基于热扩散法的青海云杉冠层导度模拟（英文）

Simulation of Canopy Conductance of Qinghai Spruce ( Picea crassifolia) Plantation based on Granier's Thermal Dissipation Probe Method

【目的】环境因子是影响林木冠层水分利用的主要因素。本研究以黄土高原高寒区主要树种青海云杉为研究对象,对其蒸散发特征进行分析,以期探讨不同冠层导度模型的适用性。【方法】于2013年6月,采用热扩散技术对青海云杉蒸散量进行定位监测,监测步长15 min,并结合彭曼—蒙特斯方程反演气孔导度(g_c),在此过程中考虑了可能存在树干液流时滞的问题。选择饱和水汽压差(D),大气温度(T)和太阳辐射(R)等3个关键气象参数,采用6个不同形式的Jarvis模型和1个多元线性模型模拟g_c,所有模型的待定系数都采用1stOpt软件进行计算,并进行交叉验证。以奇数日数据计算g_c,用偶数日的数据进行验证。【结果】青海云杉基于液流计算的冠层蒸腾对环境因子变化的响应时滞为15 min。g_c与D和T呈显著的指数函数关系(P<0.000 1),并随着其值的增大而减小,而冠层蒸散量(E_c)则与R呈显著的二次函数关系(P<0.000 1)。多元线性模型模拟g_c的回归系数为0.90,略低于6个Jarvis模型(0.91~0.92),但用多元线性模型拟合的g_c计算的日E_c的精度最高。此外,7个模型模拟的g_c/E_c值都有较高的精度。【结论】R是青海云杉冠层蒸腾的主要驱动力,但其气孔的开合却主要受到D和T的控制。7个模型都具有较高的精度,但Jarvis模型的模式较多、使用复杂,部分模式的待定系数会出现有无穷组解的现象,且不同的解之间存在较大的差异。而多元线性模型的形式简单,精度高,是模拟g_c的较优选择。

【Objective】Environmental factors are the main factors influencing canopy water use.In this study,Qinghai spruce,a main tree species in Loess Plateau,was used as the research object,and the evapotranspiration characteristics were analyzed,in order to investigate the adaptability of different canopy conductance(g c)models.【Method】In June 2013,the evapotranspiration of Qinghai spruce was monitored with Granier’s thermal dissipation probe by a time step of 15 min.The quarter-hourly g c was continuously simulated by the inversed Penman-Monteith model using the collected data by Granier’s thermal dissipation probes.Accounting for the lag time,a multivariate linear model and six Jarvis models were used to simulate the relationships between g c and three key meteorological parameters of saturated vapor pressure deficit(D),air temperature(T)and solar radiation(R).A cross-validation method was employed,that is,the data collected on odd days were used to calculate g c,and the calculated results were verified by the data collected on even days.【Result】In the studied Qinghai spruce forest,canopy transpiration lagged meteorological factors by 15 minutes.Canopy transpiration(E c)was a quadratic function of R(P<0.000 1),and g c was an exponentially decreasing function of D and T(P<0.000 1).Although multivariate linear methods yielded slightly lower regression coefficients of g c estimation(r 2=0.9)than Jarvis methods(0.91≤r 2≤0.92),they provided the best daily E c estimation from the predicted g c.Furthermore,all of the predicted g c/E c values were consistent with the measured g c/E c,indicating that all methods could predict g c with sufficiently high accuracy.【Conclusion】R was the main driving force of E c of the Qinghai spruce canopy.The 7 models all have high accuracy,but the Jarvis model has many patterns and complex applications.The undetermined coefficients of the some models can have infinite solutions which are quite different.However,the multivariate linear model is simple in form and high in precision,which is a better choice for simulating g c.