{"title":"S2-IFNet: A spatial-semantic information fusion network integrated with boundary feature enhancement for forest land extraction from Sentinel-2 data","authors":"Junyang Xie , Mengyao Zhang , Hao Wu , Anqi Lin , Marcos Adami , Abdul Rashid Mohamed Shariff , Yahui Guo","doi":"10.1016/j.jag.2025.104505","DOIUrl":"10.1016/j.jag.2025.104505","url":null,"abstract":"<div><div>Accurately extracting forest land and understanding its spatial distribution are crucial for forest monitoring and management. However, variations in tree species, human activities, and natural disturbances create diverse and distinct forest land characteristics in remote sensing images, posing challenges for precise forest land extraction. To address these challenges, we propose a spatial-semantic information fusion network (S2-IFNet) integrated with boundary feature enhancement for forest land extraction from Sentinel-2 data. S2-IFNet employs a dual-branch network to separately extract the spatial and semantic features of forest land. In the semantic branch, two modules are introduced: a boundary enhancement module using the Sobel operator to capture forest land boundary details, and an attention module to strengthen the feature representation capability. Finally, a spatial-semantic fusion module effectively combines the spatial, semantic, and boundary detail information to improve the forest land extraction accuracy. S2-IFNet was evaluated across five regions in different global climate zones, with a comparative analysis conducted against four forest land-cover products in Yunnan, China. The results show that S2-IFNet can achieve an overall accuracy exceeding 90%, demonstrating its strong forest land extraction capability. Compared to the different forest land extraction models, S2-IFNet shows a superior performance, with ablation experiments confirming the effectiveness of each module. In particular, the boundary feature enhanced spatial-semantic fusion strategy enables S2-IFNet to focus more precisely on the boundaries and range of forest land, thereby enhancing the extraction accuracy. Meanwhile, S2-IFNet can adapt to complex scenarios, including varying forest density and similar object confusion. Furthermore, S2-IFNet can achieve results that are superior to the four other products.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104505"},"PeriodicalIF":7.6,"publicationDate":"2025-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143760940","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Tatenda Dzurume , Roshanak Darvishzadeh , Timothy Dube , T.S. Amjath Babu , Mutasim Billah , Syed Nurul Alam , Mustafa Kamal , Md. Harun-Or-Rashid , Badal Chandra Biswas , Md. Ashraf Uddin , Md. Abdul Muyeed , Md. Mostafizur Rahman Shah , Timothy J. Krupnik , Andrew Nelson
{"title":"Detection of Fall Armyworm infestation in maize fields during vegetative growth stages using temporal Sentinel-2","authors":"Tatenda Dzurume , Roshanak Darvishzadeh , Timothy Dube , T.S. Amjath Babu , Mutasim Billah , Syed Nurul Alam , Mustafa Kamal , Md. Harun-Or-Rashid , Badal Chandra Biswas , Md. Ashraf Uddin , Md. Abdul Muyeed , Md. Mostafizur Rahman Shah , Timothy J. Krupnik , Andrew Nelson","doi":"10.1016/j.jag.2025.104516","DOIUrl":"10.1016/j.jag.2025.104516","url":null,"abstract":"<div><div>Fall Armyworm (FAW), Spodoptera frugiperda (J.E. Smith) (Lepidoptera: Noctuidae), poses a significant risk to global food and income security by attacking various crops, particularly maize. Early detection and management of FAW infestation are crucial for mitigating its impact on crop yields. This study investigated the effect of FAW infestation on the spectral signature of maize fields and classified infestation severity in Bangladesh using Sentinel-2 satellite imagery and Random Forest (RF) classification. Field observations on FAW infestation severity (none, moderate, and severe), collected by the Bangladesh Department of Agricultural Extension during 2019 and 2020, were used to train the RF classifier. Six thousand nine hundred ninety-eight observations were collected from 579 maize fields through weekly scouting. The Kruskal-Wallis test and Dunn’s post-hoc test were applied to identify the most significant spectral bands (P < 0.05) for detecting FAW incidence and severity across different maize growth stages. The results demonstrated that the spectral reflectance from Sentinel-2 bands varied significantly among different classes of FAW infestation, with noticeable differences observed during the early developmental stages of maize (vegetative growth stages 3 to 8). RF identified nine spectral bands and two spectral vegetation indices as important for FAW infestation discrimination. The RF classifier was evaluated using five-fold cross-validation, achieving an overall accuracy between 74 % and 84 %. The independent test set’s accuracy ranged from 72 % to 82 %. The mean multiclass AUC ranged from 0.83 to 0.95. Moreover, the results demonstrated the feasibility of detecting the severity of FAW infestation using temporal Sentinel-2 data and machine learning techniques. These findings underscore the potential of remote sensing and machine learning techniques for effectively monitoring and managing crop pests. The study provides valuable insights for classifying FAW infestation using high-resolution multitemporal data.