The review, led by Nazir Khan Mohammadi of Lanzhou University and Paktia University together with colleagues, synthesised 158 peer-reviewed publications spanning 1990 to 2025, drawn from databases including Web of Science, Scopus, Google Scholar and CAB Abstracts. The authors focused specifically on studies conducted in dryland or water-limited environments and reported outcomes on nitrogen use efficiency, yield, carbon dynamics or economic performance. Their central conclusion is striking: integrating smart agricultural tools into mixed crop-livestock systems can significantly improve nitrogen management, boost both crop and livestock productivity, and strengthen the economic resilience of farms that have historically been left behind by the technological revolution sweeping through irrigated agriculture.

At the heart of the analysis lies the problem of nitrogen. Globally, the average nitrogen use efficiency in cereal production is a mere 33 percent, meaning two-thirds of applied fertiliser never ends up in the food supply. The remainder is lost through ammonia volatilisation, nitrous oxide emissions and nitrate leaching, driving greenhouse gas emissions and water pollution. The review highlights that precision fertilisation guided by optical sensing and variable-rate application, operating at a spatial resolution of one square metre, has increased nitrogen use efficiency by more than 15 percent compared with traditional uniform application. Handheld sensors, drones and satellites such as the Copernicus Sentinel-2A/B platforms now allow farmers to tailor fertiliser inputs to the actual status and needs of their plants, rather than applying blanket rates across entire fields.

Biological nitrogen fixation offers a complementary route. Legumes such as alfalfa host symbiotic rhizobia that convert atmospheric nitrogen into biologically accessible forms, enriching the soil and reducing dependence on synthetic fertilisers. The review notes that precision agriculture techniques, including the nitrogen difference method, isotope methods and remote sensing, show great potential for diagnosing variability in symbiotic nitrogen fixation at the field level. Recent work also emphasises that below-ground nitrogen estimates must be included when quantifying fixation, otherwise the true contribution of legumes to soil fertility is systematically underestimated.

One of the most compelling farm-level strategies examined is intercropping alfalfa with silage corn. Although intercropping reduces corn yield by approximately 7 to 16 percent, it boosts alfalfa yield in the following year by 40 to 160 percent, while raising the crude protein content and improving the fermentation quality of the silage. By contrast, corn monoculture is associated with soil nitrate accumulation and inadequate ground cover. Alternative approaches, such as double cropping with rye or reseeding subterranean clover, can reduce nitrate accumulation and fertiliser requirements, though often with trade-offs in forage yield and management complexity. The review stresses that adjusting corn plant density and nitrogen application rates is essential to balance system productivity against silage quality.

The technological toolkit underpinning these gains is expanding rapidly. Internet of Things sensors deployed across fields transmit real-time data on soil moisture, temperature, pH and nutrient levels to cloud servers, where they can be integrated with geographic information systems to map soil health spatially. Field tests have validated the precision of these sensors, with R-squared values exceeding 90 percent, and some systems can trigger automated responses such as activating irrigation pumps when soil moisture falls below predetermined thresholds. Remote sensing adds another layer: the Normalized Difference Vegetation Index, combined with plant height measurements, can accurately predict biomass and nitrogen uptake in corn at different growth stages, while UAV-derived green NDVI has outperformed traditional field methods for estimating corn vigour and yield in complex smallholder systems.

Machine learning is the engine that converts this torrent of data into actionable predictions. Artificial neural networks, support vector machines and random forests integrate weather forecasts, soil sensor readings and satellite imagery to estimate crop health and forecast yields. Field-scale crop yield modelling using satellite-derived metrics alongside machine learning has accounted for over 70 percent of yield variation across different crops and agro-ecological zones, and deep neural networks have often outperformed alternative methods in comparative studies of wheat, corn and legume yield prediction. These forecasts can be generated well ahead of harvest, giving farmers and policymakers time to plan. On the livestock side, near-infrared reflectance spectroscopy can predict digestible amino acid content in animal feeds, accounting for 70 to 90 percent of variance, enabling precise feed formulation and on-farm nutritional monitoring.

Nutrient recovery closes the loop. Livestock farming generates vast quantities of nutrient-rich waste, and the review documents a growing repertoire of recovery technologies, from composting and anaerobic digestion to struvite precipitation, ammonia stripping and membrane filtration. GPS-guided manure application systems and GIS-based planning tools can optimise where and how manure nutrients are returned to fields, while the concept of manure sheds, mapping counties as sources or sinks of manure nutrients, offers a framework for recycling nutrients between animal feeding operations and nutrient-deficient cropland. Precision livestock farming, which matches nutrient supply to the requirements of individual animals through real-time sensor data, promises simultaneous gains in economic returns and reductions in environmental impact.

The economics are equally persuasive. Digital agricultural technologies have the potential to reduce fertiliser usage by up to 80 percent, cut pesticide application by 80 percent, increase crop yields by as much as 62 percent and lower labour costs by 97 percent, according to figures cited in the review. Integrated crop management systems exhibit higher net present values and cost-benefit ratios than conventional approaches, particularly in vegetable production, and intelligent platforms that combine yield data with financial analysis allow farmers to assess field profitability and negotiate land rents with confidence. Soil carbon adds a further dimension: global estimates suggest the biophysical potential for soil carbon sequestration is 4 to 5 gigatonnes of carbon dioxide per year with widespread adoption of best management practices, potentially rising to 8 gigatonnes with future technological advances.

Yet the review is candid about the obstacles. Limited internet connectivity, low digital literacy, inadequate infrastructure, high upfront costs and gender disparities all impede adoption among the smallholder farmers who stand to benefit most. A critical data gap persists: fragmented data collection and the absence of standardised protocols for integrating satellite imagery, IoT sensors, weather stations and farm records limit the development of robust, generalisable AI models, and most existing studies have been conducted under controlled conditions rather than in real-world smallholder contexts. The authors call for open-access localised datasets, long-term field trials, low-cost off-grid digital solutions and farmer-centred co-innovation, in which tools are designed collaboratively with the people who will use them. Policy support, affordable financing and digital skills training are essential complements to the technology itself. If these conditions are met, the review concludes, smart agriculture offers not merely a technological upgrade but a pathway toward resilient, efficient and equitable food systems in the world’s driest and most marginal environments.

Subject of Research: Application of smart agriculture technologies to improve nitrogen use efficiency, yield prediction and sustainability in dryland mixed crop-livestock systems

Article Title: Smart agriculture improves nitrogen use efficiency yield prediction and sustainability in dryland mixed crop livestock systems

Article References: Mohammadi, N. K., Arabzai, M. G., Inqilaabi, N. M., & Wang, Z. (2026). Smart agriculture improves nitrogen use efficiency yield prediction and sustainability in dryland mixed crop livestock systems. Discover Agriculture, 4(1), Article 316. https://doi.org/10.1007/s44279-026-00750-w

Image Credits: AI Generated

DOI: 10.1007/s44279-026-00750-w

Keywords: smart agriculture, nitrogen use efficiency, mixed crop-livestock systems, dryland farming, precision fertilisation, remote sensing, Internet of Things, artificial intelligence, yield prediction, biological nitrogen fixation, soil organic carbon, smallholder farmers

Researchers led by Sean J. Harkin of the University of Galway, working with McHale Engineering and a commercial harvesting contractor, fitted a fixed-chamber round baler-wrapper with a custom electronic control unit, a low-cost GNSS module, and a 2D LiDAR scanner, then recorded an entire five-month silage season. The result was a dataset spanning 120 fields ranging from 0.2 to 5 hectares, with yields from 7 to 33 bales per hectare and nearly 3,000 individually tracked bales. From that mountain of coordinates, the team built a deterministic algorithm that predicts, in real time, how many bales remain to be harvested.

The core insight is deceptively elegant. In a high-yield field, the baler fills its chamber quickly, so it travels a short distance between bale drops. In a sparse field, the machine must crawl over far more ground to collect the same volume of grass. Distance travelled per bale, in other words, is an inverse proxy for yield. Unlike specialised mass-flow sensors or load cells, which have never been commercialised on round balers largely because their cost rivals the machines themselves, distance comes free with any GNSS receiver.

Making that idea work required solving a harder problem first: figuring out which field the baler is actually in, without any prior map. The researchers noticed that operators almost always begin a job by circling the field perimeter to clear the headlands, creating turning room for later manoeuvres. The algorithm exploits this habit. It constructs a coverage polygon from the machine’s 4.8-metre working width, sweeping it along the recorded GPS track, and detects the moment the extended path intersects its own earlier coverage, signalling the first completed boundary loop. The enclosed area then serves as a proxy for the true field area, checked against safeguards that discard loops shorter than 100 metres or intersections caused by reversing.

