Abstract
High-resolution urban surface information, e.g., the fraction of impervious/pervious surface, is pivotal in studies of local thermal/wind environments and air pollution. In this study, we introduced and validated a domain adaptive land cover classification model, to automatically classify Google Earth images into pixel-based land cover maps. By combining domain adaptation (DA) and semi-supervised learning (SSL) techniques, our model demonstrates its effectiveness even when trained with a limited dataset derived from Gaofen2 (GF2) satellite images. The model's overall accuracy on the translated GF2 dataset improved significantly from 19.5% to 75.2%, and on the Google Earth image dataset from 23.1% to 61.5%. The overall accuracy is 2.9% and 3.4% higher than when using only DA. Furthermore, with this model, we derived land cover maps and investigated the impact of land surface composition on the local meteorological parameters and air pollutant concentrations in the three most developed urban agglomerations in China, i.e., Beijing, Shanghai and the Great Bay Area (GBA). Our correlation analysis reveals that air temperature exhibits a strong positive correlation with neighboring artificial impervious surfaces, with Pearson correlation coefficients higher than 0.6 in all areas except during the spring in the GBA. However, the correlation between air pollutants and land surface composition is notably weaker and more variable. The primary contribution of this paper is to provide an efficient method for urban land cover extraction which will be of great value for assessing the urban surface composition, quantifying the impact of land use/land cover, and facilitating the development of informed policies.
Similar content being viewed by others
Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. There is no custom code produced during the data collection and model evaluation process. The models are mainly constructed with the semantic segmentation model DeepLab v3 (https://github.com/fregu856/deeplabv3), and CycleGAN (https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix).
Notes
References
Aryal J, Neupane B (2023) Multi-Scale Feature Map Aggregation and Supervised Domain Adaptation of Fully Convolutional Networks for Urban Building Footprint Extraction. Remote Sens 15:488. https://doi.org/10.3390/rs15020488
Badrinarayanan V, Kendall A, Cipolla R (2017) SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation. IEEE Trans Pattern Anal Mach Intell 39:2481–2495. https://doi.org/10.1109/TPAMI.2016.2644615
Chen W, Zhou Y, Wu Q et al (2020) Urban Building Type Mapping Using Geospatial Data: A Case Study of Beijing, China. Remote Sens 12:2805. https://doi.org/10.3390/rs12172805
Chen L-C, Papandreou G, Schroff F, Adam H (2017) Rethinking Atrous Convolution for Semantic Image Segmentation. arXiv preprint arXiv:1706.05587. https://doi.org/10.48550/arXiv.1706.05587
Chen Y, Li W, Sakaridis C, et al (2018) Domain adaptive faster R-CNN for object detection in the Wild. Proceedings of the IEEE conference on computer vision and pattern recognition 3339–3348. https://doi.org/10.48550/arXiv.1803.03243
Diakogiannis FI, Waldner F, Caccetta P, Wu C (2020) ResUNet-a: A deep learning framework for semantic segmentation of remotely sensed data. ISPRS J Photogramm Remote Sens 162:94–114. https://doi.org/10.1016/j.isprsjprs.2020.01.013
Du H, Wang D, Wang Y et al (2016) Influences of land cover types, meteorological conditions, anthropogenic heat and urban area on surface urban heat island in the Yangtze River Delta Urban Agglomeration. Sci Total Environ 571:461–470. https://doi.org/10.1016/j.scitotenv.2016.07.012
