Working with Noise
The first consideration in the design of a seismometer installation is the quality of the site where the sensor is to be installed. The best way to determine this is deploy a sensor with very low self-noise and measure the site noise over a period of at least several months, however this is not always practical.
Minimum Earth Noise
The standard limit curves for typical earth noise spectral density were constructed by Peterson (1993). From a collection of seismic data from 75 sites located around the world, Peterson found that there is a minimum level of earth noise; noise levels below this are never - or extremely rarely - observed. This noise spectral density is known as the New Low Noise Model or NLNM. At the same time, he found that for a properly-installed sensor in a well-constructed vault, there is a maximum noise level which is typically observed when there is no other cultural or seismic activity. This is known as the New High Noise Model or NHNM.
Peterson's low-noise models serve well as an indication of the best and worst background noise levels to be expected at a well-constructed site, but these two limits cover a wide range. Even when every precaution is taken for good site design, it is extremely rare that performance as good as the NLNM is achieved.
Typical Earth Noise
More recently McNamara and Buland (2004) published a survey of approximately sixty sites scattered across the continental United States. From this survey they computed a more realistic low-noise model using Probability Distribution Functions (PDFs). They have named this model the PDF Mode Low Noise Model or MNLM. They conclude that "For a vast majority of stations within the United States, such low levels of noise [as the NLNM] are unattainable, suggesting that for routine monitoring purposes our MLNM represents a more realistic noise threshhold".
Your Site's Noise
Even if it is impossible to perform site reconnaissance with a broadband instrument, it is still a good idea to estimate the expected site noise. Is there a nearby existing permanant station? Is it reasonable to expect this site to be as good as the best in the world? Are there factors which will raise the earth noise such being in an urban area, nearby transportation corridors or local geology? Given these considerations estimate how many dB better or worse than the MLNM you expect the site to be.
Once you have determined the minimum earth noise level you are likely to encounter at a prospective site you should decide what the maximum seismic event is that you are likely to see and which events you are interested in.