Once the boundary area is known, a regression model converts it into an estimate of total swathe distance, the length of ground the machine will cover while actually picking up grass. Trained on the historical data, the relationship between field area and engaged distance proved remarkably tight, registering a coefficient of determination of 0.996. The team forced the regression through the origin for logical consistency, zero area must mean zero distance, and verified with an unconstrained model that the constraint did not distort the underlying geometry. The slope of roughly 0.185 metres of swathe per square metre of field held across fields of wildly different shapes and sizes.

The second half of the algorithm refines itself as the job progresses. Each time a formed bale transfers from the baling chamber to the wrapping tray, the distance travelled since the previous drop is logged. The analysis confirmed the hypothesis: boxplots of bale distance grouped by field showed a clear downward trend as yield increased, and crucially, each field’s bale distances behaved as a statistically distinct population rather than a slice of one season-wide distribution. A rolling average over the most recent eight bales, excluding the first to avoid cold-start artefacts, tracks the field’s local productivity and feeds into the running prediction.

Evaluated across all 120 fields, the algorithm converged from an initial mean absolute error of 2.7 bales, with a standard deviation of 4.1, down to 1.2 bales, plus or minus 2.1, by the time 90 percent of the harvest was complete, a relative error of just 4.8 percent. Both the rolling and expanding average methods outperformed a naive season-wide median of 95 metres per bale, which lagged behind at every stage. Errors also proved evenly spread across the full yield range, meaning the method works as well on lush fields as on exhausted ones, and accuracy improved most in the large fields where contractors have the most time and money at stake.

The study is honest about its limits. The boundary detection stumbled on a 5-hectare field where the operator split the work into sections, producing a 34 percent area error, and it cannot anticipate unbaled patches around obstacles or waterlogged ground. The 1 Hz GPS sampling rate may alias during fast headland turns, and the LiDAR thresholds used to label crop engagement were calibrated for grass and may not transfer to other crops. The authors also note that predictions assume a consistent operator and uniform bale chamber filling, assumptions that held across this season but may not generalise across crews.

Even so, the implications for the contracting industry are considerable. Contractors typically charge per bale, with Irish rates reaching up to 18 euro per bale for baling and wrapping, and they routinely harvest multiple fields in a single day. A prediction available after a single loop of the perimeter, refined continuously as bales accumulate, lets them match machine capacity to the job, update schedules dynamically, and quote customers with confidence, all without satellite imagery blocked by clouds, drone flights, or a preliminary site visit. Earlier research had used distance travelled to map yields after the fact in sugarcane and grapes; this is the first application, the authors say, to predicting crop remaining during a live operation.

The team plans to release the underlying dataset and benchmark metrics publicly, and future work will extend the coverage map across the whole field to correct for unharvested zones, integrate path-planning algorithms to update remaining swathe distance, and feed the predictions into dynamic fleet scheduling. For now, the study stands as a proof that sometimes the most valuable sensor on a modern farm is the one already installed, quietly logging coordinates while the real work happens behind it.

Subject of Research: Real-time round bale yield prediction using field boundary detection and distance-based analysis from GNSS machinery data

Article Title: Field boundary detection and distance-based analysis for bale yield prediction in round bale harvesting

Article References: Harkin, S. J., Crotty, T., Warren, J., Shanahan, C., Jones, E., Glavin, M., & Byrne, D. (2026). Field boundary detection and distance-based analysis for bale yield prediction in round bale harvesting. Smart Agricultural Technology, 15, Article 102470. https://doi.org/10.1016/j.atech.2026.102470

Image Credits: AI Generated

DOI: 10.1016/j.atech.2026.102470

Keywords: precision agriculture, yield prediction, GNSS, round baling, silage, field boundary detection, machine learning, agricultural machinery, swathe distance, bale distance, smart farming, coverage mapping

The study, published in Smart Agricultural Technology, draws on four growing seasons of winter wheat yield trials conducted from 2017 to 2021 at the Texas A&M AgriLife Research Experiment Station in Bushland, Texas. The trials spanned both dryland and sprinkler-irrigated conditions and represented Years 7 through 12 of the university’s wheat breeding pipeline, encompassing preliminary through advanced yield testing. Across those seasons, environmental conditions varied dramatically, including terminal drought in 2018 and 2020, optimal precipitation in 2019, and average rainfall in 2021, providing a robust test bed for models that must cope with real-world variability.

Throughout each eight-month growing season, the team flew unoccupied aerial systems over the breeding nurseries, using a DJI Phantom 4 Pro with an RGB sensor and a DJI Matrice 100 carrying a SlantRange 3P multispectral sensor. Between 20 and 32 flights were conducted per season per environment, capturing everything from early vegetative growth through senescence. From the multispectral imagery, the researchers derived three vegetation indices: the normalized difference vegetation index (NDVI) for canopy vigor, the normalized difference red-edge index (NDRE) for chlorophyll content, and the excess green index (ExG) for visible greenness. Radiometric calibration with in-field reflectance panels and survey-grade ground control points ensured that the imagery was spatially and spectrally consistent across years.

One of the study’s key methodological innovations lies in how it handles time. Most previous drone-based yield prediction studies aligned observations by calendar date or days after planting, an approach that breaks down when planting dates shift and weather patterns differ between seasons. The same calendar date can represent entirely different physiological stages in different years. Instead, the researchers aligned every flight using days relative to heading, or DRH, calculated as the flight date minus the date when 50 percent of heads in a plot reached the Feekes 10.3 stage. Heading is a biologically meaningful milestone that marks the transition into reproductive development and is closely tied to the processes that determine grain number, a major component of final yield. Because heading date is already routinely recorded in breeding programs, this alignment framework can be adopted without collecting any additional specialized trait.

With the temporally aligned dataset in hand, the team trained a diverse suite of machine learning models, including Ridge regression, Elastic Net, partial least squares regression, Random Forest, Extra Trees, XGBoost, and a fully connected neural network. Hyperparameters were tuned with Bayesian optimization, and each model was evaluated across 100 repeated random train-test splits to ensure robust performance estimates. The tree-based ensembles dominated: Extra Trees and XGBoost achieved root-mean-square errors of 6.12 and 6.13, respectively, with R-squared values of 0.953, and roughly 68 percent of their predictions fell within 10 percent of observed yield. Pairwise Wilcoxon signed-rank tests with Holm correction confirmed that these two models significantly outperformed all others, while showing no statistically significant difference from each other. Linear models clustered around RMSE values of 8.0 to 8.5, and the neural network performed far worse, with an RMSE near 19, likely because the tabular dataset was too small and structured for deep learning to shine.

But accuracy alone was not the goal. The researchers wanted to know why the models worked and when, in the crop’s development, the predictive signal was strongest. To answer this, they built an explainable AI framework with two complementary components. First, they extracted model-specific feature importance from each algorithm: impurity-based measures for the tree ensembles, absolute regression coefficients for the linear models, and latent-component weights for PLSR. Second, they validated these rankings with permutation importance, a model-agnostic technique that measures how much prediction accuracy drops when a single feature’s values are randomly shuffled in held-out test data. Because permutation importance directly evaluates changes in performance on unseen data, it is less susceptible to the biases that can distort model-specific metrics when predictors are correlated.

The results converged with striking consistency. Temporal correlation analysis showed that the relationship between vegetation indices and yield rose steadily through the vegetative stage and peaked near heading. The ExG index reached its maximum correlation with yield, an absolute r of 0.86, exactly at heading (DRH 0), while NDVI and NDRE peaked slightly earlier, between 15 and 10 days before heading, with correlations of 0.81 and 0.83. Feature importance analysis told the same story: the most influential predictors across the tree-based models were packed into the window from roughly 25 days before heading to one day after, with ExG at one day after heading emerging as the single most important feature for XGBoost and NDRE at 15 days before heading dominating for Random Forest. Irrigation also ranked as a major driver of yield variability, confirming the dominant role of water management in the Texas High Plains.

The physiological explanation for this heading-centered window is compelling. In the weeks preceding anthesis, wheat undergoes rapid spike growth and floret development, during which assimilate availability strongly influences floret survival and, ultimately, grain number. Canopies that maintain greater green leaf area, chlorophyll content, and biomass accumulation during this period have more capacity to support both spike growth and photosynthesis as the crop enters reproduction. NDRE is particularly informative here because red-edge reflectance remains sensitive to chlorophyll variation even under dense canopies, where NDVI begins to saturate. ExG, meanwhile, captures visible greenness right at the transition. Together, the three indices integrate complementary aspects of canopy condition during the very period when the crop’s reproductive sink strength is being established, which is why spectral differences detected near heading carry so much predictive weight.

