We examine for the first time in the scientific community the application of hyperspectral (HS) based anomaly detection in contrast to polarimetric (POL) based anomaly detection in the longwave infrared region of the spectrum, using a challenging dataset for the test that covers three diurnal cycles. For fairness, we standardized for both sensing modalities the characterization of the unknown background clutter through a repeated trial Binomial based random sampling approach, and attained in the process two new methods for anomaly detection. The POL method outperformed the HS method, especially in the most difficult time periods, between sunset and sunrise, by an average of 0.47 augmented performance.