Study on the Permafrost Distribution Based on RS/GIS

Transcription

1 2017 Asia-Pacific Engineering and Technology Conference (APETC 2017) ISBN: Study on the Permafrost Distribution Based on RS/GIS Kun Wang, Lichun Chen, Bin Wei and Le Wang ABSTRACT Permafrost in Qinghai-Tibet Plateau is the world s highest elevation and largest permafrost regions in the lower latitude. For its vast area and less measured data, traditional mapping methods rely on the observations of station have been largely restricted. So it is necessary to build permafrost distribution model. Based on the GIS, using the RS data as the main data source and considering the distribution characteristics of permafrost in Qinghai-Tibet Plateau, we build Multivariate analysis model. Considered the elevation, land-surface temperature, vegetation, soil moisture and surface albedo and other factors related to the permafrost distribution in Tibet Plateau, and selected the Qinghai Tibet highway as the study area to simulate the permafrost distribution. Compared with the Retrieval Model of Ground Temperature based on RS/GIS, elevation model and frost number model, it is found that simulation result of Multivariate analysis model fit best with the permafrost distribution map and it reflected the permafrost distribution well along the Qinghai-Tibet highway. INTRODUCTION Permafrost can t be observed directly by visible and infrared sensors like other parts of cryosphere, but for its wide distribution and unique hydrothermal characteristics. Its existence is closely bound to the environment and reflects in the surface information. Therefore we can use remote sensing method to obtain the surface energy, vegetation distribution and elevation information related to the permafrost distribution, to provide some surface parameters for the study on permafrost distribution, which can indirectly indicate the existence of permafrost. At present, the remote sensing information is seldom used in the simulation of permafrost distribution in Qinghai Tibet Plateau, and most of the models are based on the GIS using DEM as the main data source. But for the stability of DEM, the model has no dynamic monitoring ability and cannot reflect the change law of permafrost. The purpose of the research is to try to introduce the remote sensing technology into the study on permafrost distribution in Qinghai-Tibet Plateau, and establish permafrost distribution model suitable for Qinghai-Tibet Plateau, which can reflect the macroscopic, dynamic and convenient advantage of remote sensing in the research of permafrost distribution. Kun Wang*, Lichun Chen, Bin Wei and Le Wang, Jilin Communications Polytechnic *Corresponding author: hfxwk@163.com 1293

2 SELECTION OF MODEL FACTORS It is unrealistic to consider all the factors that affect the permafrost distribution in the model. So we need to select the main factors to build the model. Comprehensively considering the main factors that affect the permafrost distribution in Qinghai Tibet Plateau and environment information provided by RS, we selected land-surface temperature, elevation, equivalent latitude, soil moisture, albedo and NDVI as the model factors. PREPARATION FOR MODEL DATA Permafrost is the product of heat exchange between atmosphere and ground through land surface, which is restricted by climatic conditions and regional geological conditions. Under natural conditions, geological geography, vegetation cover and hydrological conditions of different regions are very different, which causes permafrost may be different in the same climatic conditions [1]. In order to reveal the variation characteristics of permafrost under normal natural conditions, selected the average state parameters of surface variables in model calculations. MODIS is one of the main remote sensing instruments in EOS series satellite, considering its characteristics of free distribution, high time resolution and covering a wide range, it is very suitable for the study on permafrost distribution on the Tibetan Plateau. This is mainly because the study area of permafrost is widely distributed, and permafrost is very sensitive to the daily temperature change. If using high resolution remote sensing image for a wide range permafrost study, not only costly, but also due to the number of requirement image is excessive, image time is difficult to reach agreement. Thus the surface temperature, albedo and NDVI are all from MODIS data products (that is MOD11A2, MCD43B3 and MOD13A3). DEM data is generated from the contour. The equivalent latitude is calculated based on the DEM. Soil moisture data is the daily soil moisture product from AMSR-E data which is downloaded from the US National Snow and Ice Center (NSIDC). The annual average ground temperature data of permafrost used in the model is mainly from 1:4 million Chinese permafrost and desert map, partly is from the published monographs and papers [2, 3], a total of 51 ground temperature observations. Because the field exploration of permafrost is currently focused on the Qinghai-Tibet highway on the Tibetan Plateau, permafrost data in other regions is very little, so the annual average temperature data of permafrost is mainly distributed along the Qinghai-Tibet highway. MULTIVARIATE ANALYSIS MODEL Multivariate analysis model is based on the Logistic model to do statistical analysis. Logistic regression model actually is the promotion of multiple linear regression models, but its dependent variable must be a binary variable, that is only two values 0 and 1. In the model established in this paper, the value 0 and 1 respectively represents absence and presence of permafrost. 1294

