摘要
为弥补广阔海面上气象参数的观测数据的不足,利用专用成像传感器SSM/I和红外辐射计AVHRR资料进行近海面气温和湿度的反演,首先分析与近海面气温和湿度关系比较密切的几个气象因子及其相关性,并采用神经网络建立近海面气温和湿度与它们之间的关系,利用训练好的网络模型反演月平均近海面气温和湿度,并与TAO和NDBC提供的浮标及观测站的实测数据进行比较,近海面气温和相对湿度的均方根差分别为0.87℃和3.73%。低纬度反演的结果精度较高,达到0.53℃(气温)和2.03%(相对湿度);较大的误差(气温1.06℃、相对湿度3.85%)主要发生在近岸和高纬度区,因为近岸的地形比较复杂,并且很容易受陆地气候的影响;高纬度地区的气候变化比较剧烈,同时目前能得到的高纬度地区的实测资料比较少,这些因素都会影响反演结果的准确度。
Parameters （SST, wind, rain rate, cloud liquid water and atmosphere water vapor） were obtained from SSM/I and AVHRR to retrieve near sea surface air temperature and humidity. An artificial neural network （ANN） was set up and trained using the data from satellites and regional observations. The results of the ANN network were compared with the observational data provided by TAO and NDBC. The root mean square （RMS） of air temperature （TA） and relative humidity （RH） were 0.87℃ and 3.73%, respectively. The results estimated in lower latitudes and open oceans were more accurate than those in higher latitude and offshore areas, at 0.53℃ and 2.03%, and those in higher latitude and offshore were 1.06℃ and 3.85%, for the RMS of TA and RH, respectively. The larger errors shown above were probably due to more continental influence in nearshore case, and scarce data and more changeable weather in high latitude case. In future, more in situ data should be collected to imorove the algorithm.
出处
《海洋与湖沼》
CAS
CSCD
北大核心
2008年第6期546-551,共6页
Oceanologia Et Limnologia Sinica
基金
国家863课题基金资助项目
2001AA633060号
中央级公益性科研院所基本科研业务费专项资金(中国水产科学研究院东海水产研究所)资助项目
2007T09号
关键词
人工神经网络
近海面气温
湿度
SSM/I
AVHRR
Artificial neural network, Near sea surface air temperature, Humidity, SSM/I, AVHRR
