A generic framework for evaluating hybrid models by reuse and composition – A case study on soil temperature simulation

We evaluate modelling solutions built composing different modelling approaches. Modelling of soil temperature was the case study, evaluating modelling solutions against a multi-year, multi-location database of field recorded data. Widely used simulation packages were re-implemented using advanced techniques, and nine new modelling solutions were built hybridizing across the original ones. Multi-metric indices were developed for modelling solution evaluation. The hybrid solution implementing the Parton's approach (surface temperature) and the SWAT (temperature along the soil profile) led to the best compromise between agreement and robustness under the explored conditions. Differently, choices of soil water redistribution models caused a modest variability with respect to the simulation of soil temperature. The methodology presented allows providing clear indications about model choice and should be considered useful practice in model development. The model libraries used to run the analysis are freely available for download, and they allow for further extension of the composition exercise. •We develop nine new modelling solutions (MSs) for soil temperature simulation.•New MSs were composed by hybridization of models from known modelling packages.•Software component libraries allowed the model composition study.•The evaluation of MSs was done on a large dataset via a multi-metric procedure.•This is a proof of concept for fine-granularity model composition and MS evaluation.