Dugord PA, Lauf S, Schuster C, Kleinschmit B (2014) Land use patterns, temperature distribution, and potential heat stress risk - The case study Berlin, Germany. Comput Environ Urban Syst 48:86–98. https://doi.org/10.1016/j.compenvurbsys.2014.07.005
Fan Y, Wang Q, Yin S, Li Y (2019) Effect of city shape on urban wind patterns and convective heat transfer in calm and stable background conditions. Build Environ 162:106288. https://doi.org/10.1016/j.buildenv.2019.106288
Fan Y, Ding X, Hang J, Ge J (2020) Characteristics of urban air pollution in different regions of China between 2015 and 2019. Build Environ 180:107048. https://doi.org/10.1016/j.buildenv.2020.107048
Fan Y, Ding X, Wu J et al (2021) High spatial-resolution classification of urban surfaces using a deep learning method. Build Environ 200:107949. https://doi.org/10.1016/j.buildenv.2021.107949
Fu B, Li S, Lao Z et al (2023a) Multi-sensor and multi-platform retrieval of water chlorophyll a concentration in karst wetlands using transfer learning frameworks with ASD, UAV, and Planet CubeSate reflectance data. Sci Total Environ 901:165963. https://doi.org/10.1016/j.scitotenv.2023.165963
Fu B, Sun X, Li Y et al (2023b) Combination of super-resolution reconstruction and SGA-Net for marsh vegetation mapping using multi-resolution multispectral and hyperspectral images. Int J Digit Earth 16:2724–2761. https://doi.org/10.1080/17538947.2023.2234340
Gao L, Wen X, Guo Y et al (2014) Spatiotemporal Variability of Carbon Flux from Different Land Use and Land Cover Changes: A Case Study in Hubei Province, China. Energies 7:2298–2316. https://doi.org/10.3390/en7042298
He K, Zhang X, Ren S, Sun J (2016) Deep residual learning for image recognition. IEEE conference on computer vision and pattern recognition (CVPR) 770–778. https://doi.org/10.48550/arXiv.1512.03385
Hoffman J, Wang D, Yu F, Darrell T (2016) FCNs in the wild: pixel-level adversarial and constraint-based adaptation. arXiv preprint arXiv:1612.02649. https://doi.org/10.48550/arXiv.1612.02649
Hoffman J, Tzeng E, Park T, et al (2018) CyCADA: Cycle-consistent adversarial domain adaptation. International conference on machine learning 1989–1998. https://doi.org/10.48550/arXiv.1711.03213
Hrebtov M, Hanjalić K (2019) River-Induced Anomalies in Seasonal Variation of Traffic-Emitted CO Distribution over the City of Krasnoyarsk. Atmosphere (basel) 10:407. https://doi.org/10.3390/atmos10070407
Hu M, Wang Y, Wang S et al (2021) Spatial-temporal heterogeneity of air pollution and its relationship with meteorological factors in the Pearl River Delta, China. Atmos Environ 254:118415. https://doi.org/10.1016/j.atmosenv.2021.118415
Ji S, Di W, Luo M (2021) Generative Adversarial Network-Based Full-Space Domain Adaptation for Land Cover Classification from Multiple-Source Remote Sensing Images. IEEE Trans Geosci Remote Sens 59:3816–3828. https://doi.org/10.1109/TGRS.2020.3020804
Jing L, Tian Y (2021) Self-Supervised Visual Feature Learning With Deep Neural Networks: A Survey. IEEE Trans Pattern Anal Mach Intell 43:4037–4058. https://doi.org/10.1109/TPAMI.2020.2992393
Klonus S, Ehlers M (2009) Performance of evaluation methods in image fusion. 2009 12th Int Conf Inf Fusion. FUSION 2009:1409–1416
Lee JH, Wu CF, Hoek G et al (2014) Land use regression models for estimating individual NOx and NO2 exposures in a metropolis with a high density of traffic roads and population. Sci Total Environ 472:1163–1171. https://doi.org/10.1016/j.scitotenv.2013.11.064
