Similarity of kinematic and diffusive waves: a comment on accuracy criteria for linearised diffusion wave flood routing. By K. Bajracharya and D.A. Barry [Journal of Hydrology, vol. 195 (1997), 200-217] R. Szymkiewicz Department of Hydraulics, Technical University of Gdan ´sk 80-952Gdan ´sk, Poland According to the title, in the referred paper an accu- racy analysis for linearised diffusion flood routing should be presented. The main purpose of this analysis was to obtain the proper values of mesh dimensions as Dt and Dx ensuring acceptable agreement between the results of computation and observed data. However, the Authors presented an accuracy analysis for kine- matic wave solved by a box scheme containing two weighting parameters. According to them the presented analysis concerns diffusive wave equation. The Authors accepted the solution of kinematic wave equation containing large numerical error as a solu- tion of diffusive wave equation. In fact the results of calculations can be similar but it does not mean that the equations giving them are also similar. For this reason I cannot agree with the title of referred paper. It contains an analysis concerning kinematic wave equa- tion or more generally an advection equation, which has nothing in common with the accuracy analysis of numerical solution of advection-diffusion equation as diffusive wave model. Consequently the presented conclusions concern the kinematic wave equation. Additionally, some of them are well known and can be formulated without any numerical calculations because they result from theory of numerical methods for partial differential equations. The following are the reasons which allow me to formulate this opinion. First let me recollect some information about the Muskingum model, which is a lumped model in the form of a linear ordinary differential equation contain- ing two parameters: X and K. The parameter K repre- sents the wave travel time between two river sections whereas the parameters X (called by Authors u ) has no physical interpretation. Cunge (1969) showed a kinematic character of Muskingum model. More- over he stated that the wave attenuation in Musk- ingum model is caused by numerical diffusion. When the trapezodial rule (v  1/2) is used to time integration, the numerical diffusion is controlled by u and D x only. Cunge proposed to accept the value u producing numerical diffusion equivalent to the hydraulic one existing in diffusive wave model. This form of Muskingum model is called Muskingum-Cunge model. It is possible to show readily that the Muskingum model is a semidiscrete form of linear kinematic wave equation obtained by space derivative approximation at point between nodes x i-1 and x i . Its location is defined by parameter u . The approximation is of II order accuracy with regard to x for u  1/2 only. For other values of u it is an approximation of I order accuracy. As the result of poor spatial approximation dissipation error in the form of numerical diffusion occurs. It ensures a flood wave attenuation by kine- matic wave model. One can add that u  1/2 is Journal of Hydrology 216 (1999) 248–251 0022-1694/99/$ - see front matter  1999 Elsevier Science B.V. All rights reserved. PII: S0022-1694(98)00276-5