Many small island developing states are adopting solar photovoltaic technologies as a solution to curtail their overreliance on fossil fuels. Some of these islands are young and of volcanic origin, characterised by intricate topographic heterogeneity. Since they are mostly small, flat land is scarce and valuable. As a result, solar photovoltaic projects are usually developed at sites with complex terrain where frequent changes in elevation, slope and shadows cast by topographic features can significantly attenuate the amount of solar irradiance received. The existence of strong local solar radiation gradients on such islands implies that any solar resource assessment must be performed at a small enough spatial and temporal resolution in order to provide a reliable basis for investment decisions. Ground-based measurements of solar irradiance are accurate but are widely dispersed geographically. Although satellite derived solar resource data have better spatial and temporal resolution than those recorded by radiometric stations, they are less accurate. Moreover, their spatial resolution is still too large with respect to the small size of the islands. In this work, a downscaling methodology is used to derive monthly and annual average global solar irradiance maps at a resolution of 250 m x 250 m using satellite hourly datasets, spanning over the period 1999 to 2015 and having a spatial resolution of 0.05 x 0.05. The combined effects of horizon blocking and sky obstruction resulting from heterogeneous topography are considered by applying a binary blocking factor and sky view factor respectively. The downscaled maps are then refined by convolving them with ground-based measurements through a kriging procedure. The methodology is validated using leave-one-out kriging procedure and illustrated successfully with the case of Reunion Island, where strong local gradients of solar irradiance were easily distinguished in regions marked by uneven topographic features. (C) 2018 Elsevier Ltd. All rights reserved.