Li S, Feng K, Li M (2017) Identifying the main contributors of air pollution in Beijing. J Clean Prod 163:S359–S365. https://doi.org/10.1016/j.jclepro.2015.10.127
Li C, Wang Z, Li B et al (2019a) Investigating the relationship between air pollution variation and urban form. Build Environ 147:559–568. https://doi.org/10.1016/j.buildenv.2018.06.038
Li Y, Yuan L, Vasconcelos N (2019b) Bidirectional learning for domain adaptation of semantic segmentation. IEEE/CVF conference on computer vision and pattern recognition (CVPR) 6929–6938. https://doi.org/10.48550/arXiv.1904.10620
Li W, Li F, Luo Y, et al (2020) Deep domain adaptive object detection: a survey. IEEE Symposium series on computational intelligence (SSCI) 1808–1813. https://doi.org/10.48550/arXiv.2002.06797
Lin Y-C, Cheng M-T (2007) Evaluation of formation rates of NO2 to gaseous and particulate nitrate in the urban atmosphere. Atmos Environ 41:1903–1910. https://doi.org/10.1016/j.atmosenv.2006.10.065
Lin L, Chan TO, Ge E et al (2020) Effects of urban land expansion on decreasing atmospheric moisture in Guangdong, South China. Urban Clim 32:100626. https://doi.org/10.1016/j.uclim.2020.100626
Liu W, You H, Dou J (2009) Urban-rural humidity and temperature differences in the Beijing area. Theor Appl Climatol 96:201–207. https://doi.org/10.1007/s00704-008-0024-6
Liu XG, Li J, Qu Y et al (2013) Formation and evolution mechanism of regional haze: A case study in the megacity Beijing, China. Atmos Chem Phys 13:4501–4514. https://doi.org/10.5194/acp-13-4501-2013
Liu B, Wu J, Zhang J et al (2017) Characterization and source apportionment of PM2.5 based on error estimation from EPA PMF 5.0 model at a medium city in China. Environ Pollut 222:10–22. https://doi.org/10.1016/j.envpol.2017.01.005
Liu Y, Xu Y, Weng F et al (2021) Impacts of urban spatial layout and scale on local climate: A case study in Beijing. Sustain Cities Soc 68:102767. https://doi.org/10.1016/j.scs.2021.102767
Liu Q, Peng J, Ning Y et al (2023) Refined Prototypical Contrastive Learning for Few-Shot Hyperspectral Image Classification. IEEE Trans Geosci Remote Sens 61:1–14. https://doi.org/10.1109/TGRS.2023.3257341
Liu C, Zeng D, Wu H, et al (2020) Urban land cover classification of high-resolution aerial imagery using a relation-enhanced multiscale convolutional network. Remote Sens 12. https://doi.org/10.3390/rs12020311
Lokoshchenko MA (2017) Urban Heat Island and Urban Dry Island in Moscow and Their Centennial Changes. J Appl Meteorol Climatol 56:2729–2745. https://doi.org/10.1175/JAMC-D-16-0383.1
Luo M, Ji S (2022) Cross-spatiotemporal land-cover classification from VHR remote sensing images with deep learning based domain adaptation. ISPRS J Photogramm Remote Sens 191:105–128. https://doi.org/10.1016/j.isprsjprs.2022.07.011
Mazza D, Pagani M (2020) Automatic Differentiation in PCF. Proc ACM Program Lang 5:1–27. https://doi.org/10.1145/3434309
National Bureau of Statistics of China (2015) China Statistical Yearbook. China Stat. Press, Beijing
Oke TR (1982) The energetic basis of the urban heat island. Q J R Meteorol Soc 108:1–24. https://doi.org/10.1002/qj.49710845502
Oke TR, Mills G, Christen A, Voogt JA (2017) Urban Climates. Cambridge University Press, Cambridge
Pan SJ, Yang Q (2010) A survey on transfer learning. IEEE Trans Knowl Data Eng 22:1345–1359. https://doi.org/10.1109/TKDE.2009.191
Peng J, Huang Y, Sun W et al (2022) Domain Adaptation in Remote Sensing Image Classification: A Survey. IEEE J Sel Top Appl Earth Obs Remote Sens 15:9842–9859. https://doi.org/10.1109/JSTARS.2022.3220875
