George Themistokleous

Accurate estimates of soil nutrients and organic carbon are essential for sustainable land management, yet the factors that most improve soil pedotransfer function (PTF) accuracy remain poorly quantified. Remote sensing (RS) is increasingly used to augment soil databases, but its benefits are often contradictory, and a genuine RS signal is rarely separated from the accuracy gained simply by adding predictors. This study framed PTF development as three practical modelling decisions: algorithm choice, feature-selection procedure, and whether to augment soil predictors with an RS group. It then quantified each decision’s effect on predictive error for nitrogen (N), phosphorus (P), potassium (K), and organic carbon (OC) across 951 agricultural sites in Cyprus, using a Bayesian hierarchical model with a permutation-based pseudo-RS control to isolate true RS signal from added dimensionality. Relative to conventional baselines, model selection had the greatest impact, with ensemble and tree-based models reducing RMSE by up to -49.8%. Feature selection was the second-largest factor, with wrapper-based methods reducing OC RMSE by about -39%, though Boruta increased it by 15.0%. Augmenting soil predictors with the RS predictor group yielded smaller, property-dependent gains (up to -4.8%); permutation controls confirmed these improvements were not chance artefacts but consistently smaller than method-related gains. This limited RS contribution is consistent with the mismatch between coarse RS spatial support and point-scale soil measurements. For point-based PTFs in Mediterranean croplands, investment in algorithm and predictor selection yields larger, more reliable returns than adding RS covariates, and the study provides a reusable, uncertainty-aware framework for such decisions.

Root system architecture and root hairs highly influence plant resource uptake, yet their simultaneous quantification at the whole-plant scale remains challenging due to the conflicting requirements of high-resolution imaging and non-destructive, repeated measurements. Here, we present a plant growth and imaging system based on A3 sized flatbed rhizotrons, combined with a non-machine-learning image analysis workflow implemented in R, that enables repeated, in situ quantification of barley root system architecture and accompanied visible projected root hair area (VP root hair area) over three weeks of growth. The system produces integrated outputs highlighting both whole-root architecture and the spatial distribution of surrounding root hairs from single images. Validation of the image analysis algorithm showed good segmentation performance, with average Matthews Correlation Coefficient values of 0.68 for projected root area and 0.65 for VP root hair area. To demonstrate its experimental applicability, the system was used to assess root growth and root hair responses under controlled environmental conditions, combining three irrigation regimes (2, 4, and 6 irrigation events per day) with three dry bulk density levels (1.4, 1.5, and 1.6 g cm ⁻³ ). In addition to whole-system metrics, the approach enables analysis of root hair expansion at individual root tips. This methodology provides a training-free method for integrated analysis of root traits e.g. projected root size, root growth, root diameter and root hairs under controlled physical conditions similar to soil, facilitating studies of root–soil interactions that require both spatial resolution and temporal continuity.

Accurate prediction of soil properties is critical for sustainable land management and agriculture. Soil pedotransfer functions (PTFs) estimate difficult-to-measure properties from easier measurements. With the emergence of remote sensing (RS), such data are increasingly used to enhance soil databases, raising the question of whether integrating RS data with conventional soil measurements improves PTF accuracy. These effects must be evaluated alongside other influential factors, particularly model and feature selection strategies, yet their relative importance remains unclear. This study evaluated the effects of model choice, feature selection, and RS-derived data integration on predictive accuracy for four key soil properties: total nitrogen (N), available phosphorus (P), exchangeable potassium (K), and organic carbon (OC). A Bayesian hierarchical framework quantified the relative importance of these factors. Model selection had the greatest impact on root mean square error (RMSE) reduction: Ensemble reduced RMSE by -25.3% for N, -32.2% for P, and -28.0% for K, while Random Forest (RF) achieved the largest improvement for OC (-49.8%). Feature selection also reduced error, particularly for OC, where Forward Recursive Feature Selection (FRFS) and Recursive Feature Elimination (RFE) reduced RMSE by -38.8% and -39.0%, respectively, while Boruta increased RMSE by 15.0%. Remote sensing-derived data provided smaller benefits, improving RMSE by -4.8% for OC, -3.4% for N, and -2.1% for P, with negligible effect for K (0.5%). Permutation tests confirmed these RS-related improvements were distinguishable from random associations. Overall, model selection and, to a lesser extent, feature selection substantially enhanced PTF performance, whereas RS integration yielded modest, variable improvements.

