At first glance, finite difference modeling is by far the simplest method to grasp. All that is necessary is to replace the continuous partial derivatives by discrete approximations. The main difficulty arises in producing an accurate approximation to the various derivatives. There are two generally accepted approaches to finding approximations. The first is based purely on some form of fitting algorithm, frequently using polynomials, wherein a set of basis functions with known derivatives approximate the function whose derivative is required. Once the fit is obtained the derivative is defined in terms of the approximating functions.
The easiest approach to finite difference approximation is to simply use a difference quotient in Equation 59 like we did when we derived the full two-way equation.
![]() | (59) |
This is what is called a first order forward difference approximation. Similarly, we have the backward difference in the form Equation 60.
![]() | (60) |
What may not be so clear is that these formulas are the result of approximating u by a straight line between x + Δx and x and between x