Public Sensing (PS) is a recent trend for building large-scale sensor data acquisition systems using commodity smartphones. Limiting the energy drain on participating devices is a major challenge for PS, as otherwise people will stop sharing their resources with the PS system. Existing solutions for limiting the energy drain through model-driven optimizations are limited to dense networks where there is a high probability for every point of interest to be covered by a smartphone. In this work, we present an adaptive model-driven PS system that deals with both dense and sparse networks. Our evaluations show that this approach improves data quality by up to 41 percentage points while enabling the system to run with a greatly reduced number of participating smartphones. Furthermore, we can save up to 81% of energy for communication and sensing while providing data matching an error bound of 1°C up to 96% of the time.