| Landslide is a common natural geological disaster in the mountainous area in our country, which is causing increasingly serious harm due to the constant development of economy. Our country today, implementing the western development strategy and especially focusing on the development of basic construction (such as water conservancy, road and railway, resettlement projects, etc.) and ecological and environmental protection; in 2008, in the face of world financial crisis, our country decided to invest 4 trillion Yuan to restructure the domestic economy, most of which will be invested in basic construction. By then, the engineering construction is going to bring enormous economic benefit to our human society, but at the same time if we do not pay attention to the protection of natural condition, as well as the positive prevention for various disasters, it will certainly bring further deterioration of the environment as well as unnecessary economic loss.For various types of geological disaster, landslide is one of the most dangerous and damage causing type. However, because of the complexity of landslide itself, in many cases, it is very difficult sometimes even impossible to predict the exact happening time for a landslide; meanwhile it is also unfeasible for a large-scale landslide mitigation. Therefore, the research trend for landslide prevention and mitigation will be the prediction of possibility for unstable slopes which would become landslides in specific region and condition, as well as the prediction of spatial distribution law for these unstable slopes, so as to help the government for land-use planning and reducing the unnecessary loss of life and property.However, many spatial prediction models for slope disaster are in fact belong to "type of post-confirmation", that is, predicting the distribution of landslides which have already occurred in the research area. But in normal situation, because of the slope deformation after sliding, the accumulation of the slope's internal stress has been released and its position has also been lowered, so the slope will be in a temporary stable state. But the real threat to life and property is the potential unstable slope which has all the objective conditions for sliding but still has not slid, and would become landslide in the future under continued external force. Therefore, for this kind of potential unstable slopes, how to select a reasonable evaluation index system and a prediction model, which can help better predicting the distribution condition, rather than a simple distribution prediction for post-landslides, is the main ideas and innovation of the thesis.The thesis is based on the Project named "Formation Mechanism and Risk Assessment of landslide in Enshi region in western Hubei" from China Geological Survey. The thesis selected some Silurian strata regions in Enshi City as the study area, by systemic survey and monitoring to collect the engineering geological and environmental geological condition, so as to identify geologic background of the region and key influencing factors for the landslide which have already occurred on Silurian stratum in Enshi area. After figuring out the evaluation factor index system by deep research about the mechanism of typical landslides which occurred on Silurian stratum; then based on application of GIS technology, combined with intelligent artificial neural network prediction model, the thesis set up a spatial prediction model for potentially unstable slopes from the region. So as to approach a technical route and a systemic method which based on GIS and aiming to evaluate slope disaster on Silurian stratum in mountainous area of western Hubei. According to the research, the following results have been acquired:(1) In Enshi City area, according to the statistics, 24.6 percent quantity of disaster occurred on Silurian stratum, that is, a quarter of the disaster has happened on Silurian Stratum. And more than 80% of these disasters are small and medium-scale soil landslides; the type of the failure is very consistent, so it is very much in favor of intelligent system to learn and remember failure samples. So according to the principle of typicality, representativeness and the extent of data acquisition, Silurian stratum in Tunbao county has been selected as a typical study area. Data collections field investigation, remote sensing data interpretation have been carried out for collecting relevant data, meanwhile digitization work for thematic maps and establishment of spatial database have also been carried out, so as to provide a reliable basic information for the coming spatial prediction analysis.(2) The precondition for intelligent prediction of slope disaster is high similarity between disaster samples and disasters which are supposed to be predicted. For this thesis, the selection of typical landslides has a very important precondition that the chosen landslides should also occurred in Silurian stratum and not far away from the prediction area, so as to ensure the similarity of the geological condition. Therefore, Baozha landslide and Qujiawan landslide which occurred in Silurian stratum in and around Enshi City region have been selected for research about formation mechanism of landslide. Based on the knowledge of regional geological environment, as well as the basic geological feature of landslide, a qualitative and quantitative research has been carried out for the material composition, as well as special geological property of bed rock, sliding zone and sliding body of each landslide.Based on that research, the thesis summed up the deformation and failure characteristic of Silurian stratum slope, including material composition, scale of damage, slide thickness, slope gradient for landslide and happening time of landslide, so as to summarize the special law of slope failure which occurred on Silurian stratum. Then more in-depth research on the relationship between six influencing factors like rock and soil property, topography, slope structure, rainfall, geological structure, engineering activities and slope stability has been carried out, including mechanism about the influence on slope stability by each single-factor, as well as the weight of each factor when they affect together. Consequently, the mechanism that how slopes on Silurian stratum evolve into landslides as well as the main factors that influence the slope stability has been figured out.