Ecological Informatics最新文献

{"title":"Remote sensing-based spatiotemporal dynamics of agricultural drought on Prince Edward Island using Google Earth engine","authors":"Fatima Imtiaz ,&nbsp;Aitazaz A. Farooque ,&nbsp;Gurjit S. Randhawa ,&nbsp;Seyyed Ebrahim Hashemi Garmdareh ,&nbsp;Xiuquan Wang ,&nbsp;Travis J. Esau ,&nbsp;Bishnu Acharya ,&nbsp;Rehan Sadiq","doi":"10.1016/j.ecoinf.2025.103073","DOIUrl":"10.1016/j.ecoinf.2025.103073","url":null,"abstract":"<div><div>Climate change is a primary factor contributing to widespread drought conditions worldwide. Therefore, assessing agricultural drought's spatial and temporal extent is crucial. This study explicitly applies remote sensing techniques to monitor drought in the cropland area of Prince Edward Island, Canada, with a particular emphasis on potato crops. The long-term drought was evaluated using MODIS for 2012–2022, while the seasonal drought at the field scale was calculated using Landsat-8 OLI/TIRS for the 2021 and 2022 crop growth seasons. The computed remote sensing drought indices include Vegetation Condition Index (VCI), Vegetation Health Index (VHI), and Temperature Condition Index (TCI), which are derived using the Google Earth Engine platform. Examining long-term drought by MODIS revealed that 2020 was the most dominant drought year, according to all three drought indices. However, the seasonal variations of VCI, TCI, and VHI at the field scale observed in the three fields in 2021 and 2022 demonstrated that June went through considerable drought in both years. August was the second critical month following June for drought conditions. CHIRPS data indicated significant rainfall anomalies relative to the long-term seasonal average for the 2021 crop season, specifically in June (−38.5 %) and August (−38.2 %), while the rainfall in the crop season in 2022 exceeded the seasonal average. Based on Pearson correlation analysis, VHI correlated strongly with VCI (CC = 0.87 for 2021 and 0.93 for 2022) and moderately with rainfall (CC = 0.68 for 2021 and 0.63 for 2022). The spatial autocorrelation analysis revealed substantial positive autocorrelation of drought for 2019, 2020, 2021 and 2022. However, 2020 has the highest spatial autocorrelation, with Moran's I of 0.54 and a z-score of 24.8. Hence, this study will optimize irrigation, decrease crop loss, sustain crop yields, and enhance food security.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103073"},"PeriodicalIF":5.8,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143421238","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluating climate-induced changes in the suitable distribution and ecological niche of Prunus mira Koehne using ensemble modeling and high-performance liquid chromatography analysis","authors":"Shuning Zhang ,&nbsp;Minglong Gao ,&nbsp;Junxing Chen ,&nbsp;Ao Dun ,&nbsp;Lin Wang ,&nbsp;Wuyun Tana ,&nbsp;Yu-e Bai ,&nbsp;Wenquan Bao","doi":"10.1016/j.ecoinf.2025.103071","DOIUrl":"10.1016/j.ecoinf.2025.103071","url":null,"abstract":"<div><div><em>Prunus mira</em> Koehne, is an economically and ecologically important woody oilseed tree species in China, but currently, it is facing great challenges from climate change and human activities, impacting its distribution area and, consequently, its growth and seed quality. Therefore, it is important to investigate the spatial distribution of <em>P. mira</em> and changes in seed kernel oil quality with respect to climate change for the conservation and utilization of this species. This study explored the influence of climate change on the geographical distribution, ecological niches, and seed oil quality of <em>P. mira</em>. Based on the Biomod2 platform and ensemble modeling, 10 species distribution model groups were incorporated to simulate potential distribution areas. High-performance liquid chromatography was used to analyze the oil composition of <em>P. mira</em> seeds. The results revealed that the current potential distribution area in China primarily comprises high-altitude regions of the Tibetan and Yunnan-Guizhou Plateaus. Future climate projections suggest limited migration of suitable areas for <em>P. mira</em>, which initially decrease before experiencing an increase in the distribution area. Furthermore, variations in <em>P. mira</em> seed oil characteristics were aligned with climate-induced changes, with seeds from more habitable regions producing improved oil qualities.