This editorial reflects on the evolution and future direction of systems intelligence for sustainability industry (SI4SI), an initiative that has grown from a workshop at the 2017 UNDP conference into a continuing international dialogue connecting systems science, artificial intelligence (AI), and sustainable development. As AI technologies rapidly democratize knowledge and capability, they simultaneously introduce systemic risks, where termination-triggering knowledge, intelligence, and actions may amplify unintended consequences at unprecedented scales. Such risks call for governance that aligns intelligence with long-term societal well-being. Inspired by a modern interpretation of Pascal’s wager, we argue that precaution and responsibility must precede large-scale deployment of advanced AI. An embedded governance paradigm integrating blockchain intelligence and decentralized autonomous organizations is proposed to enhance transparency, accountability, and collective coordination. Essentially, SI4SI advocates a shift from artificial intelligence for efficiency to prescriptive intelligence for sustainability, emphasizing that technological progress must be prescribed and controlled by human values and welfare.

African agriculture, the pillar of food security and the economy for millions of people, is on the front line when it comes to climate change. Rising temperatures, more frequent droughts and variable rainfall are threatening the yields of traditional food crops. Conventional agronomic approaches are proving inadequate in the face of this challenge. This article proposes a paradigm shift by positioning functional-structural plant models (FSPMs), coupled with artificial intelligence (AI) methods, as a strategic tool for accelerating the adaptation of African agricultural systems. We present an integrated methodological framework (FSPM-AI) that starts with the collection of high-precision agro-physiological data in the field and culminates in the creation of "digital twins" of plants (i.e., dynamic virtual replicas of real plants). These models make it possible to simulate the complex interaction between a plant’s genetics, its 3D architecture, its physiology (particularly photosynthesis), and its environment. Using AI techniques for model calibration, sensitivity analysis, and simulation of future climate scenarios (e.g., representative concentration pathways 4.5/8.5), this approach makes it possible to (1) identify the key morpho-physiological traits that confer resilience (e.g., leaf angle, biomass allocation, and photosynthetic efficiency), (2) optimise cultivation practices (density and fertilisation), and (3) guide varietal selections. Using C4 cereals such as millet and sorghum, which are essential for arid areas but scientifically under-modelled, as an example, we illustrate how this coupling of modelling and AI can fill a critical knowledge gap and provide robust decision support tools for farmers, researchers, and policy makers in Africa.

Accurate prediction of transpiration rate is crucial for precision irrigation and improved water use efficiency in greenhouses. This paper proposes a "Physics + XGBoost + Shapley additive explanation (SHAP)" framework. A physical indicator, E_phys= Cond × VpdL, is constructed using stomatal conductance (Cond) and leaf-air vapor pressure deficit based on leaf temperature (VpdL) to establish a linear physical baseline. Subsequently, an XGBoost residual correction model is built using the residual between observed transpiration rate and the predicted physical value as the learning objective, thus forming the Physics + XGBoost transpiration rate prediction framework. Results show that the coefficient of determination on the test set is approximately 0.9975, and the root mean square error is approximately 0.0583. SHAP analysis indicates that E_phys, Cond, and VpdL are the dominant factors driving residual correction. Furthermore, this paper predicts the canopy daily evapotranspiration at the leaf scale to construct an irrigation scenario, providing a generalizable technical path for water management and physical-data fusion modeling of greenhouse crops.

This paper considers the current challenges of agriculture in historical retrospect. As in all activities, agriculture is orchestrated by information governance systems. They have evolved along history following a rather regular trend of disembedding and concentration, disembedding of information from its origin in the physical world and concentration under the control of increasingly powerful market actors. The current challenges of agriculture, whether environmental, health, or culture-related, can be shown to relate to blind spots of the information systems. We consider the potential of intermediation platforms, that are increasingly in control of multi-sided markets in increasingly more sectors, to reshape the view systems have, making them more holistic, able to deal with what was previously neglected as externalities. We first consider the issue theoretically, and then illustrate with the claims of real actors towards a systemic intelligence of agriculture.