The framework also proved its mettle under the toughest test: predicting entirely unseen growing seasons. Using a leave-one-year-out validation scheme, in which each season was held out once as an independent test set, the ensemble tree models consistently outperformed linear and neural network approaches. Even under these shifted conditions, feature selection analysis revealed that the most reliable predictors clustered between 15 days before and 10 days after heading, with ExG at one day after heading repeatedly chosen across models and test years. Early-season features, by contrast, were rarely selected and contributed little. The 2019 season proved easiest to generalize to, while 2021, with its distinct weather and management profile, produced the largest errors, underscoring the importance of training on environmentally diverse datasets.

The practical implications are immediate. Rather than scheduling uniform drone flights across the entire season, breeding programs can concentrate acquisitions in the 15 days surrounding heading, where each flight delivers maximum information per unit of cost and processing effort. The authors are careful to note that the specific window reflects the environments and germplasm evaluated at a single Texas station, and they recommend future ablation studies directly comparing flight schedules, along with testing across additional locations and sensor types. But the analytical framework itself, phenology alignment by days relative to heading, multi-model machine learning, and dual explainability validation, is not site-specific. By transforming black-box predictions into physiologically interpretable, time-resolved guidance, the study offers a template for smarter, cheaper, and more trustworthy crop monitoring, one that tells breeders not just what the drones see, but exactly when they should be looking.

Subject of Research: Phenology-aligned UAS remote sensing with explainable machine learning for winter wheat grain yield prediction

Article Title: Strategically timed UAS data acquisition for wheat yield prediction using explainable AI: A multi-year phenology-aligned study

Article References: Baker, S., Islam, M. N., Bhandari, M., Chang, A., Ibrahim, A. M., Jung, J., Landivar, J., Liu, S., Mudumba, V. S., Pokharel, R., Rudd, J., Scott, J. L., & Gentimis, T. (2026). Strategically timed UAS data acquisition for wheat yield prediction using explainable AI: A multi-year phenology-aligned study. Smart Agricultural Technology, 15, Article 102606. https://doi.org/10.1016/j.atech.2026.102606

Image Credits: AI Generated

DOI: 10.1016/j.atech.2026.102606

Keywords: wheat, yield prediction, drones, UAS, explainable AI, machine learning, phenology, vegetation indices, high-throughput phenotyping, plant breeding, precision agriculture, NDVI

The answer matters because breeding programmes work on a rolling basis. Each selection cycle introduces new candidate lines and retires others, so the genotypes requiring evaluation in any given year are typically absent from the historical data available for model development. Yet many published UAV yield studies report impressive accuracy using random K-fold cross-validation, a procedure that can place replicate plots of the same line and season on both sides of the train-test split. The new research, led by Dastan Yelubayev and colleagues at the A.I. Barayev Research and Production Centre for Grain Farming, quantified just how optimistic that practice is. On identical data, random ten-fold cross-validation produced a coefficient of determination of 0.890, while holding out complete breeding lines dropped the same model to 0.587, an optimism gap of roughly 0.30 R-squared units attributable purely to the validation design.

The experiment itself was a substantial piece of field logistics. Across two seasons the team flew a DJI Phantom 4 Multispectral drone on seven and nine missions respectively, capturing five-band reflectance imagery at 2.1 to 2.9 centimetre resolution from altitudes of 30 to 45 metres. A separate LiDAR survey each year, flown on a Matrice 300 RTK with a Zenmuse L1 sensor, reconstructed the underlying terrain surface through cloth-simulation filtering, yielding static covariates such as plot slope and the fraction of each plot sitting in topographic depressions. Hourly ERA5-Land reanalysis variables, including precipitation, vapour-pressure deficit, soil moisture and solar radiation, were aggregated over line-specific phenological windows spanning emergence to heading and heading to maturity. In total, 82 candidate predictors per plot entered the analysis pipeline.

Feature selection proved humbling. Screening and recursive elimination with a gradient-boosted LightGBM learner reduced the pool to 11 predictors, dominated by eight vegetation index features, mostly computed over the grain-filling window, one ERA5-Land soil moisture descriptor and two LiDAR terrain features. But when the entire selection procedure was repeated on 200 bootstrap resamples of the 22 training breeding lines, the number of retained predictors ranged from 5 to 50, and the exact 11-predictor set reappeared in only 8 percent of repetitions. The authors are explicit that the selected set should be regarded as one outcome of an unstable process, not a uniquely identified representation, a caveat that applies to much of the small-data machine learning literature in agricultural remote sensing.

The headline result concerns ranking rather than absolute prediction. For the 31 breeding lines entirely absent from training, the correlation between predicted and observed line-mean yields was 0.612 in 2024 and 0.650 in 2025, both statistically significant under Monte Carlo permutation tests. Because only 14 and 17 independent genotypes were available in the two seasons, the genotype-level confidence intervals were wide, spanning roughly 0.65 units, but the positive association held in both years despite the enormous seasonal contrast. For breeders, this within-season ordering is the quantity that matters: a model that ranks candidates correctly can help prioritise lines for further testing, even if its absolute yield estimates carry substantial error.

The most sobering finding came from the strictly prospective experiment. The team built an entire pipeline, from screening through model configuration, using 2024 data alone, froze it, and applied it to 2025. It failed completely. Predictions were biased upward by about 11.4 centners per hectare, with a predicted mean near 19 centners for a season whose observed mean was 7.5, and rank correlations were indistinguishable from zero across all target subsets. Crucially, the failure persisted even among genotypes that had been observed in 2024, so genotype novelty was not the cause. The likely culprit is extrapolation: the grain-filling soil moisture descriptor ranged from 0.4299 to 0.4333 cubic metres per cubic metre across the 2024 training plots, while every 2025 value fell between 0.3435 and 0.3605, a gap roughly twenty times the width of the training range. Tree-based models cannot extrapolate beyond their terminal regions, and the 2025 season lay entirely outside the learned domain.

Attribution analysis reinforced the seasonal-regime interpretation. The ERA5-Land layer-2 soil moisture descriptor carried the largest mean absolute SHAP contribution at 4.84 centners per hectare, about 2.3 times the next feature, and 99.6 percent of its training variance lay between seasons. Its signed attribution swung from positive in 2024 to negative in 2025, a 9.5 centner swing, indicating the model used it as a coarse indicator separating the two years rather than as a plot-scale soil water measurement. Because the entire nursery falls within a single reanalysis grid cell, differentiation among lines arises only through their phenological windows, which slice the shared climate series differently for each genotype. When the team neutralised this line-specific alignment by rebuilding all aggregations on fixed calendar windows, the 2024 rank correlation collapsed from 0.557 to 0.172 while 2025 held at 0.628, showing how fragile one season’s signal was.

Architecture comparisons delivered a further anti-climax for the deep learning enthusiasts. Six model families, including ridge and LASSO regression, random forest, a regime-conditional ensemble and a Transformer with a year-domain-adversarial head, produced pooled test R-squared values between 0.449 and 0.596, with overlapping uncertainty intervals. The architecture with the highest pooled accuracy, the Transformer, had the weakest 2024 rank correlation, while random forest led that season’s ranking. All ten paired contrasts of rank correlation against LightGBM included zero. Modality ablation told a similar story: removing the UAV spectral block caused the largest degradation, exceeding the refitting sensitivity scale, but the full four-modality combination was separated by 0.002 R-squared units or less from simpler configurations pairing UAV data with one complementary block. No evidence emerged that the complete multimodal stack outperformed simpler alternatives.

The authors are careful about what their framework is and is not. Because several predictors derive from grain-filling observations, the ranking becomes available only late in the season, so this is not an early-season forecast tool. It is best interpreted as a secondary candidate-ranking analysis for next-stage review, complementing rather than replacing routine harvest measurements and multi-environment yield trials. With only two seasons from a single nursery complex, generalisation across locations, soil types and future climates remains untested, and the 12.2 percent of variance attributable to genotype plus genotype-by-environment interaction sits at the low end of published values, pending confirmation over more environments.

Still, the study’s methodological message may prove its most durable contribution. The 0.30 R-squared optimism gap between random and genotype-held-out validation, and the complete failure of a prospective transfer across a strong seasonal shift, are quantified warnings for a field where headline accuracies above 0.85 are routinely reported under favourable validation designs. The authors recommend that UAV cereal yield studies report a complete-line holdout alongside conventional random cross-validation, and, where multiple seasons exist, a prospective evaluation in which no information from the target season enters model development. In an era when climate variability is tightening its grip on the Eurasian breadbasket, knowing what a yield model cannot do may be as valuable as knowing what it can.

