Spatial cross-validation is not the right way to evaluate map accuracy

Ecological Modelling, Journal Year: 2021, Volume and Issue: 457, P. 109692 - 109692

Published: Aug. 12, 2021

Language: Английский

---

A review of deep learning approach to predicting the state of health and state of charge of lithium-ion batteries

Journal of Energy Chemistry, Journal Year: 2022, Volume and Issue: 74, P. 159 - 173

Published: July 2, 2022

Language: Английский

---

Artificial Intelligence in Food Safety: A Decade Review and Bibliometric Analysis

Foods, Journal Year: 2023, Volume and Issue: 12(6), P. 1242 - 1242

Published: March 14, 2023

Artificial Intelligence (AI) technologies have been powerful solutions used to improve food yield, quality, and nutrition, increase safety traceability while decreasing resource consumption, eliminate waste. Compared with several qualitative reviews on AI in safety, we conducted an in-depth quantitative systematic review based the Core Collection database of WoS (Web Science). To discover historical trajectory identify future trends, analysed literature concerning from 2012 2022 by CiteSpace. In this review, bibliometric methods describe development including performance analysis, science mapping, network analysis Among 1855 selected articles, China United States contributed most literature, Chinese Academy Sciences released largest number relevant articles. all journals field, PLoS ONE Computers Electronics Agriculture ranked first second terms annual publications co-citation frequency. The present character, hot spots, research trends were determined. Furthermore, our analyses, provide researchers, practitioners, policymakers big picture across whole process, precision agriculture through 28 enlightening

Language: Английский

---

The effect of covariates on Soil Organic Matter and pH variability: a digital soil mapping approach using random forest model

Annals of GIS, Journal Year: 2024, Volume and Issue: 30(2), P. 215 - 232

Published: Jan. 29, 2024

This research focuses on understanding the spatial variation of Soil Organic Matter (SOM) and pH levels in North Morocco. The study employs a comprehensive approach to enhance predictive modelling, incorporating Boruta algorithm for effective environmental covariates selection optimizing model parameters through hyperparameter optimization. Utilizing Random Forest (RF) with remote sensing indices topographic features, predicts SOM identify key contributors their variability. prediction saw significant success, notable correlation such as RVI, NDVI, TNDVI. These indices, indicative vegetation health productivity, emerged primary influencers SOM. In comparison, influence features like elevation, slope, aspect was found be less significant. Conversely, predicting challenging due minimal variability within dataset. Addressing this limitation could involve dataset expansion or alternative models low-correlated data handling. Despite RF model's limited efficacy prediction, an observable between identified, consistent prior research. Areas higher exhibited lower values, indicating relative soil acidification from organic matter decomposition. study's demonstrated potential using but enhancing is essential. Future may explore expansion, diverse sampling, testing better performance datasets. offers valuable insights advanced development enriches management practices.

Language: Английский

---

Open Soil Spectral Library (OSSL): Building reproducible soil calibration models through open development and community engagement

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(1), P. e0296545 - e0296545

Published: Jan. 13, 2025

Soil spectroscopy is a widely used method for estimating soil properties that are important to environmental and agricultural monitoring. However, bottleneck its more widespread adoption the need establishing large reference datasets training machine learning (ML) models, which called spectral libraries (SSLs). Similarly, prediction capacity of new samples also subject number diversity types conditions represented in SSLs. To help bridge this gap enable hundreds stakeholders collect affordable data by leveraging centralized open resource, Spectroscopy Global Good initiative has created Open Spectral Library (OSSL). In paper, we describe procedures collecting harmonizing several SSLs incorporated into OSSL, followed exploratory analysis predictive modeling. The results 10-fold cross-validation with refitting show that, general, mid-infrared (MIR)-based models significantly accurate than visible near-infrared (VisNIR) or (NIR) models. From independent model evaluation, found Cubist comes out as best-performing ML algorithm calibration delivery reliable outputs (prediction uncertainty representation flag). Although many well predicted, total sulfur, extractable sodium, electrical conductivity performed poorly all regions, some other nutrients physical performing one two regions (VisNIR NIR). Hence, use based solely on variations limitations. This study presents discusses resources were developed from aspects opening data, current limitations, future development. With genuinely science project, hope OSSL becomes driver community accelerate pace scientific discovery innovation.

Language: Английский

---

High Resolution Water Table Modeling of the Shallow Groundwater Using a Knowledge-Guided Gradient Boosting Decision Tree Model

Frontiers in Water, Journal Year: 2021, Volume and Issue: 3

Published: Sept. 1, 2021

Detailed knowledge of the uppermost water table representing shallow groundwater system is critical in order to address societal challenges that relate mitigation and adaptation climate change enhancing resilience general. Machine learning (ML) allows for high resolution modeling depth beyond capabilities conventional numerical physically-based hydrological models with respect spatial overall accuracy. For this, in-situ well proxy observations are used as training data combination covariates. The objective this study model a typical summer winter condition at 10 m over entire Denmark (43,000 km 2 ). CatBoost, state art implementation gradient boosting decision trees, employed associated uncertainties. domain has not been most prominent field applications recent ML advances due lack big data. This brings forward novel knowledge-guided framework overcome limitation by integrating simulation results from flow model. utilized (1) identify wells represent table, (2) augment missing accounting simulated level seasonality, (3) expand list curated dataset contains around 13,000 wells, 19,000 lakes, streams coastline 15 Cross validation attests generalizes mean absolute error 115 cm considering solely MAE <50 taking also into consideration. Quantile regression applied estimate confidence intervals estimated uncertainty largest moraine clay soils characterized distinct geological heterogeneity. highlights research avenue efficiently supporting predict unprecedented detail

