Permafrost carbon could produce a positive climate feedback. Until now, the ecosystem carbon budgets in the permafrost regions remain uncertain. Moreover, the frequently used models have some limitations especially regarding to the freeze-thaw process. Herein, we improved the IBIS model by incorporating an unfrozen water scheme and by specifying the parameters to estimate the present and future carbon budget of different land cover types (desert steppe, steppe, meadow, and wet meadow) in the permafrost regions. Incorporating an unfrozen water scheme reduced the mean errors in the soil temperature and soil water content by 25.2%, and the specifying leaf area parameters reduced the errors in the net primary productivity (NPP) by 79.9%. Further, the simulation results showed that steppes are carbon sources (39.16 gC/m(2)/a) and the meadows are carbon sinks (-63.42 gC/m(2)/a ). Under the climate warming scenarios of RCP 2.6, RCP 6.0, and RCP 8.5, the desert steppe and alpine steppe would assimilated more carbon, while the meadow and wet meadow were projected to shift from carbon sinks to carbon sources in 2071-2100, implying that the land cover type plays an important role in simulating the source/sink effects of permafrost ecosystem carbon in the IBIS model. The results highlight the importance of unfrozen water to the soil hydrothermal regime and specific leaf area for the growth of alpine vegetation, and present new insights on the difference of the responses of various permafrost ecosystems to climate warming.

Cold regions contain a large amount of soil organic carbon, and the warming-accelerated loss of this carbon pool could cause important feedback to climatic change. The changes of carbon budgets in cold regions are poorly quantified especially for the Qinghai-Tibet Plateau (QTP) due to limited field observation data. By considering the soil freeze-thaw process and establishing new plant functional types with localized parameters, we used the Integrated Biosphere Simulator (IBIS) model to simulate the changes of carbon budget on the QTP during 1980-2016. The model was calibrated and validated using carbon flux data from eddy covariance observations at 16 sites. The results showed that the QTP has assimilated 43.16 Tg C/yr during 1980-2016, with permafrost and non-permafrost regions accounting for approximately 15% and 85% of the carbon sink, respectively. During the past four decades, the gross primary production and ecosystem respiration have increased by 1.74 and 2.04 Tg C/ yr(2), resulting in that the carbon sink on the QTP has weakened during 1980-2016. Moreover, the weakening of carbon sink is more pronounced in the non-permafrost regions. We project that the ecosystems will release 12.30 and 24.40 Tg C by 2080-2100 under the moderate and high shared socio-economic pathways (SSP 370 and SSP 585), respectively. This could largely offset the carbon sink and even shift the carbon sink to carbon source on the QTP.