| Locust plague happens frequently in every continent except Antarctica. Only in China, it happens over 800 times since Zhou Dynasty till now. Locust is a kind of primary pests that cause severe damage to the agriculture in Xinjiang, northwest of China. The authority agency, Locust and Rodent Control Headquarters of Xinjiang, has adopted many actions in order to control and prevent the locust outbreak. However, early forecasting probable sites of locust outbreak is significant for rangeland management and agricultural protection. The forecasting does not only focus on the amount or time, but also does the spatial location of locust outbreak. Recently, “3S” technology improved the development of locust forecasting.This study aimed to predict the possible area where locust might outbreak combining with multi-criteria analysis and Markov Chain methods. Two model would be promoted: one was locust fitting model based on MCA method, which would be used to generate locust fitting map in Xinjiang from 2001 to 2013. The other one was locust forecasting model based on Markov Chain method. This model would be used to predict the locust outbreak situation from 2003 and 2014. Then the results would be compared with the real locust outbreak maps in order to verify the accuracy. In general, the main contents and conclusion in this study was shown as following:a. 6 factors including monthly average temperature, monthly relative humidity, NDVI, elevation, slope, and soil PH value have been selected from climate, vegetation, soil and topography four aspects to be the basic factors for building locust fitting model based on MCA method. Two weighting methods, AHP and average empowering weight, were tested to assign the different distribution into factors when built the model. The results showed that the locust was mainly distributed in the north and west part of Xinjiang which was highly consistentwith the actual locust distribution. In some severe locust plague areas, the average accuracy was 84.37% and the highest accuracy that appeared in Urumqi which reached to 97.25%. On the other hand, the accuracy was unsatisfied in some parts of less severe locust plague areas. Besides, the average empowering weight method was more suitable for this study as the accuracies are both over 90% in 2011 and 2012 than that using AHP method. Therefore, 13 locust fitting maps from 2001 to 2013 were generated based on the first model with average empowering weight method.b. A series of Markov indicators named the average state transition probability matrixes were calculated between every two adjacent years, which would be used to forecast the locust outbreak situation from 2003 to 2014. The predicting results showed: it was effective to predict locust outbreak extent by using the second model. The effects were particularly well in those years when locust outbreak situation haven’t changed dramatically. There were 9 out of 11 years, the accuracy were over 90% and it reached to 99% in 2006 and 2009. Even though the accuracy was a bit lower in 2007, it still reached to 76.96%. However, the prediction effects were very bad in those years that locust outbreak areas changed suddenly such as in year 2012 and 2013. Moreover, locust outbreak situation was affected somehow by that in the previous years. |
PDF Full Text Request