Journal article Open Access
“©IWA Publishing 2018. The definitive peer-reviewed and edited version of this article is published in Journal of Hydroinformatics 20 (4): 846-863, 2018, https://doi.org/10.2166/hydro.2018.027 and is available at www.iwapublishing.com.”
Prediction systems, such as the coastal ecosystem models, often incorporate complex non-linear ecological processes. There is an increasing interest in the use of probabilistic forecasts instead of deterministic forecasts in cases where the inherent uncertainties in the prediction system are important. The primary goal of this study is to set up an operational ensemble forecasting system for the prediction of the Chlorophyll-a concentration in coastal waters, using the Generic Ecological Model (GEM). The input ensemble is generated from perturbed model process parameters and external forcings through Latin Hypercube Sampling with Dependence (LHSD). The forecast performance of the ensemble prediction is assessed using several forecast verification metrics that can describe the forecast accuracy, reliability and discrimination. The verification is performed against in-situ measurements and remote sensing data. The ensemble forecast moderately out-performs the deterministic prediction at the coastal in-situ measurement stations. The proposed ensemble forecasting system is therefore a promising tool to provide enhanced water quality prediction for coastal ecosystems which, with further inclusion of other uncertainty sources, could be used for operational forecasting.