The low-cost crystal oscillators in wireless sensor networks are prone to be affected by their working conditions such as temperature, voltage, and humidity. Such problem brings two key challenges for time synchronization in wireless sensor networks (WSNs): Excessive communication overhead and the trade-off between accuracy and cost. This paper introduces a novel environment-adaptive time synchronization approach that enables nodes to estimate their clock skew by exploiting temperature information. The approach can substantially reduce communication overhead since clock skew estimation mostly relies on local information. In addition, this work proposes an environmental-adaptive interval adjustment scheme for duty-cycled clock calibration, which provides a convenient trade-off between the timing accuracy and the energy efficiency.