Bibliometric Insights into Terracing Effects on Water Resources Under Climate Change: Advances in Remote Sensing and GIS Applications

Abstract

With the increasing impacts of global climate change and the continuous expansion of the population, the scarcity of food and water resources, along with the protection of agricultural land, have become significant constraints to sustainable agricultural development. Terraces plays a vital role in controlling water loss and promoting sustainable agriculture, and they have been widely adopted across the globe. Using CiteSpace, this study conducted a bibliometric review of the literature on the application of remote sensing and GISs in terrace studies under global climate change. The dataset included publications from the Web of Science spanning the years 1992 to 2024. Based on a systematical analysis of 508 publications, we investigated major institutions, cross-author collaborations, keyword co-occurrences, and the evolution of the research focus areas regarding the applications of remote sensing and GISs in terrace studies. The results show that the prominent research themes in this domain include remote sensing, erosion, and climate change. China (132, 26%) and the United States (108, 21%) are the top contributors in terms of publication numbers, while European countries and institutions are more active in collaborative efforts. The research emphasis has transitioned from analyzing the environmental characteristics of terraces to a broader consideration of ecological factors and multi-scenario applications. Moreover, analyses of the keyword co-occurrence and temporal trends indicate a rising interest in the application of machine learning, deep learning, and luminescence dating in terrace studies. Moving forward, it is essential to advance the deployment of automated monitoring systems, obtain long-term continuous monitoring data, encourage the adoption of conservation agriculture technology, and strengthen early warning networks for extreme climate events in terrace research. Overall, this study underscores the importance of interdisciplinary approaches and collaborative efforts to address the myriad challenges faced by terraced agriculture in an era of rapid environmental change.

1. Introduction

Terraces, designed as staircase-like agricultural fields along contour lines, play a vital role in improving crop yields and controlling soil and water losses. They have earned the reputation of being “living fossils” of human ingenuity within natural environments [1]. These terraces, constructed at various times throughout history, can be found across the globe, particularly in Asia and Europe. Terraces are typically categorized into two main types: paddy terraces, which are primarily used for rice cultivation in regions with ample rainfall or effective irrigation, such as the Anshun terraces in Guizhou, the Yuanmou terraces in Yunnan, and the Ziquejie terraces in Hunan, China; and dryland terraces, which are suited for crops like wheat and maize, often located in areas with lower precipitation levels [2]. Additionally, terraces can be classified based on their construction methods into level terraces, reverse-slope terraces, and slope-following terraces. Level terraces, which are built along contour lines, effectively minimize soil erosion; reverse-slope terraces, inclined inward to gather water, are particularly suitable for arid areas; while slope-following terraces, aligned with the slope direction, are more appropriate for gentler terrains yet are more susceptible to soil erosion [2].

Historically, terraces have been recognized as a remarkable human adaptation to the challenges posed by mountainous and hilly areas. From a soil and water conservation perspective, terraces convert steep slopes into graduated platforms, significantly reducing surface runoff and preventing soil erosion, which helps maintain soil fertility in mountainous areas [3]. The design of flat cultivation layers ensures more efficient water absorption by the soil, thereby minimizing water loss, a feature that is particularly advantageous in regions with irregular rainfall and fragmented terrain [3]. Moreover, terraces enhance food production by systematically increasing the arable land area and improving the land use efficiency. Effective irrigation management in these systems not only boosts crop yields but also accommodates multiple cropping patterns, which contribute to agricultural biodiversity and ecological stability [4]. Furthermore, terraced systems facilitate the successful cultivation of a diverse range of crops, including rice, maize, and legumes, thereby significantly bolstering food security. In terms of their ecological impact, terraces play a vital role in maintaining the regional ecological balance. They create new habitats that support biodiversity and enhance carbon sequestration through vegetation cover, contributing to climate mitigation [5]. The numerous environmental benefits provided by terrace systems underscore their ongoing significance within the framework of modern sustainable development.

Globally, terraces are prevalent, with extensive distributions found in Asia, Africa, and Latin America [6]. China, recognized as one of the birthplaces of terrace technology, features renowned terrace landscapes, such as the Longji Rice Terraces in Guangxi and the Yuanyang Rice Terraces in Yunnan [2]. These spectacular formations not only bolster local food production but have also become important cultural heritage sites and tourist attractions. Other Asian countries, such as the Philippines with its Banaue Rice Terraces, exemplify the harmonious relationship between human activity and nature [7]. In Africa, particularly in the mountainous areas of East and North Africa, terraces are primarily utilized for soil erosion control and agricultural production enhancement. In countries like Ethiopia and Kenya, terrace systems help smallholder farmers stabilize their yields and sustain their livelihoods in the face of climate change, playing a crucial role in food security [8]. Furthermore, in relatively barren and arid regions, terrace systems are effective for rainwater harvesting, significantly enhancing the water use efficiency [9]. Latin America also exemplifies the significance of terrace agriculture, particularly in the Andes of Peru, where terrace systems have transformed rugged mountainous areas into sustainable agricultural regions [10]. These terraces are not only impressive feats of engineering but also serve as vital cultural relics and tourist attractions, reflecting their historical, cultural, and agricultural importance alongside their breathtaking scenery [10]. The adaptations of terracing in various regions illustrate the ingenuity of human beings in responding to unique geographical and climatic conditions, showcasing a remarkable flexibility in agricultural practices across diverse natural environments. Globally, terrace agriculture symbolizes technological advancements in farming while highlighting humanity’s innovative capacity to harmonize with nature amidst ecological and environmental challenges [6].