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104516"},"PeriodicalIF":7.6,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748404","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"D3GNN: Double dual dynamic graph neural network for multisource remote sensing data classification","authors":"Teng Yang , Song Xiao , Jiahui Qu","doi":"10.1016/j.jag.2025.104496","DOIUrl":"10.1016/j.jag.2025.104496","url":null,"abstract":"<div><div>Convolutional Neural Network (CNN) has garnered attention due to its outstanding performance in multisource remote sensing (RS) image classification. However, classical CNN-based methods primarily concentrate on information within a fixed-size neighborhood and a standard square region, neglecting long-range and global information. As non-Euclidean data, the topological structure enables flexible construction of relationships between objects, which can be served as an effective carrier of global information. Graph Neural Network (GNN), capable of extracting features from the topological structure, is considered as a solution for capturing global information. Nevertheless, GNN faces challenges as the manually defined static graph structure might not accurately capture the complexity of the data. We propose a double dual dynamic graph neural network (D<span><math><msup><mrow></mrow><mrow><mn>3</mn></mrow></msup></math></span>GNN) with dynamic topological structure refinement for multisource RS data classification. D<span><math><msup><mrow></mrow><mrow><mn>3</mn></mrow></msup></math></span>GNN generates multiple topological structures to achieve a comprehensive perception of scene features by utilizing local spatial information and distinctive data from various sources. Given the characteristics of heterogeneous-structure data, D<span><math><msup><mrow></mrow><mrow><mn>3</mn></mrow></msup></math></span>GNN implements targeted topological structure remodeling and refinement to overcome the limitations imposed by static graph, thereby enabling the network to generate feature embeddings with enhanced discriminative power. The experimental results show that D<span><math><msup><mrow></mrow><mrow><mn>3</mn></mrow></msup></math></span>GNN achieves superior performance compared to other current methods.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104496"},"PeriodicalIF":7.6,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748403","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Mengqi Lyu , Liyuan Li , Wencong Zhang , Long Gao , Yifan Zhong , Jingjie Jiao , Xiaoyan Li , Fansheng Chen
{"title":"Combining multi-source data to investigate vessel wake temperature gradients and dynamic patterns","authors":"Mengqi Lyu , Liyuan Li , Wencong Zhang , Long Gao , Yifan Zhong , Jingjie Jiao , Xiaoyan Li , Fansheng Chen","doi":"10.1016/j.jag.2025.104509","DOIUrl":"10.1016/j.jag.2025.104509","url":null,"abstract":"<div><div>The movement of a vessel generates cold and warm wake patterns with temperature gradients on the sea surface, which provide detection possibilities for satellite-based infrared detection systems. This work analyzes the temporal characteristics of ship wake dissipation on the sea surface, based on multi-source data from low Earth orbit satellite thermal infrared imaging, Automatic Identification System (AIS) data, and numerical simulation results, revealing the dynamic information contained within the thermal wake. The temperature images of sea surface thermal wakes generated by vessels at different speeds were obtained using numerical simulation methods. The thermal infrared characteristics of the surface vessel wakes were verified using images from the thermal imaging spectrometer aboard the SDGSAT-1 satellite. The simulation results reveal the patterns of generation, diffusion, and attenuation of the infrared thermal wake produced by moving vessels in the ocean. By combining simulations with infrared images from the SDGSAT-1 satellite, the thermal infrared temperature characteristics of wakes on the sea surface are summarized. This method overcomes the limitations of traditional optical monitoring techniques at night, while capturing more information on sea surface temperature variations. By deeply exploring sea surface thermal signature data, this paper provides technical support for all-weather vessel speed inversion using single-satellite data.