In this paper we model small-scale (micro) location variables using geographic information systems (GIS). These variables include information that is usually not available on a small scale from public sources, for example residential environment, noise, accessibility or elevation. The modeling of micro-location variables is mainly based on grid technology with a resolution accuracy of up to 30 meter covering the whole research area (Austria). We give several examples how to model spatial raster variables using the Python programming language. These variables are then used as explanatory covariates in hedonic house price models. We apply a generalized additive modeling (GAM) framework, where continuous covariate effects are modeled as polynomial splines and unexplained spatial heterogeneity (beyond what can be explained by location covariates) as random effects. We find that additional usage of high resolution micro-location variables improves the model quality significantly. The results are displayed on overview maps for Austria as well as in detail for some selected regions, which allows inspecting the effects of the applied grid technology in more depth. The resulting models are used for automated valuation purposes for residential real estate in a large Austrian bank.