Machine Learning in Land Use Change Detection

Version	Summary	Created by	Modification	Content Size	Created at	Operation
1		Bo Liu	--	2181	2023-05-24 03:52:38	\|
2	format correction	Dean Liu	-2 word(s)	2179	2023-05-24 04:46:10	\|

This entry is adapted from the peer-reviewed paper 10.3390/land12051050

Land use change detection (LUCD) is a critical technology with applications in various fields, including forest disturbance, cropland changes, and urban expansion.

bibliometric analysis LUCD machine learning

1. Introduction

Land use and land cover (LULC) are fundamental topics in global change research, serving as the basis for human survival and development ^[1]. Changes in land use at various scales have significant impacts on land cover, which is crucial for research on global climate change, ecological environment monitoring, and related fields ^[2]. Change detection involves identifying differences in the state of an object or phenomenon over time, providing an accurate reflection of changes that occur over an extended period ^[3]. The timely and precise detection of changes in global land use/cover is closely linked to various aspects of global change, facilitating a better understanding of the relationship between humans and natural phenomena and better resource management and utilization ^[4]^[5]^[6]^[7].

Remote sensing (RS) is the primary means of LUCD ^[8], allowing for the analysis of changes in land use characteristics and processes by applying datasets from different time periods and covering large areas. RS imagery offers several advantages, such as broad coverage, long coverage periods, and relatively convenient data processing, making it a major data source for detecting land use changes over recent decades ^[9]^[10]^[11]. Several remote sensing satellites have been launched, providing datasets with different time and spatial resolutions. The Landsat series, which was launched by the United States in 1972 and increased the satellite image resolution to 30 m in 1982 with Landsat−4, has become one of the main data sources for researchers to detect changes in land use ^[12]. Other remote sensing satellites, such as the moderate-resolution imaging spectroradiometer (MODIS) and advanced very high-resolution radiometer (AVHRR), have also been widely used due to their advantages of large coverage area and high temporal resolution. Time-series datasets extracted from these satellites, such as the normalized difference vegetation index (NDVI), are important data sources for detecting vegetation changes and extracting cultivated land information ^[13]. In recent years, Sentinel−2, Landsat−8, and synthetic aperture radar mounted on uncrewed aerial vehicles have been increasingly used thanks to their higher temporal and spatial resolutions and faster revisit periods, driving the rapid development of technology for LUCD ^[14]^[15]^[16].

The emergence of multi-platform, multi-sensor, and multi-source remote sensing images, along with the increasing complexity of land use/cover change types, has led to the development of various LUCD techniques, such as classification-based change detection ^[17], time-series change detection ^[18], model-based change detection ^[19]^[20], and deep learning-based change detection ^[21]^[22]. While monitoring changes in Earth’s surface features is crucial, different authors often reach varying conclusions due to the dependence of change detection techniques on the data source and the landscape and environment of the study area, as well as various complex factors ^[9]^[23]^[24]. Existing literature reviews primarily focus on the research on LUCD techniques without providing a comprehensive analysis of the theme, current situation, and future development trends of change detection ^[25].

Bibliometrics is a valuable tool for analyzing scientific research progress by quantifying information related to a specific research topic in online citation databases. This tool identifies authors, publication quantities, and research institutions involved in the field, helping to identify key literature, provide keywords, and quantify the current status and future trends of research topics ^[26]^[27]. HistCite ^[28], SATI ^[29], and CiteSpace ^[30] are some common bibliometric analysis tools. Bibliometrix, an open-source tool developed by Massimo Aria in 2017 ^[31], offers more literature information analysis and result visualization functions and can import and convert data from multiple database sources, such as Web of Science, Scopus, Dimensions, and Lens ^[32]. In geography-related fields, many scholars use Bibliometrix to perform quantitative analyses of literature. Xie et al. ^[33] employed data mining and quantitative analysis to investigate research papers related to land degradation from 1990 to 2019 in the Web of Science Core Collection database to reveal the current research status of global land degradation and to assess future research directions. Xu et al. ^[34] conducted a bibliometric analysis of land consolidation literature from the Web of Science Core Collection database from 2000 to 2020, aiming to identify the historical development, evolutionary trajectory, and future trends of this theme. However, the manual selection of research literature by expert knowledge remains a disadvantage of bibliometrics, as this process is time-consuming, labor-intensive, and may miss potentially related studies, leading to narrow searches ^[35]. In cases where the literature growth rate outpaces the available time for manual review, it may not be possible to conduct a comprehensive search of the field ^[36].

