The ongoing COVID-19 pandemic has caused immeasurable losses for people worldwide. To contain the spread of virus and further alleviate the crisis, various health policies (e.g., stay-at-home orders) have been issued which spark heat discussion as users turn to share their attitudes on social media. In this paper, we consider a more realistic scenario on stance detection (i.e., cross-target and zero-shot settings) for the pandemic and propose an adversarial learning-based stance classifier to automatically identify the public attitudes toward COVID-19-related health policies. Specifically, we adopt adversarial learning which allows the model to train on a large amount of labeled data and capture transferable knowledge from source topics, so as to enable generalize to the emerging health policy with sparse labeled data. Meanwhile, a GeoEncoder is designed which encourages model to learn unobserved contextual factors specified by each region and represents them as non-text information to enhance model's deeper understanding. We evaluate the performance of a broad range of baselines in stance detection task for COVID-19-related policies, and experimental results show that our proposed method achieves state-of-the-art performance in both cross-target and zero-shot settings.