Multipath is one of the prominent error sources in Global Positioning System (GPS), which adversely affects the accuracy of the user's position. This paper describes the estimation of frequency components of multipath error signal using spectral analysis and its effective mitigation using time-varying digital filters. Four types of 2nd order filters, namely, Butterworth, Type I and II Chebyshev and Elliptic filters, are examined for mitigation of multipath and their performance is compared. It is observed that by applying digital filters of different cut-off frequencies over the spectrum of the multipath, one can significantly reduce the multipath errors. The designed filters are applied to GPS data acquired from a typical urban environment site and found that Butterworth filter reduced the error most effectively. To extract the magnitude of error contribution in low and medium multipath frequency ranges, its mitigation is carried out with Butterworth filter. The performance of the filter is validated by the spectral analysis for multipath signals before and after filtering process. However, this spectral analysis followed by digital filtering is different from the approaches reported elsewhere in the open literature. This new approach is very useful in identifying least-affecting multipath sites for GPS applications. [ABSTRACT FROM AUTHOR]