Subject of Research: UAV-based prediction of yield ranking for spring wheat breeding lines absent from model training across two contrasting seasons in northern Kazakhstan

Article Title: Yield ranking of spring wheat breeding lines absent from model training within two contrasting seasons in northern Kazakhstan

Article References: Yield ranking of spring wheat breeding lines absent from model training within two contrasting seasons in northern Kazakhstan. (n.d.). https://doi.org/10.1016/j.atech.2026.102588

Image Credits: AI Generated

DOI: 10.1016/j.atech.2026.102588

Keywords: spring wheat, UAV multispectral imaging, yield prediction, breeding lines, Kazakhstan, machine learning, LightGBM, ERA5-Land, LiDAR terrain, cross-validation, genotype-by-environment, drought

The research, published in the Journal of Agriculture and Food Research, grew the semi-dwarf Basmati variety PB 1692 during the kharif seasons of 2022 and 2023 in a split-plot experiment. Five nitrogen regimes were tested against four zinc fertilization strategies, from conventional prilled urea at the full recommended dose of 130 kilograms of nitrogen per hectare to reduced doses supplemented with foliar sprays of liquid nano-urea, nano-zinc oxide, and even a cyanobacterial formulation based on Anabaena torulosa. From 300 plot-level observations, the team compiled 22 variables spanning plant height, tiller density, dry matter production, panicle characteristics, and nitrogen and zinc concentrations and uptake in straw, hull, bran, and white rice kernels.

Those variables became the training ground for five machine learning algorithms: multiple linear regression, support vector regression, k-nearest neighbors, gradient boosting, and random forest. The researchers applied recursive feature elimination with 10-fold cross-validation to select the most informative predictors and used exhaustive grid-search optimization to tune the random forest’s hyperparameters, fixing the number of trees at 500 after confirming error convergence. A strict bias-variance criterion, requiring the difference in R-squared between training and cross-validation to stay below 10 percent, guarded against overfitting, and final performance was confirmed on a held-out test set using Lin’s concordance correlation coefficient.

The random forest emerged as the clear champion. It achieved R-squared values of 0.97 in 2022 and 0.96 in 2023, with root mean square errors of just 0.08 tonnes per hectare and mean absolute errors of 0.063, outperforming support vector regression, which averaged an R-squared of 0.92, and leaving k-nearest neighbors, the weakest performer at R-squared values between 0.73 and 0.80, far behind. Taylor diagrams, which simultaneously display correlation, standard deviation, and error, visually confirmed the random forest’s dominance in both training and prediction stages across both seasons.

More revealing than raw accuracy was what the model considered important. Using two feature-importance metrics, the percentage increase in mean squared error and the increase in node purity, the random forest ranked total nitrogen uptake, grain nitrogen concentration, dry matter production, and total zinc uptake as the dominant drivers of yield. Plant height and tiller counts, traits farmers and agronomists often watch closely, carried relatively little predictive weight. The message is biologically coherent: fertilizer application alone guarantees nothing unless the crop actually captures, assimilates, and remobilizes nitrogen into grain, processes that depend on root architecture, photosynthetic capacity, and source-sink coordination during grain filling.

The field results reinforced that insight. The full recommended dose of 130 kilograms of nitrogen per hectare produced the highest grain yields, 4.79 and 4.52 tonnes per hectare in 2022 and 2023, gains of 25.4 and 28.0 percent over the unfertilized control, along with nitrogen uptake of roughly 120 kilograms per hectare and zinc uptake about 35 percent higher than in unfertilized plots. But the treatment combining 97.5 kilograms of soil nitrogen with two foliar sprays of nano-urea produced statistically equivalent yields, demonstrating that a 25 percent reduction in mineral nitrogen input could maintain productivity when paired with the nanoscale foliar supplement.

Zinc fertilization delivered its own measurable benefits. Foliar sprays of 0.1 percent nano-zinc oxide produced the strongest yield attributes, raising panicle weight by 9.6 percent and grains per panicle by 4.1 percent on a two-year mean basis compared with no zinc, and adding roughly 0.2 tonnes of grain per hectare. Zinc uptake reached nearly 600 grams per hectare under the nano-zinc treatment, a 10.7 percent increase over the control. Because zinc deficiency remains a widespread human nutritional problem, agronomic biofortification of rice grain through such foliar strategies carries significance well beyond the yield column.

The climate accounting added a compelling twist. Using the CCAFS Mitigation Option Tool, the researchers estimated greenhouse gas emissions in carbon dioxide equivalents, incorporating both fertilizer production and field-induced emissions. The 97.5-kilogram nitrogen plus nano-urea treatment cut absolute emissions by 188.9 kilograms of carbon dioxide equivalent per hectare, a 6.55 percent reduction, and lowered emission intensity by 3.22 to 4.62 percent relative to the full recommended dose, all without a yield penalty. Deeper cuts to 65 kilograms of nitrogen, whether with nano-urea or conventional urea, saved more emissions, 377.2 and 323.5 kilograms per hectare respectively, but at a significant cost in grain, defining the agronomic boundary of an acceptable trade-off. Strikingly, the unfertilized control had the worst emission intensity per kilogram of grain, showing that yield and climate goals are inseparable in irrigated rice.

The study’s authors are candid about limitations. Weather variables were excluded because the two experimental seasons were climatically stable, yet weather remains a major determinant of crop performance, and future models should incorporate it. The work spanned only two years and a single rice variety, and several of the most powerful predictors, including total nutrient uptake and biomass, require destructive sampling and laboratory analysis that are impractical for routine farm use. The researchers argue that the next step is to build simplified, transferable frameworks using readily available inputs such as weather, soil data, fertilizer records, and remotely sensed vegetation indices, potentially augmented by drone observations and process-based crop models.

Even with those caveats, the study marks a meaningful advance in precision agriculture. Most machine learning yield studies lean on satellite or meteorological data; this one integrated field-measured agronomic, physiological, and nutrient variables within a single interpretable framework, and the model’s importance rankings aligned with established crop physiology rather than statistical artifacts. For a crop on which billions of people depend, the combination of a random forest that explains 96 to 97 percent of yield variability and a fertilizer strategy that trims nitrogen and emissions without sacrificing harvest offers a rare, quantified win for both food security and the climate.

Subject of Research: Machine learning prediction of rice grain yield and yield-emission trade-offs under nitrogen and zinc fertilization in a two-year field experiment

Article Title: Rice yield prediction using machine learning models and yield-emission trade-offs from field experiment

Article References: Reddy, K. S., Shivay, Y. S., Prasanna, R., Kumar, D., Peramaiyan, P., Mandi, S., Nayak, S., Baral, K., Alekhya, G., Reddy, K. S., & Borate, R. B. (2026). Rice yield prediction using machine learning models and yield-emission trade-offs from field experiment. Journal of Agriculture and Food Research, 31, Article 103309. https://doi.org/10.1016/j.jafr.2026.103309

Image Credits: AI Generated

DOI: 10.1016/j.jafr.2026.103309

Keywords: rice, machine learning, random forest, nitrogen uptake, nano-urea, zinc biofortification, greenhouse gas emissions, yield prediction, precision agriculture, nutrient use efficiency, Basmati rice, sustainable farming

The research team, led by Renival Almeida de Carvalho and José Luiz Rodrigues Torres, set out to test a hypothesis that had received surprisingly little attention in the precision agriculture literature: that crop phenology, the sequence of developmental stages a plant passes through, is a major driver of how well satellite data can predict both how much cane a field produces and how much recoverable sugar it contains. To do so, they worked across five commercial sugarcane fields totaling 98.1 hectares, all planted with the widely adopted CTC-04 variety and covering second- and third-ratoon crops on soils ranging from clayey to sandy. The fields sit roughly 800 meters above sea level in a tropical savanna climate classified as Aw under the Köppen–Geiger system, with about 1,500 millimeters of annual rainfall and a mean annual temperature of 21 degrees Celsius.

Field data collection was deliberately grounded in the realities of commercial production rather than experimental manipulation. No fertilizer rates, irrigation regimes, or management treatments were imposed; instead, the natural variability of the five fields, including differences in soil texture, crop age, canopy development, and yield potential, served as the source of variation the models had to capture. Using a geographic information system, the researchers distributed 51 georeferenced sampling points across the fields, each corresponding to a three-square-meter unit consisting of two one-meter rows of stalks spaced 1.5 meters apart. Every stalk within each unit was harvested and weighed with a digital scale accurate to 10 grams, allowing the team to compute total cane yield, expressed in tonnes per hectare, and net cane yield after stripping leaves and apical portions. Ten stalks per point were then sent to the Coruripe Plant laboratory for determination of total recoverable sugar, the industry-standard measure of technological quality used for commercial payment and harvest planning.

