Prediction Model Of Debris Flow Based On Bayes Discriminant Analysis In Miyun County

Aiming at the frequent occurring of debris flow in the upper stream of Miyun Reservoir, which is located in Miyun county, Beijing city, according this research can set up a forecasting model of debris flow, which provides a convenient and feasible method for the instantaneous forecast and supports for disaster of debris flow control, which can most decrease the harm of debris flow.According to investigation of typical debris flow valley system elements and laboratory analysis, it is showed that physiognomy and landform are prone to happening of debris flow in this area; the frequency of modern debris flow happening is increase gradually, and the characteristics of it including high frequency of regional outbreaks,low frequency of single gully,large scale outbreak,activities short and serious impact on water resources; the happening of debris flow consistent with precipitation distribution, which is mostly happened at July and August.Using the system science and debris flow prediction theory, the external environmental conditions of debris flow system was considered only. Accumulated precipitation of preceding 15 days (x₁) and intraday precipitation (x₂) are introduced as model indexes based on the Principal Component Analysisabout precipitation in research area.The debris flow forecasting model of all valley and three different drainage area prediction models (three kinds of watershed with drainage area from 0.30 to 2.00, from 2.00 to 5.00, and from 5.00 to 15.42 respectively) were established based on Bayes discriminant analysis. The general formula as follows,If f₁ > f₂ , the debris flow may happen under certain precipitation, and if f₁ < f₂ , the debrisflow will not happen.According to self-validation and cross-validation, it was showed that the accuracy of forecast models exceed 82.4% , which can be used for debris flow forecast in the study area.Then a prediction system was established, which was based on computer language of C++. The study area can achieve forecast of debris flow by the value of the observed rainfall input into the prediction system.