Simulation models are valuable tools for estimating ecosystem response to environmental conditions and are particularly relevant for investigating climate change impacts. However, because of high computational requirements, models are often applied over a coarse grid of points or for representative locations. Spatial interpolation of model output can be necessary to guide decision-making, yet interpolation is not straightforward because the interpolated values must maintain the covariance structure among variables. We present methods for two key steps for utilizing limited simulations to generate detailed maps of multivariate and time series output. First, we present a method to select an optimal set of simulation sites that maximize the area represented for a given number of sites. Then, we introduce a multivariate matching approach to interpolate simulation results to detailed maps for the represented area. This approach links simulation output to environmentally analogous matched sites according to user-defined criteria. We demonstrate the methods with case studies using output from (1) an individual-based plant simulation model to illustrate site selection, and (2) an ecosystem water balance simulation model to illustrate interpolation. For the site selection case study, we identified 200 simulation sites that represented 96% of a large study area (1.12 × 106 km2) at a ~1-km resolution. For the interpolation case study, we generated ~1-km resolution maps across 4.38 × 106 km2 of drylands in North America from a 10 × 10 km grid of simulated sites. Estimates of interpolation errors using cross validation were low (<10% of the range of each variable). Our point selection and interpolation methods, which are available as an easy-to-use R package, provide a means of cost-effectively generating detailed maps of expensive, complex simulation output (e.g., multivariate and time series) at scales relevant for local conservation planning. Our methods are flexible and allow the user to identify relevant matching criteria to balance interpolation uncertainty with areal coverage to enhance inference and decision-making at management-relevant scales across large areas.