Ren J, Hao Y, Simayi M et al (2021) Spatiotemporal variation of surface ozone and its causes in Beijing, China since 2014. Atmos Environ 260:118556. https://doi.org/10.1016/j.atmosenv.2021.118556
Ronneberger O, Fischer P, Brox T (2015) U-Net: Convolutional Networks for Biomedical Image Segmentation. Lect Notes Comput Sci (including Subser Lect Notes Artif Intell Lect Notes Bioinformatics) 9351:12–20. https://doi.org/10.1007/978-3-319-24574-4_2
Scott GJ, Marcum RA, Davis CH, Nivin TW (2017) Training Deep Convolutional Neural Networks for Land-Cover Classification of High-Resolution Imagery. IEEE Geosci Remote Sens Lett 14:1638–1642. https://doi.org/10.1109/LGRS.2017.2722988
Sha C, Wang X, Lin Y et al (2018) The impact of urban open space and ‘lift-up’ building design on building intake fraction and daily pollutant exposure in idealized urban models. Sci Total Environ 633:1314–1328. https://doi.org/10.1016/j.scitotenv.2018.03.194
Sokhi RS, Singh V, Querol X, et al (2021) A global observational analysis to understand changes in air quality during exceptionally low anthropogenic emission conditions. Environ Int 157. https://doi.org/10.1016/j.envint.2021.106818
Stewart ID, Oke TR (2012) Local climate zones for urban temperature studies. Bull Am Meteorol Soc 93:1879–1900. https://doi.org/10.1175/BAMS-D-11-00019.1
Tao J, Zhang L, Cao J et al (2017) Source apportionment of PM2.5 at urban and suburban areas of the Pearl River Delta region, south China - With emphasis on ship emissions. Sci Total Environ 574:1559–1570. https://doi.org/10.1016/j.scitotenv.2016.08.175
Tao H, Xing J, Zhou H et al (2018) Impacts of land use and land cover change on regional meteorology and air quality over the Beijing-Tianjin-Hebei region, China. Atmos Environ 189:9–21. https://doi.org/10.1016/j.atmosenv.2018.06.033
Toldo M, Maracani A, Michieli U, Zanuttigh P (2020) Unsupervised Domain Adaptation in Semantic Segmentation: A Review. Technologies 8:35. https://doi.org/10.3390/technologies8020035
van Engelen JE, Hoos HH (2020) A survey on semi-supervised learning. Mach Learn 109:373–440. https://doi.org/10.1007/s10994-019-05855-6
Volpi M, Tuia D (2017) Dense semantic labeling of subdecimeter resolution images with convolutional neural networks. IEEE Trans Geosci Remote Sens 55:881–893. https://doi.org/10.1109/TGRS.2016.2616585
Wang Y, Li L, Chen C et al (2014) Source apportionment of fine particulate matter during autumn haze episodes in Shanghai, China. J Geophys Res Atmos 119:1903–1914. https://doi.org/10.1002/2013JD019630
Wilson G, Cook DJ (2020) A Survey of Unsupervised Deep Domain Adaptation. ACM Trans Intell Syst Technol 11:1–46. https://doi.org/10.1145/3400066
Wong SC, Gatt A, Stamatescu V, McDonnell MD (2016) Understanding Data Augmentation for Classification: When to Warp? 2016 Int Conf Digit Image Comput Tech Appl DICTA 2016. https://doi.org/10.1109/DICTA.2016.7797091
Xu D, Zhou D, Wang Y et al (2019) Field measurement study on the impacts of urban spatial indicators on urban climate in a Chinese basin and static-wind city. Build Environ 147:482–494. https://doi.org/10.1016/j.buildenv.2018.10.042
Xu H, Chen H, Zhou X et al (2020) Research on the relationship between urban morphology and air temperature based on mobile measurement: A case study in Wuhan, China. Urban Clim 34:100671. https://doi.org/10.1016/j.uclim.2020.100671
Xu M, Wu M, Chen K et al (2022) The Eyes of the Gods: A Survey of Unsupervised Domain Adaptation Methods Based on Remote Sensing Data. Remote Sens 14:4380. https://doi.org/10.3390/rs14174380
