River network solution for a distributed hydrological model and applications Raghunath Jha,* Srikantha Herath and Katumi Musiake IIS, University of Tokyo, 7-22-1 Roppongi, Minato-ku, Tokyo 106, Japan Abstract: A simultaneous solution for one-dimensional unsteady ¯ow routing for a network of rivers has been developed, which can be used either with a complete distributed hydrological model, a simple rainfall-runo model or as a stand alone river routing model. Either dynamic or kinematic solution schemes can be selected to simulate the river ¯ows. The river network is either generated from the Digital Elevation Model (DEM) or directly input to the model. The model can handle any number of upstream channels and computational points. A sparse matrix solution algorithm is used to solve the 2N  2N matrix resulting from N nodes in the network. A submodule generates the initial water depth and discharge at each computational point from equilibrium discharge in the absence of observed initial conditions. The model is applied in three sub-catchments of the Chao Phraya river basin, Thailand, considering three dierent conditions. The simulated results show good agreement with observed discharges and provide insight to water level ¯uctuations, especially where tributaries join the main channel. Copyright # 2000 John Wiley & Sons, Ltd. KEY WORDS distributed hydrological model; ¯ood forecasting; river network solution; Thailand INTRODUCTION Channel network routing is used to predict the changing magnitude, speed and shape of the ¯ood wave as it propagates through water ways, such as canals and rivers, which are frequently encountered in natural river basins. Yen and Osman (1976) and Yen (1986) have de®ned the channel network as composed of channel segments arranged in a bracing con®guration, with individual segments connected at junctions to form loops and treelike dendritic structures. In sequential type modelling (Cung et al., 1980) traditionally applied for ¯ood routing through a channel network, the diculty caused by the mutual backwater eect of channels does not arise because the backwater eect from downstream is ignored. The routing starts from most upstream channels and is carried downstream, channel by channel in a sequential manner, satisfying the ¯ow continuity requirements at junctions. The sequential models give unrealistic solutions when the downstream backwater eect is signi®cant. For subcritical ¯ow in an open channel network, mutual backwater eects exist among the channel branches joining at a junction. Therefore, the branches cannot be treated individually when a dynamic wave model (Amein and Chu, 1975) is adapted to route ¯oods in an open channel network. Many investigators in the past have simulated channel networks by considering these systems to be a combination of several independent channels or to be a main channel having tributaries with distributed lateral in¯ows (Abbott and Basco, 1988; Fread, 1973). These simpli®ed solution algorithms have been used in an attempt to avoid solving complicated solution algorithms resulting from a full dynamic approach. Among the simpli®ed methods, the overlapping Y-segment method suggested by Sevuk and Yen (1973) and Akan and Yen (1981) leads to an ecient solution scheme. This method is based on the CCC 0885±6087/2000/030575±18$17 . 50 Received 1 December 1998 Copyright # 2000 John Wiley & Sons, Ltd. Accepted 22 February 1999 HYDROLOGICAL PROCESSES Hydrol. Process. 14, 575±592 (2000) *Correspondence to: Dr R. Jha, Institute of Industrial Science, University of Tokyo, 5th Department, 7-22-1 Roppongi, Minato-ku, Tokyo 106-8558, Japan. E-mail: jha@incede.iis.u-tokyo-ac.jp