Temporal And Spatial Evolution Of Rice Phenology In The Three Gorges Reservoir Area And Its Surrounding Areas Under The Data Mining Algorithm

Phenology is the direct indicator of changes in environmental conditions.Its evolution can reflect the characteristics of water ecosystem,carbon and nitrogen cycle,and it is of important theoretical and practical significance to study the climate change.Rice is the most important food crops in China.So it is important agricultural decisionmaking reference data and theoretical basis to master rice phenological evolution process and law accurately and timely,and that is good for the estimation of crop yield,agricultural forecast and monitoring crop growth.In this study,phenological spatio and temporal processes of one season rice during main growth period(sowing,seedling stage,tillering stage and maturity)and important growth stages(growth season and tillermaturity),as well as their impact factors in the Three Gorges Reservoir area and the surrounding areas,were examined.Data obtained from 27 monitoring sites for long term agricultural phenology observation between year 1991 to 2010 were used.The methods applied in this study included multivariate regression algorithm,K-means clustering algorithm and support vector machine model data mining analysis.This study will give supports for further investigations of agricultural and ecological effects triggered by the Three Gorges Project,and decision-making on agricultural production in the area.The main conclusions are as follows:(1)The method of time scale linear regression and spatial K-means clustering are used for observing 27 stations.The sowing data of reservoir areas and its western region is in advance averagely 1.33 days / decade(d/dec);western region has more significant trend than eastern region;the emergence period is not changing obviously;most of the sites at tillering stage are ahead of the average 1.45d/dec,and these sites mainly distributed in the reservoir region and adjacent to the reservoir area;85% of the site’s mature period presents the trend of delay in average of 2.37d/dec.(2)The earlier sowing period and more delayed mature period lead to the extension of growth duration in most sites.The growth duration of reservoir(Jiangjin station,Fengdu station and Wanzhou station)and its northern region is advanced obviously.Both sides of the reservoir area are not changing and essentially the same.Southern region of the growth period has a shorten trend.The clustering results of period between tillering and mature were basically consistent with the law of growth season.(3)The temperature and annual precipitation in the study area showed an upward trend,while the sunshine hours decreased.And the climate change trend of the growing season was the same as that of the whole year.Through multiple regression analysis and the support vector machine model,the results showed that influence of the climate on the season rice phenology is not obvious.So,temperature,precipitation and sunshine hours may not the main factors influencing the growing duration of one season rice in this area.