The Internet-of-things ecosystem has been a driving force in the creation of smart communities where a variety of physical phenomena can be monitored continuously, e.g., air quality, traffic conditions on roads, energy consumption in buildings, etc. In this paper, we address how IoT can be quickly and effectively deployed for short-term and sporadic events (e.g., fire spread in a wildland area and flood propagation), where monitoring the evolving event is critical. In particular, we propose QuIC-IoT, a model-driven planning platform that aims to temporarily deploy a custom IoT infrastructure for monitoring short-term events, where phenomena-spread is driven by models that are physics-based. Our driving usecase event is a quasi-planned prescribed fire or RxFire - this is a wildfire resilience technique where intentional small fires are ignited apriori by forestry personnel to destroy fuel and help contain the spread of actual wildfires. Anomalies that may occur during these quasi-planned events must be rapidly captured by the IoT deployment, e.g., escaped RxFires can escalate to catastrophic wildfires under unpredictable conditions of wind, vegetation, etc. QuIC-IoT incorporates domain expert-developed models to guide IoT deployment; the event area is partitioned into subregions and a criticality metric that quantifies the likelihood of anomalies at each location is computed. QuIC-IoT allows us to mix fixed and quasi-mobile IoT devices to flexibly deploy IoT in challenging terrain and as the phenomena (RxBurn) evolves. We evaluate QuIC-IoT in two real-world forest settings (large and small) in Blodgett Forest, CA, USA, with concrete burn plans developed by wildfire experts. Our experimental results reveal that QuIC-IoT enables over 3X improvement in cost-effectiveness and performance (timely detection of anomalies) as compared to baseline IoT deployment algorithms.