广东近40年来5～9用气候变化及与主要农作物关系

本文以广东省４７个测站近４０年来６～９月降水量、平均气温、总日照时数和一些主要农作物逐年单产量资料为基础，用主分量分析、周期回归等方法，分析广东全省性的气候时空变化特征；分析一些主要农作物逐年单产的变化规律．用多元逐步回归分析方法探讨了广东全省性气候变化与同期及后期（１～３年）主要农作物单产量间的关系并作了预报．

IHE CHANGE OF THE CLIMATE FORM MAY TO SEPTEMBER FOR THE YEARS 1954-1990 WITH RELATION TO THE MAIN CROP YIELDS IN GUANGDONG

limatic data from May to Septemfor for the years 1954 - 1990 in 47 stations over Guangdong Provi8nce are used.in an empirical orthogonal tunctions (also referred to eigenvectors or principal componentS) analysis of the change patterns of temperafure, precipitation .and sunshine- duration. The first eigenvector of each variable describes the same phase change pattern; the others describe gradients in some directions. The most prominent features in the amplitude of the first eigenvector of each varable are periodic changes, especially. 5, 8 years intemperatore, and 11, 18 years in precipitation; but no evident signals of linear trends.The characters in auntal yields (per mu) of three crops (rice,peanut, sugar cane ) forthe years 1954- 1990 over the whole province are: their trenbant yields (mainly controlled byman-made factors) are all nonlinear, but approximately, rice yield may be taken as an linear one and their climate yields (possibiely effected by Climate factors) have no periodic signals. The analyses, by the ways of cross correlations and Stepwise multiple repression analysis the relationship between the amplitudes of eigenvectors of ail climate variables and the main crop yields in four Cases, show that: the first mode of precipitation is the most important factor effecting the crop yields, besides, for rice, sunshine factors are the introduced into the togression equations, so are temperafure factors for other crops. The climate factors in the prediction equations are very different from those of the simultaneous equations. But the predictands of all crop yields for 1, 2 and 3 years lag are basically effective, and surprisingly the multiple correlation coefficients in the prediction equations of multiple regression analysis are even botter than those in the simultaneous equations.