Xue Y, Tian H, Yan J et al (2016) Temporal trends and spatial variation characteristics of primary air pollutants emissions from coal-fired industrial boilers in Beijing, China. Environ Pollut 213:717–726. https://doi.org/10.1016/j.envpol.2016.03.047
Xue R, Wang S, Li D et al (2020) Spatio-temporal variations in NO2 and SO2 over Shanghai and Chongming Eco-Island measured by Ozone Monitoring Instrument (OMI) during 2008–2017. J Clean Prod 258:120563. https://doi.org/10.1016/j.jclepro.2020.120563
Yan H, Dong L (2015) The impacts of land cover types on urban outdoor thermal environment: the case of Beijing, China. J Environ Heal Sci Eng 13:1–7. https://doi.org/10.1186/s40201-015-0195-x
Yang H, Chen T, Lin Y et al (2020) Integrated impacts of tree planting and street aspect ratios on CO dispersion and personal exposure in full-scale street canyons. Build Environ 169:106529. https://doi.org/10.1016/j.buildenv.2019.106529
Yang Y, Xu Z (2020) Rethinking the value of labels for improving class-imbalanced learning. Advances in neural information processing systems 33:19290–19301. https://doi.org/10.48550/arXiv.2006.07529
Yoo JM, Jeong MJ, Kim D et al (2015) Spatiotemporal variations of air pollutants (O3, NO2, SO2, CO, PM10, and VOCs) with land-use types. Atmos Chem Phys 15:10857–10885. https://doi.org/10.5194/acp-15-10857-2015
Zhang W, Huang B, Luo D (2014) Effects of land use and transportation on carbon sources and carbon sinks: A case study in Shenzhen, China. Landsc Urban Plan 122:175–185. https://doi.org/10.1016/j.landurbplan.2013.09.014
Zhang Q, Xu D, Zhou D et al (2020) Associations between urban thermal environment and physical indicators based on meteorological data in Foshan City. Sustain Cities Soc 60:102288. https://doi.org/10.1016/j.scs.2020.102288
Zhang J, Xu S, Sun J et al (2022) Unsupervised Adversarial Domain Adaptation for Agricultural Land Extraction of Remote Sensing Images. Remote Sens 14:1–20. https://doi.org/10.3390/rs14246298
Zhao J, Zhao X, Liang S et al (2020a) Assessing the thermal contributions of urban land cover types. Landsc Urban Plan 204:103927. https://doi.org/10.1016/j.landurbplan.2020.103927
Zhao Y, Chew LW, Kubilay A, Carmeliet J (2020b) Isothermal and non-isothermal flow in street canyons: A review from theoretical, experimental and numerical perspectives. Build Environ 184:107163. https://doi.org/10.1016/j.buildenv.2020.107163
Zheng S, Guldmann JM, Liu Z, Zhao L (2018) Influence of trees on the outdoor thermal environment in subtropical areas: An experimental study in Guangzhou, China. Sustain Cities Soc 42:482–497. https://doi.org/10.1016/j.scs.2018.07.025
Zhu J-Y, Park T, Isola P, Efros AA (2017) Unpaired image-to-image translation using cycle-consistent adversarial networks. IEEE International conference on computer vision (ICCV) 2242–2251. https://doi.org/10.48550/arXiv.1703.10593
Funding
The support of the grant from the National Natural Science Foundation of China (NSFC) (No. 51908489) is acknowledged. Research is funded by “Pioneer” and “Leading Goose” R&D Program of Zhejiang (2023C03152), and "Zhejiang University Global Partnership Fund” (100000–11320/209).
Author information
Authors and Affiliations
Contributions
Xiaotian Ding: Conceptualization, Methodology, Software, Formal analysis, Writing-Original Draft. Yifan Fan: Conceptualization, Resources, Writing-Review & Editing, Supervision. Yuguo Li: Writing-Review & Editing. Jian Ge: Resources, Supervision.