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104509"},"PeriodicalIF":7.6,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748402","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"SAM-CTMapper: Utilizing segment anything model and scale-aware mixed CNN-Transformer facilitates coastal wetland hyperspectral image classification","authors":"Jiaqi Zou , Wei He , Haifeng Wang , Hongyan Zhang","doi":"10.1016/j.jag.2025.104469","DOIUrl":"10.1016/j.jag.2025.104469","url":null,"abstract":"<div><div>Accurate and effective coastal wetland classification using hyperspectral remote sensing technology is crucial for their conservation, restoration, and sustainable development. However, the large scale variance of land covers in complex wetland scenes poses challenges for existing methods and leads to misclassifications. Additionally, existing methods encounter difficulties in practical wetland classification tasks due to the high cost of hyperspectral wetland data labeling. This paper introduces SAM-CTMapper, a coastal wetland classification framework that incorporates a scale-aware mixed CNN-Transformer (CTMapper) to precisely identify wetland cover types using hyperspectral images, and the advanced segment anything model (SAM) to save labor costs in data labeling. Specifically, a novel scale-aware mixed CNN-Transformer layer is designed in CTMapper to effectively leverage local and long-range spectral–spatial features from the whole HSI to reduce misclassification. This layer comprises a multi-head scale-aware convolution layer to capture local land-cover details, a multi-head superpixel self-attention layer for extracting long-range contextual features, and a dynamic selective module to facilitate effective aggregation of local and long-range information. Additionally, we devise a SAM-based semi-automatic labeling strategy to construct two PRISMA hyperspectral wetland (PRISMA-HW) datasets over Liaoning Shuangtai and Shanghai Chongming for evaluation purposes. Experimental results on two PRISMA-HW datasets and two publicly available hyperspectral wetland datasets demonstrate the effectiveness of CTMapper method in terms of both accuracy metrics and visual quality. For the sake of reproducibility, the PRISMA-HW datasets and the related codes of SAM-CTMapper framework will be open-sourced at: <span><span>https://github.com/immortal13</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104469"},"PeriodicalIF":7.6,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748401","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Chuanwu Zhao , Haishuo Wei , Gudina Legese Feyisa , Thiaggo de Castro Tayer , Gelilan Ma , Hanyi Wu , Yaozhong Pan
{"title":"Evaluating spectral indices for water extraction: Limitations and contextual usage recommendations","authors":"Chuanwu Zhao , Haishuo Wei , Gudina Legese Feyisa , Thiaggo de Castro Tayer , Gelilan Ma , Hanyi Wu , Yaozhong Pan","doi":"10.1016/j.jag.2025.104510","DOIUrl":"10.1016/j.jag.2025.104510","url":null,"abstract":"<div><div>With the intensification of climate change and human activities, water resource shortages, floods, and water quality anomalies are becoming increasingly serious. It is urgent to ensure the effective realization of water resource management, flood monitoring, and water quality assessment through fine-scale monitoring of water body spatial distribution and dynamics. Currently, various technologies have been applied to water information monitoring, with remote sensing-based spectral index methods being widely used due to their simplicity, low cost, and large-scale observation capabilities. However, the wide variety of existing water spectral indices, each suited to different scenarios and objectives, makes it challenging for ordinary users to select the most appropriate index and determine its optimal usage (i.e., threshold settings). This study addresses these challenges by evaluating the performance and applicability of 15 widely used water detection indices, using Sentinel-2 imagery across 14 representative global regions. The results revealed that the performance of water indices varied across different scenarios. Common issues include misidentification in high-reflectance backgrounds (e.g., buildings and snow), low-reflectance backgrounds (e.g., shadows), and omission errors for water bodies with high chlorophyll content. On this basis, this study provides a recommended table for water body index selection in different scenarios and a recommended range table for index thresholds, and point out key directions for future development of water body indices. This study offers valuable guidance for the selection and use of water body indices in practical application, helping to enhance the accuracy and efficiency of fine-scale monitoring of water bodies.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104510"},"PeriodicalIF":7.6,"publicationDate":"2025-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143725658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ziyi Suo , Yingcheng Lu , Lijian Shi , Bin Zou , Qing Wang , Ling Li , Jun Tang , Weimin Ju , Manchun Li