Predicting crop yield by integrating molecular basis for crop improvement attracts increasing interest but remains a great challenge due to complex biological and physiological processes across various scales. Crop physiological models have been developed and successfully used in predicting crop growth, development, and grain yield formation in response to various environments for more than half century. The crop model realized a chain of non-linear processes from canopy development, canopy photosynthesis, biomass growth and allocation, grain yield formation, and nitrogen, water, and energy balance based on detailed physiological and agronomic experiments. On the other hand, leaf biochemical photosynthesis was also modelled with relatively simple equations four decades ago, which receives wide attention due to underlying explicit biochemical mechanisms. Apart from that, the model is based on photosynthetic events occurring on leaf cellular chloroplast, which more closely represents the fundamental molecular regulation. Molecular regulatory network of photosynthesis is made clear, but such achievement in improving grain yield has not been translated well in that the molecular basis is not directly reflected in whole plant performance and lacking predicting ability from cellular level to whole plant level. Consequently, connecting both types of models will greatly accelerate the efforts of realizing the prediction of molecular outcomes in grain yield formation. Here we reviewed development of leaf and canopy photosynthesis models, and discussed the connection of both types of models to support crop improvement.

This paper presents XPlantGL, a Matlab-based software for functional-structural plant model (FSPM) derived from the GreenLab model, designed for addressing the practical limitations of existing FSPM tools (e.g., poor interactivity, limited support for branching/stochastic crop growth, and complex parameter calibration). XPlantGL enables precise simulation of plant architectures and growth dynamics via sink-source relationships, and integrates a generalized least square method (GLSM) for single-stage or multi-stage model parameter estimation from measured data. The software is validated on over ten crops (lettuce, cucumber, tomato, maize, rice, etc.), covering field crops, horticultural crops, and herbaceous plants. As a userfriendly FSPM tool, XPlantGL provides a reliable platform for researchers and educators in plant growth modelling, digital phenotyping, and smart agriculture research. Its core innovations include enhanced interactivity, native support for stochastic/branching crop growth, and streamlined parameter fitting workflows. Future research will focus on environment factor integration and high-fidelity visualization optimization.

Multi-view image fusion for three-dimensional (3D) object detection in agricultural scenarios has received limited research attention. To address this issue, we propose Tractor-RoadBEV, a multi-stage multi-view perception framework. The proposed framework first performs two-dimensional (2D) object detection and mask enhancement to extract object-aware features from images. The multi-view features are then transformed into a unified bird’s eye view (BEV) representation through a depth mapping module and a spatial unification module. Subsequently, a temporal fusion module integrates the current BEV feature map with historical BEV features to enhance temporal consistency. Finally, the fused BEV feature map is decoded to obtain 3D object detection results. Experimental results on the constructed tractor road dataset show that TractorRoadBEV achieves F1 scores of 0.8307 and 0.8166 for person detection and vehicle detection, respectively, which significantly outperforms the classical monocular detection method Deep3DBox. In addition, the backbone network of TractorRoadBEV demonstrates strong generalization capability in BEV scene segmentation, achieving an intersection over union (IoU) of 0.602, which is 0.151 higher than that of the classical lift-splat-shoot (LSS) method. These results indicate that the proposed method provides a low-cost, efficient, and scalable perception solution for autonomous agricultural machinery in tractor road environments.

Sugar beet is a major sugar crop in temperate regions and rapid, high-throughput, and accurate estimation of field phenotypes is essential for variety selection and production optimization. In this paper, ten commercial sugar beet varieties adapted to high latitudes are investigated using unmanned aerial vehicle (UAV) based red-green-blue (RGB), multispectral, and thermal infrared imaging across multiple growth stages. Canopy structural, texture, spectral, and temperature features are extracted, and three machine learning algorithms, random forest (RF), partial least squares (PLS), and support vector machine (SVM), are used to predict sugar content, root fresh weight, and yield. The results show that all three methods estimate sugar content well, with relative root mean square error (rRMSE) values below 11.0%, while RF and PLS outperform SVM. Multispectral features provide higher accuracy than RGB features, and multi-sensor feature combinations generally improve sugar content prediction compared with single-sensor inputs. For root fresh weight, SVM slightly outperforms RF and PLS, and RGB features are more informative than multispectral features. The integration of thermal infrared features does not notably improve RF or PLS models, but the combination of multispectral and thermal infrared features achieves the best SVM performance ( R²=0.58, RMSE = 75.3 g, and rRMSE = 23.7%). For yield estimation, RF achieves the highest accuracy, with rRMSE values ranging from 15.4% to 18.8%. Yield prediction accuracy increases as the time of image acquisition approaches harvest, and combining multi-temporal data from periods close to harvest further improves model performance. Overall, multi-sensor UAV data can effectively estimate sugar content, root fresh weight, and yield in sugar beet, providing a useful approach for phenotypic analysis, precision management, and variety selection.

