An open online simulation strategy for hydrological ensemble forecasting

Hydrological ensemble forecasting is crucial for regional or urban flood forecasting. The development of real-time hydrological ensemble forecasting methods and early warning systems that produce accurate and timely forecasts and flood warnings is of paramount importance for effective disaster risk reduction and the mitigation of loss of life.

However, the majority of current hydrological ensemble forecasting systems are centralized, requiring researchers to collect data, download executable programs for models and related methods, and configure the runtime environment on local computers based on specific scenarios (e.g., simulation and forecasting of a specific city or watershed). This method is extremely time-consuming and labour-intensive, and there is a high level of coupling between modelling resources such as data, models (or methods), and parameters. When researchers simulate other scenarios, the models used in certain hydrological processes may not be applicable to the new environment due to changes in the natural environment, and new models may need to be implemented (for instance, the models for runoff yield under saturated storage and runoff yield under excess infiltration conditions are distinctly different). Substantial amounts of time and effort must be invested in recollecting and deploying forecasting resources in local computer, which leads to repetitive labour. This involves downloading models, configuring the operating environment for each ensemble forecasting process, collecting pertinent data, compiling data and model adaptation methods, designing optimization schemes and evaluating the model based on results.

Therefore, to change the current complicated download and installation usage patterns associated with hydrological ensemble forecasting and to facilitate the seamless replacement and integration of various hydrological process model components, we propose an open online simulation strategy. This strategy utilizes a service-oriented web architecture to support the online sharing, invocation, integration, and optimization of simulation resources at the three perspectives: model, input data, and model parameters. Specifically, we explore (1) a service-oriented hydrological ensemble forecasting model sharing method and a document-based model service integration and management method, (2) a hydrological ensemble forecasting data sharing and Python-based data adaptation method, and (3) an online optimization and recommendation method for model parameters. By applying the strategy proposed in this paper to hydrological ensemble forecasting, it is possible to reduce the cost of using models, encourage the sharing of hydrological resources and the exchange of knowledge, and ultimately improve the accuracy of flood forecasting.