In this paper, which represents a continuation of the previous author's work, an inconstancy of GPS residual error ANOVA estimates and their variances are presented. For the purpose of the analysis, fixed solutions for all of the three coordinates, e (eastwards), n (northwards) and u (upwards), obtained by using ionosphere-free (L0) linear combination of double-difference phase observations in the processing of GPS data, were employed. The aim of the research was to consider the behaviour of variances of GPS residual error ANOVA estimates in time because there has not been any paper dealing with that issue so far. Herein, it turned out a seasonal pattern in related time series was present. In addition, it was concluded there was a difference in ANOVA estimate extreme values obtained when one considered daily data subsets compared to those obtained in approaches considering monthly data of the fixed solutions. GPS data collected at ending stations of a baseline of 40 km in length within a four-year period, involving the lowest and increased solar activity, were used in calculations.

Key words: ''far-field'' multipath, tropospheric and ionospheric effect, ANOVA, GPS positioning, seasonal pattern