On the remote sensing side, the study harnessed two complementary satellite systems. Optical imagery came from the Sentinel-2 Level-2A surface reflectance collection, processed in Google Earth Engine with cloud and cloud-shadow pixels masked using the sensor’s quality assessment layers. From Sentinel-2’s 10-meter bands, the team calculated five vegetation indices: NDVI, EVI, NDRE, GNDVI, and VARI, the last of which uses only visible bands and therefore represents a potential low-cost alternative built on RGB imagery alone. Radar observations came from the ALOS/PALSAR-2 sensor operating in ScanSAR mode at roughly 25-meter resolution, providing HH and HV polarization backscatter and their difference, metrics sensitive to canopy structure, biomass, and water status and, crucially, unaffected by cloud cover. Images were acquired at two contrasting phenological stages: February 2023, during active vegetative development characterized by rapid canopy expansion and biomass accumulation, and the pre-harvest maturation window between May and July 2023, when growth slows and sucrose accumulates in the stalks.

The agronomic data revealed just how heterogeneous commercial sugarcane production can be. Mean total cane yield across the five fields ranged from 82.46 to 130.28 tonnes per hectare, while net cane yield spanned 70.66 to 100.64 tonnes per hectare. At the level of individual sampling points, total cane yield varied from 71.00 to a remarkable 252.50 tonnes per hectare, with coefficients of variation of 29.49 percent for total yield and 28.76 percent for net yield, figures that underscore why conventional field-based assessment struggles to capture within-field variability. Total recoverable sugar, by contrast, was far more stable, averaging 163.63 kilograms per tonne with a coefficient of variation of only 6.60 percent, a difference that would prove important for interpreting the modeling results.

Using multiple linear regression with best-subset predictor selection, variance inflation factor screening for multicollinearity, and leave-one-out cross-validation, the team fitted separate models for each response variable at each phenological stage. The results delivered a clear message about timing. For total cane yield, the best model, based on NDVI acquired during vegetative development, explained 60.3 percent of the observed variation, with a root mean square error of 20.64 tonnes per hectare and a mean absolute percentage error of 18.17 percent. The corresponding pre-harvest NDVI model managed only an R² of 0.401, with the error climbing to 24.99 tonnes per hectare. The same pattern held for net cane yield: the vegetative-stage NDVI model reached an R² of 0.548 with an RMSE of 18.24 tonnes per hectare, whereas the best pre-harvest models fell to R² values between 0.404 and 0.474 with larger errors. In other words, the same satellite, the same index, and the same fields produced meaningfully better predictions simply because the images were captured while the crop was actively growing.

The physiological explanation lies in what the canopy is doing during each stage. During vegetative development, rapid leaf area expansion, high chlorophyll content, and intense photosynthetic activity strengthen the link between spectral reflectance and biomass, allowing indices like NDVI to track the canopy traits most closely tied to final stalk yield. By the pre-harvest stage, the picture becomes muddier: dense canopies push NDVI toward its well-known saturation point, where additional biomass produces diminishing changes in red and near-infrared reflectance, while senescent leaves and accumulated crop residues further decouple canopy reflectance from stalk mass. Interestingly, the RGB-based VARI index, though generally outperformed by the near-infrared indices, still produced significant yield models, suggesting that freely available visible-band imagery could serve as an accessible fallback when multispectral data are unavailable, even if it remains more vulnerable to illumination conditions and soil background effects.

Total recoverable sugar told a strikingly different story. Unlike yield, sugar concentration is governed by internal biochemical processes, including carbon assimilation, assimilate translocation, respiration, sink strength, and the partitioning of photoassimilates among plant organs, processes only indirectly visible to a satellite measuring canopy reflectance. Accordingly, TRS models showed moderate coefficients of determination, around 0.550 during vegetative development using NDVI alone and 0.504 during pre-harvest using a combined NDRE plus NDVI model. Yet the prediction errors were remarkably low, with mean absolute percentage errors below 4 percent at both stages, and the pre-harvest model achieving the lowest RMSE at 7.19 kilograms per tonne. The authors emphasize that this contrast between moderate R² values and low errors illustrates why goodness-of-fit and error metrics must be read together: because TRS varies within a narrow range, a model of moderate explanatory power can still deliver operationally accurate estimates. The success of the NDRE and NDVI combination at pre-harvest likely reflects the red-edge band’s resistance to saturation and its sensitivity to chlorophyll status in dense canopies, capturing physiological conditions associated with sucrose accumulation that NDVI alone misses. Notably, RGB-only indices failed to produce significant TRS models, confirming that visible bands alone cannot detect the biochemical processes behind sugar content.

Radar metrics, despite faithfully following the seasonal arc of crop development, delivered weaker predictive relationships than the optical indices, a limitation the authors attribute to SAR backscatter’s entanglement with vegetation architecture, water content, soil moisture, surface roughness, incidence angle, and sensor configuration, as well as its own saturation behavior in high-biomass crops. Still, the team argues that SAR remains essential in tropical sugarcane regions where persistent cloudiness can blind optical sensors precisely during critical windows, and that future monitoring systems will benefit from integrating optical and radar data streams rather than choosing between them. The authors are candid about the study’s limits: validation relied on leave-one-out cross-validation without an independent external dataset, the work covered a single growing season, one farm, and one cultivar, and only interpretable linear models were tested, leaving comparisons with machine-learning approaches and the integration of soil, weather, and management data to future research. Even so, the core conclusion stands as a practical guide for the industry: successful satellite-based sugarcane monitoring demands not just the right sensor and the right index, but the right moment, with vegetative-stage imagery favoring yield forecasts and pre-harvest red-edge combinations favoring sugar quality estimates. As Brazil’s sugar and ethanol sectors push ever deeper into the Cerrado, aligning the satellite calendar with the crop’s internal clock may prove one of the cheapest accuracy upgrades available.

Subject of Research: Influence of crop phenology on the accuracy of satellite-based remote sensing predictions of sugarcane yield and total recoverable sugar in commercial fields of the Brazilian Cerrado.

Article Title: Crop phenology drives remote sensing prediction accuracy of sugarcane yield and quality in the Brazilian Cerrado

Article References: de Carvalho, R. A., Torres, J. L. R., Loss, A., Abreu, D., da Silva Vieira, D. M., & Pereira, D. P. (2026). Crop phenology drives remote sensing prediction accuracy of sugarcane yield and quality in the Brazilian Cerrado. Smart Agricultural Technology, 15, Article 102464. https://doi.org/10.1016/j.atech.2026.102464

Image Credits: AI Generated

DOI: 10.1016/j.atech.2026.102464

Keywords: sugarcane, remote sensing, crop phenology, NDVI, Sentinel-2, Brazilian Cerrado, yield prediction, total recoverable sugar, precision agriculture, SAR, vegetation indices, Smart Agricultural Technology

The platform, called AgroYachay—’yachay’ meaning knowledge in Quechua—is described in the journal SoftwareX and released under the MIT licence with an archived DOI, version 1.0.1. It is the cloud counterpart of a companion tool, AgroCommish, which handles the upstream work of commissioning ESP32 sensor nodes: flashing firmware, discovering sensor pins, provisioning WiFi credentials and registering each verified device with the cloud. Together, the two tools form a complete, reproducible device-to-decision pipeline that a small research group or agricultural programme can deploy entirely on open-source software and commodity hardware, with no per-seat licensing or proprietary cloud lock-in.

Architecturally, AgroYachay follows a three-tier edge-cloud design. At the edge, ESP32 microcontrollers wired to a DHT11 sensor for air temperature and humidity and an FC-28 probe for soil moisture post JSON readings over WiFi every ten seconds. The backend is a Flask 3 application of roughly 7,300 lines of Python, organised into controllers, feature modules and a service layer, with state persisted across fifteen normalised tables in PostgreSQL 17. Authentication relies on JSON Web Tokens with bcrypt password hashing and optional Google OAuth. The frontend is a React 18 single-page application of about 13,000 lines styled with Tailwind CSS, charting live sensor series with Recharts. More than ninety REST API routes tie the system together.

The intelligence layer is deliberately hybrid. Image-based pest diagnosis runs on a self-hosted open multimodal model—qwen2.5-VL-3B, served through Ollama—which keeps farmers’ photographs on local hardware rather than shipping them to a third-party image service. A grower photographs an affected leaf; the vision model, prompted as a phytopathologist specialising in Andean crops, returns a structured verdict covering health status, disease name, confidence, severity, causes, treatment and prevention. A follow-up text-only call to a cloud large language model hosted by Groq converts that verdict into an actionable management plan: urgency, step-by-step actions, chemical products with doses available in Peru, an organic alternative, and guidance on when to consult a human agronomist. The prompt targets potato, quinoa, maize, faba bean, oca and cañihua, the staple crops of the region.

