Abstract
In a period in which climate change is significantly varying rainfall regimes and their intensity all over the world, river-flow prediction is a major concern of geosciences. In recent years there has been an increase in the use of deep-learning models for river-flow prediction. However, in this field we can observe two main issues: i) many case studies use similar (or the same) strategies without sharing the codes, and ii) the application of these techniques requires good computer knowledge. This work proposes to employ a Google Colab notebook called CleverRiver, which allows the application of deep-learning for river-flow predictions. CleverRiver is a dynamic software that can be upgraded and modified not only by the authors but also by the users. The main advantages of CleverRiver are the following: the software is not limited by the client hardware, operating systems, etc.; the code is open-source; the toolkit is integrated with user-friendly interfaces; updated releases with new architectures, data management, and model parameters will be progressively uploaded. The software consists of three sections: the first one enables to train the models by means of some architectures, parameters, and data; the second section allows to create predictions by using the trained models; the third section allows to send feedback and to share experiences with the authors, providing a flux of precious information able to improve scientific research.