Accurate and large-scale monitoring of pine wilt disease (PWD) remains a critical challenge in forest management due to limited coverage, high operational cost, and insufficient detection reliability in conventional survey methods. To address these limitations, this study develops a real-time intelligent monitoring system that integrates a fixed high-altitude sensing platform with a YOLOv5-based deep learning detector. The system provides continuous panoramic visual acquisition, while an edge-cloud collaborative architecture enables real-time image preprocessing and automatic identification of infected pine trees. A high-resolution dataset collected from the Qinba Mountain region of Chongqing, China, was constructed to train and evaluate multiple YOLOv5 variants. Experimental results demonstrate that the optimal model (YOLOv5x) achieves a mean average precision (mAP) of 68.8% and an overall detection accuracy of 93.78%. Compared with conventional unmanned aerial vehicle (UAV) based monitoring, the proposed system provides larger monitoring coverage (approximately 12 km²), lower operational cost per 700 km² monitoring area (reduced by approximately 76.00%), and improved identification reliability. These results demonstrate that stable high-altitude observation combined with deep learning-based detection offers a scalable and cost-efficient solution for large-area forest epidemic surveillance and intelligent ecological management.

Contract farming is a vital organizational model for mitigating systemic risks in agricultural supply chains, yet it is persistently hindered by severe trust deficits and prohibitive legal enforcement costs. Traditional centralized credit evaluation systems often fail to encompass dispersed smallholder farmers, leaving their historical compliance locked as illiquid, implicit social capital. To address these structural vulnerabilities, this paper proposes a novel blockchain-enabled agricultural contract fulfillment mechanism grounded in decentralized identity (DID) and verifiable credentials. Rather than treating reputation as a direct transactional currency, we design a bidirectional anchoring architecture where on-chain reputation acts as a cryptographic trust anchor to validate the authenticity of verifiable credentials (VCs). These VCs transparently prove a farmer’s actual production capacity and off-chain agricultural records. Through the autonomous execution of smart contracts, this endogenous governance system links these authenticated capacity credentials directly to agricultural transactions. Successful fulfillment naturally yields a reputation premium, whereas opportunistic default, or the issuance of falsified credentials, triggers the immediate, systemic destruction of the responsible parties’credit capital. A multi-agent transaction simulation validates the proposed framework, demonstrating its efficacy in filtering market risks based on verifiable capacity and overcoming the cold-start dilemma for new rural entrants.

Deep reinforcement learning (DRL) shows great potential in irrigation scheduling. However, deploying complex DRL models on field hardware remains difficult. This paper proposes a deployable smart irrigation system. We integrate a biophysical crop growth model with an agent. A modular Python-Flask backend enables efficient real-time inference. We decoupled inference from training to reduce computational overhead. The system uses a bidirectional long short-term memory encoder to extract temporal features. We developed a browser/server (B/S) architecture dashboard for real-time monitoring and explainable decision heatmaps. System-level validation was conducted using historical data from Xinjiang cotton fields. The results show sub-10 ms inference latency and a 0.45 s total system response time. The system provides a 1.9% yield and 5.9% irrigation water use efficiency (IWUE) boost over standard soft actor-critic (SAC) algorithms, proving its algorithmic efficacy. This architecture transforms theoretical gains into executable irrigation schedules for real-world farming. It ensures stable and failure-free operation across continuous growing seasons.

Low-cost plant phenotyping for individual plants is increasingly important for plant digital twins, breeding, and continuous crop monitoring. This review surveys representative studies on image-based plant phenotyping from 2015 to 2025 and organizes them into two major categories: deep-learning-based trait estimation from RGB/RGB-D images and low-cost 3D phenotyping based on multi-view reconstruction. The literatures indicate that deep learning methods are advantageous for rapid, automated, and high-throughput estimation of specific traits, but they often depend on annotated data and may show limited crossscenario generalization. In contrast, 3D reconstruction methods usually provide richer structural information and broader trait coverage, but they require more complex acquisition workflows and higher computational burden. Therefore, low-cost plant phenotyping should be understood not only in terms of sensor price, but also in terms of the overall trade-off among sensing, computation, labor, and deployability. Finally, this review summarizes current challenges and identifies future directions, including cost-aware benchmarking, lightweight image-based models, robust field 3D reconstruction, and tighter integration with plant digital twin systems.