Researchers and natural resource managers need predictions of how multiple global changes (e.g., climate change, rising levels of air pollutants, exotic invasions) will affect landscape composition and ecosystem function. Ecological predictive models used for this purpose are constructed using either a mechanistic (process-based) or a phenomenological (empirical) approach, or combination. Given the accelerating pace of global changes, it is becoming increasingly difficult to trust future projections made by phenomenological models estimated under past conditions. Using forest landscape models as an example, I review current modeling approaches and propose principles for developing the next generation of landscape models. First, modelers should increase the use of mechanistic components based on appropriately scaled ‘‘first principles’’ even though such an approach is not without cost and limitations. Second, the interaction of processes within a model should be designed to minimize a priori constraints on process interactions and mimic how interactions play out in real life. Third, when a model is expected to make accurate projections of future system states it must include all of the major ecological processes that structure the system. A completely mechanistic approach to the molecular level is not tractable or desirable at landscape scales. I submit that the best solution is to blend mechanistic and phenomenological approaches in a way that maximizes the use of mechanisms where novel driver conditions are expected while keeping the model tractable. There may be other ways. I challenge landscape ecosystem modelers to seek new ways to make their models more robust to the multiple global changes occurring today.