International journal of applied earth observation and geoinformation : ITC journal最新文献

{"title":"Refining multi-modal remote sensing image matching with repetitive feature optimization","authors":"Yifan Liao ,&nbsp;Ke Xi ,&nbsp;Huijin Fu ,&nbsp;Lai Wei ,&nbsp;Shuo Li ,&nbsp;Qiang Xiong ,&nbsp;Qi Chen ,&nbsp;Pengjie Tao ,&nbsp;Tao Ke","doi":"10.1016/j.jag.2024.104186","DOIUrl":"10.1016/j.jag.2024.104186","url":null,"abstract":"<div><div>Existing methods for matching multi-modal remote sensing images (MRSI) demonstrate considerable adaptability. However, high-precision matching for rectification remains challenging due to differing imaging mechanisms in cross-modal remote sensing images, leading to numerous non-repeated detailed feature points. Additionally, assuming linear transformations between images conflicts with the complex aberrations present in remote sensing images, limiting matching accuracy. This paper aims to elevate matching accuracy by implementing a detailed texture removal strategy that effectively isolates repeatable structural features. Subsequently, we construct a radiation-invariant similarity function within a generalized gradient framework for least-squares matching, specifically designed to mitigate nonlinear geometric and radiometric distortions across MRSIs. Comprehensive qualitative and quantitative evaluations across multiple datasets, employing substantial manual checkpoints, demonstrate that our method significantly enhances matching accuracy for image data involving multiple modal combinations and outperforms the current state-of-the-art solutions in matching accuracy. Additionally, rectification experiments employing WorldView and TanDEM-X images validate our method’s ability to achieve a matching accuracy of 1.05 pixels, thereby indicating its practical utility and generalization capacity. Access to experiment-related data and codes will be provided at <span><span>https://github.com/LiaoYF001/refinement/</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104186"},"PeriodicalIF":7.6,"publicationDate":"2024-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142422893","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":"Estimation of district-level spring barley yield in southern Sweden using multi-source satellite data and random forest approach","authors":"Xueying Li ,&nbsp;Hongxiao Jin ,&nbsp;Lars Eklundh ,&nbsp;El Houssaine Bouras ,&nbsp;Per-Ola Olsson ,&nbsp;Zhanzhang Cai ,&nbsp;Jonas Ardö ,&nbsp;Zheng Duan","doi":"10.1016/j.jag.2024.104183","DOIUrl":"10.1016/j.jag.2024.104183","url":null,"abstract":"<div><div>Remote sensing observations and artificial intelligence algorithms have emerged as key components for crop yield estimation at various scales during the past decades. However, the utilization of multi-source satellite data and machine learning for estimating aggregated crop yield at the regional level in Europe has been only scarcely explored. Our study aims to bridge this research gap by focusing on the district-level spring barley yield estimation in southern Sweden from 2017 to 2022. We developed an estimation method with the random forest (RF) approach using four satellite-derived products along with two climate variables. These variables were used individually and in combinations as inputs for the RF approach. The results showed that vegetation indices (VIs) outperformed solar-induced chlorophyll fluorescence (SIF) in barley yield estimation, while combining VIs and SIF variables achieved the highest model performance (R² = 0.77, RMSE = 488 kg/ha). The inclusion of climate variables generally had little added contributions to the model performance. Importantly, barley yield prediction could be achieved two months prior to harvest, using monthly VIs and SIF data from April and May. Our study demonstrated the feasibility of using freely accessible satellite data and the machine learning approach for estimating crop yield at the pan-European regional level. We expect that our proposed methodology can be extended to different crop types and regional-scale crop yield estimation in Europe, benefiting national and local authorities in making agricultural productivity decisions.