Purpose: Paediatric anterior cruciate ligament (ACL) injuries are becoming increasingly frequent. Surgical management remains challenging due to the risk of growth plate disturbance. Physeal‐sparing techniques aim to restore stability while minimizing this risk. The purpose of this study is to evaluate the clinical, functional and radiological outcomes of an all‐inside, all‐epiphyseal ACL reconstruction in skeletally immature patients, with particular attention to complications, return to sport (RTS) and the influence of associated injuries. Methods: Seventy‐eight patients (48 males and 30 females; mean age 13.1 years, range: 9–16) underwent physeal‐sparing ACL reconstruction between 2015 and 2024. All procedures were performed using hamstring autograft with all‐epiphyseal sockets under fluoroscopic guidance. Pre‐ and postoperative assessments included plain radiographs, magnetic resonance imaging (MRI) and clinical evaluation. Outcome measures were Lysholm, Paediatric International Knee Documentation Committee (Pedi‐IKDC) and visual analogue scale (VAS) scores, recorded preoperatively and at a mean follow‐up of 48 months (range: 24–82). Return‐to‐play rates, complications and growth disturbances were documented. Subgroup analyses investigated the effect of meniscal/chondral lesions, gender and surgical delay. Results: Significant postoperative improvements were achieved in all clinical scores: Lysholm (53.4–94.6, p < 0.01), Pedi‐IKDC (53.5–90.6, p < 0.01) and VAS (5.6–1.8, p < 0.01). No growth arrest, angular deformities, or leg length discrepancies were identified. Complications included one infection, two re‐ruptures and one residual laxity. RTS was achieved in 94.8% of patients, with a mean recovery period of 11.1 months. Meniscal and chondral lesions were associated with inferior outcomes and delayed return‐to‐play. Delayed surgery significantly increased the risk of meniscal injury (odds ratio [OR]: 5.1, p = 0.001). Conclusion: All‐inside physeal‐sparing ACL reconstruction demonstrates excellent mid‐ to long‐term outcomes in skeletally immature patients. The technique offers high RTS rates and minimal growth‐related complications. Early surgical intervention may reduce the risk of secondary meniscal and chondral damage. Level of Evidence: Level IV, retrospective case series.

Agriculture significantly contributes to greenhouse gas (GHG) emissions, yet fluxes from irrigated semi-arid systems remain poorly quantified. This study investigates CO₂, CH₄, N₂O, and NH₃ fluxes in a short-term lettuce experiment under semi-arid conditions. The objective was to quantify flux variability and identify key environmental and management drivers. High-frequency soil gas flux measurements were conducted under three treatments: irrigated soil (I), irrigated soil with plants (IP), and irrigated soil with plants plus NH₄NO₃ fertilizer (IPF). Environmental factors, including solar radiation, soil temperature, water-filled pore space, and relative projected leaf area, were monitored. A Random Forest model identified main flux determinants. Fluxes varied with plant function, growth, and fertilization. IP exhibited net CO₂ uptake through photosynthesis, whereas I and IPF showed net CO₂ emissions from soil respiration and fertilizer-induced disruption of plant function, respectively. CH₄ uptake occurred across treatments but decreased with plant presence. Fertilization in IPF triggered episodic N₂O (EF = 0.1%) and NH₃ emissions (EF = 0.97%) linked to nitrogen input. Vegetated semi-arid soils can act as CO₂ sinks when nitrogen is optimally managed. Excess or poorly timed nitrogen delays CO₂ uptake and increases reactive nitrogen losses. Methanotrophic activity drives CH₄ dynamics and is influenced by plants and fertilization. Maintaining crop vigor and applying precision nitrogen management are essential to optimize productivity while mitigating GHG and NH₃ emissions in semi-arid lettuce cultivation.

Accurate and spatially detailed soil information is essential for supporting sustainable land use planning, particularly in data-scarce regions such as Cyprus, where soil degradation risks are intensified by land fragmentation, water scarcity, and climate change pressure. This study aimed to generate national-scale predictive maps of key soil health descriptors by integrating satellite-based indicators with a recently released geo-referenced soil dataset. A machine learning model was applied to estimate a suite of soil properties, including organic carbon, pH, texture fractions, macronutrients, and electrical conductivity. The resulting maps reflect spatial patterns consistent with previous studies focused on Cyprus and provide high resolution insights into degradation processes, such as organic carbon loss, and salinization risk. These outputs provide added value for identifying priority zones for soil conservation and evidence-based land management planning. While predictive uncertainty is greater in areas lacking ground reference data, particularly in the northeastern part of the island, the modeling framework demonstrates strong potential for a national-scale soil health assessment. The outcomes are directly relevant to ongoing soil policy developments, including the forthcoming Soil Monitoring Law, and provide spatial prediction models and indicator maps that support the assessment and mitigation of soil degradation.