摘要
在收集资料极为困难的条件下，应用投影寻踪技术成功地定量预报出１９９８年长江三峡年最大洪峰，预见期２个多月：投影寻踪回归（ＰＰＲ）模型预留１０年的检验合格率达９０％，预报值６４５００ｍ３·ｓ－１；投影寻踪细密自回归（ＰＰＤＡＲ）模型预留１０年的检验合格率达８０％，预报值６２０００ｍ３·ｓ－１。用实测值６３６００ｍ３·ｓ－１验证，长期定量预报的误差分别为＋１．４％（ＰＰＲ）和－２．５％（ＰＰＤＡＲ）。
Under the very difficult conditions of the data shorting,we forecast quantitatively annual maximun flood peak flow with two months forecasting lead time of Three Gorges in 1998 successfully.The passing rate of pre reserved ten years data testing is 90 percent and forecast value is 64500m 3·s -1 with Projection Pursuit Regression (PPR) model,80 percent,62000m 3·s -1 with Projection Prusuit Detailed Auto Regression (PPDAR) model.Through verified using observed value with 63600m 3·s -1 ,the long range quantitative forecast errors of PPR and PPDAR model are +1.4% and -2.5% respectively.
出处
《新疆农业大学学报》
CAS
1998年第4期312-315,共4页
Journal of Xinjiang Agricultural University