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104183"},"PeriodicalIF":7.6,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142422518","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":"A two-dimensional bare soil separation framework using multi-temporal Sentinel-2 images across China","authors":"Jie Xue ,&nbsp;Xianglin Zhang ,&nbsp;Yuyang Huang ,&nbsp;Songchao Chen ,&nbsp;Lingju Dai ,&nbsp;Xueyao Chen ,&nbsp;Qiangyi Yu ,&nbsp;Su Ye ,&nbsp;Zhou Shi","doi":"10.1016/j.jag.2024.104181","DOIUrl":"10.1016/j.jag.2024.104181","url":null,"abstract":"<div><div>Accurate and detailed spatial–temporal soil information is crucial for soil quality assessment worldwide, particularly in the countries with large populations and extensive agricultural areas. Using remote sensing technology to generate bare soil reflectance composites has been shown as a prerequisite for effectively modeling soil properties. However, most bare soil extraction methods rely on the single-period satellite imagery, making it difficult to produce a complete bare soil map. Although some developed methods have explored the advantages of multitemporal images, single indicators (e.g., Normalized Difference Vegetation Index and Normalized Burn Ratio 2) are prone to misidentifying bare soil as other land cover types such as impervious surface. Additionally, these methodologies were designed for specific areas and coarse spatial resolution images, leaving their generalizability to other areas or larger scales underexplored. Therefore, we proposed a Two-Dimensional Bare Soil Separation (TDBSS) framework to generate the bare soil composites of Chinese cropland at 10-m spatial resolution using multi-temporal Sentinel-2 images. This method employs the Normalized Difference Red/Green Redness Index and Soil Adjusted Vegetation Index as bidimensional indicators. We identified optimal thresholds for these indicators by analyzing ecoregion-specific samples and then implemented them across nine major agricultural zones in China. Additionally, we evaluated the framework against three prevalent bare soil extraction methods (i.e., Barest Pixel Composite, Soil Composite Mapping Processor, and Geospatial Soil Sensing System) based on spatial accuracy. The results showed that TDBSS outperformed the others with the highest overall accuracy of 78.28% and the lowest omission error of 0.198. The findings indicated that the TDBSS algorithm is competent in mapping bare soil at a national scale. The produced composite map of bare soil reflectance is particularly valuable for retrieving soil attributes in Chinese cropland. The TDBSS method can be easily implemented across broad areas with computational efficiency, contributing to land management, food security, and the development of policies for precision agriculture.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104181"},"PeriodicalIF":7.6,"publicationDate":"2024-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142330103","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":"Transfer learning reconstructs submarine topography for global mid-ocean ridges","authors":"Yinghui Jiang ,&nbsp;Sijin Li ,&nbsp;Yanzi Yan ,&nbsp;Bingqing Sun ,&nbsp;Josef Strobl ,&nbsp;Liyang Xiong","doi":"10.1016/j.jag.2024.104182","DOIUrl":"10.1016/j.jag.2024.104182","url":null,"abstract":"<div><div>Mid-ocean ridges are unique, tectonically active geographical units on Earth that profoundly control the ocean environment and dynamics at the global scale. However, high-resolution topographic data from mid-ocean ridges are rarely available due to the difficulty in detecting ocean floors, which further limits ocean research at the global scale. Here, we divide the global mid-ocean ridge system into 2805 tiles and reconstruct their high-resolution topography by using a transfer learning approach with freely available low-resolution digital elevation models (DEMs) and limited high-resolution DEMs. A high-frequency terrain feature-based deep residual network is proposed to generate high-resolution global mid-ocean ridge DEMs. In this network, topographic knowledge related to mid-ocean ridges is integrated and quantified to improve the learning efficiency and reconstruction quality of the network. A series of verifications and evaluations demonstrate the reliability of reconstructed topographies for submarine topography research. We observe that reconstructed topography can achieve good environmental understanding and information acquisition in the global mid-ocean ridge range. We find that the complexity of the previous terrain environment is underestimated by 26.63% in terms of the slope gradient and by 14.95% in terms of terrain relief, while a 101.10% information improvement can be obtained for the reconstructed topography. The reconstructed topography indicates that diverse and intricate topographical environments of mid-ocean ridges exist among different ocean regions. The proposed transfer learning method for reconstructing high-resolution mid-ocean ridge topographies is valuable and can be utilized for reconstructing information in regions that are difficult to observe directly and lack sufficient data.