When analyzing the truncation error of numerical format, we often encounter the form of truncation error term, and this p(q) is the format accuracy respectively. For example, if the first derivative is replaced by the intentional error, the truncation error term will be generated. If there is no low-order error term, then the total error is the second-order accuracy in space. The significance of error convergence rate is the speed of error reduction, which has nothing to do with the actual size of error. That is to say, the accuracy of format A is 1, and the accuracy of format B is 2. Is the error of format B smaller than that of format A on the same grid? I do not think this is necessarily the case. What really reflects the accuracy of format is the rate at which the error of comparison format B decreases when the grid scale decreases gradually on the encrypted grid. In addition, you said that in the case of two dimensions, it is a bit complicated because there will be such a mixed term, but I think the step size is equivalent except the time step size (? ), can be regarded as. Spatial accuracy and time accuracy are treated differently, because most of the schemes now discretize the equation into ODE about time, and then integrate it in time with RK scheme (Euler, AB, implicit, etc.). ). RK has different accuracy of 2~5 orders, which can basically be the same as or higher than spatial accuracy, so spatial accuracy is generally the limit of format accuracy. In the papers I have read, there is little theoretical analysis of format accuracy, and it is usually calculated by numerical method, that is, the error is seen on the gradually subdivided grid. Theoretically, the error is truncated, but it is difficult for us to know the spatial distribution of the error function in numerical calculation, so we can only use a numerical value to represent the error. This value is usually expressed by the norm in the functional, f is the numerical solution, f0 is the analytical solution, and the commonly used three norms are L 1, L2 and infinite norm Linf.