Abstract
Corrosion on a steel plate is not totally random, but has spatial pattern. We propose a statistical model of steel corrosion depths considering spatial auto-correlation, with the help of spatial modeling theories developed in regional analysis. Distribution parameters are derived through moment method and maximum-likerihood method. The statistical model is applied to corroded steel plates cut out from a steel pier. The case study shows that the proposed model fits to the corroded plates better than an ordinal Weibul distribution model neglecting spatial auto-correlation. The parameter of spatial decay of the auto-correlation was stable for selveral number of the applied cases.