Language: Английский

---

Progress in research on site-specific nutrient management for smallholder farmers in sub-Saharan Africa

Field Crops Research, Journal Year: 2022, Volume and Issue: 281, P. 108503 - 108503

Published: March 2, 2022

Increasing fertilizer access and use is an essential component for improving crop production food security in sub-Saharan Africa (SSA). However, given the heterogeneous nature of smallholder farms, application needs to be tailored specific farming conditions increase yield, profitability, nutrient efficiency. The site-specific management (SSNM) approach initially developed 1990 s generating field-specific recommendations rice Asia, has also been introduced rice, maize cassava cropping systems SSA. SSNM shown Yield gains with SSA were on average 24% 69% when compared farmer practice, respectively, or 11% 4% local blanket recommendations. there need more extensive field evaluation quantify broader benefits diverse environments. Especially should expanded rainfed systems, which are dominant further take into account soil texture water availability. Digital decision support tools such as RiceAdvice Nutrient Expert can enable wider dissemination locally relevant reach large numbers farmers at scale. One major limitations currently available requirement acquiring a significant amount farm-specific information needed formulate scaling potential will greatly enhanced by integration other agronomic advisory platforms seamless digital soil, climate improve predictions reduced on-farm data collection. Uncertainty included future solutions, primarily better varying prices economic outcomes.

Language: Английский

---

The Role of Remote Sensing Data and Methods in a Modern Approach to Fertilization in Precision Agriculture

Remote Sensing, Journal Year: 2022, Volume and Issue: 14(3), P. 778 - 778

Published: Feb. 7, 2022

The precision fertilization system is the basis for upgrading conventional intensive agricultural production, while achieving both high and quality yields minimizing negative impacts on environment. This research aims to present application of modern prediction methods in by integrating agronomic components with spatial component interpolation machine learning. While were a cornerstone soil past decades, new challenges process larger more complex data have reduced their viability present. Their disadvantages lower accuracy, lack robustness regarding properties input sample values requirements extensive cost- time-expensive sampling addressed. Specific (ordinary kriging, inverse distance weighted) learning (random forest, support vector machine, artificial neural networks, decision trees) evaluated according popularity relevant studies indexed Web Science Core Collection over decade. As shift towards increased accuracy computational efficiency, an overview state-of-the-art remote sensing improving precise was completed, accent open-data global satellite missions. State-of-the-art techniques allowed hybrid predict sampled supported such as high-resolution multispectral, thermal radar or unmanned aerial vehicle (UAV)-based imagery analyzed studies. representative approaches performed based 121 samples phosphorous pentoxide (P2O5) potassium oxide (K2O) common parcel Croatia. It visually quantitatively confirmed superior retained local heterogeneity approach. concludes that significant role agriculture today will be increasingly important future.

Language: Английский

---

Comparing the prediction performance, uncertainty quantification and extrapolation potential of regression kriging and random forest while accounting for soil measurement errors

Geoderma, Journal Year: 2022, Volume and Issue: 428, P. 116192 - 116192

Published: Oct. 25, 2022

Geostatistics and machine learning have been extensively applied for modelling predicting the spatial distribution of continuous soil variables. In addition to providing predictions, both techniques quantify uncertainty associated with although geostatistics is more developed in this respect. Despite increased use these techniques, most algorithms ignore that measurements are not error-free. Recently, concern has also arisen about extrapolation risk be it geographic space, feature or both. paper, regression kriging (RK) random forest (RF) were compared respect their ability deliver accurate predictions prediction uncertainties, while accounting measurement errors data. The sensitivity results models was evaluated, as well potential. This done a case study Cameroon where pH, clay organic carbon mapped from obtained using conventional proximal sensing methods. showed produced comparable ranges maps predicted values properties interest. Compared RF, RK outperformed RF by presenting generally higher Model Efficiency Coefficient (MEC), lower Root Mean Squared Error (RMSE) better performance. improvement RMSE 10, 12 2 % MEC on average 5, 22 1 SOC, respectively Overestimation local observed larger than shown accuracy plots, indicating uncertainties quantified model. Better performance derived at unsampled locations cross-validation metrics scatter particularly when used extrapolation. effects incorporating significant due fact calibration data had same error variance. comparison should go beyond common validation only evaluate but must account locations.

Language: Английский

---

A global soil spectral calibration library and estimation service

Soil Security, Journal Year: 2022, Volume and Issue: 7, P. 100061 - 100061

Published: April 1, 2022

There is growing global interest in the potential for soil reflectance spectroscopy to fill an urgent need more data on properties improved decision-making security at local scales. This driven by capability of estimate a wide range from rapid, inexpensive, and highly reproducible measurement using only light. However, several obstacles are preventing wider adoption spectroscopy. The biggest large variation analytical methods operating procedures used different laboratories, poor reproducibility analyses within amongst laboratories lack physical archives. In addition, hindered expense complexity building spectral libraries calibration models. Global Soil Spectral Calibration Library Estimation Service proposed overcome these providing freely available estimation service based open, high quality diverse library extensive archives Kellogg Survey Laboratory (KSSL) Natural Resources Conservation United States Department Agriculture (USDA). initiative supported Network (GLOSOLAN) Partnership Spectroscopy Good network, which provide additional support through dissemination standards, capacity development research. public good stands benefit assessments globally, but especially developing countries where resources conventional most limited.

Language: Английский

---