To streamline the process of screening potentially relevant literature in the field of LUCD, researchers employed a machine learning algorithm combined with bibliometric analysis. Specifically, the machine learning algorithm was used to select relevant literature, followed by a comprehensive analysis and evaluation of LUCD-related publications in the Web of Science Core Collection from 1985 to 2022 using the Bibliometrix R package.

2. Future Research Directions for LUCD

Regarding data sources, common remote sensing data used for LUCD include AVHRR, MODIS, Landsat, and Sentinel−2. Over time, remote sensing imagery has become increasingly higher in spatial resolution, resulting in clearer land cover conditions. However, the improved spatial resolution also highlights the shadows of land cover features and exacerbates the problem of cloud cover, leading to the loss of important land cover information and adversely affecting LUCD ^[37]^[38]. Furthermore, for large-scale applications, the high spatial resolution of satellite data results in limited coverage and high computational demands, posing challenges to data analysis. Google Earth Engine (GEE) is a platform that can collect and process large amounts of already published data products, and its advanced computing capabilities surpass traditional computers and servers ^[39]. Researchers can utilize this platform to greatly expand the temporal and spatial scales of their research, significantly improving the accuracy and efficiency of LUCD.

In the domain of change detection, precise geometric registration and atmospheric correction or normalization between multi-temporal images are crucial factors that determine the success of the project ^[23]. The use of multiple sensors in the pre-processing stage requires the resolution of the issues arising from differences in spectral characteristics and geometric registration errors, especially in cases where images need to be joined or cropped. As generations of remote sensing satellites continue to upgrade, algorithms have been developed to make acquired images have similar brightness or gradually increasing brightness, allowing for a more natural joining of two images. Additionally, establishing a uniform geographic reference system can reduce registration errors between multi-source remote sensing images.

The accuracy assessment of LUCD results is crucial for decision-making. Commonly used methods for accuracy assessment include overall accuracy, Kappa coefficient, recall rate, and IOU index. However, the accuracy of the Kappa coefficient has been questioned by some scholars ^[40]. Additionally, the current accuracy evaluation is mainly based on the pixel-based approach, and the accuracy assessment methods for object-oriented and feature-based LUCD require further research.

In terms of methods, there are numerous types of available change detection methods for land use, each of which is suitable for different scenarios. Despite the development of various change detection technologies, it remains difficult to select the appropriate method for accurate change detection, particularly for specific research purposes or fields. With the popularity of high-resolution and sub-meter resolution remote sensing images, the consideration of different land cover types and change characteristics increases, requiring more effective detection methods to address these issues. With the advent of object-based analysis methods and models, most scholars have applied these methods to LUCD, achieving high accuracy. The most critical step in object-oriented change detection is image segmentation, and numerous scholars have proposed various segmentation methods, each with specific applicable scenarios and conditions. In future research on object-oriented LUCD, image segmentation techniques still require further in-depth research. Furthermore, threshold selection must be considered in the change detection method, and more convenient methods should be developed to quickly select the appropriate threshold and further improve the accuracy of LUCD.

There has been an increasing amount of research focused on using deep learning methods, such as convolutional neural networks, recurrent neural networks, and transformers, to improve the accuracy of land use classification. However, the automated mining of data and the acquisition of spatiotemporal features still present some difficulty. For training change detection methods, particularly in parameter selection, a significant amount of time and effort is required. Thus, future development should focus on creating lightweight modules that can enhance the performance of network models and reduce the time required for model training. As multi-source remote sensing data, deep learning, big data, and artificial intelligence continue to develop and be applied, researchers should provide a flexible framework that can integrate critical issues related to data pre-processing, feature extraction, land use type interpretation, and accuracy evaluation, in order to construct an intelligent change detection system.

The following directions are suggested for future research in the field of LUCD, based on the analysis of past issues and existing research methods:

(1): Expanding the range of image data acquisition can be achieved by combining multiple data sources.
(2): Cloud platforms should be utilized to conduct more precise, long-term, and large-scale land use change detection studies.
(3): Further research is needed on the geometric registration and spectral differences of multi-source remote sensing images during the preprocessing stage.
(4): Accuracy evaluation should be improved, and object-oriented and feature-based accuracy evaluation methods should be developed.
(5): Future research should focus on studying optimal, adaptive, and full-scale image segmentation and threshold selection techniques.
(6): Based on deep learning, LUCD methods have demonstrated great potential in recent years through their multi-level and deep network structures.