The team validated the vision module with unusual candour, evaluating it as a plant-health triage aid rather than a fine-grained diagnostic tool. Using 509 images—300 laboratory photographs from PlantVillage and 209 field images from PlantDoc—they found perfect sensitivity: every one of the 409 diseased leaves was flagged as diseased, the property that matters most for an early-warning system. Specificity was lower at 62 percent, meaning the model over-flags healthy leaves as diseased, a conservative bias favouring missed nothing over missed disease. Species-level discrimination was weaker, with late blight frequently confused with early blight and a macro-F1 of 0.55. The authors state plainly that the module supports binary field triage, not reliable differential diagnosis, and that performance on real Andean field imagery remains untested. Repeated queries at the deployed temperature of zero produced identical classifications in ten out of ten trials, confirming reproducibility.

Beyond diagnosis, a context-aware conversational assistant answers free-form questions in Spanish, conditioned on the farmer’s registered crops, region and latest sensor values. Separate services translate the five-day OpenWeather forecast into a risk level, a weekly activity plan and optimal-day recommendations for spraying, sowing and irrigation. A crop management module records species, variety, parcel area, GPS location, sowing date and phenological stage, ensuring that every reading, alert, diagnosis and estimate is tied to a specific crop and device. An inputs module calculates fertiliser and agrochemical requirements from parcel area and target yield, while an advisory module queues requests for human agronomists, keeping an expert in the loop when the automated tools should not act alone.

The platform’s most distinctive feature may be its economic layer. Rather than a black-box regressor, the yield-and-revenue estimator is a transparent multiplicative factor model requiring no training data, which suits the sparse-data reality of Andean deployment. A crop-specific base yield—15.5 tonnes per hectare for potato, 1.8 for quinoa, 8.5 for oca—is adjusted by climate factors computed from mean temperature, humidity and accumulated rainfall within phenological windows, by a phenological factor that grows with crop progress, and by a piecewise area factor. The projected tonnage is multiplied by a regional reference price to give expected gross revenue. Because the model is purely multiplicative, its sensitivity is fully transparent: a 10 percent error in any climate factor moves the estimate by exactly 10 percent, and every contributing factor is displayed to the user. Results include pessimistic, probable and optimistic scenarios with a confidence score, and can be exported as styled PDF reports suitable for credit or crop-insurance applications.

The multilingual interface is treated as a first-class feature rather than an afterthought. Of 1,488 UI strings, 94 percent are localised into both Quechua and Aymara, with a complete English and Spanish locale; only technical terms such as ‘IoT’ and ‘ESP32’ are deliberately left untranslated. This matters because a substantial share of Andean smallholders operates primarily in indigenous languages, and generic farm-management platforms assume Spanish or English literacy. The authors note an honest limitation here: while the interface is multilingual, the LLM-generated advice, pest verdicts and exported reports are currently produced in Spanish only, so a Quechua-speaking user still encounters Spanish text at the moment of advice. Extending the prompts and report templates to indigenous languages is the next inclusion step.

Performance measurements on the production server show that core operations are genuinely interactive: lightweight REST calls complete in 2.6 milliseconds at the median, dashboard reads in under 10 milliseconds, and sensor ingestion in 19 milliseconds. A 22-minute bench test of a real ESP32 node delivered 132 of 134 expected sampling cycles, a 98.5 percent success rate with a maximum reconnection gap of 20 seconds. The exception is the vision module: on the CPU-only demonstration server a single diagnosis takes minutes, whereas on a consumer GPU it completes in seconds, so GPU-backed deployment is recommended for interactive use. The conversational assistant, dominated by external inference, averages around two seconds per response.

The developers, based at the Universidad Nacional del Altiplano in Puno, drew requirements from direct field experience in a smallholder economy centred on potato, quinoa, cañihua and Andean livestock, where commercial precision-agriculture suites are cost-prohibitive. They self-tested the platform on their own plantings and received positive informal feedback from local farmer demonstrations, though a formal quantitative user-acceptance study remains future work. So do calibration of the yield estimator against actual harvest records, an expert-labelled Andean field image benchmark, an on-device offline fallback for pest screening, and security hardening measures such as per-device authentication, database encryption at rest and automated secret rotation. Even without the LLM layer, sensor monitoring, alerts, the factor-model estimator and reporting already function. As an early but complete contribution, AgroYachay demonstrates that the entire chain—from flashing a sensor in a highland workshop to issuing a financed-harvest report—can now run on open-source software, potentially reshaping who gets to participate in precision agriculture across Latin America and other developing regions.

Subject of Research: An open-source IoT and large-language-model platform supporting agronomic and economic decision-making for Andean smallholder farmers.

Article Title: AgroYachay: An open-source IoT and large-language-model platform supporting agronomic and economic decision-making for Andean smallholders

Article References: Torres-Cruz, F., Vilca-Solorzano, R. A., Yana-Yucra, D. M., Ibañez-Quispe, V., & Fuentes-Navarro, E. L. (2026). AgroYachay: An open-source IoT and large-language-model platform supporting agronomic and economic decision-making for Andean smallholders. SoftwareX, 36, Article 103037. https://doi.org/10.1016/j.softx.2026.103037

Image Credits: AI Generated

DOI: 10.1016/j.softx.2026.103037

Keywords: AgroYachay, precision agriculture, IoT, ESP32, large language models, smallholder farming, Andean agriculture, plant disease diagnosis, open-source software, yield prediction, Quechua, Peru

The research, led by Abiodun Fatai Olayinka of the West Africa Centre for Crop Improvement at the University of Ghana and the International Institute of Tropical Agriculture in Ibadan, Nigeria, set out to develop a rapid phenotyping protocol for cassava using the Normalized Difference Vegetation Index, or NDVI. NDVI is calculated from the difference between near-infrared and red light reflectance divided by their sum, a formula that exploits the way healthy plant canopies strongly absorb red light for photosynthesis while reflecting near-infrared radiation from their internal leaf structure. Because NDVI rises with vegetative vigour, it has long been used from satellites and drones to monitor crop growth. But those platforms are costly and technically demanding for many resource-constrained breeding programs, so the team turned to the Trimble GreenSeeker, an affordable handheld meter that emits red light at 660 nanometres and near-infrared light at 780 nanometres and reads back the reflectance in real time.

The scale of the trial was substantial. The researchers evaluated 453 cassava accessions, including clonal evaluation lines and commercial checks, across two contrasting agroecological zones in Nigeria during the 2021/2022 planting season. The first site, Mokwa, lies in the Southern Guinea Savannah at 180 metres above sea level, with sandy loam soils, bimodal rainfall of roughly 1200 millimetres per year and temperatures ranging from 28 to 35 degrees Celsius. The second, Onne, sits in the Humid Forest zone just 6.8 metres above sea level, with clay loam soils, abundant rainfall near 2700 millimetres per year and cooler temperatures of 24 to 32 degrees. Using an augmented block design with two replications per location, the team collected NDVI readings at three, six and nine months after planting, holding the sensor one metre above the canopy at a 45-degree angle and averaging three readings per plot, alongside ground-truth measurements of 26 agronomic traits covering yield, plant architecture and root quality.

The genetic parameters that emerged from the trial were encouraging for breeders. Broad-sense heritability estimates—the proportion of observed variation attributable to genetics—were moderate to high for the traits that matter most: 0.56 for fresh root yield, 0.61 for dry matter content, 0.61 for starch content and 0.66 for harvest index. These figures indicate that selection should deliver real genetic gains. Heritability ranged from zero for NDVI at three months up to 0.87 for height at first branch at the same stage. Genotypic coefficients of variation reached nearly 49 percent for the number of lodged plants per plot, and the highest genetic advance as a percentage of the mean, 71.57 percent, was recorded for that same lodging trait, meaning the trait would respond dramatically to selection pressure applied to the top five percent of the population.

Perhaps the most striking biological finding concerned plant architecture. Genotypic correlations revealed strong negative relationships between lodging—the tendency of plants to lean or collapse—and key yield components: fresh root yield correlated at minus 0.56 and harvest index at minus 0.72. At Mokwa, the correlation between harvest index and the number of lodged plants per plot reached minus 0.73, while lodging showed strong positive correlations with plant height at six and nine months, at 0.87 and 0.79 respectively. In other words, tall plants tend to fall over, and falling over costs yield. Path coefficient analysis reinforced the picture: at Mokwa, root weight exerted a direct positive effect of 1.0 on fresh root yield, harvest index boosted root weight by 0.69, and plant height at nine months contributed positively through root weight, while the number of lodged plants and plant height at six months dragged root weight down.

