Old forecasts vs actual shoreline evolution: Assessing model's performance and projections accuracy

Abstract

Nowadays there are numerous computational tools to forecast the evolution of the shoreline position in sandy coast environments. However, few studies have waited for the time horizon of the predictions to actually check the accuracy of the forecasted scenarios. Thus, this work proposed to assess the accuracy of medium-term predictions made with Long-Term Configuration (LTC) and GENEralised model for SImulating Shoreline change (GENESIS) numerical models for the time horizon of 15-year and evaluate the models’ performance in forecasting the actual status of a complex sandy coastal stretch in 2018, the targeting year. The projection premises were already presented in Baptista et al. (2014) on which observational data used to calibrate and blind hindcast future scenarios of shoreline development was discussed. The case study focused on an area located downdrift of Poço da Cruz groin, on the northwest Portuguese coast, and considered two beach sectors in accordance with Baptista et al. (2014): sector AC, about 3.5 km long, and sector AB, about 2.4 km long and included in the first one. The whole coastal stretch is composed of sandy beaches, and past trends highlight erosion downdrift of Poço da Cruz groin due to the interruption of sediment transport. A monitoring program performed by the University of Aveiro between 2003 and 2018 (corresponding to the period of simulation presented in Baptista et al., 2014), including six topographic surveys, was considered for comparison with the projections previously made. An additional topographic survey performed in 2018, within COSMO programme, was also considered. Simulations results have proved divergent performances of LTC and GENESIS models, despite the equity of inputs parameters and modelling processes. It is concluded that GENESIS and LTC, considering the typical (TYP) wave climate (randomly generated to represent the typical frequencies of wave height and direction during five years), create an envelope (GENESIS below and LTC above) to the real average results. For TYP, GENESIS underestimates the mean retreat rates up to 23 m in sector AB and up to 23 m in AC; LTC overestimates the retreat up to 33 m in sector AB and up to 13 m in AC. Conversely, observed (OBS) wave climate (shorter characterisation period but maintaining the registered data sequence) projections overestimate the shoreline retreat in both coastal stretches, suggesting that a larger characterisation period should be more important than maintaining the sequence of recorded waves. For OBS, the models overestimate the mean retreat rates in both sectors: GENESIS overestimates up to 90 m in sector AB and up to 51 m in AC; LTC overestimates up to 60 m in AB and up to 60 m in AC. Also, the consideration of the artificial nourishment operations executed during the simulation period leads to better results in the LTC, which was overestimating the erosion rate. This showed that if all man-made interventions could be anticipated, the average retreat in the main sector, AC, would only be up to 5 % different from the projections made by Baptista et al. (2014) with LTC model and TYP wave climate. • Checked accuracy of 15-year forecasts. • Checked performance of LTC and GENESIS shoreline evolution models. • The wave sampling period used in calibration has relevant impact on forecasts made. • LTC and GENESIS create an envelope to the real average retreat surveyed. • Better projections would be achieved if planed interventions were known in advance.