This work introduces two interaction techniques for examining multi-attribute spatial data for visual analytic applications: Layer Toggling and Visibility-Preserving Lenses. With the rise of open city data and services, gathering information at the city scale, like transportation, modeling, and predicting city events such as crime and traffic, has become feasible. Visually assessing the quality of these models requires correlating features, predictions, and ground truth both globally and locally, which is challenging with multi-attribute data due to occlusion. To address this, we introduce Layer Toggling for instantaneously changing layer visibility using a physical button box, allowing comparison of spatially coherent views using retinal persistence. Visibility-preserving lenses dynamically adjust to the density of revealed features, facilitating the exploration of spatial (2D) and quantitative data (1D), such as temporal attributes. We validate our approach on a use case visualizing urban data from São Paulo city across multiple data layers. Our methods support user exploration and expert analysis tasks, especially in validating and interpreting prediction algorithm outcomes at both global and local scales.