Challenges of modelling climate change impacts on hydrology and water resources: AI is the game changer—a review

There has been remarkable progress over the past 20 years to support hydrological analysis in climate change context. This study reviewed literature to identify key challenges and provide information for improving understanding of areas that entail knowledge gaps. The challenges are of both traditional and emerging nature. Some identified challenges include complexity in climate modelling, issues of downscaling, choosing fixed or flexible modelling approach, complexity in hydrological modelling, uncertainties in hydrological and climate models, and hydrological analysis in data-scarce catchments. Prominently, there is a notable shift towards the application of artificial intelligence (AI) for tackling these challenges. For instance, the integration of data assimilation and AI is a promising advance for regional analysis of climate change impacts. However, the increasing integration of AI in hydrology aggravates the challenge of ‘black box problem’ in which a modeller has no clue on relationships used to derive outputs from the given inputs. To tackle this, revolutionizing and adopting explainable AI in hydrology is imperative. Model complexity control is a vital procedure to encompass the systematic balance of intricacy with both quality and quantity of available model inputs. Additionally, the choice of a model amid the required flexibility and complexity should be linked to the overall cost and benefits based on the object of the analysis. Finally, to comprehensively identify, characterize, quantify and communicate uncertainties to stakeholders, uncertainty analysis should be integrated with management decision making. This requires recognition of the need for science-policy interfacing tailored for planning climate change adaptation measures.