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104182"},"PeriodicalIF":7.6,"publicationDate":"2024-09-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142330067","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":"A fine crop classification model based on multitemporal Sentinel-2 images","authors":"Tengfei Qu ,&nbsp;Hong Wang ,&nbsp;Xiaobing Li ,&nbsp;Dingsheng Luo ,&nbsp;Yalei Yang ,&nbsp;Jiahao Liu ,&nbsp;Yao Zhang","doi":"10.1016/j.jag.2024.104172","DOIUrl":"10.1016/j.jag.2024.104172","url":null,"abstract":"<div><div>Information on the sowing areas and yields of crops is important for ensuring food security and reforming the agricultural modernization process, while crop classification and identification are core issues when attempting to acquire information on crop planting areas and yields. Obtaining information on crop planting areas and yields in a timely and accurate manner is highly important for optimizing crop planting structures, formulating agricultural policies, and ensuring national economic development. In this paper, a fine crop classification model based on multitemporal Sentinel-2 images, CTANet, is proposed. It comprises a convolutional attention architecture (CAA) and a temporal attention architecture (TAA), incorporating spatial attention modules, channel attention modules and temporal attention modules. These modules adaptively weight each pixel, channel and temporal phase of the given feature map to mitigate the intraclass spatial heterogeneity, spectral variability and temporal variability of crops. Additionally, the auxiliary features of significant importance for each crop category are identified using the random forest-SHAP algorithm, enabling the construction of classification datasets containing spectral bands, spectral bands with auxiliary features, and spectral bands with optimized auxiliary features. Evaluations conducted on three crop classification datasets revealed that the proposed CTANet approach and its key CANet component demonstrated superior crop classification performance on the classification dataset consisting of spectral bands and optimized auxiliary features in comparisons with the other tested models. Based on this dataset, CTANet achieved higher validation accuracy and lower validation loss than those of the other methods, and during testing, it attained the highest overall accuracy (93.9 %) and MIoU (87.5 %). When identifying rice, maize, and soybeans, the F1 scores of CTANet reached 95.6 %, 95.7 %, and 94.7 %, and the IoU scores were 91.6 %, 91.7 %, and 89.9 %, respectively, significantly exceeding those of some commonly used deep learning models. This indicates the potential of the proposed method for distinguishing between different crop types.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104172"},"PeriodicalIF":7.6,"publicationDate":"2024-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142324166","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":"A novel BH3DNet method for identifying pine wilt disease in Masson pine fusing UAS hyperspectral imagery and LiDAR data","authors":"Geng Wang ,&nbsp;Nuermaimaitijiang Aierken ,&nbsp;Guoqi Chai ,&nbsp;Xuanhao Yan ,&nbsp;Long Chen ,&nbsp;Xiang Jia ,&nbsp;Jiahao Wang ,&nbsp;Wenyuan Huang ,&nbsp;Xiaoli Zhang","doi":"10.1016/j.jag.2024.104177","DOIUrl":"10.1016/j.jag.2024.104177","url":null,"abstract":"<div><div>Pine Wilt Disease (PWD) is a forest infectious disease that inflicts substantial economic losses to China’s forestry. Its rapid spread and the significant challenges associated with its control make early detection of infected trees crucial for disaster prevention. Unmanned aerial systems (UASs) hyperspectral imaging (HSI) and light detection and ranging (LiDAR) technologies provide high-resolution spectral diagnostic information coupled with intricate three-dimensional structural data, which has potential for fine grained monitoring of PWD. However, how to fuse HSI and LiDAR data to identify the early infected individual trees is still a challenge. This study presents a novel instance segmentation network, BH3DNet, to identify individual trees at different PWD-infected stages by extracting high-level abstract features based on the fusion of drone HSI and LiDAR data. BH3DNet introduces the PointNet++ model as the base network, and incorporates a shared encoder and twin parallel decoders to align semantic category prediction and instance segmentation of individual trees in an end-to-end approach. By applying an enhanced point cloud dataset that fuses drone HSI and LiDAR point cloud data, this model facilitates the identification of PWD infection stages at the individual tree scale. We evaluated the proposed model in a Masson pine forest stand sparsely mixed with broadleaf trees in a variety of infection states ranging from healthy to severely infected by PWD, and compared the performance of the model using the RGB bands, full HSI bands and screened bands as inputs, respectively. BH3DNet achieves an overall accuracy of 89.65 % with a Kappa <strong>×</strong> 100 of 87.29 for identifying individual trees using screened HSI bands and LiDAR point cloud, significantly outperforming the Mask R-CNN using only HSI data (overall accuracy: 70.81 %, Kappa × 100: 64.16). Moreover, BH3DNet’s accuracy at the early infection stage reaches 83.75 %. It proves that fusing HSI and point cloud data reflects the information of individual trees distribution and infection status, and the BH3DNet is suitable for high-precision monitoring of PWD.