The predictive heart of the study lay in the regression models linking NDVI to agronomic traits. In the Mokwa trial, NDVI measured at six months after planting predicted fresh root yield, dry yield and root weight with a coefficient of determination of 0.90—an exceptionally strong result for a single handheld reading. Plant height was predicted with even greater precision at early stages, with NDVI at three months explaining 94 percent of the variation in plant height at three months. Harvest index was moderately predicted at around 0.7 using NDVI at six and nine months, and quality traits such as dry matter content showed low error margins, with root mean squared errors between 1.2 and 1.6 percent. For cassava breeders who currently must wait until harvest, often nine to twelve months after planting, to know what a genotype can deliver, a mid-season spectral reading that forecasts yield this accurately is transformative.

But the story was not uniformly rosy. At Onne, the humid forest site, prediction accuracy dropped to moderate levels, with R-squared values around 0.50 for fresh root yield, dry yield and root weight predicted from NDVI at nine months, and error terms nearly doubled—root mean squared error rose to 15.3 tonnes per hectare and mean absolute error to 13 tonnes, compared with 8.3 and 7.4 tonnes per hectare at Mokwa, roughly a 50 percent error reduction in the savanna site. The explanation lies in the atmosphere and the soil. Mokwa enjoys higher solar irradiance and lower atmospheric humidity, conditions that favour stable spectral reflectance, while Onne’s persistent cloud cover and heavy rainfall introduce noise and interference that degrade the consistency of NDVI data. Similar patterns have been reported elsewhere, with prediction accuracy declining under wet, saturated field conditions.

An intriguing paradox also surfaced in the genetic analysis. Despite NDVI’s superb predictive performance, its broad-sense heritability across all timepoints was essentially zero, driven by very low genotypic variance—peaking at just 0.00029—and high error variance. The observed variation in NDVI, the authors conclude, was largely non-genetic, reflecting the sensor’s sensitivity to micro-environmental noise such as light variability, soil exposure and inconsistent canopy closure. This positions NDVI not as a heritable trait in its own right but as an environmentally responsive estimator of phenotypic performance, valuable precisely because it integrates the crop’s response to its growing conditions. The practical implication is clear: handheld NDVI readings are excellent for rapid phenotyping and mid-season selection decisions, but their reliability depends heavily on site conditions, and breeders should treat predictions from different environments with appropriate caution.

The broader significance of the study extends well beyond cassava. Strong genotype-by-environment interaction was evident for most traits, echoing earlier Nigerian and South African trials, and the authors argue that genetic improvement remains achievable through multi-environment testing, selection for broad or specific adaptation, and predictive models that explicitly incorporate these interactions. Because dry matter content is highly correlated with starch content, selection for one can drive gains in the other, and the moderate heritability combined with a genetic advance of roughly 20 percent for starch content suggests meaningful improvement is within reach for industrial-quality roots. With error reductions of this magnitude available from a sensor that costs a tiny fraction of drone or satellite systems, and with data and code openly archived on Zenodo, the study offers breeding programs across the tropics a practical, low-cost route to faster variety development—provided they respect the environmental limits of what a single greenness reading can tell them.

Subject of Research: NDVI-based high-throughput phenotyping for accelerating cassava genetic improvement

Article Title: Accelerating cassava genetic improvement through NDVI-based high-throughput phenotyping

Article References: Accelerating cassava genetic improvement through NDVI-based high-throughput phenotyping. (n.d.). https://doi.org/10.1186/s44399-026-00037-x

Image Credits: AI Generated

DOI: 10.1186/s44399-026-00037-x

Keywords: cassava, NDVI, high-throughput phenotyping, plant breeding, genotype-by-environment interaction, heritability, yield prediction, remote sensing, plant architecture, Nigeria, food security, GreenSeeker

The stakes could hardly be higher. Wheat is a cornerstone of China’s national food security, yet escalating global warming has made extreme climate events considerably more frequent and severe, threatening the stability of production. Traditional tools for forecasting yields have struggled in exactly those years when forecasts matter most. Mechanistic crop models, which simulate plant growth using genotype parameters, weather and soil data, often perform poorly under extreme weather because of heavy data demands and structural rigidity. Statistical models, meanwhile, typically assume linear relationships between vegetation indices and yield, blinding them to the nonlinear dynamics that govern crop responses to stress.

Machine learning approaches such as Random Forest, Support Vector Regression and neural networks have improved on these limits, but they frequently fail to extract spatiotemporal dynamics from complex datasets, and most existing yield studies focus on late growth stages, delaying predictions until it is nearly too late to act. Deep learning architectures have begun to close this gap. Convolutional neural networks excel at extracting spatial features, while Long Short-Term Memory networks, with their memory cells and gating mechanisms, are adept at modeling the time-series character of crop growth. Earlier hybrid CNN-LSTM models outperformed either architecture alone, but they relied mainly on one-dimensional convolutions, ignored static variables such as soil, and—critically—did not account for extreme climate events at all, introducing systematic bias into their predictions.

MBF-HybridNet addresses all three shortcomings at once. The model adopts a multi-branch parallel architecture with three distinct modules: a Dynamic Variables Module that processes daily remote sensing and meteorological data, a Dynamic ECIs Module that handles yearly-scale extreme climate index data, and a Static Variables Module that ingests soil properties. The dynamic module stacks three two-dimensional convolutional layers with asymmetric 1 by 2 kernels, each followed by batch normalization and ReLU activation, and inserts a Self-Attention mechanism after the first two layers to focus on the most informative features. The extracted features are then reshaped into sequences and fed into a two-layer LSTM network with 256 units per layer and dropout to curb overfitting. The static module processes soil organic carbon, cation exchange capacity, pH, sand and clay content through its own convolutional stream. A staged fusion strategy then concatenates these streams through successive fully connected layers to produce the final yield estimate.

A key innovation lies in how the data are prepared. Rather than relying on county-level averages, the team used the Thiessen Polygon method to partition the winter wheat planting area into 288 uniform analysis cells centered on evenly distributed sampling points, preserving fine-scale environmental information while avoiding contamination from non-planting surfaces such as mountains and water bodies. The data were then organized as pseudo-2D image tensors, with rows representing time steps, columns representing feature variables and a third dimension representing the aggregated subregions. The growing season was divided into three progressively cumulative windows: the vegetative growth phase from sowing to pre-jointing, the vegetative-reproductive phase spanning jointing to heading, and the reproductive phase from heading to maturity.

To capture the fingerprint of extreme weather, the researchers computed nine extreme climate indices covering heat, frost and precipitation extremes—hot days, heat stress intensity, consecutive hot days, frost days, cold stress intensity, consecutive cold days, heavy precipitation days, consecutive wet days and consecutive dry days—calculated for each growth stage. Because feeding all 21 growth-stage indices into the model risked dimensionality and overfitting, a genetic algorithm was deployed to select the most informative subset for each stage: three indices for the vegetative phase, five for the vegetative-reproductive window, and eleven for the full season.

The performance gains were substantial. Validated with leave-one-year-out cross-validation across 2004 to 2019, MBF-HybridNet achieved R-squared values of 0.756 to 0.765 across the three cumulative growth stages, with mean absolute percentage errors around 4.2 percent, compared to the baseline LSTM’s R-squared values of 0.652 to 0.671 and errors approaching 4.9 percent. Relative to the baseline, the new model cut root mean square error by up to 55.67 kilograms per hectare and mean absolute error by up to 48.12 kilograms per hectare. An ablation study confirmed that the gains arise from the complementary contributions of spatial representation, self-attention and temporal modeling rather than any single component: CNN alone reached an R-squared of 0.646, adding self-attention lifted it to 0.674, and the full CNN-SA-LSTM stack reached 0.763.

The extreme climate indices proved especially valuable in anomalous years. When the study years were split into normal and extreme groups—with 2006, 2013, 2014 and 2019 flagged as extreme—the model without the indices showed visibly degraded accuracy in extreme years, with R-squared values dropping to around 0.72 to 0.74. Adding all indices raised extreme-year R-squared values to as high as 0.799, and the genetically optimized subsets performed even better, reaching 0.803 in the vegetative-reproductive window while reducing computational cost. Across the full record, the GA-based models improved R-squared by 1.6 to 2.1 percentage points over the index-free model. SHAP interpretability analysis revealed a clear phenological pattern: pre-flowering low-temperature events such as frost days and cold stress intensity dominated early stages, while post-flowering heat and water stress—consecutive hot days, heavy precipitation days and consecutive dry days—took over as the key drivers during grain filling. Wind speed, temperature, precipitation and vegetation indices, particularly solar-induced chlorophyll fluorescence, rounded out the most influential predictors.

