The noise in electromagnetic (EM) surveys that originates in time-dependent magnetic fields is quantified as a function of geographic location, season, and polarization. An overall variability of 10 4 is observed. The variance of the observed geophysical data caused by this noise is quantified in terms of the noise bandwidth of the signal-processing system. While a time-domain EM (TEM) system has a wide bandwidth for the target (i.e., synchronous) signal, it has a narrow bandwidth for the asynchronous EM noise. A procedure is developed to compute the magnetic moment required to provide a specified ratio of signal to noise (S/N) in the survey data for frequency-domain EM (FEM) and both step-response and impulse-response TEM. This procedure indicates that magnetic moments in the range 10 5 to 5 X 10 6 A.m 2 are necessary for moving transmitter systems, and up to 3 X 10 7 A.m 2 for fixed-loop systems, for both frequency and time domain.A number of models of the overburden are established, and we show that the half-space response is strongly dependent upon geology and weathering history. A variation of three orders of magnitude in the half-space response occurs throughout the world. A procedure is defined to calculate the smallest target response detectable in the presence of a specified level of geologic noise. From this response we show that the overburden in the tropical and arid zones of the Earth strongly desensitizes the EM method. This is particularly severe for those styles of mineralization yielding short decay time constants. We conclude it would be impossible to detect targets with time constants <3 ms in substantial portions of the world's arid zone using EM alone. An overall survey design procedure is defined that determines whether a specified target would be detectable in the presence of the prevailing geologic noise, and the magnetic moment that would be required to detect it in the presence of the anticipated EM noise. The manner in which the primary field corrupts both frequency- and time-domain systems is also analyzed, and we conclude that the TEM method is essentially free of primary field effects, while the simpler forms of FEM can be corrupted by primary field effects which obscure targets with secondary responses that are less than 10 percent of the primary field.