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103071"},"PeriodicalIF":5.8,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143453725","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"DAMI-YOLOv8l: A multi-scale detection framework for light-trapping insect pest monitoring","authors":"Xiao Chen ,&nbsp;Xinting Yang ,&nbsp;Huan Hu ,&nbsp;Tianjun Li ,&nbsp;Zijie Zhou ,&nbsp;Wenyong Li","doi":"10.1016/j.ecoinf.2025.103067","DOIUrl":"10.1016/j.ecoinf.2025.103067","url":null,"abstract":"<div><div>Insect pest detection plays a crucial role in agricultural production for accurate and early pest control, thus significantly reducing crop damage and increasing yields. However, currently the small size and multi-scale characteristics of insect pests pose significant challenges for accurate object detection using computer vision technology. To address this issue, we propose a novel framework called DAMI-YOLOv8l to detect pest in images collected by a light-trapping device. The DAMI-YOLOv8l model integrates three key innovations: the Depth-wise Multi-Scale Convolution (DMC) module, the Attentional Scale Sequence Fusion with a P2 detection layer (ASF<img>P2) neck structure, and a novel bounding box regression loss function named Minimum Point Distance inner Intersection over Union (MPDinner-IoU). The DMC module improves multi-scale feature extraction to enable the effective capture and merging of features across different detection scales while reducing network parameters. The ASF-P2 neck structure enhances the fusion of multi-scale features while preserving critical local information related to small-scale features. Additionally, the MPDinner-IoU loss function optimizes feature measurement for small insect pest datasets by introducing geometric correction capabilities. By leveraging these innovations, the results demonstrate that the proposed framework improves many metrics, such as mAP₅₀ from 74.5 % to 78.2 %, mAP_50:95 from 52.5 % to 57.3 %, and FPS from 109.89 to 121.12, compared with those of YOLOv8l model on the proposed LP24 dataset. Furthermore, we validate its robustness on two other public datasets related to small objects, Pest24 and VisDrone2019.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103067"},"PeriodicalIF":5.8,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143402726","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AgriDeep-net: An advanced deep feature fusion-based technique for enhanced fine-grain image analytics in precision agriculture","authors":"Rakesh Chandra Joshi ,&nbsp;Radim Burget ,&nbsp;Malay Kishore Dutta","doi":"10.1016/j.ecoinf.2025.103069","DOIUrl":"10.1016/j.ecoinf.2025.103069","url":null,"abstract":"<div><div>With the vast diversity and rapidly evolving nature of agricultural landscapes, the need for cutting-edge technological solutions has become increasingly apparent. Addressing the complex challenges of fine-grained agricultural image classification, AgriDeep-Net is introduced as an innovative multi-model deep-learning framework, strategically incorporating advanced techniques to navigate complexities in the field. This precision-driven methodology distinguishes AgriDeep-Net, offering a strategic approach to extract salient and discriminative features from diverse deep-learning models involving highly similar agricultural images marked by low inter-class visual discrimination. Each model is characterized by unique architectural configurations, enabling strategic feature fusion that empowers AgriDeep-Net to capture nuanced semantic information within multi-class agricultural images. The framework adeptly manages the hurdles posed by uneven data distribution, intra-class diversity, and the demands of multi-class classification. Rigorous experimentation underscores AgriDeep-Net's exceptional performance, achieving a testing accuracy of 93.29 % for the ACHENY dataset and an even more impressive 98.44 % for the Indian Basmati seeds dataset. Benchmarking against state-of-the-art deep neural networks, AgriDeep-Net proves its efficacy across diverse datasets collected under real-world and controlled conditions. This framework stands out as a beacon of efficiency and accuracy, eliminating the need for extensive image pre-processing operations and showcasing its potential to empower farmers with precision tools for optimizing crop yields, resource allocation, and swift responses to emerging agricultural challenges.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103069"},"PeriodicalIF":5.8,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143421237","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep sea spy: An online citizen science annotation platform for science and ocean literacy","authors":"Marjolaine Matabos ,&nbsp;Pierre Cottais ,&nbsp;Riwan Leroux ,&nbsp;Yannick Cenatiempo ,&nbsp;Charlotte Gasne-Destaville ,&nbsp;Nicolas Roullet ,&nbsp;Jozée Sarrazin ,&nbsp;Julie Tourolle ,&nbsp;Catherine Borremans","doi":"10.1016/j.ecoinf.2025.103065","DOIUrl":"10.1016/j.ecoinf.2025.103065","url":null,"abstract":"<div><div>The recent development of deep-sea observatories has enabled the acquisition of high temporal resolution imagery for studying the dynamics of deep-sea communities on hourly to multi-decadal scales. These unprecedented datasets offer valuable insight into the variation of species abundance and biology in relation to changes in environmental conditions. Since 2010, camera systems