Perhaps the most striking result is the model’s early-warning capability. Prediction accuracy improved as seasonal information accumulated but plateaued at the vegetative-reproductive stage, meaning that winter wheat yields can be reasonably estimated approximately 30 days before harvest. The model also corrected a persistent weakness of simpler networks: the tendency to overestimate low yields and underestimate high ones, a bias rooted in the imbalanced distribution of yield samples concentrated in the 5000 to 7000 kilograms per hectare range. Residual analysis showed most county-level errors stayed within plus or minus 400 kilograms per hectare, with the strongest performance in medium- and high-yield areas.

The researchers caution that the framework, built and tested in Qingdao’s temperate monsoon climate, would need regional recalibration elsewhere: extreme-climate thresholds should be adjusted for arid or subtropical zones, growth-stage windows redefined by local phenology, and topographic factors such as elevation and slope added for hillier terrain. Management variables—irrigation, fertilization and cultivar choice—were not explicitly modeled and may explain some residual uncertainty. Still, because every input is drawn from public datasets, the model’s architecture offers a transferable blueprint. As climate extremes intensify, tools like MBF-HybridNet could give farmers and policymakers the lead time they need to protect harvests before the damage is done.

Subject of Research: A multi-branch fusion deep learning model that integrates remote sensing, meteorological, soil and extreme climate index data to estimate winter wheat yields under extreme climate events in Qingdao, China.

Article Title: Develop a multi-branch fusion deep learning model to estimate the winter wheat yields under extreme climate events

Article References: Jiang, X., Kong, D., Zhang, J., Zhang, S., Ma, Z., Yu, L., Yang, S., Bai, Y., Ali, S., & Ullah, H. (2026). Develop a multi-branch fusion deep learning model to estimate the winter wheat yields under extreme climate events. Artificial Intelligence in Agriculture. https://doi.org/10.1016/j.aiia.2026.09.001

Image Credits: AI Generated

DOI: 10.1016/j.aiia.2026.09.001

Keywords: winter wheat, yield prediction, deep learning, extreme climate events, remote sensing, LSTM, convolutional neural network, self-attention, extreme climate indices, Thiessen polygons, food security, genetic algorithm

The review, published in Smart Agricultural Technology, synthesises 199 peer-reviewed studies published between 2014 and 2026 that applied unmanned aerial vehicle remote sensing to rice phenotyping, the systematic measurement of plant traits. Led by Xiaoao Yang of Guangdong Province and colleagues including Spyros Fountas of Greece’s Agricultural University of Athens, the team screened 2,327 records from Scopus and Web of Science, ultimately distilling a corpus they call Core-199. Their goal was to build a rice-specific framework connecting platforms, sensors, data-processing methods and agronomic tasks, from nitrogen diagnosis to yield prediction, in a field that has grown explosively but unevenly.

The case for drones rests on a simple observational gap. Manual field surveys are labour-intensive and subjective, demanding an estimated 200 to 500 man-hours per hectare, while destructive sampling rules out repeated monitoring of the same plants. Satellites, meanwhile, revisit fields at best every five days and deliver pixels of 10 to 30 metres, far coarser than the 0.01 to 0.1 metre detail needed to evaluate individual breeding plots. UAVs slot neatly between these extremes, offering centimetre-scale imagery on schedules chosen by the operator, carrying anything from cheap RGB cameras to hyperspectral imagers and laser scanners.

Platform choice matters. Multirotor drones dominate rice phenotyping because they hover stably, need little takeoff space and suit the small, fragmented paddies typical of Asian rice farming. Fixed-wing aircraft cover more ground faster but cannot hover and demand launch infrastructure, so they remain rare in plot-level work. The review also situates UAVs within a three-tier observation system: satellites for broad regional mapping, drones for sub-decimetre plot detail, and handheld or tractor-mounted proximal sensors for calibration and real-time decisions. Economic analyses cited in the review suggest the tiers are complementary, with break-even areas for satellite-based nitrogen management ranging from about 2.5 to 13 hectares depending on imagery resolution.

Each sensor family contributes something distinct. RGB cameras, cheap and sharp, excel at structural traits: plant height from digital surface models, canopy cover, panicle counting and lodging assessment. Multispectral cameras capture the red-edge and near-infrared bands that power vegetation indices such as NDVI, supporting routine retrieval of leaf area index, chlorophyll, nitrogen status and yield. Hyperspectral imagers, with hundreds of narrow bands, detect subtle biochemical signals and early stress but bring high costs and data redundancy. LiDAR actively probes the three-dimensional canopy, and thermal infrared cameras, used in only two of the 199 studies, reveal canopy temperature linked to water status and heat tolerance, a capability the authors flag as a high-priority research frontier as flowering-stage heat stress intensifies.

The application chapters reveal both striking progress and stubborn caveats. Nitrogen monitoring is the most mature task, with reported coefficients of determination ranging from about 0.49 to 0.94 depending on sensor, trait and model, and one machine-learning precision-nitrogen strategy raising yields by 7 to 15 percent and economic returns by 4 to 16 percent. Chlorophyll retrieval has reached R-squared values as high as 0.97 in multistage hyperspectral frameworks. Yet the review repeatedly warns that these numbers cannot be compared across studies, because target traits, units, growth stages, sensors and validation designs differ so widely, and the authors decline to claim any general superiority for multisource fusion over simpler single-sensor approaches.

Lodging and disease monitoring showcase the field’s move toward real-time, on-board intelligence. Semantic segmentation networks now delineate lodged rice at pixel level with mean intersection-over-union above 90 percent, and one edge-computing workflow on an Nvidia Jetson Xavier NX processed imagery at nearly 14,418 square metres per second, covering roughly 10 square kilometres in an 80-minute flight. On the disease front, thermal and optical fusion allowed researchers to identify infection a remarkable 72 hours before visible lesions appeared, with an FPGA implementation consuming just 0.076 watts per classification. A lightweight false-smut detector built on YOLOv12n cut parameters by 25 percent while maintaining a mean average precision of 80.7 percent.

Yield prediction has evolved from single-date vegetation indices into multi-temporal, multisource and even process-coupled models that assimilate drone-derived nitrogen into crop simulation frameworks such as CERES-Rice. Organ-level phenotyping offers an alternative route: detecting and counting panicles from aerial imagery, with one framework classifying yield levels at 83.63 percent accuracy and another reporting yield-estimation errors between 1.4 and 11.7 percent across test plots. The review also cautions that some eye-catching near-perfect R-squared values in the literature describe proxy traits such as panicle counts or plant height rather than direct grain-yield prediction, and one segmentation-based study with R-squared of 0.98 carried relative errors of 21 to 31 percent.

Perhaps the most sobering statistic concerns genetics. Only six of the 199 reviewed studies linked UAV-derived traits to genetic association analysis, three using genome-wide association studies and three using QTL mapping. Those that did recovered known genes such as sd1, Ghd7.1 and TAC1, and one drought study across 240 accessions identified 111 significant loci, but no drone-derived QTL has yet been independently validated by another research group. The authors frame this as the principal evidence gap between high-throughput phenotyping and breeding impact.

Cross-regional generalisation emerges as the field’s central technical challenge. Environmental background interference, especially the standing water, sun glint and mixed pixels of flooded paddies, and site or cultivar bias are the best-documented causes of model failure when models move between regions, years or seasons. The review proposes a staged roadmap: standardised multi-environment public datasets with rich metadata as the foundation, interpretable hybrid models combining radiative-transfer physics with machine learning in the transition, and eventually closed-loop systems connecting drones, cloud processing and field decisions. Until then, the authors conclude, the technology’s promise depends less on fancier algorithms than on standardised data, honest external validation and independent confirmation of genetic findings, the unglamorous infrastructure that will decide whether drone phenotyping graduates from academic demonstration to routine agricultural practice.

Subject of Research: UAV remote sensing for high-throughput rice phenotyping

Article Title: Advances in UAV remote sensing for high-throughput rice phenotyping: a systematic review

Article References: Yang, X., Zhou, Z., Huang, H., Wei, X., Kong, X., Fountas, S., & Tang, Y. (2026). Advances in UAV remote sensing for high-throughput rice phenotyping: a systematic review. Smart Agricultural Technology, 15, Article 102561. https://doi.org/10.1016/j.atech.2026.102561

Image Credits: AI Generated

DOI: 10.1016/j.atech.2026.102561

Keywords: UAV remote sensing, rice phenotyping, precision agriculture, high-throughput phenotyping, hyperspectral imaging, LiDAR, yield prediction, nitrogen monitoring, disease detection, lodging monitoring, GWAS, edge computing