Overview
The Surrogate Engine for Crop Simulations Framework (SECSF) is a deep-learning framework designed to emulate the process-based ECroPS model. Its primary function is to provide spatially explicit, probabilistic, and rapid crop-risk information for anticipating climate-related crop stress. This framework operates solely on daily minimum temperature, daily maximum temperature, and daily precipitation data. SECSF was developed to facilitate large seasonal forecast and climate-scenario ensembles.
Research Context
Anticipating climate-related crop stress necessitates crop-risk information characterized by spatial explicitness, probabilistic nature, and sufficient speed. Such information is crucial for processing extensive seasonal forecast and climate-scenario ensembles. The existing process-based ECroPS model is a tool for crop simulation, but its computational demands can be significant, especially when applied across large ensembles.
Approach
The SECSF deep-learning framework was trained using ECroPS simulations that were forced by ERA5 data. The training focused on two specific crops: grain maize and spring barley. The input variables for SECSF were limited to daily minimum temperature, daily maximum temperature, and daily precipitation. This approach aimed to create a surrogate model capable of replicating the outputs of the more complex process-based model using a reduced set of inputs and significantly lower computational requirements.
Findings
The SECSF framework demonstrated a close reproduction of daily crop-growth dynamics and harvest timing when compared to the process-based ECroPS model. A notable outcome of using SECSF is a reduction in computational cost by approximately four orders of magnitude. This computational efficiency enables ensemble-scale inference, making the framework suitable for both research and operational pipelines.
When SECSF was forced with seasonal forecast data, it effectively captured spatially coherent crop-risk patterns across Europe during the high-impact year 2022. These patterns showed consistency with independent monitoring data, supporting SECSF's utility for developing probabilistic Areas of Concern.
Under CMIP6 scenarios, SECSF identified the Mediterranean basin as a hotspot for maize-risk signals projected through mid-century. Signals for central and northern Europe were observed to be more mixed within these same scenarios.
Why This Matters
The computational efficiency and accuracy of SECSF make it suitable for operational pipelines, including agricultural early warning systems and adaptation planning under uncertainty across seasonal-to-climate timescales. Its ability to identify regional crop-risk hotspots under future climate scenarios can inform strategic planning for agricultural resilience.