{"title":"Discrimination between sea ice and clouds over the Chukchi Sea via China’s ultraviolet–visible-infrared observations","authors":"Ziyi Suo , Yingcheng Lu , Lijian Shi , Bin Zou , Qing Wang , Ling Li , Jun Tang , Weimin Ju , Manchun Li","doi":"10.1016/j.jag.2025.104508","DOIUrl":"10.1016/j.jag.2025.104508","url":null,"abstract":"<div><div>The distribution of Arctic sea ice is an important direct indicator of climate change, and spaceborne optical remote sensing represents one primary technique for sea ice monitoring due to its high spatiotemporal resolution and wide swath coverage. However, this process is often impeded by heavy cloud cover, which shares similar visual and spectral features with sea ice. To address these limitations, this study proposes a novel methodological framework for discriminating between sea ice and different cloud types (cirrus and cumulus) via the ultraviolet–visible-infrared observations from China’s Haiyang-1C/D (HY-1C/D) satellites, and the ultraviolet (UV) data from the onboard Ultraviolet Imager (UVI) are used to study sea ice and clouds over the Chukchi Sea for the first time. The spectral properties are characterized by the top-of-atmosphere (TOA) reflectance (<em>ρ</em><sub>TOA</sub>) in both UV and visible and near-infrared (VNIR) wavelengths. This indicates that the 355 nm UV band has the optimal sensitivity to the presence of sea ice and clouds, with cirrus clouds composed of high-altitude ice crystals exhibiting extremely high UV reflectivity. A hybrid threshold is subsequently determined to separate sea ice and cloud pixels. In comparison to the MODIS MOD29 sea ice product, which masks cloud pixels with brightness temperature (BT) differences, this algorithm can effectively reduce the misclassification resulting from surface temperature inversions in polar regions. The ice/cloud identification results have been further applied to sea ice concentration (SIC) estimation, and extensive trials of this UV-based ice/cloud detection approach in the Arctic Passages demonstrates its potential applicability.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104508"},"PeriodicalIF":7.6,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143725752","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Runjie Huang , Li Zhuo , Na Jie , Jixiang Zheng , Haiyan Tao
{"title":"Improving spatial and temporal accuracy in coastal reclamation mapping using satellite image time series","authors":"Runjie Huang , Li Zhuo , Na Jie , Jixiang Zheng , Haiyan Tao","doi":"10.1016/j.jag.2025.104499","DOIUrl":"10.1016/j.jag.2025.104499","url":null,"abstract":"<div><div>Driven by urban expansion and economic growth, China has carried out extensive coastal reclamation in the past four decades, leading to negative ecological impacts. Accurate mapping of reclamation dynamics helps analyze its environmental effects and supports the government in regulating reclamation projects. Nevertheless, existing methods for monitoring coastal reclamation dynamics face two major limitations: overestimating reclamation areas and low temporal accuracy. This study proposes a novel method to automatically distinguish artificial reclamation from natural expansion to reduce the overestimation of reclaimed areas and improve temporal dynamic accuracy through time-series analysis of long-term Landsat satellite imagery. In China’s coastal regions, our method achieves an overall classification accuracy of over 94 % for artificial reclamation and natural expansion. The estimation accuracies of the year of reclamation are 91.60 % (one-year tolerance) and 96.44 % (two-year tolerance), which are more accurate than the post-classification comparison method based on the Global Surface Water (GSW) dataset. From 1990 to 2022, China’s coastal reclamation totaled 5,121.1 km2, with a slowdown after 2014 and a shift toward green development after 2018. Reclamation patterns varied regionally, with most projects concentrated in economically developed bays, driven primarily by urbanization and industrialization. The results prove that the proposed method can improve the spatiotemporal accuracy of coastal reclamation mapping, hence providing a more reliable data foundation for project siting, ecological impact assessment, and coastal management.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104499"},"PeriodicalIF":7.6,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143725751","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Siying Wang , Shuangyin Zhang , Dawei Wang , Weifeng Li