3 CORRELATION ANALYSIS In the paper used elevation, surface temperature, NDVI, soil moisture, equivalent latitude and surface albedo to build model. While the number of existing sample points that can be used to determine the presence of permafrost is only 51, it is far short of the modeling requirements. In order to obtain a sufficient number of sample points, collected sample points every 0.01 degree along the Qinghai-Tibet highway, whose ground temperatures are calculated from annual average ground temperature model. The purpose of calculating the ground temperature is mainly used to determine the existence of permafrost. Used ground temperature 0.5 as the boundary, less than 0.5 represents presence of permafrost, the value is 1. Otherwise there is no permafrost, the value is 0. Thus we established a binary variable, which is as the dependent variable of Logistic model, and the surface variables obtained from RS data are as independent variables. In the paper we selected 291sample points, 208 points located in the permafrost zone, while the other 83 points located in the non-distribution permafrost zone. So the maximum number variables that can be analyzed in the model are 8. Before the establishment of model, firstly to have a correlation analysis between the permafrost distribution and various surface variables. Since the variables used for correlation analysis have qualitative data (that is presence or absence of permafrost), so we used the Spearman rank correlation analysis. Through the analysis, when the significance level is 0.01, elevation, surface temperature, soil moisture, albedo and NDVI are significantly associated with the permafrost distribution. The NDVI and elevation are the most significant. BUILDING MODEL After meeting the basic requirement of Logistic modeling, regression analysis was performed in the SPSS The cut-off point of model prediction is at 0.5, the model will predict the variables based on this cut-off point. After setting, established the Logistic regression equation as follows: P 1/1 exp( 0.009x x x x ) (1) Where P is the probability of presence of permafrost x1, x2, x3, x 4 respectively represents the elevation, soil moisture, NDVI and albedo. SIMULATION RESULTS Built the model in the Modeler module of ERDAS, and calculated the probability of the presence of permafrost using equation (1), then obtained the results along the Qinghai-Tibet highway. Respectively selected the permafrost existence probability P 0.5, P 0.6, P 0.7, P 0.8, P 0.9, and simulated the permafrost distribution along the Qinghai-Tibet highway, then obtained the simulation results of corresponding probabilities. With the P value increasing, simulation results haven t changed much at both ends of the highway. Only the difference in the middle of highway is growing. Compared the permafrost distribution area with the permafrost map along the highway, when P 0.8, the simulation results are closest to the map, but in the middle of the highway the difference is relatively large (Fig.1). Overlaying the result with the geographical base map, it can be seen that melt zones are mostly near the 1295

4 area of large river, such as Chumaer river and section from Tongtian river to Tuotuo river, where are perennial rivers and have large flow. This is consistent with the published literatures that described there are melt zones in the riverbed of Chumaer river, Tongtian river and Tuotuo river [4]. Therefore it is reasonable that there are melt zones in the simulation results, which is consistent with the actual distribution of permafrost. So we took the simulation result obtained when P>0.8 as the final result of Multivariate analysis model. Figure 1. Simulation result of permafrost (P>0.8). Figure 2. Overlaying simulation result with permafrost map. OTHER PERMAFROST SPACE MODELS Recently a number of permafrost space models have been developed based on Qinghai-Tibet Plateau, such as MAGT model [5], ground temperature inversion model based on RS/GIS [6, 9, 10], elevation model [7, 8], frost number model [1, 8] and TTOP model [1, 3] and so on. In order to evaluate the advantages and disadvantages of each model, to assess the study results, based on existing conditions, selected ground temperature inversion model based on RS/GIS, elevation model and frost number model to simulate the permafrost along the Qinghai-Tibet highway, then compared the results with the model built in this paper. VERIFICATION AND ANALYSIS OF SIMULATION RESULTS VERIFICATION OF SIMULATION RESULTS Simulation results mentioned above represents the distribution of permafrost in the plane, so they are verified mainly from the two parameters: distribution area and distribution range. 1296