The rapid and accurate acquisition of information about terraces—such as the area, type, and spatial distribution—is crucial for a comprehensive assessment of their ecological functions. Historically, the statistics regarding terraces focused primarily on area measurement, heavily depending on manual operations. This traditional approach is not only time-consuming and labor-intensive but also suffers from inefficiencies. However, the advent of remote sensing satellite imagery has enabled attempts to extract terrace information using advanced technologies [11]. Unfortunately, the extraction results have often been subpar, primarily due to typical terrace widths being less than 20 m, compounded by limitations in the image quality and processing techniques [12]. The increasing availability of high-resolution remote sensing satellite imagery, coupled with the widespread application of GISs, offers the potential to acquire comprehensive data on terraces, including their location, area, width, embankment height, and slope [11]. These technologies represent a significant advancement in terrace surveying and dynamic monitoring capabilities. In contrast to traditional field survey methods, remote sensing technology allows for the rapid coverage of extensive areas while substantially reducing the workload and costs associated with field investigations [12]. In the early stages of remote sensing development, the limited spatial resolution posed challenges in extracting terrace information, often relegating related research to manual visual interpretations. However, the deployment of high-resolution remote sensing satellites and significant advancements in image processing technologies have led to breakthroughs in extracting terrace information. Furthermore, with the swift development of machine learning technologies and the expanding application of model algorithms, deep learning-based semantic segmentation methods are increasingly being employed in image segmentation, text extraction, and object recognition [10,11,12]. Currently, numerous researchers are utilizing deep learning to classify and recognize terrace backgrounds in imagery, enabling the automatic learning of relevant features for precise segmentation between terrace and non-terrace regions.

The terrace identification accuracy and precision are directly dependent on the resolution of remote sensing images. High-resolution imagery, obtained from sources such as QuickBird and WorldView, provides detailed information that is essential for accurately identifying and extracting terrace morphological features. However, these high-resolution images typically cover smaller areas and are cost-prohibitive, making them impractical for large-scale monitoring and long-term studies [13]. In contrast, medium- to low-resolution imagery from sources like Landsat and MODIS offers broader coverage and longer time-series data but often lacks the spatial detail required to precisely resolve terrace structures and detect changes. This deficiency can lead to identification errors or omissions in terrace mapping [14]. Therefore, selecting the appropriate resolution of remote sensing imagery for terrace recognition is a critical consideration in research design, necessitating a careful balance between the spatial and temporal scales.

GISs play a significant role in terrace research, although their adaptability and limitations become particularly pronounced in complex terrains [11]. GIS technology excels at managing vast spatial datasets, conducting terrain analysis, and identifying spatial patterns. However, the intricate topographical features of terraces require high data accuracy and complex modeling. In regions characterized by rugged terrain with diverse slope and aspect variations, conventional GIS analysis tools can struggle, particularly when high-precision digital elevation models (DEMs) are required. The existing accuracy of DEMs may not suffice to accurately represent the microtopographical features of terraces [15]. Additionally, data collection in such complex terrains often comes with high costs and logistical challenges, prompting the need for innovative techniques and methodologies to address these limitations.

Terrace research reveals distinct focal areas and methodological differences across various regions, closely tied to the social, economic, and environmental contexts of these areas. In Asia, particularly in countries like China, the research primarily centers on soil and water conservation alongside the ecosystem services provided by terraces. Scholars in this region utilize a broad array of remote sensing and geospatial analysis tools to assess both the environmental benefits and cultural heritage value of terraces [16]. Conversely, in Africa, the terrace research often emphasizes food security and water resource management, typically employing a combination of ground surveys and satellite observations to evaluate the terrace productivity and adaptability to climate change [17]. In Latin America, the research often focuses on the heritage value and sustainable development of terraces, employing interdisciplinary methodologies that integrate social science surveys with natural science analyses [18].

The integration of interdisciplinary perspectives has significantly advanced terrace research [1]. An ecological perspective frames terraces within the broader context of ecosystems, analyzing their impacts on biodiversity, water cycles, and carbon budgets [19,20]. Agricultural sciences examine the stability of terrace agricultural yields, crop management, and soil quality, thereby aiding in the optimization of land use and the enhancement of the production efficiency [8,21]. From a geographical standpoint, researchers investigate the spatial distribution, topographical characteristics, and land use relationships of terraces. Geographers leverage remote sensing and GIS technologies to thoroughly analyze the impacts of terraces on land use changes and regional development [2,22]. Overall, interdisciplinary collaboration enriches both the depth and breadth of terrace research, providing a comprehensive theoretical foundation and practical applications for sustainable management and conservation strategies for terraces.