Corresponding author
Ethics declarations
Ethical approval
This manuscript does not contain any individual person’s data in any form (including individual details, images or videos). There is no consent need for this manuscript.
Consent to participate
This manuscript does not contain any individual person’s data in any form. There is no consent need for this manuscript.
Consent for publication
All of the authors have read and approved the paper for publication.
Competing interests
The authors have no relevant financial or non-financial interests to disclose.
Additional information
Responsible Editor: Marcus Schulz
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A
See Table 5
Appendix B. The locations of meteorological and air quality monitoring stations
The locations of meteorological and air quality monitoring stations in Beijing, Shanghai, and GBA are demonstrated in Fig.
9, and
10. The meteorological monitoring stations are national surface meteorological monitoring stations and location information is provided by China National Meteorological Information Centre. The air quality monitoring network is built by The China National Environmental Monitoring Center (CNEMC).
Appendix C. the impact of the misalignment in Google dataset
The Google dataset is built with image from Google Earth and ground truth from GF2 dataset, as shown in Fig.
11(a1–a3). The misalignments exist between the Google Earth and GF2 images mainly because they were obtained at different acquisition times and off-nadir angles. Therefore, after being clipped into smaller tiles, tiles with severe misalignment (Fig. 11(b1–b3)) caused by the large land cover change were excluded. As a result, the target dataset now consists of 31 km2 of Google images with ground truth. Misalignments caused by different shading or off-nadir view angle are hardly evitable, as shown in the red frame in Fig. 11(c1–c3). Therefore, the classification accuracy evaluated on the target dataset will be underestimated and can only be used for reference.
With minor land cover change and camera angle difference the classification overall accuracy can reach a high level (Fig. 12 (a1–a4)). However, severely underestimated score will occur when there is a land cover change or camera angle difference, even if the model has correctly classified the images. As shown in Fig.
12(b1–b4) and Fig. 12(c1–c4), the underlying land cover has been well recognized, but the overall accuracy score is significantly low because of the misalignment between Google image and ground truth caused by land cover change and difference caused by off-nadir view angles. Although we have excluded image tiles with significant land cover changes, misalignment caused by different camera angles still exists in most test images. This is because buildings in the study areas were mostly high-rise buildings with more than seven floors. Therefore, the evaluation result on Google dataset is underestimated.
Appendix D. Detailed settings for the training of deep learning models
The whole process was coded in the PyTorch (v1.10) (Mazza and Pagani 2020) framework. To reduce the dependence on the computer memory size, each training sample was furtherly divided into 64 patches [256m × 256m for one patch and 8 × 8 patches (2048/256 = 8)]. 80% of the training dataset was used for training and the remaining 20% was reserved for performance evaluation. Data augmentation is a common strategy used in deep learning training (Wong et al. 2016; Diakogiannis et al. 2020), especially when the training dataset is not sufficient. In this study, training samples were augmented with the transform modules implemented in torchvison(Mazza and Pagani 2020) including RandomHorizontalFlip, RandomVerticalFlip, ColorJitter, RandomAdjustSharpness, and RandomAutocontrast. Following Li’s work (Li et al. 2019b), the initial learning rate for training the classification model is 2.5 × 10−4 and decreased with a ‘poly’ learning rate policy with power as 0.9. The batch size is set to 4 and the CycleGAN model was trained for 15 epochs. The training process was conducted on a computer platform with an Intel(R) Core i7 9700K CPU, 64 GB of RAM, and two Nvidia RTX 2080Ti graphics processing units (GPUs).
Appendix E. Correlation analysis result in 2020
Correlation analysis was conducted using meteorological parameters from 2020. Seasonal and annual average values were used for analysis. Additional analysis for air quality parameters hasn't been conducted, because the available air quality stations in Shanghai are all located in compact urban areas which are inappropriate for analysis. As listed in Table
6, the results indicate a small seasonal variation for most of the correlation coefficients except for two cases: first, the relative humidity is correlated with artificial surfaces in the summer of Beijing which is not shown in other seasons; Second, in the autumn of Beijing and GBA, the correlation coefficients is different from other seasons.
