کاربست چندجمله‌ای‌های هرمیت مکعبی در ساخت روش نیمه‌لاگرانژی یکنوا برای معادله فرارفت

از آنجا که روش های نیمه لاگرانژی محدودیت پایداری به‌شرط کورانتفردریکسلوی (cfl) ندارند، به صورت وسیعی در مدل های گردش کلی به کار برده می شوند. در غالب صورت های روش نیمه لاگرانژی، برای محاسبه کمیت حل شونده (مانند سرعت باد، رطوبت، دما و جرم) درون‌یابی در فاصله بین نقاط شبکه ای لازم است. از مشکلات مواجه در کاربرد روش نیمه لاگرانژی تولید نوسان اضافی در مناطق با گرادیان شدید است. دو رویکرد برای حذف نوسان های اضافی در مناطق با گرادیان شدید به کار گرفته می شود: الف) ترکیب یک روش با مرتبه بالا با روش پادجریان سو، ب) استفاده از روش های درون‌یابی یکنوا. در این پژوهش از روش درون‌یابی گزینشی با چندجمله ای های هرمیت مکعبی نایکنوا و یکنوا با صافی مشتق برای درون‌یابی مقادیر کمیت حل شونده در فاصله بین نقاط شبکه ای استفاده شده است. نتایج نشان می دهد که حل معادله فرارفت با روش مذکور نوسان اضافی در مناطق با گرادیان شدید ایجاد نمی کند و میزان میرایی آن نیز ناچیز است. عدم تولید مقادیر منفی با استفاده از درون‌یاب یکنوای هرمیت در حل میدان های ذاتاً نامنفی مانند رطوبت و جرم از دیگر نتایج این پژوهش است.

Using cubic Hermite polynomials in constructing monotone semiLagrangian methods for advection equation

SemiLagrangian methods have been widely applied in general circulation models of the atmosphere as they do not suffer from a Courant–Fredericks–Levy (CFL) constraint for computational stability. Ease of application, high accuracy and speed of execution in general circulation models are other reasons for the popularity of semiLagrangian methods. Two fundamenta lissues in semiLagrangian methods are related to the trajectory computation and interpolation from the regular grid to departure points. If sufficiently accurate schemes are used to solve for trajectories with interpolations, one can expect good performance from the semiLagrangian scheme in solving the equations of motion of the atmosphere. Two general methods of solving the trajectory equation are the forward and backward methods. Most semiLagrangian methods use backwardtrajectory schemes for estimating positions of the air parcels that arrive at the grid points in the future time step. Solving the trajectory equation is carried out by iteration. In the research reported, two iterations are used for trajectory computation. The fundamental difference between the forward and backward trajectory scheme rests in the calculation of advective quantity at the departure and destination points. While in the backward solution procedure, it is necessary to make interpolation from the regular grid to departure points; in the forward scheme, it is necessary to make interpolation from the irregular grid of destination points to the regular grid. The usually used interpolation methods in the semiLagrangian method include piecewise cubic Lagrange and Hermite, cascades, and monotone Hermite. Increasing the degree of polynomial interpolation leads to a higher degree of formal accuracy, but it leads to the generation of unwanted oscillation in regions with severe gradients of the transported quantities. Eliminating the unwanted oscillations is done through a variety of methods which generally increase the computational cost and reduce the accuracy of the scheme. To address the issue, in this research, a new selective monotone semiLagrangian method is developed and tested along with two standard methods based on the Lagrange and Hermite interpolations. The Lagrange polynomials have been considered by researchers for the high speed of computation in operational models. The fictitious oscillations produced at the edges of sharp gradients of the advected quantities are the main shortcoming of this method. The fictitious oscillations cannot be eliminated by increasing the degree of interpolation polynomials, which can only lead to a reduction in the wavelength of the oscillations. The results presented on increasing the degree of interpolation polynomials clearly show that the removal of the fictitious oscillations requires the use of monotone polynomials for interpolation. It is important to note that the Hermite interpolation polynomials are not inherently monotone. To make them monotone, one needs to manipulate the derivatives at the grid points appropriately. This process, however, may lead to a substantial deteriration of accuracy. For this reason, in this paper, a selective interpolation method is desined to obtain the best accuracy in solution of the advection equation, while preserving monotnonicity and removing the issue with the fictitious oscillations. In the selective method, first the interpolation is done by the nonmonotonic cubic Hermite and then a properly designed slope function is calculated at each grid interval. If the slope function takes values outside the range, it indicates that a fictitious oscillation has occurred in the interpolantion. To remove the oscillation, the nonmonotone interpolation is abandoned and the monotone interpolation is performed by limiting the derivative to the monotone region. This technique can minimize the error caused by the changes in the derivatives. Results are shown to demonstrate the working and superiority of the seclective montone scheme.