3. Advantages and Uncertainties

It is demonstrated the effectiveness of utilizing machine learning models and software tools such as ASReview and Bibliometrix for screening articles in the field of LUCD. This approach can significantly reduce the time and labor costs involved in screening large databases of articles, making it an attractive option for researchers. Furthermore, machine learning models can be continuously improved by incorporating feedback from experts and adding relevant and non-relevant articles to the training data. Although labeling training data is a time-consuming process, ASReview significantly reduces this burden by actively identifying relevant articles and excluding irrelevant ones during the training process. Moreover, ASReview is an open-source tool that does not require any programming knowledge, making it accessible to a broad range of researchers. The use of machine learning models and software tools such as ASReview and Bibliometrix can potentially enhance the efficiency and accuracy of literature reviews in various research fields ^[41].

First, the bibliometric analysis results are highly dependent on the selected databases, which may not include all relevant publications in the field. In addition, researchers only focused on a specific topic using the Web of Science Core Collection database. Although it is one of the most influential databases, incorporating other databases could provide a more comprehensive global perspective. Lastly, further research is needed to integrate bibliometrics with machine learning methods to improve the efficiency of systematic reviews and reduce labor costs while uncovering more interesting results. Future studies should explore more advanced machine learning techniques and data sources to further enhance the effectiveness of this approach.

A noteworthy concern arises regarding the bibliometric analysis process employed to obtain the development trends of LUCD by analyzing the article quantity and citation rate of journals. The chosen approach may be subject to influence from the publishing method of the journals, such as gold open access or subscription, and limitations on the number of articles published annually by certain journals. The comparison of journals and articles with varying publishing methods poses a challenging task and necessitates careful consideration of the underlying assumptions supporting these methods. A fundamental assumption in this context is that subscription-based journals offer superior quality control and peer review mechanisms, as publishers have a financial motivation to uphold high standards. Conversely, open access journals may be perceived as having lower barriers to entry, which can lead to lower standards of peer review and editorial oversight. However, this premise is not always accurate, as some open access journals have rigorous peer review procedures and maintain a high degree of quality control. Another presumption is that open access journals are more accessible and inclusive than subscription-based journals, as they do not require payment for article access. Nonetheless, this assertion does not fully address the reality that open access journals may levy article processing fees, which could limit access to articles for authors from low-income countries or institutions.

Furthermore, an underlying assumption posits that the free and accessible nature of open access articles may engender increased citation rates and a wider dissemination of research findings. However, limitations on the number of articles published annually by certain journals may affect the citation rates of individual articles. Should a journal be able to publish only a finite number of articles each year, authors may find it increasingly arduous to secure publication within said journal, potentially impeding the attention and citation their research receives. Nevertheless, this does not always hold true, as some journals with limited publishing capacity are highly discerning and esteemed, thereby augmenting the impact and visibility of the research they publish. Additionally, the factors influencing citation rates and impact are intricate and multivariate, variably influenced by factors such as research quality, author and journal reputation, employed methodology and techniques, as well as novelty and significance of the findings.

