华北平原农田生态系统的Noah-MP模型参数敏感性分析

Parameter Sensitivity Analysis of Noah-MP Model for Farmland Ecosystem in the North China Plain

农田生态系统每时每刻都与大气进行着热量、水分与二氧化碳的交换,消耗可利用水资源的同时生产人类生存所需粮食。陆面模式能定量刻画土壤-植被-大气间能量、动量和水汽交换,但其应用受到参数不确定性的制约。选取禹城站与栾城站2个华北平原典型农业生态系统试验站,采用拉丁超立方抽样法(LHS)基于Sobol方差分析检测了Noah-MP模型的42个参数对潜热通量(LE)、显热通量(H)、净生态系统碳交换量(NEE)、第1层土壤含水率(SM1)及第2层土壤含水率(SM2)5个输出变量的全局敏感度,筛选出高敏感参数。结果表明:若不考虑敏感性指标的精确排序,为节约计算成本,拉丁超立方抽样法的抽样次数可以取为参数数量的5倍左右;对这5个模型输出而言,敏感参数在站点间及不同气孔阻抗的土壤水分限制因子参数化方案下的表现较为一致,主要包含5个植被相关参数(LTOVRC、QE25、MRP、VCMX25、SLA)和4个土壤相关参数(BEXP、SMCMAX、SMCREF、SMCWLT),而土壤参数SMCMAX、SMCREF、SMCWLT与BEXP的全局敏感度在华北平原的不同站点间变异性较大;对比禹城站,在更为干旱的栾城站应用不同的土壤水分限制因子参数化方案时,BEXP的全局敏感度发生显著变化,因此干旱地区应重视对土壤参数BEXP的校准。

Farmland ecosystems exchange heat,water,and carbon dioxide with the atmosphere continuously,consuming available water resources while producing the food necessary for human survival.Land surface models can quantitatively describe the energy,momentum,and water vapor exchange between soil,vegetation,and atmosphere,but their application is constrained by input parameter uncertainty.This study selected two typical agricultural ecosystems experimental stations,e.g.,Yucheng Station and Luancheng Station in the North China Plain,and used Latin hypercube sampling(LHS)based on Sobol method to detect the global sensitivity of five output variables,e.g.,latent heat flux(LE),sensible heat flux(H),net ecosystem carbon exchange(NEE),first-layer soil moisture(SM1),and second-layer soil moisture(SM2),to 42 input parameters of the Noah-MP model,and screened out highly sensitive parameters.The results showed that if the exact ranking of sensitivity indicators is not considered,Latin hypercube samples can be about five times the number of parameters to save computational costs.For these five model outputs,highly sensitive parameters were consistent between sites and under different stomatal resistance parameterization schemes,including five vegetation parameters(LTOVRC,QE25,MRP,VCMX25,SLA)and four soil parameters(BEXP,SMCMAX,SMCREF,SMCWLT).In arid areas,special attention should be paid to the calibration of BEXP.