Fast Simulation for Computational Sustainability Sequential Decision Making Problems

Solving sequential decision making problems in computational sustainability often requires simulators of ecology, weather, fire, or other complex phenomena. The extreme computational expense of these simulators stymie optimization and interactive visualization of decision rules (policies). This work presents our results in creating an interactive visualization for a wildfire management problem whose simulator normally takes several hours to run. We successfully generate visualizations for a landscape’s development over 100 year time spans within 17 seconds, when the original simulator took several hours.