Based on the comprehensive analysis of the research and application at home and abroad, the theoretical principles of regional geological disaster early warning and forecasting based on meteorological factors can be preliminarily divided into three categories, namely implicit statistical forecasting method, explicit statistical forecasting method and dynamic forecasting method.
4.2. 1 implicit statistical forecasting method
Implicit statistical forecasting method implies the role of geological environment factors in rainfall parameters, and only rainfall parameters are considered in the early warning criterion of a certain area to establish a model. Implicit statistical prediction method can be called the first generation prediction method, which is more suitable for small areas with simple geological environment model. Because this method only involves one or one kind of parameters, it can be used regardless of the research level in the field of early warning, so it is widely used at home and abroad and is the easiest method to popularize. This method is especially suitable for areas with limited space and little change in geological environment conditions, such as Hongkong, China, where granite and its weathered residues are mainly distributed. This method has been studied, applied and deepened for many years.
The rainfall parameters considered in this method include annual rainfall, quarterly rainfall, monthly rainfall, multi-day rainfall, daily rainfall, hourly rainfall and 10 minute rainfall. In practical application, only 1 ~ 3 parameters are generally involved as forecast standards, such as critical rainfall, rainfall intensity, effective rainfall or equivalent rainfall.
The rainfall criterion early warning method for critical process of sudden geological disasters grasps the key aspects of geological disasters induced by meteorological factors, but the early warning accuracy is bound to be limited by the size of early warning area, the number of samples of sudden geological events, the complex stability of geological environment and the state of regional social activities, and the representativeness of a single critical rainfall index as an early warning criterion is limited.
Representative research results mainly include:
Onodera et al. (1974) studied a large number of landslides in Japan, and put forward that the cumulative rainfall exceeds 150 ~ 200mm, or the rainfall intensity per hour exceeds 20 ~30mm. Nilsen et al. (1976) found that when the accumulated rainfall exceeds180mm, landslides will occur frequently in Alameda and California. Oberste-lehn( 1976) thinks that when the accumulated rainfall reaches about 250mm, landslides will occur in San Benito, California, USA. Guidicini and Iwasa( 1977) analyzed the landslide records and rainfall data in nine regions of Brazil, and reached the conclusion that landslides will occur when the rainfall exceeds 8% ~ 17% of the annual average rainfall. If it exceeds 20%, a catastrophic landslide will occur. Kane (1980) comprehensively summarized the existing data in the world, and gave the relationship between rainfall intensity, duration and landslides in different regions. Of course, this relationship cannot be applied to all parts of the world (crozet proved this in 1997), and it is still a milestone to explore the critical rainfall value of landslides.
According to the research of Brand et al. (1984) in Hongkong, China, most landslides are induced by local high-intensity short-term rainfall, and the previous rainfall is not the main factor, unless it is a small landslide. Ng Heshi (1998) thinks that continuous rainfall is also an important factor to induce landslides. If it is predicted that the rainfall in China Mainland and Hongkong will reach 175mm within 24 hours or the rainfall in the urban area will exceed 70mm within 60 minutes, it is considered that the landslide prediction threshold has been reached and the government will issue an announcement. On average, China and Hong Kong issue early warning of flash floods and landslides three times a year.
Ganuti et al. (1985) put forward the concept of critical rainfall coefficient (CPC), and concluded that when CPC >0.5, landslides with return period of 10a would occur. When CPC >0.6, there will be a landslide with a return period of 20 years.
Glade( 1997) established three models to determine the critical rainfall value of landslides, and verified them in Wellington area in the south of North Island, New Zealand. The basic data needed by the three models are: daily rainfall, landslide occurrence date and potential daily evaporation of soil (calculated by Thornthwaite method). Glade( 1997) model of rainfall intensity critical value 1- daily rainfall model, simply analyze the daily rainfall (Glade, 1998) with or without landslide, and get the minimum critical value and the maximum critical value, that is, no landslide will occur below the minimum critical value; If the maximum critical value is exceeded, landslides will inevitably occur. The rainfall grade is divided into 20mm as a grade; Glade( 1997) model 2 is a daily rainfall model without detection, which considers the influence of previous rainfall. He believes that there are two main factors that determine the early stage: the duration of early rainfall and the speed of soil moisture reduction; Glade( 1997) model 3, the critical value of soil moisture state-the first soil moisture state model, he thinks that in addition to the previous rainfall, soil moisture content and potential evaporation have great influence on landslides.
Liu Chuan is presiding over the national meteorological early warning of geological disasters in May 2003. Based on the rainfall before geological disasters 15d, the critical process rainfall in landslide and debris flow areas is established, and the early warning criterion model diagram is created. Combined with specific areas (28 areas in 2003 and 74 areas after 2004), the correction method is adopted. This method adapts to the requirements of three-level forecast, and defines the α line and β line as the early warning grade boundary. The release, inspection and application of flood warning results for more than three years have proved that this method has scientific basis. However, due to the large area of early warning, too few basic data and statistical samples of geological disasters, the accuracy needs to be improved, which fully demonstrates the urgency of carrying out the research on geological disaster data integration.
In addition, Chengdu Institute of Mountain Disaster and Environment of Chinese Academy of Sciences and other institutions have made unremitting research on monitoring and early warning modeling of single debris flow for many years, and achieved representative results.
4.2.2 Explicit statistical prediction method
Explicit statistical prediction method is a method to establish an early warning criterion model considering the change of geological environment and the superposition of rainfall parameters, which is transformed from geological disaster risk zoning and spatial prediction (Carrara A.,1983; Chunshan. & Kawakami. , 1984; Bezak &CorominasJ。 , 1996; Carrara. ，CardinaliM。 &GuzzettiF。 , 199 1; Liu, 2004; Yin Kunlong, 2005).
Risk assessment and risk zoning of regional geological disasters are still the mainstream of current research, and using them for real-time early warning and release of geological disasters is mostly in the exploratory stage. This method can fully reflect the changes of geological environment elements in the early warning area, and with the improvement of investigation and research accuracy, the spatial early warning accuracy of geological disasters can be improved. Explicit statistical prediction method can be called the second generation prediction method, which is being explored and is more suitable for large areas with complex geological environment models.
The theory and method of space-time early warning of sudden geological disasters based on spatial analysis of geological environment are synthesized according to the results of unit analysis, which overcomes the limitation of relying only on a single critical rainfall index. However, there are still many problems in the expression of critical inducing factors, the selection of early warning indicators and quantitative classification that need further study.
Therefore, in order to realize the early warning of regional sudden geological disasters in a completely scientific sense, it is necessary to establish a theoretical method of coupling model of critical process rainfall criterion and spatial analysis of geological environment-generalized explicit statistical model geological disaster prediction method, and the early warning grade index (W) is an internal and external dynamic simultaneous equations. that is
Regional early warning method of geological disasters in China and its application
Where: W is the early warning grade index; A is the gravitational force of extraterrestrial bodies, including the tidal force of the sun and the moon, and the influence of sunspots, surface flares and solar wind on the earth's surface, a=f(a 1, a2, …, an); B is the dynamic action within the earth, mainly manifested in fault activity, earthquake and volcanic eruption, and b=f(b 1, b2, …, bn); C is the dynamic action outside the earth's surface, including rainfall, seepage, erosion, weathering, plant root cracking, storm, temperature, drying and freezing and thawing. ，c=f(c 1，c2，…，cn)； D is the influence of human social engineering and economic activities, including the geological environment changes caused by resources, energy development and engineering construction, and d=f(d 1, d2, …, dn).
In 1970s, taking the landslide sensitivity map of san mateo County, San Francisco, California as a representative, the regional landslide disaster prediction map was obtained by weighted (or unweighted) superposition of multi-parameter maps.
In the 1980s, carrara. (1983) The multivariate statistical analysis and prediction methods are introduced into the regional landslide spatial prediction, which has been rapidly developed and popularized worldwide. For example, Haruyama H. & Kawakami H. (1984) used mathematical statistics theory to evaluate the risk of landslide disaster caused by rainfall in Japan's volcanic active areas. Bates communicates. & Corominasj。 (1996) Using statistical discriminant analysis model to evaluate the sensitivity of shallow landslide, the correct prediction rate of slope instability reached 96.4%, which strongly demonstrated the applicability of statistical prediction. Cardinal carrara. & Guzzetif。 Et al. (199 1) combined the statistical model with GIS and applied it to the landslide risk assessment of a small watershed in central Italy, realizing the automation from data collection to analysis and management. The results show that the comprehensive application of statistical analysis and GIS is a fast, feasible and low-cost method for regional landslide risk assessment and mapping.
Since the middle and late 1990s, with the rapid development of computer technology and information science, and the joint application of "3S" technologies such as RS, GIS and GPS, it is possible to quickly process massive geological environment data, and there is a new trend in the application research of spatial prediction model and method of geological disasters, which gradually combines the risk assessment and early warning application of geological disasters.
Liu et al. (2004) created and published the theory and method of spatio-temporal progressive analysis of regional geological disaster evaluation and early warning, which is referred to as "four-degree progressive analysis method" (AMFP), and it has been applied in the Three Gorges reservoir area (54 175km2) and Ya 'an experimental area of geological disaster early warning in Sichuan.
Li Changjiang et al. (2004) combined GIS with ANN (Artificial Neural Network), and considered different geological, geomorphological and hydrogeological backgrounds, established a regional LAPS landslide probability prediction (alarm) system with given rainfall in Zhejiang Province.
Song Guangqi et al. (2004) established a GIS-based geological disaster prediction system for given rainfall in Sichuan Province according to the probability distribution of topography, lithology and geological structure.
Yin Kunlong et al. (2005) discussed the early warning and prediction of sudden geological disasters based on WebGIS, taking Zhejiang Province as an example.
At present, China's early warning and exploration work is in the forefront of the world, because the China government has implemented a regional geological disaster early warning and forecasting mechanism nationwide.
4.2.3 Dynamic prediction method
Dynamic prediction method is a method of establishing early warning criterion equation considering the dynamic change process of soil mass under the coupling effect of ground and atmosphere during rainfall, which is essentially an analytical method. The prediction result of dynamic prediction method is deterministic, which can be called the third generation prediction method. At present, it is only applicable to a single experimental area or a particularly important local area. Based on the transformation mechanism of rainfall infiltration in slope before, during and after rainfall, this method describes the corresponding relationship between the dynamic change of slope groundwater and the whole process state and stability of slope. By monitoring groundwater level dynamics, pore water pressure and slope stress-displacement, the real-time dynamic response of groundwater in the slope before, during and after rainfall is revealed, and the relationship between the physical properties of the whole slope and its deformation and failure process is revealed. Under the conditions of fully considering water content, matrix suction, pore water pressure, seepage water pressure, formation of saturated zone and transformation of landslide and debris flow, the mathematical and physical equations are selected to study and analyze the relationship between the change law of groundwater dynamic field and slope stability, determine the multi-parameter early warning threshold, and realize real-time dynamic prediction of geological disasters.
At present, this method is limited to the research and exploration stage of the test site or single slope, and it must rely on the three-dimensional monitoring network (ground-air coupling) with real-time monitoring, real-time transmission and real-time data processing functions to realize real-time prediction. Because of the high demand for theory, technology and funds, this method is more suitable for research monitoring and early warning in important small areas or units.
According to the research, when the six-hour rainfall in San Francisco Bay Area reaches 4 inches (10 1.6 mm), a large area of debris flow may be triggered. In order to monitor the change of groundwater pressure during rainfall, the researchers set up several pore water pressure gauges to observe the change of groundwater pressure in the slope. The real-time regional landslide early warning system in San Francisco Bay Area includes the experience and analysis relationship between rainfall and landslide, real-time rainfall monitoring data, rainfall forecast of National Meteorological Service Center and sketch of landslide-prone areas.
In China, Liu et al. (2004) monitored the layered response of atmospheric precipitation and the change of water content in slope rock and soil in Ya 'an regional geological disaster monitoring and early warning experimental zone in Sichuan, and found that the water content in slope rock and soil changed obviously under different rainfall processes and rainfall intensities, which can be used to study the seepage process of rainfall in slope rock and soil until landslides and mudslides occur.
August 23-25, 2003 is a typical rainfall process that caused various geological disasters and caused casualties, which can be used as an analysis example. With the water content of 15 on August 9th as the background value, the rainfall processes on August 23rd, 24th and 25th correspond to the water content of 96, 120 and 144h respectively. The recording curves of four horizons clearly reflect that the water content of slope rock and soil increases sharply with the increase of accumulated rainfall, and the first and second horizons reach the state of supersaturation, and the water content increases sharply. The peak value of water content in each layer appears at 15 1h, which is close to the regional landslide outbreak time (0: 00 on 26th, corresponding to 153h). Groundwater seepage along the fault is not considered in this analysis (Figure 4. 1).
Fig. 4. 1 sangshupo monitoring point in ya' an, Sichuan1curve of water content in layer ~ 4 with time.
Through the analysis and comparison of implicit statistical prediction method, explicit statistical prediction method and dynamic prediction method, we think that the future direction is to explore the joint application method of implicit statistics, explicit statistics and dynamic early warning model of geological disasters to meet the early warning needs of geological disasters at different levels. The research contents include the statistical model of critical rainfall, the statistical model of superposition of geological environment factors and the mathematical and physical model of real-time change of geological bodies (hydrodynamic force, stress, strain, thermal field, geomagnetic field, etc.). ), etc. The joint application of the three models not only adapts to particularly important regions or small watersheds, but also provides decision-making basis for dynamic early warning and emergency response of a single geological disaster.