deployed at hydrothermal vents have acquired over 11 terabytes (TB) of data that cannot be processed by research labs only. Although deep learning offers an alternative to human processing, training algorithms requires large annotated reference datasets. The Deep Sea Spy project allows citizens to contribute to the annotation of pictures acquired with underwater platforms. Based on approximately 4000 photos, each annotated 10 times by independent participants, we were able to develop a data validation workflow that can be applied to similar databases. We compared these annotations with expert-annotated data and analysed the agreement rate among participants for each of the 15,000 annotated individual organisms to optimise the robustness and confidence level in non-expert citizen science. The optimal number of repeat annotations per photo was also analysed to guide the definition of a trade-off between the accuracy and amount of data. An agreement rate of 0.4 (i.e., 4 out of 10 participants detecting one given individual) was established as an efficient threshold to reach counts similar to that obtained from an expert. One important result lies in the robustness of the temporal trends of species abundance as revealed by time-series analyses. Regarding the number of times a photo needs to be annotated, results varied greatly depending on the target species and the difficulty of the associated task. Finally, we present the communication tools and actions deployed during the project and how the platform can serve educational and decision-making purposes. Deep Sea Spy and the proposed workflow have a strong potential to enhance marine environmental observation and monitoring.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103065"},"PeriodicalIF":5.8,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143421179","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Cetacean feeding modelling using machine learning: A case study of the Central-Eastern Mediterranean Sea","authors":"Carla Cherubini ,&nbsp;Giulia Cipriano ,&nbsp;Leonardo Saccotelli ,&nbsp;Giovanni Dimauro ,&nbsp;Giovanni Coppini ,&nbsp;Roberto Carlucci ,&nbsp;Carmelo Fanizza ,&nbsp;Rosalia Maglietta","doi":"10.1016/j.ecoinf.2025.103066","DOIUrl":"10.1016/j.ecoinf.2025.103066","url":null,"abstract":"<div><div>Investigating environmental drivers of cetacean feeding behaviour is essential for effective marine resource management, especially in the Mediterranean Sea, a biodiversity hotspot heavily impacted by human activities and climate change. This study realized a pioneer assessment of feeding activity related to the marine environment for three cetacean species - striped dolphin, common bottlenose dolphin, and Risso's dolphin - in the Gulf of Taranto (Northern Ionian Sea, Central-eastern Mediterranean) using an innovative Machine Learning (ML) approach. Behavioural data from April 2016 to October 2023, coupled with 20 environmental variables from Copernicus Marine Service and EMODnet-bathymetry datasets, were used to build Cetacean Feeding Models (CFMs) for the target species using Random Forest and RUSBoost algorithms. Multiple subsets of environmental predictors—physiographic, physical, inorganic, and bio-chemical—were employed to develop and evaluate ML models tailored to feeding prediction. Risso's dolphin resulted to be the best modelled species, with the bio-chemical model based on the RUSBoost algorithm achieving a Balanced Classification Rate (BCR) of 94 %, primarily influenced by 3D chlorophyll-a concentrations, a close proxy for prey availability. The second-best model was the physical one for the common bottlenose dolphin with a BCR of 72 %, influenced by salinity, currents speed, and temperature. These differences in predictive performance might reflect the distinct trophic niches of the studied odontocetes. Finally, simulated predictive maps of Risso's dolphin feeding habitats for summer months were realized in the Gulf of Taranto, providing actionable insights for conservation and sustainable management. The developed CFMs enhance understanding of cetacean feeding preferences and offer a versatile framework for integrating behavioural processes into species distribution models to inform area-based conservation measures, with significant potential for application across other Mediterranean areas.