RESPONSE OF PERMAFROST TO GLOBAL CHANGE ON THE QINGHAI-XIZANG PLATEAUÑ A GIS-AIDED MODEL Li Xin, Cheng Guodong, Chen Xianzhang State Key Laboratory of Frozen Soil Engineering, Lanzhou Institute of Glaciology and Geocryology Chinese Academy of Sciences, Lanzhou 730000, P. R. China e-mail: wdcdgg@ns.lzb.ac.cn Abstract On the basis of three-dimensional rules for high altitude permafrost, the Òaltitude modeló, a Gaussian distribution function used to describe the latitudinal zonation of permafrost, was adopted to develop a quantitative mathematical model. This model is executed within a Geographic Information System (GIS). The digital elevation data and the land-types were first converted to 0.5¼X0.5 grids; then the Òaltitude modeló was used to calculate the permafrost distribution on the Qinghai-Xizang Plateau. The third step was to determine the simulation accuracy by using spatial analysis in GIS. The last step was to simulate the response of high altitude permafrost to global change according to GCM forecasts of climatic change scenarios. The results show that the Òaltitude modeló can simulate the high altitude permafrost distribution under present climatic conditions and under various scenarios of climate change. GIS in Permafrost research WHY GIS? In recent years, Geographic Information Systems (GIS) have been used more and more in the field of permafrost research (Keller, 1992; Anisimov, 1996; Haeberli et al., 1993). This is because permafrost models require high-resolution spatial data such as elevation, land-type and climatic data and GIS can manipulate these data efficiently. Since some permafrost models integrate so well with GIS data models, GIS methods have became the essential components of some models. GIS-based permafrost models have the following advantages: 1. The GIS data model is suitable for spatial simulation of permafrost. Zonal parameters are managed by the GIS grid or TIN data models, so they can be resampled, interpolated or displayed efficiently. These can be overlain by non-zonal parameters, such as large river melting areas, if these parameters are also properly georeferenced. 2. Modern GIS often integrates closely with remote sensing; therefore it can provide some important physical parameters, such as surface temperature, water content and snow amount for permafrost models. 3. GIS spatial analysis methods can help in testing the simulation processes and in modifying the models; sometimes, they also can be the essential components of permafrost models. 4. GIS is useful for visualizing in maps, animation and three-dimensional views. GIS OF THE CHINESE CRYOSPHERE In order to promote the application of GIS in cryospheric research, a GIS of the Chinese cryosphere, CCREIS (Chinese cryosphere resource and environment information system), has been established at the LIGG (Lanzhou Institute of Glaciology and Geocryology). One of its sub-databases covers the Qinghai-Xizang Plateau. The sub-database includes a DEM (Digital Elevation Model) with a resolution of 2Õ30Ó in latitudinal orientation and 345 in longitudinal orientation, thematic maps of the cryospheric and background information on a scale of 1:4,000,000, and meteorological data for 87 stations for more than 30 years. For both computational and cartographic purposes, the CCREIS data are stored in two types of map projections, geographic and albers conic equal-area projection, so that all the areal distribution phenomena can be retained and the system will be compatible with GCM scenarios The CCREIS is managed by ARC/INFO for UNIX and by ERDAS IMAGINE. A grid GIS base-class, CGridGIS also has been developed using the object-oriented method; therefore, the permafrost models can inherit from it and be naturally embedded in GIS. GIS-based Permafrost model From a GIS viewpoint, simulating permafrost response to global change can be divided into the following steps: 1. Model Selection; the models can be integrated with GIS. Li Xin, et al. 657
Equation 1 is called the Òaltitude modeló. This model differs greatly different from linear mathematical models. At the first stage of the curve, the lower limit of permafrost rises with increasing latitude, and reaches its extreme value of 5078m at latitude 25 22ÕN. The limit then descends as latitude increases (Figure 1). The features of the function are closely related to the mean latitudinal distribution of the earthõs radiation budget. Clearly, the magnitude of the altitude decrease in the permafrost limit south of 25 N is not correct, but this is not important here as the simulation is carried out for areas north of this latitude. Figure 1. Altitude model. 2. Present all spatial parameters, climate change scenarios and the current permafrost distribution with the GIS data model. 3. Use the models to simulate the current permafrost distribution, selecting a model providing good results. 