Interpolation of a spatially correlated random process is used in many scientific domains. The best unbiased linear predictor (BLUP), often called kriging predictor in geostatistical science, is sensitive to outliers. The literature contains a few attempts to robustify the kriging predictor, however none of them is completely satisfactory. In this article, we present a new robust linear predictor for a substitutive error model. First, we derive a BLUP, which is computationally very expensive even for moderate sample sizes. A forward search type algorithm is used to derive the predictor resulting in a linear likelihood-weighted mean procedure that is robust with respect to substitutive errors. Monte Carlo simulations support the theoretical results. The new predictor is applied to the two SIC2004 data sets and is evaluated with respect to automatic interpolation and monitoring.