Regional flood risk analysis for agricultural crops: Insights from the implementation of AGRIDE-c in central Italy

The development of reliable models for different exposed sectors is key for obtaining more comprehensive flood damage and risk assessments. This study discusses the implementation of the recent conceptual damage model for crops, AGRIDE-c, to a central region of Italy, in order to demonstrate its potential for application and related challenges from a flood risk assessment perspective at the river basin scale. The process for adaptation of model's components to the local characteristics of the investigated area is first described, including the extension of the analytical flood loss model to annual vegetables and perennial crops (here exemplified for grapevine) not covered in the original AGRIDE-c. Moreover, being a multi-variable damage model, with also some input parameters of uncertain estimation, this study examines the implications of modelling assumptions on damage and risk results to obtain general insights from a modeller's perspective. Finally, the issue of spatial transferability is discussed by comparing model outcomes for cereals in the analysed region and in the Po Plain, for which AGRIDE-c was originally developed. The results highlight the importance, in practical applications, (i) to determine a confidence band for loss estimates that can help for more informed decision making processes and (ii) when transferring models, to verify the similarity between the physical and economic contexts of implementation (and, if necessary, adjust model's input and/or assumptions), in order to prevent inaccurate or distorted damage and risk estimations.

Automated summary

This study discusses the application potential and challenges of a crop damage model in a central region of Italy, as well as the process of adapting the model to the local characteristics. It examines the implications of modelling assumptions on damage and risk results, and emphasizes the importance of verifying similarity when transferring models.