摘要
基于通量网中23个站点下草地、农田、郁闭灌丛、落叶阔叶林、常绿针叶林、开阔灌丛、湿地7种植被类型2001-2007年气象和潜热数据,运用人工神经网络方法进行潜热通量/蒸散量的不同时空尺度的模拟,以期探究模型输入因子重要性及对人工神经网络法在蒸散发估算中时空尺度外推能力进行验证.结果表明:以温度、相对湿度、太阳辐射、土壤含水量、归一化植被指数、风速、土壤热通量为主的输入因子对不同植被类型蒸散量的影响程度不同:太阳辐射对各植被类型影响都相对较大,风速的影响与植被类型有关,归一化植被指数对植被生长期影响有差异;蒸散量模拟效果具有较大的时空异质性,时间尺度上的模拟效果比空间尺度上的好.其中,草地与农田的模拟效果最好.模拟结果中纳什系数多为0~1,个别小于0,模拟结果与观测值的相关系数均在0.6以上.基于人工神经网络法的潜在蒸散估计可以进行一定的区域外推,模拟效果的好坏受控于局地的气候状况.
Based on 2001-2007 meteorological and latent heat data from 23 sites including grassland, cropland, closed shrubland, deciduous broadleaf forest, evergreen needle-leaf forest, open shrubland and permanent wetland, the artificial neural network method was used to simulate the latent heat flux/evapo- transpiration in different spatial and temporal scales, in order to evaluate the importance of input factors and validate the extrapolation capability of the artificial neural network. The research indicated that tem- perature, relative humidity, net radiation, soil water content, normalized difference vegetation index, wind speed, soil heat flux should be chosen as the main input factors, whose influence on the evapotranspira- tion of different vegetation types was in different degrees, with the net radiation＇s influence being consid- erable, that of wind speed being connected with the vegetation type, and the normalized difference vegeta- tion index being varied in the growing stage of vegetation. The simulation results of the latent heat flux/ evapotranspiration were in great temporal and spatial heterogeneity, and the simulation results in different temporal scales were better than in the spatial scales. The Nash coefficients were mostly between 0 and 1,the individual was less than 0, but the correlation coefficients between the simulation results and the ob- served data were all above 0.6, indicating that the artificial neural network can be used for extrapolation in certain areas. The simulation results were related with the climatic conditions in different regions.
出处
《兰州大学学报（自然科学版）》
CAS
CSCD
北大核心
2017年第2期186-193,共8页
Journal of Lanzhou University(Natural Sciences)
基金
国家自然科学基金项目(31370467
41571016)
关键词
通量数据
人工神经网络
蒸散量
时空尺度
flux data
artificial neural network
evapotranspiration
spatial and temporal scale
