tune the system's recording and processing parameters until the network has acquired sufficient events and operational time with natural noise, man-made seismic noise, and the earthquake signals at the sites in the network and until the parameters which have to be tuned are understood. Therefore, tuning a network often takes months of systematic work. Because of this time requirement for this task, the system's manufacturer can not do it. Only the network operator can correctly tune the network and respond to changes in the goals of the network. Due to the addition of new stations, re-tuning of the network will probably be required from time to time. In reality, tuning a seismic network is an ongoing task, which cannot be done 'once and for all'. The most common hardware and real-time processing parameters that need to be adjusted are:

Seismic gain at individual stations.

Signal conditioning filter parameters.

Pre-trigger band-pass filter parameters.

Trigger algorithm parameters, which usually include:

Trigger and release threshold values.

Trigger windows' duration.

Individual station weights in a coincident trigger algorithm.

Sub region definition for coincidence trigger algorithm.

Pre and post-event time duration.

Minimum and maximum runtime duration.

Propagation window length.

Note that not all these parameters exist in every seismic network but that some adjustments may be missing from this list.

The following are some of the offline seismic analysis software issues which have to be adapted and organized for efficient routine observatory work:

Files containing information about data acquisition parameters.

Files containing data about geometrical configuration of seismic stations and station parameters as a function of time.

Sensor calibration data.

Earth model parameters.

Automatic phase picker parameters.

Magnitude scaling parameters.

Routine, everyday housekeeping programs for analysis of seismic signals. Keeping a network failure free and in perfect working order, waiting to record earthquakes year after year requires hard and responsible work and a lot of discipline. This goal is generally not simple to achieve. Seismic observatory staff will have to operate in a highly professional manner and accomplish the following:

Clearly defined tasks associated with the routine operation of the network and everyday analysis and archiving activities,

Regular maintenance of hardware and software,

Continuous verification of all tasks and hardware operation,

Precise record keeping of all activities, changes in network operational parameters,

Maintenance works and repair activities, and in the seismic data archives, and

Well-defined personal responsibility with respect to network alteration and strict obedience to the established procedures.

Regular processing of seismic data requires that the details of how data is processed and stored are well planned.

Generally, network recording parameters should be changed only if there is an important and well thought out reason for the change. A change of recording parameters changes the network's detection threshold. Thus changes are avoided as much as possible. Some changes are inevitable from time to time but the changes should be kept within reasonable limits carefully documented. This careful and continuous documentation of network operation parameters in a logbook, log file, or in the seismic database itself, is essential. This documentation should contain all information about data acquisition parameters and their changes, all station calibrations, a precise track of stations' downtime, a descriptions of technical problems and solutions, and descriptions of maintenance and service work. The exact times of changes must be properly. The information is an integral part of seismic data archive because only the signals and the precise conditions at the time of the recording, can be properly interpreted.

Maintaining a seismic network is a continuous activity that inevitably requires well-trained personnel. Nowadays, the state of remote seismic stations are monitored by modern, high-end seismic systems with duplex data transmission links. Critical parameters include: backup battery voltage, charging voltages, software or communication problems, time keeping, equipment's health, potential water intrusion, etc. These utilities significantly reduce the need for field service work and lower the cost of network operation. However, regular visits to the stations are still necessary, though less frequently than in the past. Once per year seems a minimum.

Note that it is a mistake to put off visits of remote seismic stations until the station fails. Periodic visual checks of cables and equipment, potential corrosion problems, grounding and lightning systems, intrusion of water and small animals, batteries,lightning protection elements, and cleaning the vaults and solar panels will efficiently eliminate technical problems before they occur. When something does go wrong, technical staff must be prepared to respond quickly with the trained personnel and spare parts. One should always maintain a good stockpile of the most common spare parts and have a well-trained technician available.

Batteries require special attention. Lightning damages are the most frequent source of equipment failures during normal operation but battery failures are the number one reason for station failure. It should be noted that the output voltage alone of a battery provides little information about its overall capacity. Many types of batteries may still have adequate output voltage while at the same time their power output is reduced to a small fraction of the original value. Batteries in this condition will not do the job in case of a long duration power failure, which is almost certain to occur after damaging earthquakes. Note that backup batteries for solar-powered stations traditionally fail in December at the time of the ensolation minimum. If December access is a problem due to weather, battery replacement should anticipate the problem by battery testing or simple annual replacement. A hundred-dollar battery may be far cheaper than a snowcat hire.

Seismological observatories should calibrate the sensors in their seismic system regularly - ideally, once a year. Strictly speaking, only the seismic signals recorded between two successive sensor calibrations that show no significant change in sensor frequency response function are completely reliable. Even the most reliable and simple sensor will degrade over time and should be checked for proper function. The most egregious example of this is some vertical seismometers whose weight is supported by a spring. As the spring ages, the sensor loses dynamic range and finally the weight settles against the lower stop resulting in a very quiet seismic station!

After several years of operation a seismic network, the scientific value of a seismic archive is extremely high. Therefore, attention must be paid to data archives and failsafe backups of the data. Lost seismic data can never be regenerated. The backups should be kept in a different physical location, no matter whether they are on paper, tape, disk, CD or other memory medium. Medium obsolesence is a non-trivial aspect of backups. Eight-inch floppy disks and 556 bpi 2400- foot tapes are not easily used today as will esoteric DVD formats not be current in the future.

When a seismic network is started, one needs to think thoroughly the organization of the data that is recorded. Often, this crucial aspect of seismic system organization is overlooked. Not only must the data be saved, but it must be retrieved to be useful. The proverbial 'write-only' medium has to be avoided. Nevertheless, filename coding of events must be planned to avoid confusion and/or file name duplications. Fortunately Y2K is behind us, but similar difficulties hang over the head of poorly designed file-organization schemes. Special data bases which have been developed in seismological community for the needs of seismology, seem to be the best choice at the moment. Always keep the unprocessed seismic data in the archive along with documentation about the recording conditions. Processing and seismic analysis methods will advance as time passes. Future generations will appreciate having raw seismic data available to further their research and knowledge.

Cooperation in the dissemination of seismic data is facet of the operation of any seismic network. Data sharing is the best way to obtain feedback about the quality of the work. Everyone can greatly improve their own quality by observing and comparing the phase reading residuals, the event locations, the magnitude determinations and the source mechanism results, with the results of others. The sharing of data may be restricted by the network owner and permission must be obtained to publish bulletins or earthquake catalogs. However, the purpose of any network is most often achieved by dissemination of factual data.