Peroxyacetyl nitrate (PAN) is a NOx reservoir compound that has the ability to transport NOx to remote environments, allowing for NOx photochemistry and/or deposition of nitrogen to these clean locations. Measurements of PAN have been made at Alert, Nunavut, and Summit, Greenland, aimed at understanding the impact of PAN chemistry on atmospheric nitrogen in the Arctic. These measurements show concentrations of PAN that are only slowly varying, even during ozone depletion events at polar sunrise, when free radical photochemistry is relatively active. We used a zero-dimensional photochemical model incorporating known gas-phase chemistry to simulate the observed behavior of PAN at Alert, Nunavut, and Summit, Greenland. The model simulations suggest a substantial net production rate for PAN over sunlit surfaces, which is inconsistent with the measured gas-phase concentrations. These observations thus indicate a fundamental problem with our understanding of PAN chemistry in low-temperature, snow-covered environments. We explore the possibility that we are missing an important sink for atmospheric PAN above snow-covered surfaces. If the loss is caused by snowpack deposition, the data result in calculated deposition velocities ranging from 0.05 to 1 cm s-1, which would represent a significant fraction of the unidentified total nitrate input to the snowpack and glacial ice at Summit, Greenland.