Abstract All systems have causes and effects that can be appreciated at different spatial scales. Understanding and representing the complexity of multi‐scale patterns in maps and spatial models are key research objectives. We describe the use of three types of correlation analyses: (1) a standard Pearson correlation coefficient, (2) a 'global' multi‐scale correlation, and (3) local geographically weighted correlation. These methods were applied to topographic and vegetation indices in a small catchment in Honduras that is representative of the country's hillsides agro‐ecosystem which suffers from severe environmental degradation due to land‐use decisions that lead to deforestation, overgrazing, and unsustainable agricultural. If the geographical scale at which topography matters for land‐use allocation can be determined, then integration of knowledge systems can be focused. Our preliminary results show that: (1) single‐scale correlations do not adequately represent the relationship between NDVI and topographic indices; (2) peaks in the global multi‐scale correlations in agricultural areas coincided with the median farm size, but there was no evidence of any community or larger‐scale land‐use planning or optimization; and (3) local multi‐scale correlations varied considerably from the global results at all scales, and these variations have a strong spatial structure which may indicate local optimization of land use. Keywords: ScaleGeographically weighted statisticsNDVITopography Acknowledgements We thank Annie Jones, for her editing and proofreading, and three anonymous reviewers for helpful and constructive comments on previous drafts. This research was undertaken as part of a 4‐year project entitled 'Methodologies for integrating data across geographic scales in a data rich environment: Examples from Honduras', co‐ordinated by the International Center for Tropical Agriculture (Centro Internacional de Agricultura Tropical, CIAT). The primary funding for this project came from the Ecoregional Fund to Support Methodological Initiatives, a Dutch initiative to address problems of natural‐resource management across a broad range of agro‐ecological and socio‐economic conditions and scales. The Beowulf PC cluster used to run the multi‐scale models was funded by the UK Economic and Social Research Council (RES‐474‐25‐0007: Geographical Analysis and Modelling of Populations).