Studies On Comparation Of Forecast Methods Of Rodent Population Dynamic

This thesis collected the data that included the species, quantity and reproduction of rodent populations with line trap-day method, and also the meteorological data of Helingeer County in Huhhot suburb"the Ministry of Agriculture Shaerqing Township forage grass resources observations from key field scientific experimental station"(original Huhhot test site of Grassland Research Institute of Chinese Academy of Agricultural Science) and Tuoketuo County Yongshengyu Township from 1984 to 2004. The Markov forecast model, time series method (triple exponential smoothing method and a combined of triple exponential smoothing method and Markov model) and stepwise regression forecast method were choosed, and the basic principles and processes of the three models were analyzed and compared. Based on above, the three methods were used to predict population dynamic of Cricetulus barabansis and Meriones unguiculatus. The characteristic, accuracy and applicability of three methods were analysed and compared. The results showed that, (1) Markov model belongs to a middle-long time forecast model, and the result was interval forecast, but it reached the requirement of forecasting the harmful population dynamic of rodents. Through the forecasted population dynamic of Cricetulus barabansis and Meriones unguiculatus in 2004, the results were accurate; this meant that Markov model was a good forecast model to Cricetulus barabansis and Meriones unguiculatus population. (2) Triple exponential smoothing method and the combined of triple exponential smoothing method and Markov model both belong to time series method. Triple exponential smoothing method strongerly dependenced to short-term data. The combined method needed data which had more accurate; it must be the data of population quantity of many years continuously. The two methods had accurate results in forecasting Cricetulus barabansis, but not in Meriones unguiculatus. It was mainly because Meriones unguiculatus population was affected by environment factors at random, and the dynamic tendency was not obvious. It meant that the two time series methods were accurate for the populations which were little affected by environment factors at random. (3) Stepwise regression forecast method was a common method for the rodent population quantity forecast. This method requested the data which not only need the data of population quantity of many years continuously, and also need factors which influenced population. The forecast result was more accurate in 2004, and had concrete value. In brief, the rodent populations were affected by many factors; therefore, for the forecast of population quantity, it was difficult to find a most effective and most method at present. Therefore, how to found a suitable forecast method should need to study population quantity characteristic and the method continuously.