Dung deposition is an important pathway of nutrient return and redistribution in alpine grasslands of the Qinghai-Tibetan Plateau. To date, information on the greenhouse gas emissions of yak dung on alpine grasslands, especially where there are large amounts of rock fragments, is limited. Our aim, therefore, was to evaluate variations in N2O, CH4, and CO2 emissions from yak dung (CCD), and compare it to dung placed on rock fragments (RCD), alpine steppe soil (CSD), and a soil and rock fragment mixture (RSD) over a 30-day incubation period. The results showed that the total N2O emissions from treatments without soil were significantly (P < 0.05) lower than those from treatments with soil. The highest total CH4 emissions were detected in the CSD treatment, while CH4 losses from treatments without rock fragments were significantly (P < 0.05) greater than those with rock fragments. The total CO2 emissions from the RSD treatment was 6.30%-12.0% lower than those in the other three treatments. The soil beneath yak dung pats elevated the global warming potential (GWP), while the addition of rock fragments to the soil significantly (P < 0.05) decreased the GWP by reducing emissions of the three greenhouse gases. We therefore suggest that interactions between rock fragments and alpine steppe soil are effective in decreasing yak dung greenhouse gas emissions. This finding indicates that rock fragments are an effective medium for reducing greenhouse gas emissions from dung pats, and more attention should therefore be paid to evaluate its ecological impact in future studies. These results should help guide scientific assessments of regional GHG budgets in agricultural ecosystems where the addition of livestock manure to soils with large amounts of rock fragments is common.