References

Song, X.-P.; Hansen, M.C.; Stehman, S.V.; Potapov, P.V.; Tyukavina, A.; Vermote, E.F.; Townshend, J.R. Global land change from 1982 to 2016. Nature 2018, 560, 639–643.
Coppin, P.; Jonckheere, I.; Nackaerts, K.; Muys, B.; Lambin, E. Review ArticleDigital change detection methods in ecosystem monitoring: A review. Int. J. Remote Sens. 2004, 25, 1565–1596.
Singh, A. Review article digital change detection techniques using remotely-sensed data. Int. J. Remote Sens. 1989, 10, 989–1003.
Zhang, M.; Kafy, A.-A.; Ren, B.; Zhang, Y.; Tan, S.; Li, J. Application of the optimal parameter geographic detector model in the identification of influencing factors of ecological quality in Guangzhou, China. Land 2022, 11, 1303.
Saha, M.; Kafy, A.-A.; Bakshi, A.; Almulhim, A.I.; Rahaman, Z.A.; Al Rakib, A.; Fattah, M.A.; Akter, K.S.; Rahman, M.T.; Zhang, M. Modelling microscale impacts assessment of urban expansion on seasonal surface urban heat island intensity using neural network algorithms. Energy Build. 2022, 275, 112452.
Yang, K.; Hou, H.; Li, Y.; Chen, Y.; Wang, L.; Wang, P.; Hu, T. Future urban waterlogging simulation based on LULC forecast model: A case study in Haining City, China. Sustain. Cities Soc. 2022, 87, 104167.
Siddik, M.S.; Tulip, S.S.; Rahman, A.; Islam, M.N.; Haghighi, A.T.; Mustafa, S.M.T. The impact of land use and land cover change on groundwater recharge in northwestern Bangladesh. J. Environ. Manag. 2022, 315, 115130.
Sobhani, P.; Esmaeilzadeh, H.; Mostafavi, H. Simulation and impact assessment of future land use and land cover changes in two protected areas in Tehran, Iran. Sustain. Cities Soc. 2021, 75, 103296.
Hussain, M.; Chen, D.; Cheng, A.; Wei, H.; Stanley, D. Change detection from remotely sensed images: From pixel-based to object-based approaches. ISPRS J. Photogramm. Remote Sens. 2013, 80, 91–106.
Chen, G.; Hay, G.J.; Carvalho, L.M.; Wulder, M.A. Object-based change detection. Int. J. Remote Sens. 2012, 33, 4434–4457.
Singh, P.; Kikon, N.; Verma, P. Impact of land use change and urbanization on urban heat island in Lucknow city, Central India. A remote sensing based estimate. Sustain. Cities Soc. 2017, 32, 100–114.
Zhu, Z.; Wulder, M.A.; Roy, D.P.; Woodcock, C.E.; Hansen, M.C.; Radeloff, V.C.; Healey, S.P.; Schaaf, C.; Hostert, P.; Strobl, P. Benefits of the free and open Landsat data policy. Remote Sens. Environ. 2019, 224, 382–385.
Verbesselt, J.; Hyndman, R.; Newnham, G.; Culvenor, D. Detecting trend and seasonal changes in satellite image time series. Remote Sens. Environ. 2010, 114, 106–115.
Yu, Y.; Guan, H.; Li, D.; Gu, T.; Wang, L.; Ma, L.; Li, J. A hybrid capsule network for land cover classification using multispectral LiDAR data. IEEE Geosci. Remote Sens. Lett. 2019, 17, 1263–1267.
Kamps, M.T.; Bouten, W.; Seijmonsbergen, A.C. LiDAR and orthophoto synergy to optimize object-based landscape change: Analysis of an active landslide. Remote Sens. 2017, 9, 805.
Roy, D.P.; Wulder, M.A.; Loveland, T.R.; Woodcock, C.E.; Allen, R.G.; Anderson, M.C.; Helder, D.; Irons, J.R.; Johnson, D.M.; Kennedy, R. Landsat-8: Science and product vision for terrestrial global change research. Remote Sens. Environ. 2014, 145, 154–172.
Chen, X.; Chen, J.; Shi, Y.; Yamaguchi, Y. An automated approach for updating land cover maps based on integrated change detection and classification methods. ISPRS J. Photogramm. Remote Sens. 2012, 71, 86–95.
Lin, Y.; Zhang, L.; Wang, N. A new time series change detection method for landsat land use and land cover change. In Proceedings of the 2019 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images (MultiTemp), Shanghai, China, 5–7 August 2019; pp. 1–4.
Zhu, Z.; Woodcock, C.E. Continuous change detection and classification of land cover using all available Landsat data. Remote Sens. Environ. 2014, 144, 152–171.
Kennedy, R.E.; Yang, Z.; Cohen, W.B. Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr—Temporal segmentation algorithms. Remote Sens. Environ. 2010, 114, 2897–2910.
Feizizadeh, B.; Mohammadzade Alajujeh, K.; Lakes, T.; Blaschke, T.; Omarzadeh, D. A comparison of the integrated fuzzy object-based deep learning approach and three machine learning techniques for land use/cover change monitoring and environmental impacts assessment. GISci. Remote Sens. 2021, 58, 1543–1570.