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104177"},"PeriodicalIF":7.6,"publicationDate":"2024-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142327722","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":"Spectral domain strategies for hyperspectral super-resolution: Transfer learning and channel enhance network","authors":"Zhi-Zhu Ge ,&nbsp;Zhao Ding ,&nbsp;Yang Wang ,&nbsp;Li-Feng Bian ,&nbsp;Chen Yang","doi":"10.1016/j.jag.2024.104180","DOIUrl":"10.1016/j.jag.2024.104180","url":null,"abstract":"<div><div>As the network structures continue to innovate and evolve, significant achievements have been achieved in hyperspectral image super-resolution tasks. However, how to further explore the spectral domain potential from prior knowledge and channel-enhanced structures to achieve better performance has inspired the following two works: Firstly, to systematically compare prior knowledge of spectral with spatial domain for HSI-SR tasks, four transfer learning strategies are proposed. The superior performance of the Relevant Channel/Random Space (RCRS) strategy reveals the importance of spectral feature reconstruction in HSI-SR tasks. Meanwhile, an interesting phenomenon has been observed that even without training on real datasets, the model can already exhibit a core or even decent super-resolution capability based solely on prior knowledge of above four strategies. Secondly, a dual-branch channel network with complementary channel feature extraction (CCFE) and adjacent channel feature extraction (ACFE) module is designed for spectral feature enhancement, which demonstrate superior performance compared to state-of-the-art methods on six datasets. To conclude, the effectiveness of RCRS strategy with pseudo prior channel knowledge on seven dual-input and eight single-input networks, as well as superiority of the proposed channel-enhanced network indicate the importance of spectral properties for HSI-SR tasks.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104180"},"PeriodicalIF":7.6,"publicationDate":"2024-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142324164","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":"High-accuracy bathymetric method fusing ICESAT-2 datasets and the two-media photogrammetry model","authors":"Yifu Chen ,&nbsp;Lin Wu ,&nbsp;Yuan Le ,&nbsp;Qian Zhao ,&nbsp;Dongfang Zhang ,&nbsp;Zhenge Qiu","doi":"10.1016/j.jag.2024.104179","DOIUrl":"10.1016/j.jag.2024.104179","url":null,"abstract":"<div><div>Improving the accuracy of nearshore bathymetric measurements is essential for understanding coastal environments, resource management, and navigation. The Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) is the first laser satellite that uses the photon-counting technique. The ICESat-2 is equipped with the Advanced Topographic Laser Altimeter System (ATLAS), which enables higher-accuracy measurements of water, ice, and land elevation on Earth. Two-media photogrammetric bathymetry is a type of nearshore bathymetric technology that uses the geometrical characteristics of light rays. With this technique, the accuracy and reliability mainly depend on eliminating systematic errors and ensuring accurate spatial photogrammetric positioning relative to the object being measured. To improve the bathymetric accuracy of two-media photogrammetry, we integrated high-accuracy elevation data from photon datasets as constraining and control parameters. The improved method effectively eliminated systematic errors in two-media photogrammetry during the established joint-block adjustment model. To improve its accuracy and reliability, we employed multispectral WorldView-2 stereo images in our experiments. Furthermore, the bathymetric results were validated and assessed using in situ and photon data. The experimental results show that the highest accuracy achieved with the bathymetric measurements in our study area was a root mean square error (RMSE) of 0.96 m and a mean absolute error of 0.57 m. Using the proposed fusion method, the bathymetric accuracy (as measured using the RMSE) was 1 m higher than that of two-media photogrammetry without the photon datasets.