Compositional and Machine Learning Tools to Model Plant Nutrition: Overview and Perspectives

Horticulturae, Journal Year: 2025, Volume and Issue: 11(2), P. 161 - 161

Published: Feb. 3, 2025

The ceteris paribus assumption that all features are equal except the one(s) being examined limits reliability of nutrient diagnosis and fertilizer recommendations. objective is to review machine learning (ML) compositional data analysis (CoDa) tools make management feature specific. average accuracy ML methods was 84% across crops. additive orthogonal log ratios CoDa reduce a D-parts soil composition D-1 variables, alleviating redundancy in predictive models. Using Brazilian onion (Allium cepa) database, combined returned crop response patterns, allowing feature-specific recommendations be made. centered ratio (clr) diagnoses plant nutrients as (CND). Quebec database vegetable crops, mean variance clr variables (VAR¯) allowed comparing total among species growth stages. While summation equally weighted dual ratios, may show unequal importance regarding performance. RReliefF scores or gain can provide weighting coefficients for each ratio. widely contrasting (wlr) improved models muck database. models, VAR¯ wlr, advanced improve diagnosis.

Language: Английский

---

Combining field measurements and process‐based modelling to analyse soil tillage and crop residues management impacts on crop production and carbon balance in temperate areas

Soil Use and Management, Journal Year: 2024, Volume and Issue: 40(3)

Published: July 1, 2024

Abstract Crop residues management is an important issue in the context of climate change. They might be kept on field and restituted to soil enhance its fertility or exported for other uses such as production energy through biomethanization. Furthermore, choices regarding tillage operations impact potential incorporate residues, which turn affects physical (e.g. structure, water retention), biological organic matter, microorganisms) chemical nutrient release mineralization) fertility. We combined measurements from a 14‐year experiment Hesbaye loamy region Belgium simulation with STICS soil‐crop model investigate impacts crop production, characteristics carbon balance. Four treatments were compared, where all combinations incorporation versus exportation conventional reduced tested. The comparison observations simulations proved that adequate explore contrasted management. analysis data outputs showed was positively influenced by but unresponsive fate. Reduced led clear stratification observed SOC content topsoil (0–30 cm), also increase simulated stocks (0–26 cm). This gain greater retention under tillage. Moreover, both treatments, incorporating increased despite associated augmentation heterotrophic respiration. Finally, importance environmental conditions balance suggests modelling very useful specific agro‐pedoclimatic contexts, especially when facing

Language: Английский

---

Feature-specific nutrient management of onion (Allium cepa) using machine learning and compositional methods

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: March 12, 2024

Abstract While onion cultivars, irrigation and soil crop management have been given much attention in Brazil to boost yields, nutrient at field scale is still challenging due large dosage uncertainty. Our objective was develop an accurate feature-based fertilization model for crops. We assembled climatic, edaphic, managerial features as well tissue tests into a database of 1182 observations from multi-environment fertilizer trials conducted during 13 years southern Brazil. The complexity cropping systems captured by machine learning (ML) methods. RReliefF ranking algorithm showed that the split-N micronutrients S were most relevant predict bulb yield. decision-tree random forest extreme gradient boosting models yield predictors (R 2 > 90%). As shown gain ratio, foliar standards nutritionally balanced high-yielding specimens producing 50 Mg ha −1 set apart ML classification differed among cultivars. Cultivar × environment interactions support documenting local diagnosis. controllable feature run future universality assess models’ ability generalize growers’ fields.

Language: Английский

---

Vegetable Response to Added Nitrogen and Phosphorus Using Machine Learning Decryption and the N/P Ratio

Horticulturae, Journal Year: 2024, Volume and Issue: 10(4), P. 356 - 356

Published: April 3, 2024

The current N and P fertilization practices for vegetable crops grown in organic soils are inaccurate may potentially damage the environment. New models needed. Machine learning (ML) methods can combine numerous features to predict crop response fertilization. Our objective was evaluate machine predictions marketable yields, offtakes, N/P ratio of crops. We assembled 157 multi-environmental fertilizer trials on lettuce (Lactuca sativa), celery (Apium graveolens), onion (Allium cepa), potato (Solanum tuberosum) documented 22 easy-to-collect soil, managerial, meteorological features. random forest returned moderate substantial strength (R2 = 0.73–0.80). Soil managerial were most important. There no added null independent universality tests. offtakes impacted by P-related features, indicating N–P interactions. mass ratios harvested products generally lower than 10, suggesting excess that would trigger plant acquisition possibly alter soil C cycles through microbial processes. Crop prediction ML ex post diagnosis proved be useful tools guide management decisions soils.

Language: Английский

---

Calibrating primary crop parameters to capture undersown species impacts

European Journal of Agronomy, Journal Year: 2025, Volume and Issue: 169, P. 127676 - 127676

Published: May 10, 2025

Language: Английский

---

Simulating Maize Response to Split-Nitrogen Fertilization Using Easy-to-Collect Local Features

Nitrogen, Journal Year: 2023, Volume and Issue: 4(4), P. 331 - 349

Published: Nov. 9, 2023

Maize (Zea mays) is a high-nitrogen (N)-demanding crop potentially contributing to nitrate contamination and emissions of nitrous oxide. The N fertilization generally split between sowing time the V6 stage. right rate apply at minimize environmental damage challenging. Our objectives were (1) predict maize response added using machine learning (ML) models; (2) cross-check model outcomes by independent on-farm trials. We assembled 461 trials conducted in Eastern Canada 1992 2022. dataset grain yield comprised dosage, weekly precipitations corn heat units, seeding date, previous crop, tillage practice, soil series, texture, organic matter content, pH. Random forest XGBoost predicted accurately stage (R2 = 0.78–0.80; RSME MAE 1.22–1.29 0.96–0.98 Mg ha−1, respectively). Model accuracy up was comparable that full-season prediction. patterns simulated varying doses showed started plateau 125–150 kg total ha−1 eight out ten independently. There great potential for economic gains from ML-assisted fertilization.

Language: Английский

---

Simulating phosphorus dynamics between the soil and the crop with the STICS model: Formalization and multi-site evaluation on maize in temperate area

European Journal of Agronomy, Journal Year: 2024, Volume and Issue: 164, P. 127475 - 127475

Published: Dec. 20, 2024

Language: Английский

---