DT-Agro, a digital twin of the Greek AgroHydroSystem, development and testing

Developing operational Digital Twins for large-scale agro-hydrological systems presents significant challenges regarding data heterogeneity, computational efficiency, and the integration of Earth Observation (EO) with process-based modeling. This study presents the development and initial testing of DT-Agro, a spatially explicit Digital Twin of the Greek agro-hydro-system, designed to support sustainable water management and agricultural planning at the national scale.

DT-Agro integrates a high-resolution spatial database, a hybrid meteorological forcing scheme, and a distributed agro-hydrological model (AgroHydroLogos) recoded in C++ and Python for enhanced performance. A key innovation in the process simulation is the development of a novel, impervious-aware SCS-CN formulation. Unlike traditional lumped approaches, this method explicitly decomposes each grid cell into pervious and impervious fractions using high-resolution Copernicus Imperviousness Density data. This allows for a physically consistent representation of runoff generation in mixed landscapes, capturing the hydraulic response of small impervious patches that are often lost in standard gridded models.

Furthermore, to address the chronic fragmentation of ground-based monitoring networks, the system introduces a "virtual station" meteorological framework. Recognizing that raw global reanalysis products (e.g., AgERA5) often exhibit significant biases in Greece’s complex terrain, we developed a hybrid correction workflow. AgERA5 time series are sampled at the locations of historical stations and bias-corrected using station-specific regressions. This creates a network of "virtual stations" that provide continuous, homogenized daily records, filling temporal gaps while preserving local climatological characteristics. These records drive a dynamic spatial interpolation scheme that accounts for temperature and precipitation gradients, ensuring physically consistent meteorological forcing across the national domain.

We present results from the initial national-scale application of the system. The testing phase focused on quantifying irrigation water abstractions and their spatial-temporal drivers. Initial simulations estimate the long-term average national irrigation abstraction at approximately 6,600 hm³/year, with significant inter-annual variability (6,000–7,800 hm³) driven by climatic conditions. Validation against theoretical net irrigation requirements for major crops (maize, cotton, alfalfa) yielded consistent depths (380–420 mm), confirming the biophysical realism of the model core. These results demonstrate DT-Agro’s capability to provide a robust, evolving representation of the Greek agro-hydro-system for climate adaptation planning.