使用一个太平洋海盆尺度环流模式模拟氚的分布,并分别采用二阶中央差(CD)、通量修正输送法(FCT)、分段抛物线法(PPM)、多维正定平流输送法(MPDATA)及二阶矩法(SOM)计算流场对氚的平流输送,通过结果比较考察算法间的差异。模拟试验设定为完整物理过程情形(CASE1)及仅含平流、对流情形(CASE2)。CASE1中,不同试验模拟的高低纬间氚沿等位密度面的分布特征基本一致。但CD结果在赤道、副热带次表层特征以下深度出现了负值,影响了侧向通风输送的计算,致使120°W次表层深度的氚偏多。负值也使得CD结果中125°W断面200m以下的氚分布不合理,而FCT、PPM、MPDATA及SOM则不存在负值的影响,但在125°W上体现出垂直方向的数值耗散差异。CASE2中,除CD试验出现严重失真外,其他算法的结果均在合理的数值范围内。但在西边界延伸流及南极绕流区域,FCT、PPM、MPDATA及SOM在数值耗散及稳定性上存在差异。综合看来,PPM及SOM在数值稳定性上均优于FCT、MPDATA,计算解相对较光滑。相比PPM,SOM引入的数值耗散更弱,易于模拟出合理的强梯度特征,但SOM计算速度慢,因此在考察海洋示踪物的长时间输送演变过程上PPM更具优势。
The centered difference scheme for advection transport may cause unphysically negative values when used in an ocean model to simulate oceanic passive tracers.Selecting positive definite schemes can suppress the generation of negative values.In order to examine the difference of advection schemes,second-order Centered Difference(CD),Flux-Corrected Transport algorithm(FCT),Multidimensional Positive Definite Advection Transport Algorithm(MPDATA),Piecewise Parabolic Method(PPM) and Second-Order Moments advection scheme(SOM) are chosen to simulate the distribution of oceanic tritium,respectively.Two cases are set to compare the differences between these schemes.The physical fields are exactly same for these runs with different advection schemes for tritium in a basinwide ocean general circulation model of the Pacific Ocean.In Case 1 that contains full physical processes,all runs give similar horizontal distributions along isopycnals from the subarctic to tropical regions.However,many negative values arise in the CD run,while for other runs much weaker negative values are generated in some grids.During model years 1968-1976,negative values in the CD run gradually become comparable in magnitude to main signals in the subsurface layer of the eastern Tropical Pacific.Those negative noises beneath the depth of maximum tritium concentration within the subsurface layer can disturb the lateral ventilation along isopycnals,leading to much larger subsurface values near 120°W.At 125°W below 200-m depth,negative values also exist,making the CD results meaningless,while the runs with other schemes all give acceptable values.Both FCT and SOM runs show a weaker numerical dissipation in the vertical direction than PPM and MPDATA runs.In Case 2,which does not have physical dissipation,the CD run fails to produce a valid solution,while other runs all succeed in getting nonnegative distributions.The SOM run gives the strongest meridional gradient near the western boundary in the North Pacific,while the PPM run prod