Four spatial interpolation methods are compared using rainfall data from a network of thirteen rain gauges on Norfolk Island (area 35 km(2)). The purpose is to obtain spatially continuous rainfall estimates across the island, from point measurements and for different integration times, by the most effective means. The more computationally demanding method of kriging provided no significant improvement over any of the much simpler inverse-distance, Thiessen, or areal-mean methods, In order to assimilate some of the characteristics of spatially varying rainfall, and based on the comparisons performed, the inverse-distance method is recommended for interpolations using spatially dense networks. (C) 1998 Elsevier Science B.V. All rights reserved.