Moncrieff, G.R. Continuous land cover change detection in a critically endangered shrubland ecosystem using neural networks. Remote Sens. 2022, 14, 2766.
Lu, D.; Mausel, P.; Brondizio, E.; Moran, E. Change detection techniques. Int. J. Remote Sens. 2004, 25, 2365–2401.
Chughtai, A.H.; Abbasi, H.; Ismail, R.K. A review on change detection method and accuracy assessment for land use land cover. Remote Sensing Applications Society and Environment 2021, 22, 100482.
Zhang, M.; Kafy, A.-A.; Xiao, P.; Han, S.; Zou, S.; Saha, M.; Zhang, C.; Tan, S. Impact of urban expansion on land surface temperature and carbon emissions using machine learning algorithms in Wuhan, China. Urban Clim. 2023, 47, 101347.
Rousseau, D.M. The Oxford Handbook of Evidence-Based Management; Oxford University Press: Oxford, UK, 2012.
Crane, D. Invisible Colleges: Diffusion of Knowledge in Scientific Communities; University Of Chicago Press: Chicago, IL, USA, 1972.
Garfield, E. From the science of science to Scientometrics visualizing the history of science with HistCite software. J. Informetr. 2009, 3, 173–179.
Zhuang, J.; Nie, Q.; Man, W.; You, C. Visualization applications of the bibliometric software SATI and Histcite. In Proceedings of the Proceedings of the 2022 2nd International Conference on Business Administration and Data Science (BADS 2022), Xinjiang, China, 28–30 October 2022.
Wang, W.; Lu, C. Visualization analysis of big data research based on Citespace. Soft Comput. 2020, 24, 8173–8186.
Aria, M.; Cuccurullo, C. bibliometrix: An R-tool for comprehensive science mapping analysis. J. Informetr. 2017, 11, 959–975.
Gagolewski, M. Bibliometric impact assessment with R and the CITAN package. J. Informetr. 2011, 5, 678–692.
Xie, H.; Zhang, Y.; Wu, Z.; Lv, T. A bibliometric analysis on land degradation: Current status, development, and future directions. Land 2020, 9, 28.
Xu, X.; Chen, Q.; Zhu, Z. Evolutionary overview of land consolidation based on bibliometric analysis in web of science from 2000 to 2020. Int. J. Environ. Res. Public Health 2022, 19, 3218.
Van de Schoot, R.; de Bruin, J.; Schram, R.; Zahedi, P.; de Boer, J.; Weijdema, F.; Kramer, B.; Huijts, M.; Hoogerwerf, M.; Ferdinands, G. An open source machine learning framework for efficient and transparent systematic reviews. Nat. Mach. Intell. 2021, 3, 125–133.
Van den Bulk, L.M.; Bouzembrak, Y.; Gavai, A.; Liu, N.; van den Heuvel, L.J.; Marvin, H.J. Automatic classification of literature in systematic reviews on food safety using machine learning. Curr. Res. Food Sci. 2022, 5, 84–95.
Jing, C.; Zhou, W.; Qian, Y.; Yu, W.; Zheng, Z. A novel approach for quantifying high-frequency urban land cover changes at the block level with scarce clear-sky Landsat observations. Remote Sens. Environ. 2021, 255, 112293.
Zhu, Z.; Woodcock, C.E. Automated cloud, cloud shadow, and snow detection in multitemporal Landsat data: An algorithm designed specifically for monitoring land cover change. Remote Sens. Environ. 2014, 152, 217–234.
Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sens. Environ. 2017, 202, 18–27.
Pontius, R.G., Jr.; Millones, M. Death to Kappa: Birth of quantity disagreement and allocation disagreement for accuracy assessment. Int. J. Remote Sens. 2011, 32, 4407–4429.
Nosek, B.A.; Alter, G.; Banks, G.C.; Borsboom, D.; Bowman, S.D.; Breckler, S.J.; Buck, S.; Chambers, C.D.; Chin, G.; Christensen, G. Promoting an open research culture. Science 2015, 348, 1422–1425.

© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license; additional terms may apply. By using this site, you agree to the Terms and Conditions and Privacy Policy.

Upload a video for this entry

Information

Subjects: Remote Sensing

Contributors MDPI registered users' name will be linked to their SciProfiles pages. To register with us, please refer to https://encyclopedia.pub/register :

Bo Liu

Wei Song

Zhan Meng

Xinwei Liu

View Times: 301

Update Date: 24 May 2023

Table of Contents

Video Upload Options

Confirm

1. Introduction

2. Future Research Directions for LUCD

3. Advantages and Uncertainties

References