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104179"},"PeriodicalIF":7.6,"publicationDate":"2024-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142324163","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":"Scale matters: How spatial resolution impacts remote sensing based urban green space mapping?","authors":"Zhongwen Hu ,&nbsp;Yuqiu Chu ,&nbsp;Yinghui Zhang ,&nbsp;Xinyue Zheng ,&nbsp;Jingzhe Wang ,&nbsp;Wanmin Xu ,&nbsp;Jing Wang ,&nbsp;Guofeng Wu","doi":"10.1016/j.jag.2024.104178","DOIUrl":"10.1016/j.jag.2024.104178","url":null,"abstract":"<div><div>Urban green spaces (UGS) provide ecological and habitat benefits such as carbon sequestration, oxygen production, humidity increase, noise reduction, and pollution absorption. UGS maps derived from remote sensing images serve as the fundamental data for urban planning and carbon sequestration assessments. However, the spatial resolution of remote sensing image and the pattern of urban structures significantly influence UGS mapping, making it challenging to obtain accurate UGS maps. To investigate the impact of spatial resolution on UGS mapping, this study utilized five different spatial resolution datasets: Gaofen2 (1 m, 4 m), Sentinel2 (10 m), and Landsat8 (15 m, 30 m). Random forest, LightGBM, and support vector machine were employed to map UGS, and the accuracies of UGS maps at different spatial resolutions were compared. Subsequently, the spatial distribution patterns of uncertainties in UGS maps were analyzed from both overall and urban functional zone perspectives. Furthermore, the uncertainty analysis of UGS mapping was conducted considering different landscape patterns in urban functional zones. The results indicate: (1) UGS map varies at different spatial resolution. Higher uncertainties associated with coarser spatial resolutions. Medium and coarse spatial resolution images inadequately capture the fine-grained distribution of urban green spaces. (2) Uncertainty in UGS mapping at different spatial resolutions is generally consistent in spatial distribution. From a functional zoning perspective, the accuracy of green space mapping over non-natural zones is sensitive to spatial resolution. (3) The distribution pattern of UGS patches affects the accuracy of UGS mapping. Uncertainty can be reduced in UGS mapping at medium and coarse spatial resolutions based on UGS landscape pattern indices by multiple linear regression, random forest and LightGBM model. This study comprehensively reveals that uncertainties in mapping UGS from multi-spatial resolution remote sensing images vary across urban functional zones and landscape pattern indices, and it is the first attempt to propose methods for UGS area correction based on landscape pattern indices. The results of this study will facilitate the application of remote sensing data at different spatial resolutions in urban areas.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104178"},"PeriodicalIF":7.6,"publicationDate":"2024-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142318673","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":"An improved geographic pattern based residual neural network model for estimating PM2.5 concentrations","authors":"Heng Su ,&nbsp;Yumin Chen ,&nbsp;Huangyuan Tan ,&nbsp;John P. Wilson ,&nbsp;Lanhua Bao ,&nbsp;Ruoxuan Chen ,&nbsp;Jiaxin Luo","doi":"10.1016/j.jag.2024.104174","DOIUrl":"10.1016/j.jag.2024.104174","url":null,"abstract":"<div><p>Accurate and continuous PM_2.5 data is essential for effective prevention of PM_2.5 pollution. Despite the achievements of deep learning methods in estimating PM_2.5 concentrations, existing neural network models have relied too much on the self-learning capability and have ignored geographic patterns of PM_2.5. Few have taken a geographic perspective when modeling PM_2.5, resulting in lower model interpretability. In this paper, rather than inputting spatiotemporal information directly into the networks, we propose an improved geographic pattern based residual neural network (IGeop-ResNet) for estimating PM_2.5 concentrations in the Beijing-Tianjin-Hebei region (BTH) of China considering spatial heterogeneity and spatial autocorrelation by introducing spatial eigenvector and attention mechanism, as well as the encoding and embedding methods for temporal categorical variables. A DEM-weighted loss function was introduced to enhance the spatial predictive ability, particularly in high-altitude regions. The results show that the IGeop-ResNet model achieves excellent spatial predictive abilities (R² of 0.925 in terms of station-based cross-validation) and offers a certain level of interpretability compared to the Ori-STResNet (ordinary directly inputs temporal and spatial information in the ResNet model) and the Geop-ResNet model (without the DEM-weighted loss function). Continuous maps derived from the IGeop-ResNet model suggest the PM_2.5 concentrations in the BTH region exhibited a downward trend from 2015 to 2018 and experienced a sharp drop in 2017. The results indicate that NO₂ is the Granger cause of PM_2.5, while the relationship between SO₂ and PM_2.5 is insignificant.</p></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"134 ","pages":"Article 104174"},"PeriodicalIF":7.6,"publicationDate":"2024-09-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S1569843224005302/pdfft?md5=105431b6064ef059c232701a3e987868&pid=1-s2.0-S1569843224005302-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142271457","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}