4. Couple the selected model with climatic change scenarios, and forecast the permafrost change. After comparing different permafrost models, the Òaltitude modeló was adopted for use in this paper. ALTITUDE MODEL Cheng has pointed out (Cheng, 1984, 1992) that there is an obvious three-dimensional zonation in the distribution of high altitude permafrost, namely vertical, latitudinal and aridity (or longitudinal) zonations. By using the method of curve fitting, an empirical correlation between the lower limit of high-altitude permafrost (H) and latitude (j) has been obtained. It is a Gaussian distribution function, and can be expressed as: H=3650 exp[-0.003(j-25.37) 2 ]+1428 [1] Because the Òaltitude modeló takes the lower limit as the main criterion of the high altitude permafrost distribution, and geographic coordinates can be obtained from the digital elevation model (DEM), the DEM then can be used to calculate the lower limits for every grid and compare it to the elevation of the grid to conclude if permafrost exists. The judgment function can be expressed as: ì1, h> H P = í î0, h H Where, P is a Boolean variable. When P=1, permafrost exists; when P=0, permafrost does not exist; h is the elevation (m) of the grid. GIS data processing DATA FLOW CHART The use of the Òaltitude modeló, to simulate the permafrost distribution, can be divided into the procedures shown by the data flow chart in Figure 2. This model does not consider the influence of nonregional factors such as terrain, terrestrial heat and large rivers. In principle, if some quantitative criteria become available, these non-zonational factors can be [2] Figure 2. Data flow chart of Òaltitude modeló. 658 The 7th International Permafrost Conference
included in the model by overlaying, buffering and other GIS methods. DIGITAL ELEVATION DATA The original DEM of the Qinghai-Xizang Plateau has a high spatial resolution. It was resampled to a coarse resolution of 0.5 degree. The resampling method used was cubic convolution because it preserves the statistical properties of the original DEM. PERMAFROST DISTRIBUTION MAP The permafrost map complied by Li Shude and Cheng Guodong (1996) was digitized to provide the base data, because it reflects the most up-to-date permafrost observations. The map was transferred to a grid map with a resolution of 0.5 by ARC/INFO software, so that it could be spatially registered with the DEM. GCM The GCM model HADCM2 (Viner, 1996) was adopted for climate scenarios. It was developed at the Hadley Center for Climate Prediction and Research in Britain. At present the Hadley CenterÕs high-resolution, coupled ocean-atmosphere (O/A) GCMs have a spatial resolution of 2.5 latitude 3.75 longitude. This is the highest O/A GCM currently operational for transient climatic change experiments. The HADCM2 can provide 30 climatic variables. Among all the variables, only air temperature in the HADCM2GHS, which includes the effects of both greenhouse gases and sulfate aerosols (Hadley Center, 1997), was used in the response model of high altitude permafrost to climatic change. In order to preserve the original air temperature forecast results in the HADCM2, the nearest-neighbor method was used to resample the air temperatures for the years 2009, 2049 and 2099 into 0.5 X0.5 grids compatible with the DEM. The maps of air temperature change show that, for the above three time steps, the air temperature would increase by 0.51, 1.10 and 2.91 C, respectively, on the Table 1. Permafrost changes on the Qinghai-Xizang Plateau Qinghai-Xizang Plateau. The maximum air temperature increases would be 1.62, 2.99 and 5.45 C respectively. The regions with maximum air temperature increases are in the southern foothills of the Himalaya Mountains, the western Qinghai-Xizang Plateau centered around Shiquanhe, the southern edge of the Talimu and Chaidamu Basins and other very high mountains. These regions are on the outside margin of the Qinghai-Xizang Plateau. The air temperature increase would be relatively small in the hinterland of the Qinghai-Xizang Plateau. SIMULATION RESULTS OF THE PERMAFROST DISTRIBUTION ON THE QINGHAI-XIZANG PLATEAU In order to compare the simulation results using the Òaltitude modeló with the actual permafrost distribution on the Qinghai-Xizang Plateau, 1938 spatial samples were taken for regression analysis. The results show that the correlation is 0.92 and the coefficient of determinacy is greater than 80%. The simulation results show that Òaltitude modeló can describe the permafrost distribution on the Qinghai-Xizang Plateau very well. Permafrost change on Qinghai-Xizang plateau ASSUMPTIONS Because there are no climatic variables in the Òaltitude modeló, some assumptions must be provided for permafrost forecasting under climate change. These assumptions are: 1. That the Gaussian function that describes high altitude permafrost distribution will not change during the climatic warming. Because the latitudinal distribution of the earthõs radiation budget is relatively stable, this assumption is reasonable. 2. If the air temperature increases 1 C, the latitudinal zonation at the north of the curveõs extreme point (25 22ÕN) will shift to the south by 1 degree; the latitudinal zonation at the south of the curveõs extreme point K Li Xin, et al. 659
