Advancing Knowledge-guided AI To Develop Agricultural Digital Twins For Singapore’s Climate Resilience

Climate change poses significant challenges to most aspects of the human and natural environment. Singapore, in particular, faces high vulnerabilities and external pressures driven by the need to secure food supplies and achieve food security. In response to this concern, the nation is boosting efforts in food production and advancing the agricultural industry domestically and in the ASEAN region.

Advancing Knowledge-guided AI To Develop Agricultural Digital Twins For Singapore’s Climate Resilience (KGAI4Ag) addresses this challenge by developing the first-of-its-kind digital twins (DT) for ASEAN agroecosystems through two primary approaches:

  1. Foundational AI (FAI): Innovating in foundational AI methodology (i.e., KGAI) to be broadly applied to agriculture and other domains
  2. Use-Inspired Research (UIR): Utilizing KGAI to develop digital twin templates for agroecosystems and food trade systems

With the KGAI-empowered digital twins, the research program aims to develop versatile DT templates that are adaptable across different crops, geographical locations, and related scientific domains.

CHARACTERIZING
PHYSICAL REALITY

Thrust 1: Characterizing Physical Reality

Multi-scale data collection and remote sensing across ASEAN agroecosystems

The observational foundation for the programme includes a network of low-cost IoT field sensors and cameras to capture ground-level measurements of soil conditions, crop biophysical properties, and farm management practices. These are integrated with satellite remote sensing data, including imagery from geostationary satellites that provide daily, cloud-free coverage across Southeast Asia, and using AI-driven data fusion methods to produce consistent, analysis-ready datasets at field to regional scales.

DEVELOPING DIGITAL TWINS

Thrust 2: Developing Digital Twins

Model-data fusion and virtual representation of field and regional agroecosystems

The development of field-level and regional digital twins lies at the core of KGAI4Ag. At the field scale, dynamic 4D plant models are coupled with biogeochemical modeling through KGAI-based data assimilation, allowing continuous integration of sensor and satellite observations to constrain model states and parameters. At the regional scale, a Bridge Model framework enables efficient upscaling of field-level digital twin outputs to extend high-fidelity simulations across thousands of heterogeneous farm sites while substantially reducing computational cost.

DECISION-MAKING

Thrust 3: Decision-Making

AI-driven decision support for farm management, food trade, and environmental assessment

This research component translates digital twin outputs into operational tools for decision-making at farm, national, and regional scales. At the farm level, reinforcement learning methods are used to identify adaptive management strategies that balance crop productivity, greenhouse gas emissions, and economic outcomes. At the regional level, a digital twin for agricultural trade integrates computable general equilibrium models with KGAI to simulate the impacts of climate extremes and policy interventions on ASEAN food supply chains. A complementary module supports measurement, reporting, and verification (MRV) of agricultural carbon outcomes in support of sustainability and green finance initiatives.