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103066"},"PeriodicalIF":5.8,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143378400","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Integrating direct observation and environmental DNA data to enhance species distribution models in riverine environments","authors":"Luca Carraro","doi":"10.1016/j.ecoinf.2025.103056","DOIUrl":"10.1016/j.ecoinf.2025.103056","url":null,"abstract":"<div><div>The recent advances in both theoretical and modeling approaches (species distribution models) and molecular techniques (environmental DNA) offer new opportunities to advance the assessment of biodiversity. This is particularly the case for riverine environments, whose biodiversity is disproportionately under peril, but also whose dendritic connectivity allows a spatial interpretation of eDNA samples, which reflect a biodiversity signal averaged over a certain upstream area. Conversely, traditional, direct observation surveys provide localized information on taxon density. Here, I propose a framework to leverage both data types to improve estimates of a taxon’s spatial distribution. Specifically, I expand the eDITH model (which allows estimating the spatial distribution of taxa based on spatially replicated stream eDNA data) to include direct observations, and upgrade the <span>eDITH</span> R-package to allow a broad implementation of such method. Moreover, I propose optimized sampling strategies for both eDNA and direct sampling, with algorithms (included in the upgraded <span>eDITH</span> package) that mathematically translate rule-of-thumb criteria to maximize the spatial coverage of sites’ arrangement in a riverscape based on the peculiar features of each data type. Finally, I test such framework by means of an in-silico experiment, whereby I show that optimized sampling strategies outperform random-based strategies in the ability to reconstruct a taxon’s spatial distribution. When eDNA and direct sampling sites are spatially arranged in an optimized fashion, the highest prediction skill for a fixed total number of sampling sites deployed is reached when both data types are included in the model fitting. The optimal trade-off between eDNA and direct sampling observations depends on both characteristics of the investigated taxon (e.g., the spatial heterogeneity in its distribution) and the level of uncertainty in the observed data. These results will contribute to designing efficient strategies for integrated biomonitoring in river networks.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103056"},"PeriodicalIF":5.8,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143395916","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Remote sensing of total suspended matter of inland waters: Past, current status, and future directions","authors":"Hui Tao ,&nbsp;Kaishan Song ,&nbsp;Zhidan Wen ,&nbsp;Ge Liu ,&nbsp;Yingxin Shang ,&nbsp;Chong Fang ,&nbsp;Qiang Wang","doi":"10.1016/j.ecoinf.2025.103062","DOIUrl":"10.1016/j.ecoinf.2025.103062","url":null,"abstract":"<div><div>Total suspended matter (TSM) serves as an important water quality parameter, often carrying nutrients, micro-pollutants, and heavy metals, thereby closely influencing the ecological health of aquatic ecosystems. With the recent advancements in remote sensing technology, artificial intelligence algorithms, and cloud platforms, understanding remote sensing is crucial for TSM monitoring, especially in water resources management and decision-making. This review aims to summarize research advancements in TSM remote sensing of inland waters while addressing current conditions' limitations, outlining future directions, and providing recommendations. The technology for remote sensing utilized to capture inland TSM has its origins in the 1970s. In the last five decades, approximately eight hundred pertinent studies have been carried out by researchers, progressing from the development of algorithms to their applications in science. The band ratio algorithm, bio-optical model, and machine learning algorithm are increasingly recognized as the predominant methodologies. The red band, near-infrared band, and their combinations are typically chosen as sensitive bands for detecting TSM in turbid waters, whereas the blue and green bands are generally utilized for clear waters. Analysis of bibliometric data indicates that empirical and semi-empirical algorithms comprise the largest share at 72 %, with semi-analytical algorithms following at 9 %. The TSM is co-influenced by the composition, particle size, and refractive index. Considering these parameters to develop high-precision TSM inversion algorithms remains a challenge. Researchers frequently utilize the Landsat series sensors and MODIS for retrieving TSM concentrations across regional, national, and global scales, representing28 % and 16 % of the total publications, respectively, while Sentinel follows closely with 8 %. The Taihu Lake, Poyang Lake, Ebinur Lake, Vembanad Lake, Amazon River, Yangtze River, and Mississippi River have emerged as hot spot regions for research on TSM by scholars form various countries. However, for complex and variable inland water bodies, the atmospheric corrections, adjacency effects and limited resolutions of current sensors, as well as model transferability remain challenges, and many attempts should be made in the future.