Predictors of the peak width for networks with exponential links

Abstract

We investigate optimal predictors of the peak (S) and distance to peak (T) of the width function of drainage networks under the assumption that the networks are topologically random with independent and exponentially distributed link lengths. Analytical results are derived using the fact that, under these assumptions, the width function is a homogeneous Markov birth-death process. In particular, exact expressions are derived for the asymptotic conditional expectations ofS andT given network magnitudeN and given mainstream lengthH. In addition, a simulation study is performed to examine various predictors ofS andT, includingN, H, and basin morphometric properties; non-asymptotic conditional expectations and variances are estimated. The best single predictor ofS isN, ofT isH, and of the scaled peak (S divided by the area under the width function) isH. Finally, expressions tested on a set of drainage basins from the state of Wyoming perform reasonably well in predictingS andT despite probable violations of the original assumptions.