will shift to the north by 1 degree. Thus, the lower limit of permafrost would increase. But the air temperature increase will be different at different latitudes. The change in the lower limit is 170m at 40 N, 156m at 35 N and 86m at 30 N. This increase is in accord with field observations (Cui, 1980; Xie, 1996). Thus, this assumption is also reasonable. 3. That the lakes, glaciers and deserts will not change. Although climatic warming does cause land-type changes, the above features occupy only a small areal extent on the Qinghai-Xizang Plateau. The influence of land-type changes on permafrost distribution can be ignored. Based on the above assumptions, and if only the air temperature increase is taken into account, the permafrost distribution during the future 20-100 years can be forecast using the Òaltitude modeló. SIMULATION RESULTS The simulation results in Figure 3 show that for the above three time periods, if the present simulation result is regarded as the baseline (with an area of 1,294,376 km 2 ), the spatial distribution changes of permafrost on the Qinghai-Xizang Plateau can be obtained. Table 1 shows the air temperature increases and related permafrost changes and the main regions where permafrost will disappear. In summary, the permafrost area on the eastern Qinghai-Xizang Plateau and the permafrost area in the foothills surrounding very high mountains facing the Plateau decreases more rapidly than in other regions. The trend for permafrost degradation is from the outside towards the inner plateau. Figure 3. ÒAltitude modeló simulation result of permafrost change on Qinghai-Xizang Plateau. 660 The 7th International Permafrost Conference
Discussion and Conclusions 1. The Òaltitude modeló was tested and found to be suitable for the prediction of the high altitude permafrost distribution on the Qinghai-Xizang Plateau. However, using this model to forecast the changes is somewhat difficult, because the effect of global warming on the lower limit of permafrost must first be known. With certain assumptions, the Òaltitude modeló can be used to forecast the permafrost response to global change. 2. The simulation results shows that the permafrost on the Qinghai-Xizang Plateau will change significantly if the air temperature increases by an average of 2.91 C. The decrease in the area of permafrost will exceed 31%, but the permafrost on the northwestern Qinghai-Xizang Plateau, Himalaya Mountains, FenghuoShan Mountains will not disappear. 3. The method used in this paper is suitable for meso to continental-scale permafrost simulation and forecasting. For local-scale permafrost forecasting, more factors exist to influence the distribution of permafrost and their influence is more complex. The development of a more detailed GIS using physical models based on the thermal balance is necessary for detailed permafrost forecasting. 4. The decrease in the area of permafrost is not the unique criterion reflecting permafrost degradation. Global warming also results in changes in the engineering properties and the stability of permafrost. Developing models to evaluate the engineering properties and the stability of permafrost is very important. These models will rely on long-term remote sensing monitoring, and integrated GIS, so that the spatial distribution of some very important physical parameters, such as surface temperature, water content and snow amount can be obtained and be included in the GISbased permafrost models. Acknowledgments The authors wish to thank Prof. Kang Xincheng for providing 30 years (1961-1990) of mean monthly air temperature data on the Qinghai-Xizang Plateau. Also, to thank Dr. David Viner at the University of East Anglia, for calculating and providing monthly global air temperature, surface temperature, precipitation and snow amount data from the year 1990 to 2099. Thanks to Professor Max C. Brewer, Mr. Fred Wright and Professor M. W. Smith, for their critical review and editing of the paper. This work is supported by the Chinese Academy of Sciences as part of the project ÒFundamental Research of the Cryosphere Ó. References Anisimov, O. A. and Nelson, F. E. (1996). Permafrost distribution in the Northern Hemisphere under scenarios of climate change. Global and Planetary Change, 14, 59-72. Cheng Guodong and Dramis, F. (1992). Distribution of mountain permafrost and climate. Permafrost and Periglacial Process, 3(2), 83-91. Cheng Guodong (1984). Problems on zonation of high-altitude permafrost. ACTA Geographica Sinica, 39(2), 185-193. Cui Zhijiu (1980). Periglacial phenomena and environmental reconstruction in the Qinghai-Tibet Plateau. In: Collection of Geological Research Papers for the International Exchange. Written for the 26th session of the International Geology Conference, Geological Publishing House, pp. 109-115. Hadley Center (1997). http://www.cru.uea.ac.uk/link/pubs.html#mitchell. Haeberli, W., Cheng, G. Gorbunov, A. P. and Harris, S. A. (1993). Mountain permafrost and climate change. Permafrost and Periglacial Process, 4(2), 165-174. Keller, F. (1992). Automated mapping of mountain permafrost using the program PERMAKART within the geographic information system ARC/INFO. Permafrost and Periglacial Process, 3(2), 139-142. Li Shude and Cheng Guodong (1996). Permafrost distribution map on the Qinghai-Xizang Plateau. Gansu Culture Press. Viner, D. (1996). The climate impacts LINK projects: data sets available for climate change research. http://www.nerc.ac.uk/ukgeroff/globe24.htm. Xie Youyu (1996). Effects of Climate Change on Permafrost in China. Institute of Geography, Chinese Academy of Sciences, Global Change Study No. 2, Series publication. Li Xin, et al. 661