5 1 DISTRIBUTION AREA Seen from the area statistics table, the simulation result of Multivariate analysis model is the closest to the permafrost map, whose relative error is 1.98%, ground temperature inversion model is 16.7% and elevation model is 17.6%. Table 1. Comparison of simulation results and permafrost map (unit: 10 4 km 2 ). Permafrost area Non-permafrost area Relative error Permafrost map Ground temperature inversion model % Multivariate analysis model % elevation model % frost number model % 2 DISTRIBUTION RANGE Overlaid the simulation results of four models with the vector of permafrost distribution, it can be seen that at the northern end of the highway, the simulation results and permafrost range are particular good fit, there is no big difference. Big difference is mainly concentrated in the island permafrost regions in the southern end of the highway. Simulation results of ground temperature inversion model and elevation model in the island permafrost region are larger than the permafrost map, while frost number model is significantly less than the map. The result of multivariate analysis fits the best with the map in the southernmost tip of the highway, although there are still differences in some places. In the simulation results of ground temperature inversion model and multivariate analysis model, there are some melt zones in the middle of the highway, and the multivariate analysis model is more obvious. ANALYSIS OF SIMULATION RESULTS Melt zones have been well reflected in the simulation results, which are mostly near the area of large river, such as Chumaer river and section from Tongtian river to Tuotuo river. The simulation result is consistent with the actual distribution of permafrost. This is due to the model established in the paper, which is based on the surface variables obtained from RS. Remote sensing information is a natural surface condition, which objectively reflects the real, so the model is more reasonable. Elevation model and frost number model only considers a single factor that affects the distribution of permafrost, it can t reflect the real situation of the surface, so the melt zones can t be well represented. CONCLUSIONS 1. Multivariate analysis model mainly used the statistical relationship of surface variables to infer the presence of permafrost indirectly. For the analysis sample points are collected along the highway, and in the island permafrost is also dotted with a number of sample points. So the sample points are representative. As long as 1297

6 the multivariable analysis model includes a reasonable choice of sample points, it can have a good simulation of the study area. 2. Compared with the previous models those only consider a single factor of elevation and temperature, model established in this paper comprehensively considers the factors that affect the permafrost distribution in Qinghai-Tibet Plateau. And they can be obtained quickly and accurately from remote sensing data, which will make up for the lack of permafrost data in Qinghai-Tibet Plateau, and it will be convenient to promote the model. Melt zones have been well reflected in the simulation results, which is more consistent with the actual distribution of permafrost. 3. Used statistical methods to build model, and selected model factors based on the factors that affect the permafrost distribution in Qinghai-Tibet Plateau, so model established in this paper can only be used in Qinghai-Tibet Plateau, which is not suitable for the other areas. 4. Factors considered in the model are some natural factors. But along the Qinghai-Tibet Plateau highway, engineering factors in some local areas also play an important role. Due to the lack of relevant data, at least it is difficult to consider them now. REFERENCES [1]. Z.T. Nan, S.X. Li, G.D. Cheng. Prediction of permafrost change in the next 50 and 100 year. Science in China (Series D). 34(6): (2004). [2]. Y.W. Zhou, D.X. Guo, G.Q. Qiu. Frozen soil in China. Beijing: Science Press, 1-2(2000). [3]. Wu Q. B. Adaptability study on permafrost environment change and engineering under human engineering activities. Lanzhou: Chinese academy and science institute of cold and arid environment and engineering, 20-24(2000). [4]. G.Q. Qiu, D.X. Guo. Discussion of melt zones along Qinghai-Tibet highway [C]. Research papers of Qinghai-Tibet Plateau permafrost. Beijing: Science Press, 30-37(1983). [5]. Z.T. Nan, S.X. Li, Y.Z. Liu. Permafrost distribution mapping and application based on mean annual ground temperature. Glacier permafrost, 24(2): (2002). [6]. K. Wang. Study on permafrost distribution based on RS/GIS. Changchun: Jilin University, 47-52(2009). [7]. X. Li, G.D. Cheng. High altitude permafrost in response to global change model. Science in China (Series D), 29(2): (1999). [8]. X. Li. Cryosphere information system and its application. Lanzhou Institute of Glaciology and Cryopedology. Chinese Academy of Sciences. Lanzhou, 65-78(1998). [9]. K. Wang, Q.G. Jiang. Study on permafrost distribution in Qinghai-Tibet Plateau based on MODIS data [J]. International Conference on Remote Sensing, Environment and Transportation Engineering, (RSETE 2013). [10]. K. Wang, L.C. Chen, B. Wei. Study on permafrost distribution in Qinghai-Tibet highway based on ASTER data. Hydraulic Engineering: