Critical Rainfall Of Debris Flows Based On Rough Sets

Debris flow is of a significant kind of natural disaster that bursts in a very short time with great destructiveness which occurs more frequently in mountainous areas.It is the most important secondary geological disaster after the Wenchuan Earthquake which leads to directly threatens to the security of human lives and their properties and to the infrastructure construction.Although debris flow bursts with very fast speed and current research can not predict its time and magnitude accurately before its occurrence,critical rainfall can be obtained by monitoring the debris-prone area through the comparison to the record of debris flow.If we get the critical rainfall of each district,more effective protects to the lives and properties will be done successfully.However,two common problems are consequently suffered,one is the data missing with bad weather and equipment failure,the other is most of the critical rainfall models are linear models about the rainfall intensity and duration or the rainfall intensity and accumulation rainfall.Not only that,the nonlinear model about rainfall intensity and accumulation rainfall proposed by Chyan-Deng Jan and Ming-Hsi Lee was applied by Chen long and Yin guo-long in the three typical districts of Wenchuan Earthquake with less success than in Taiwan.So,new warning model should be established in order to modify the performance of the debris warning system.Rough set theory proposed by Pawlak,a new ideal set theory different from the classical set theory proposed by Cantor,can analyze any kind of data such as the incomplete data and imprecise data and the analysis and process to the data just depends on the data itself only without any other information.In other words,the rough set approach has the ability to classify data obtained from observation or measurements and reasoning from imprecise data,more specifically,discovering relationships in data.The rough set concept overlaps in many aspects many other mathematical ideas developed to deal with the imprecision and vagueness and proves to be very useful in practice.Based on the analysis above,achieved the following main results and conclusions:(1)An algorithm based on the rough set theory to complete the incomplete data was proposed in this paper and the missing data was filled up with high rate of reinforcement.(2)In this paper,more than ten factors are chosen as the influence factors of debris flow hazard.Attribute reduction based on discernable matrix and attribute significance of the classical rough set theory was presented to find the core attributions of the critical rainfall in detail.To take a step further,the grey correlation analysis method was applied to get the important factors.(3)A new data processing method was proposed to deal with the rainfall parameters.Most debris flows occur in abnormal rainfall process,in other words,the rainfall of specific area has exceeded its normal extent.The daily mean rainfall in rain season was used to normalize the rainfall intensity and the accumulation rainfall and new relevant parameters were got with less variation caused by different patterns of rainfall in different districts.(4)Based on the results of attribute reduction and the data processing,a new warning model was set up with different rainfall intensity combined with 5min,10 min and 60 min rainfall intensity which normalized by the daily mean rainfall in rain season and with the general environment index.(5)This paper presents a series of improved models based on the RTI model with the mean annual precipitation(MAP)and the daily mean rainfall(DMR)normalizing the accumulation rainfall respectively,rainfall intensities especial the 5min,10 min and 60 min rainfall intensity were used respectively.These models were used for the determination of the critical rainfall for the different water shed of the Wenchuan earthquake and the corresponding results were obtained.(6)A method based on rough set theory for rainfall intensity anomaly detection with low overhead and high efficiency was presented and applied to monitor the abnormal behavior of 5min rainfall intensity.The method is capable of extracting a set of detection rules to form a normal behavior model from the record of system.This method,compared with other methods,requires a smaller size of training data set,less efforts to collect training data and more suitable for real-time detection.When accumulation rainfall has achieved to some extent,the rainfall intensity anomaly is more important to debris occurrence.By comparing the application results in the three typical districts of Wenchuan earthquake area and those obtained from geology monitoring,these models are proved to be credible.