Appendix F. Correlation analysis result for air pollution
The PM2.5 and PM10 concentrations are positively correlated with artificial land cover fraction in Shanghai, whereas this relationship is insignificant in Beijing and GBA (Table
7). The strong background PMs pollution and regional transport of pollutants can weaken the significance of intra-urban differences (Wang et al. 2014; Tao et al. 2017; Liu et al. 2017; Fan et al. 2020). Studies have shown that the air pollution in Beijing and GBA is significantly affected by its surrounding area (Liu et al. 2013; Xue et al. 2016; Tao et al. 2017). The local industrial activities and vehicle emission are the major contributors to PMs in Shanghai (Wang et al. 2014), which may explain its stronger correlation than the other two areas.
There is a ubiquitous correlation (0.26 to 0.38) between NO2 concentrations and artificial land cover in all areas (Table 7). The traffic in urban areas with dense populations is usually heavier than that in rural areas as a major source for NO2 in urban areas (Lee et al. 2014; Fan et al. 2020; Xue et al. 2020). The NO2 concentrations of those sites vary significantly even with a similar artificial surface fraction (Fig.
13(c, e)). This can be explained by the fact that the traffic intensity can be significantly different even when the fraction of artificial land cover is similar. Furthermore, other local factors like ship emissions can be an important factor affecting the correlation result (Xue et al. 2020).
Traffic is also a major factor in total CO emissions (Hrebtov and Hanjalić 2019). The results show that the CO concentration also has a distinct variation even in areas with a similar artificial surface fraction (Fig.
14 a, d). However, the correlation between CO concentration and artificial land cover in GBA is weaker (− 0.02 to 0.23) than that of NO2. This is probably because the concentration of CO is also affected by agricultural and/or residential heating-related biomass burning in rural areas (Sokhi et al. 2021).
A negative correlation (− 0.42 to − 0.46) exists between the fraction of artificial surface and SO2 concentrations in the summer of Shanghai and Beijing. The major sources of SO2 are heavy industries and coal-fired power stations (Yoo et al. 2015; Xue et al. 2016, 2020; Li et al. 2017), and these heavy industries are normally located in suburban or rural areas. Therefore, the SO2 emissions in rural industrial land can be larger than in urban residential areas. For instance, with a heavier industrial intensity, the SO2 emission in the northern region of Shanghai is larger than in the denser urban central region (Xue et al. 2020). The local concentration of SO2 also depends on the wind direction. In the summer of Shanghai, the prevailing southeast monsoon from the sea brings cleaner air, causing a lower SO2 concentration in urban areas (Xue et al. 2020). However, in other seasons, the north wind from more polluted inland may lead to a similar SO2 concentration throughout the city, making it a regional pollution problem and reducing the intra-urban variability. Therefore, the correlation in Spring can be weak. The correlation between SO2 and artificial surfaces in GBA is weaker than in Shanghai and Beijing. Unlike Shanghai and Beijing, there still are heavy industries and electric power stations (Hu et al. 2021) in the center of GBA (Foshan city), which can weaken urban–rural differences.
There is no clear correlation shown for the O3 concentration, which may be because O3 has a more complex photochemical generation mechanism than other pollutants (Ren et al. 2021). The formation of O3 can be affected by factors like air temperature and reduced solar radiation in rainy seasons. Therefore, the O3 concentration shows a strong seasonal variation (Fan et al. 2020).
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Ding, X., Fan, Y., Li, Y. et al. Urban surface classification using semi-supervised domain adaptive deep learning models and its application in urban environment studies. Environ Sci Pollut Res 30, 123507–123526 (2023). https://doi.org/10.1007/s11356-023-30843-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11356-023-30843-8