基于UCAR公布的电离层F2层最大电子密度数据NmF2,利用人工神经网络技术,构建局部地区NmF2模型。以年积日DOY、当地时LT、经度LON、纬度LAT和F10.7太阳活动指数FLUX为网络输入,以NmF2为网络输出,提供磁平静期NmF2模型值作为参考背景,通过模型值与观测值的比较,发现2008年5月12日汶川7.9级地震前震中附近上空NmF2在震前第6—4天(6—8日)减小约30%,震前第3~2天(9—10日)明显增大约40%。
On the basis of the F2 layer peak electron density (NmF2) from University Corporation for Atmos- pheric Research( UCAR), we constructed a Back Propogation(BP) artificial neural network(ANN) in order to de- tect pre-earthquake anomalies for the first time. The ANN provides NmF2 model value with five parameters:DOY, local time (LT), longitude ( LON), latitude (LAT) and solar activity index of F10.7 (FLUX). We compare the model value with observations during the Wenchuan earthquake. It is found that NmF2 around the forthcoming epi- center decreased remarkably in the afternoon period of day 6 - 4 before the earthquake, but enhanced day 3 - 2 be- fore the earthquake.