TerrSysMP–PDAF (version 1.0): a modular high-performance data assimilation framework for an integrated land surface–subsurface model

摘要: Abstract. Modelling of terrestrial systems is continuously moving towards more integrated modelling approaches, where different compartment models are combined in order to realise a sophisticated physical description water, energy and carbon fluxes across boundaries provide view on processes. While such can effectively reduce certain parameterisation errors single models, model predictions still prone uncertainties regarding input variables. The resulting be tackled by data assimilation techniques, which allow one correct with observations taking into account both the measurement uncertainties. steadily increasing availability computational resources makes it now increasingly possible perform also for computationally highly demanding system models. However, as burden well techniques quite large, there an need efficient frameworks that run make use massively parallel resources. In this paper we present framework land surface–subsurface part Terrestrial System Platform (TerrSysMP). TerrSysMP connected via memory-based coupling approach pre-existing library PDAF (Parallel Data Assimilation Framework). This provides fully modular environment performing surface subsurface compartment. A simple synthetic case study (0.8 million unknowns) used demonstrate effects assess scaling behaviour system. Results show corrects states parameters reference values. Scaling tests evidence  > 30 k processors. Simulations large problem size (20 forward were efficiently handled proposed useful simulating estimating predicted over spatial scales at high resolution utilising