Applications Of Permafrost Distribution Models On The Qinghai-Tibet Plateau

As one of the major components of cryosphere, permafrost can not be ignored in the global climate system. Permafrost is widely distributed, and has unique hydro-thermal properties. It is a very important factor in the Earth land surface processes. The Qinghai-Tibet Plateau (QTP), the highest and the largest permafrost area in low-middle latitudes all over the world, has a significant effect on the formation, development, and evolution of the climate system in East China and even East Asia, and therefore is known as the alarming or startup region. With increasing human activities, such as the operation of Qinghai-Tibet railway, a correct assessment of the permafrost distribution in the QTP is not only the base of investigating and solving practical engineering problems, but also the requirement for stable and sustainable development of ecological environment and local economy.Because of high altitudes and vast areas of the QTP, permafrost data are difficult to obtain using traditional regional investigation methods. In this case, permafrost distribution modeling is an effective method. Although there have been a number of permafrost distribution models available in the literature, not all of them can be applied to the QTP. Applicability analysis should be conducted before we use them.As a main parameter of the energy balance on the Earth surface, land surface temperature (LST) is one of the boundary conditions to numerical permafrost models and is also an important content of permafrost observations. Traditionally we obtain LST through meteorological stations. However, because such station represents only a single point situation, the traditional approach usually may not reflect average condition well in a larger area, especially when LST is so variable to land covers. With the development of remotely sensed technology, land surface temperature observation based on the remote sensing data becomes an improtant method. However, due to the complicated terrain surface, the data acquisition, processing, and analysis might not be accurate enough. Further analysis prior to use is necessary. In this thesis, we examine the MODIS land surface temperature product. The applicability of the MODIS surface temperature data over the QTP are carefully investigated with various levels of analyses, namely single-point, areal, and model analyses. The single-point comparison is conducted by comparing the surface (0 cm) temperature data observed at 69 meteorological stations with the MODIS surface temperature data at the same location. It is observed that the two time series data sets have a similar trend but obvious absoluate errors. With the areal validation approach, a regression with latitiude, longitude and elevation, and the Kriging interpolation methods are applied to firstly generate areal LST distribution over the QTP, which are then compared to the the MODIS time series. The evaluation shows that interpolated areal distribution presents consistent in trend with MODIS, and also some variations in terms of magnitude and spatial pattern. We also discovered that the Kriging interpolation performs worse than the regression method. Meanwhile, both the single-point and the areal validations have revealed that the two data sets have a larger difference and a smaller correlation during the warm season (May-August) than during the cold season (September-April). A semi-physical, semi-empirical permafrost distribution model, the Top Temperature Of Permafrost (TTOP) model is used for model validation purpose. Measured temperature data and the MODIS LST are input to the TTOP model respectively, to evaluate the performance of the two data sets in a simulation model. Area statistics and the Kappa agreement coefficients show that the simulation with MODIS LST is closer to the survey-based QTP permafrost map than that with measured data.In this paper, we also applied four permafrost distribution models to the QTP, including the TTOP model, a modified surface frost model originally developed by Nelson, the Kudryavtsev model, and the mean annual ground temperature (MAGT) model. Taking the QTP permafrost map, developed by survey data and expertise, as the reference, it is observed that the TTOP model performs best, the surface frost model followed, and then the MAGT model, and the Kudryavtsev model. The MODIS LST derives better simulation with all the 4 models than measured surface (0 cm) data. The Kappa analysis shows that, in the north and the hinterland of the QTP covered by continuous permafrost, the simulation is always better; in south QTP and the Himlayas area with discontinuous permafrost, the simulation is worse. The agreement analysis based on stratifying altitude sand relief amplitudes show obvious vertical zonality characteristics of the high-altitude permafrost discoved in earlier works.