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103062"},"PeriodicalIF":5.8,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143395918","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A digital twin-enabled fog-edge-assisted IoAT framework for Oryza Sativa disease identification and classification","authors":"Goluguri N.V. Rajareddy ,&nbsp;Kaushik Mishra ,&nbsp;Satish Kumar Satti ,&nbsp;Gurpreet Singh Chhabra ,&nbsp;Kshira Sagar Sahoo ,&nbsp;Amir H. Gandomi","doi":"10.1016/j.ecoinf.2025.103063","DOIUrl":"10.1016/j.ecoinf.2025.103063","url":null,"abstract":"<div><div>The integration of agri-technology with the Internet of Agricultural Things (IoAT) is revolutionizing the field of smart agriculture, particularly in diagnosing and treating <em>Oryza sativa</em> (rice) diseases. Given that rice serves as a staple food for over half of the global population, ensuring its healthy cultivation is crucial, particularly with the growing global population. Accurate and timely identification of rice diseases, such as Brown Leaf Spot (BS), Bacterial Leaf Blight (BLB), and Leaf Blast (LB), is therefore essential to maintaining and enhancing rice production. In response to this critical need, the research introduces a timely detection system that leverages the power of Digital Twin (DT)-enabled Fog computing, integrated with Edge and Cloud Computing (CC), and supported by sensors and advanced technologies. At the heart of this system lies a sophisticated deep-learning model built on the robust AlexNet neural network architecture. This model is further refined by including Quaternion convolution layers, which enhance colour information processing, and Atrous convolution layers, which improve depth perception, particularly in extracting disease patterns. To boost the model's predictive accuracy, the Chaotic Honey Badger Algorithm (CHBA) is employed to optimize the CNN hyperparameters, resulting in an impressive average accuracy of 93.5 %. This performance significantly surpasses that of other models, including AlexNet, AlexNet-Atrous, QAlexNet, and QAlexNet-Atrous, which achieved respective accuracies of 75 %, 84 %, 89 %, and 91 %. Moreover, the CHBA optimization algorithm outperforms other techniques like CSO, BSO, PSO, and CJAYA and demonstrates optimal results with an 80–20 % training-testing parameter split. Service latency analysis further reveals that the Fog-Edge-assisted environment is more efficient than the Cloud-assisted model for latency reduction. Additionally, the DT-enabled QAlexNet-Atrous-CHBA model proves to be far superior to its non-DT counterpart, showing substantial improvements in 18.7 % in Accuracy, 17 % in recall, 19 % in Fβ-measure, 17.3 % in specificity, and 13.4 % in precision, respectively. These enhancements are supported by convergence analysis and the Quade rank test, establishing the model's effectiveness and potential to significantly improve rice disease diagnosis and management. This advancement promises to contribute significantly to the sustainability and productivity of global rice cultivation.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103063"},"PeriodicalIF":5.8,"publicationDate":"2025-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143511127","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Spatial autocorrelation in machine learning for modelling soil organic carbon","authors":"Alexander Kmoch,&nbsp;Clay Taylor Harrison,&nbsp;Jeonghwan Choi,&nbsp;Evelyn Uuemaa","doi":"10.1016/j.ecoinf.2025.103057","DOIUrl":"10.1016/j.ecoinf.2025.103057","url":null,"abstract":"<div><div>Spatial autocorrelation, the relationship between nearby samples of a spatial random variable, is often overlooked in machine learning models, leading to biased results. This study compares various methods to account for spatial autocorrelation when predicting soil organic carbon (SOC) using random forest models. This kind of systematic comparison has not been done previously. Five models incorporating spatial structure were compared against baseline models with no added spatial components. Cross-validation showed slight improvements in accuracy for models considering spatial autocorrelation, while Shapley Additive Explanations confirmed the importance of spatial variables. However, no decrease in spatial autocorrelation of residuals was observed. Random Forest Spatial Interpolation emerged as the top performer in capturing spatial structure and improving model accuracy. Raster-based models exhibited enhanced prediction detail. The findings emphasize the value of incorporating spatial autocorrelation for better prediction of SOC with machine learning. Considerations such as the spatial distribution of predictions and computational complexity should help guide the selection of suitable approaches for specific spatial modelling tasks.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"86 ","pages":"Article 103057"},"PeriodicalIF":5.8,"publicationDate":"2025-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143436486","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}