Conference Agenda

The European Green Deal’s Farm to Fork strategy mandates a 20% reduction in fertiliser use by 2030. In heterogeneous forage systems, this requires shifting from uniform to precision management. This study develops a site-specific decision framework in INIAV Elvas innovation hub (Elvas, Portugal), integrating soil zoning with agronomic response to model a Variable Rate Application (VRA) strategy. High-resolution apparent electrical conductivity (ECa) data defined Management Zones (MZs) of contrasting potential, which were cross-referenced with N response curves (0, 120, 200 kg N ha⁻¹) and legume-inclusion trials. Results showed that high inputs (200 kg N ha⁻¹) caused diminishing returns and low N use efficiency, whereas 120 kg N ha⁻¹ maximised profitability. Crucially, legume-rich mixtures showed superior economic margins. The proposed model demonstrates that restricting synthetic N to <100 kg ha⁻¹ in low-potential zones (sandy/low-ECa)—relying on biological fixation—while targeting 120 kg N ha⁻¹ in high-potential zones (clay/high-ECa), significantly reduces total fertiliser usage without compromising returns. Prioritising legume mixtures in marginal soils and optimising N-doses in productive areas decouples productivity from excessive fertiliser use. This framework offers a practical pathway to reduce chemical dependency and nutrient loss, transforming soil heterogeneity into an efficiency opportunity.

Producing uniform and structurally stable nursery mats is essential for reliable rice mechanical transplanting. Seedling performance is strongly influenced by the composition of growing media and the seeding rate. This study evaluated one rice variety (F67) under mechanized nursery conditions at Aipe, Colombia, using four growing media: A (2/3 silt + 1/3 humus), B (1/3 rice husk + 2/3 silt), C (1/2 silt + 1/4 humus + 1/4 husk), and D (3/4 silt + 1/4 husk), combined with four seeding rates (60, 90, 120, 150 g tray⁻¹). Three nursery samplings (early, intermediate, and transplant‑age) were evaluated. Stage‑1 assessed early vigor using the relative growth rate of plant length between the intermediate and transplant‑age samplings (ANCOVA including emergence as covariate). Stage‑2 evaluated transplant‑age quality using seedling length, tray stability (average per tray), chlorophyll index and leaf number. Substrate significantly affected early vigor (p = 0.004), while seeding rate and interaction did not. At transplant age, seedling length showed a substrate‑level trend (p = 0.060), and tray stability was significantly influenced by substrate (p = 0.027); in both cases, medium C combined with 90 g achieved the highest overall performance. Thus, the recommended combination for F67 is C - 90.

Light management is a key environmental factor influencing crop growth and productivity in controlled environment agriculture. Despite varying light requirements across growth stages, many commercial systems use fixed lighting throughout the cultivation period. This study evaluated whether a stage-specific LED lighting recipe that adjusts the light spectrum enhances the growth of butterhead lettuce compared with a white-light control. Butterhead lettuce was cultivated in a closed, refrigerator-type aeroponic cultivation system (20°C, 60–70% RH). The control group was positioned in the upper compartment and the experimental group in the lower compartment, with eight plants cultivated in each group. Both the treatment and control groups were exposed to the same 16-hour photoperiod and stepwise PPFD levels (75, 150, and 200 μmol m⁻²s⁻¹). The treatment group received a stage-specific lighting recipe with adjusted spectral ratios, whereas the control group was illuminated with a white LED spectrum. Results showed that the treatment significantly improved all growth parameters compared with the control (Welch’s t-test, p < 0.01), including fresh weight (+144%), leaf area (+90%), LAI (+136%), leaf number (+33%), and stem elongation (+128%). These findings demonstrate that stage-specific regulation of light spectrum can substantially enhance biomass production and canopy development in indoor cultivation systems.

Climate change is intensifying drought frequency, threatening crop productivity worldwide. Non-destructive indicators for crop water status are therefore urgently needed for optimal irrigation management. Leaf water content (LWC) is a key physiological indicator, and short-wave infrared (SWIR) hyperspectral imaging has emerged as an effective monitoring tool due to its sensitivity to water absorption features. However, conventional whole-leaf spectral averaging fails to capture within-leaf spatial heterogeneity in water distribution. This study developed an unsupervised region-of-interest (ROI) segmentation strategy to capture within-leaf spectral substructures for improved LWC prediction. SWIR images (900-1700 nm) were acquired from 244 sweet basil (Ocimum basilicum L.) leaves under progressive drought stress, with LWC ranging from 62.50% to 93.15%. Pixel-level spectra were transformed via Principal Component Analysis (PCA) and residual PCA, then clustered using K-means and Gaussian Mixture Models (GMM) into three subregions. Mean cluster spectra were used as ROI inputs for five regression models. GMM Cluster 0 with Gaussian Process Regression (GPR) achieved the highest performance (R² = 0.92), outperforming the whole-leaf baseline (R² = 0.91). These results confirm that unsupervised ROI segmentation enhances LWC prediction for precision irrigation. Acknowledgement: This research was supported by a grant(D2510010) from Gyeonggi Technology Development Program funded by Gyeonggi Province.

Sensor-based yield data from self-propelled forage harvesters often contain errors, which must be pre-processed to obtain realistic application maps for site-specific grassland management The aim of this study was to develop an automated, universally applicable workflow for processing those data from Bavarian grassland areas to create accurate and reliable yield and application maps.

The workflow comprises four steps: removing points outside farmer-mapped field boundaries; deleting zero/NA values in yield and moisture content attributes; setting upper and lower moisture content limits based on manufacturers’ near-infrared spectroscopy sensors calibration models limits; elimination of point values exceeding typical maximum yields per cut according to each plot’s cutting regime. The workflow was implemented in R-Studio (v2026.01.0) and tested on data from a trial area in northern Bavaria (4.72 ha).

On average 5.95% of data points in a dataset were removed (3.92% removed in step four). Adjusted datasets exhibited no major spatial gaps. The mean RMSE of interpolated yield maps decreased from 3.56 to 0.35 after preprocessing, and maps displayed realistic yield ranges.

The workflow reliably processes multiple large datasets and eliminates significantly plausibility-impairing errors for yield maps. Preset limits may exclude naturally occurring extremes. The Workflow requires validation with datasets from other locations.

This study developed a numerical method to evaluate the tree vigor of Japanese pear using 3D point cloud data, aiming to replace traditional reliance on farmers' intuition. Precise pruning is essential for pear cultivation, but objective standards have been lacking. The researchers used a 3D laser scanner to collect data from 87 trees in Tottori Prefecture, Japan. To isolate the tree structures, a custom algorithm was applied to remove noise from the ground, trellis wires, and other environmental factors. The study focused on the correlation between two key metrics: the estimated crown area, calculated using the Quickhull algorithm, and the estimated total length of pruned branches, derived from branch weight. The analysis yielded a regression line, y = 2.7002x (R2 = 0.88), which serves as the benchmark for vigor. Trees plotted above this line were classified as "strong" in vigor. The results showed that vigor varies significantly even among trees of similar size, with 31.3% of young trees and over 52% of mature/old trees judged as vigorous. This system enables data-driven, individual tree management, allowing for optimized pruning strategies that can improve overall orchard productivity and sustainability.