Input reduction for long-term morphodynamic simulations

Abstract

Input reduction is imperative to long-term (> years) morphodynamic simulations to avoid excessive computation times. Here, we discuss the input-reduction framework for wave-dominated coastal settings introduced by Walstra et al. (2013). The framework comprised 4 steps, viz. (1) the selection of the duration of the original (full) time series of wave forcing, (2) the selection of the representative wave conditions, (3) the sequencing of these conditions, and (4) the time span after which the sequence is repeated. In step (2), the chronology of the original series is retained, while that is no longer the case in steps (3) and (4). The framework was applied to two different sites (Noordwijk, Netherlands and Hasaki, Japan) with multiple nearshore sandbars but contrasting long-term offshore-directed behavior: at Noordwijk the offshore migration is gradual and not coupled to individual storms, while at Hasaki the offshore migration is more episodic, and wave chronology appears to control long-term evolution. The performance of the model with reduced wave climates was referenced to a simulation with the actual (full) wave-forcing series. It was demonstrated that input reduction can dramatically affect long-term predictions, even to such an extent that the main characteristics of the offshore bar cycle are no longer reproduced. This was particularly the case at Hasaki, where all synthetic series that no longer capture the initial chronology (steps 3 and 4) lead to rather unrealistic long-term simulations. At Noordwijk, synthetic series can result in realistic behavior, provided that the time span after which the sequence is repeated is not too large; the reduction of this time span has the same positive effect on the simulation as increasing the number of selected conditions in step 2. It was further demonstrated that, although storms result in the largest morphological change, conditions with low to intermediate wave energy must be retained to obtain realistic long-term sandbar behavior. The input-reduction framework must be applied in an iterative fashion as to obtain a reduced wave climate that simulates long-term sandbar sufficiently accurately within an acceptable computation time. Given its potential huge impact on the actual simulation, we believe it is imperative to consider input reduction as an intrinsic part of model set-up, calibration and validation.