(3) Set up a targeted evaluation index system. The complexity and difficulty of the slope disaster prediction is that it is a complex system which consist of numerous and strongly uncertain factors. So select major and controlling factors from all those factors against the Silurian stratum slope failure and give up those less relevant or even not relevant factors, is the key point for setting up a suitable evaluation index system.In the process of factor selection, it is required to consider that, on the one hand, the influence factor in terms of Silurian stratum landslide mechanism, that is, whether the factor is relating to slope failure; On the other hand, the influence factor in terms of using factors for spatial prediction of slope disaster, that is, whether the factor has enough "influence" to the disaster, not just whether or not related, as well as whether it is possible to collect large-scale regional spatial data and digitize all the layers. Based on these two considerations, the thesis further analyzed the selected seven factors one by one and finally selected slope, slope structure, rainfall, engineering activities as the elements of evaluation index system for spatial prediction of Silurian stratum landslide.The rainfall factor is a temporal variable, but the prediction of the thesis is for spatial distribution of disaster, that is to say, indicators should have feature of spatial distribution, and will not changed by time. So the thesis introduced a concept namely "catchment area": that is, after the occurrence of rainfall, due to the difference of slope landform, slopes have different capacity of accumulating surface water which comes from the rainfall. Based on the concept "catchment area", the temporal rainfall variable can be translated into capacity of accumulating surface water of slopes, and thus associate rainfall as a spatial evaluation indicator with the stability of slopes indirectly. And the calculation result of "catchment area" distribution, not only be able to accurately describe the distribution of normal water system; but also be able to express a very good surface runoff distribution after rainfall, as well as distribution of gully landform where there is no perennial surface water but also plays a crucial role for the slope stability.After all the influencing factors have been selected, then the thesis used different options to quantify these selected factors, such as continuous variable, linear variable and discrete variable. On the one hand, it can prevent "inflated values" of certain variables due to different dimension; on the other hand, it can provide standardized input parameters for the intelligent prediction system for next step.Finally in this chapter, the thesis also discussed weight assignment of influencing factors. There are now two main ideas about analysis for weight of prediction index system: the first one is that, based on landslide hazard map and overlay map of factors, use quantitative and semi-quantitative way to determine the sensitivity of each landslide indicator, and then the weight can be determined; the other way is based on the theory analysis about relationship between influencing factors of landslide hazard and landslide hazard, then use expert marking or rating method to assign weight for each factor. However, the thesis selected an intelligent prediction system, which based on the sample law learning and memory by neural network, as well as a prediction process that simulates the thinking of a human brain. So during the calculation, the prediction model assigns the weight index automatically and dynamically by self-study of samples, which can also effectively avoid the human subjectivity that comes from the export weight assignment.(4) Research on theory and method of intelligent prediction. Because of the complexity of geological environment in western Hubei and non-linear geological disasters, based on consideration of latest development in this field, and more in-depth discussion about the theory and method of prediction, the thesis introduced a non-linear intelligent neural network method which provides a new method for spatial unstable slope prediction GIS system.For training sample selection, based on dominant idea of the thesis, the chosen sample contains both stable and unstable slopes in the study region; so as to let the neural network learn and remember different slope stability state separately. The division of slope stability state is confirmed by field investigation, the main division criteria based on following apparent slope deformation: such as whether the slope has tension and shear cracks on the surface, or around the slope or even in the residential area; whether the cracks are still further developing; whether the groundwater is rich in content according to the investigation of vegetation and surface water distribution, and so on. For the so called "unstable slope", it means the slope that with objective situation for further deformation and damage but has not completely destroyed yet; and in the future, with the passage of time or the outside continuous influence, which might turn into a landslide or a secondary failure on a post-landslide. So the intelligent prediction system can focus on sample study which is one of most important point of the thesis.According to the deep research on structure, study formula, study method and steps of neural network, a neural network model has been designed for the spatial slope disaster prediction. Also according to the research on selection of input layer number, hidden layer number, hidden layer's neuron number, output layer number, weight and threshold value, as well as a special research on selection of training function and neuron number of hidden layer, ultimately the thesis designed a suitable intelligent neural network model for the spatial slope disaster prediction.(5) Integration of intelligent spatial disaster prediction system. First of all, according to field investigation and specific condition in the research area, the thesis processed the thematic maps based on evaluation index system, as well as digitized these thematic maps, and then set up a spatial property database for these various thematic maps which all managed by GIS system. Besides, different influencing factors are managed through separate layers, according to the research on relationship between slope stability and influencing factors, the thesis classified these factor layers by certain laws, and then established the spatial property database for various thematic layers about the research region. Subsequently, based on GIS analysis of digital elevation models, digital terrain analysis, buffer analysis, hydrology analysis and spatial overlay analysis, the process for basic data has been finished. Finally, based on the designed neural network prediction model, the technology route and method system for intelligent spatial slope disaster prediction has been achieved.(6) Apply the intelligent spatial disaster prediction model on the Silurian stratum in Tunbao County in Enshi City region. The prediction result shows the distribution of potential unstable slopes, which are slopes with objective condition for further deformation and damage but have not completely destroyed yet; and in the future, with the passage of time or the outside continuous influence, which might turn into landslides.Subsequently, the thesis compared the result map from remote sensing interpretation with the result map from intelligent spatial prediction system, and finally made a qualitative and quantitative assessment about the accuracy of this intelligent spatial prediction system.
|
PDF Full Text Request