Abstract
This paper proposes a fitting method for discharge rating curves using Bayes spline regression. At present, these curve-fittings are conducted empirically. Operators usually employ piece-wise least-squares method for estimating these curves with linearizing height-quantity relationship via taking the square root of flow quantity. However, this method has a serious fault; difference in the form of curves among operators. In our method, arbitrariness of operators is excluded with determination of the number of knots via the generalized model selection that minimizes ABIC and EIC. Furthermore, spline curves that sustain continuity evacuate conflict of lines near boundaries and guarantee continuity of discharge rating curve. To verify validity of our model, we adapt our method to the cross-sections of the Toyohira and the Tama rivers. As a result of the analysis, we can obtain favorable fitting curves in the both rivers.