{"title":"Spatial-and-local-aware deep learning approach for Ground-Level NO2 estimation in England with multisource data from satellite-based observations and chemical transport models","authors":"Siying Wang , Shuangyin Zhang , Dawei Wang , Weifeng Li","doi":"10.1016/j.jag.2025.104506","DOIUrl":"10.1016/j.jag.2025.104506","url":null,"abstract":"<div><div>High-resolution near-surface NO<sub>2</sub> data are crucial for monitoring air pollution dynamics. Satellite-based machine learning models are commonly used to estimate NO<sub>2</sub> concentrations, but tailoring advanced deep learning techniques to specific environmental problems remains challenging. This study applies a two-stage deep learning approach to estimate ground-level NO<sub>2</sub> concentrations in England at a 1 km spatial resolution from 2019 to 2021. Initially, we imputed the TROPOMI NO<sub>2</sub> column density to a continuous 1 km resolution. We then developed an efficient spatial-and-local-aware deep learning network (SLNet) for NO<sub>2</sub> mapping by integrating the imputed TROPOMI NO<sub>2</sub> data with multi-source information from meteorology, chemical transport model (CTM) simulations, and other auxiliary predictors. To address the translation invariance of convolutional neural networks (CNNs), we combined a local channel to identify spatial heterogeneity in the model. Our imputed TROPOMI NO<sub>2</sub> surfaces, which initially covered only 34.12 % of valid data, achieved full coverage with reliability and continuity at 1 km spatial resolution. Cross-validation confirmed that the SLNet model outperformed other state-of-the-art methods in estimating ground-level NO<sub>2</sub>. The prediction model achieved R<sup>2</sup> values of 0.914, 0.919, and 0.887 for 2019, 2020, and 2021, respectively, and performed well in urban regions. Additionally, the Shapley Additive Explanations (SHAP) method revealed that features such as satellite and CTM NO<sub>2</sub>, precipitation, green space, and road density significantly contributed to estimations through both spatial and local channels. The mapping results closely aligned with ground-level observations and accurately captured spatial variations. This study advances NO<sub>2</sub> concentration estimation by applying adaptable deep learning techniques and enhances the understanding of air pollution dynamics.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104506"},"PeriodicalIF":7.6,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143725753","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Emanuele Barca, Maria Clementina Caputo, Rita Masciale
{"title":"Building the optimal hybrid spatial Data-Driven Model: Balancing accuracy and complexity","authors":"Emanuele Barca, Maria Clementina Caputo, Rita Masciale","doi":"10.1016/j.jag.2025.104478","DOIUrl":"10.1016/j.jag.2025.104478","url":null,"abstract":"<div><div>Mapping environmental variables is crucial for natural resource management. Researchers and scholars have continually advanced this field with modern techniques such as Integrated Nested Laplace Approximation (INLA), Deep Learning (DL), and Graph Neural Networks (GNN) models. While effective, these models often present a significant challenge due to their <em>black</em> nature, which obscures the process of generating final maps from raw data. Recent theoretical breakthroughs have shown that white/grey-box models can achieve the same level of accuracy as these advanced techniques, debunking the belief that complex models are necessarily the most accurate. Based on these findings, we have developed a methodology that employs a series of statistical tests and data analytics to identify essential features hidden in spatial data in order to assess the predictive model (of white/grey kind) that best approximates underlying spatial processes. This methodology profiles the model that better adapts to the data, aiding in the selection of the simplest model that achieves the desired accuracy, functioning similarly to a recommender system for model selection. Furthermore, the set of permissible models includes only regressive-like ones to clarify the data’s contribution to map construction and can be applied to a wide range of datasets. By reducing complexity, this approach enhances the transparency of the model’s results. Real-world dataset demonstrates this methodology’s remarkable ability to produce highly accurate results.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"139 ","pages":"Article 104478"},"PeriodicalIF":7.6,"publicationDate":"2025-03-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143697510","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}