Multi-dimensional optical remote sensing in agriculture: Spectral, angular, and spatial scaling for crop stress monitoring

Early and accurate detection of crop stress is essential for sustainable agriculture and food security, particularly as climate change and environmental degradation intensify agricultural challenges. This comprehensive review examines advanced crop stress monitoring strategies that leverage multi-dimensional optical remote sensing approaches, specifically integrating spectral, angular, and spatial perspectives across diverse observation scales. We systematically analyze how biotic stresses (diseases, pests) and abiotic stresses (drought, nutrient deficiency, temperature extremes) manifest through detectable changes in plant spectral signatures, from chlorophyll degradation in the visible spectrum to water content variations in shortwave infrared regions. Our review encompasses sensing technologies spanning RGB, multispectral, hyperspectral, thermal infrared, and chlorophyll fluorescence sensors deployed across three complementary scales: proximal ground-based systems for detailed physiological assessment, unmanned aerial vehicles (UAVs) for field-scale monitoring, and satellites for regional surveillance. A key innovation of this work is the emphasis on multi-angle remote sensing, which captures bidirectional reflectance distribution function (BRDF) effects that reveal stress-induced changes in canopy structure and leaf orientation invisible to conventional nadir-only observations. We demonstrate how viewing geometry significantly affects vegetation indices (NDVI, PRI) and sun-induced fluorescence (SIF) measurements, requiring sophisticated angular correction methods for accurate stress assessment. Through synthesis of 138 recent studies spanning 12 major crop types, we identify critical research gaps including: (1) inconsistent angular reflectance modeling across stress types, (2) inadequate sensor calibration protocols for variable field conditions, and (3) lack of standardized frameworks for integrating multi-source, multi-scale data streams. Our analysis reveals that advanced machine learning approaches particularly deep learning and transformer networks show exceptional promise for extracting meaningful stress signatures from complex, high-dimensional datasets while maintaining interpretability for agricultural decision-making. We propose a hierarchical monitoring architecture supported by physics-aware artificial intelligence models that address three fundamental challenges: temporal optimization for capturing stress progression dynamics, spatial integration across observation scales, and angular standardization for consistent stress quantification. This framework aims to transform crop stress monitoring from reactive management to predictive intervention, enabling real-time diagnostics suitable for diverse agricultural systems ranging from high-value specialty crops to extensive grain production. The review concludes with a strategic roadmap for operational implementation, addressing economic constraints, technological limitations, and knowledge transfer requirements necessary for widespread adoption. Our findings indicate that successful deployment requires service-based delivery models, simplified decision support interfaces, and staged implementation approaches that demonstrate incremental value while building organizational capacity. The literature selection was conducted using Scopus, Web of Science, and IEEE Xplore databases, covering publications from 2018 to 2024. Search terms included “crop stress monitoring,” “spectral remote sensing,” “multi-angle sensing,” and “UAV agriculture.” A total of 138 peer-reviewed studies meeting relevance and methodological rigor criteria were included. These studies span 12 major crop types: wheat, maize, rice, soybean, cotton, sugarcane, potato, grapevine, tomato, barley, sorghum, and rapeseed, ensuring broad coverage across cereal, legume, fiber, tuber, and horticultural crops.