Despite these advancements, significant challenges remain in terrace research. A primary issue is the insufficiency of data and the limitations of contemporary technologies. Terraces are often located in remote mountainous and hilly areas, complicating field survey efforts and resulting in a scarcity of high-resolution, long-term data. Additionally, the existing remote sensing datasets are frequently constrained by their spatial, temporal, and spectral resolutions, hindering a comprehensive representation of terraced landscapes [23].

On a technical level, while remote sensing and GISs provide powerful tools for monitoring and analyzing terraces, many regions and researchers struggle to harness these technologies effectively due to the high costs, limited access to sophisticated equipment, and a shortage of technical expertise. Moreover, the current data processing tools and algorithms often fail to deliver the necessary accuracy and efficiency when addressing the complexities of terrace topography, thereby hindering data interpretation and restricting the potential applications of various models [24].

The utilization of remote sensing and GISs in the context of terrace research, alongside their implications for agriculture and environmental management, have been comprehensively reviewed by a number of scholars. For instance, Deng et al. (2021) provided an extensive analysis of the deployment of remote sensing and GIS technologies for evaluating terrestrial ecosystem services. Their research underscores the significant possibility of these technologies in evaluating both agricultural and ecological impacts, particularly within the unique contexts of terraced landscapes, and offers insights into future research trajectories [4]. Similarly, AbdelRahman (2023) investigated the role of remote sensing and GISs in promoting sustainable agricultural development, with a focus on terraced regions. The review indicates that these technological tools are invaluable for monitoring land use, managing water resources, and enhancing environmental conservation efforts, ultimately contributing to the sustainability of terraced agriculture [25]. Collectively, these findings underscore the efficacy of remote sensing and GISs in evaluating the interconnections between environmental and agricultural variables, thereby offering critical insights for effective land management and informed policymaking in terraced ecosystems. Nonetheless, traditional review articles often encounter limitations pertaining to their breadth of scope, and the presence of subjective biases may compromise the impartiality and accuracy of evaluations within this field. Such constraints can impede the formulation of a comprehensive perspective on developments in terrace research.

Despite significant advancements in remote sensing and GIS applications for terrace studies, critical research gaps remain in addressing the unique challenges of terraced landscapes, specifically (1) a lack of systematic evaluation: no comprehensive, quantitative analysis exists to assess the effectiveness of different remote sensing and GIS approaches across diverse terrace types (e.g., bench vs. contour terraces); (2) the absence of standardized protocols: the current methodologies lack guidelines for selecting appropriate technologies based on topographic and climatic conditions; and (3) the limited integration of emerging technologies: the synergy between traditional remote sensing and GIS methods and advanced tools (e.g., Internet of Things [IoT] and artificial intelligence [AI]) for terrace management remains underexplored.

Bibliometric analysis has emerged as a powerful tool to address these gaps, overcoming the limitations of traditional literature reviews, which often suffer from incomplete coverage and subjective biases. In contrast to conventional reviews, bibliometric analysis employs quantitative techniques to systematically catalog and synthesize dispersed research findings. This approach enables researchers to discern trends, identify significant challenges, and recommend future directions in a specific domain. Through a radical examination of vast literature, bibliometric analysis reveals key research nodes and networks, enabling academic dialogue, enhancing resource distribution, and encouraging innovative developments. The findings gained through this process greatly enrich the continuity and depth of scientific research and our relationship to it. Researchers worldwide have utilized CiteSpace to investigate a diverse range of topics, from soil erosion modeling [26] to GISs [27]. While these studies have illuminated conservation practices in multiple aspects, they have often neglected the intricate interactions between remote sensing, GISs, and terrace research—an area that this study seeks to explore in depth.

We applied CiteSpace (version 6.2.R4) in this study, a professional software that facilitates the bibliometric analysis and visualization of scientific literature [28]. By leveraging CiteSpace to explore the bibliometric landscape, this research aims to reveal significant trends and identify areas where knowledge is lacking on the use of remote sensing and GISs in terraced farmland studies. Citation network visualization will illuminate the evolution of this research domain, showcasing key contributions, patterns of collaboration, and emerging research trends. Such findings play an essential role in steering future research endeavors and developing effective strategies for sustainable soil management, thereby fostering resilience in agricultural systems. This comprehensive bibliometric analysis will provide a deeper insight into the significant application of remote sensing and GISs in terraced farmland research. By systematically identifying key research elements and visualizing the interconnectedness of the scholarship in this domain, we aim to facilitate informed decision making and to encourage collaborative efforts that are necessary for addressing global agricultural challenges.

2. Materials and Methods

2.1. Data Collection

We performed a thorough search of the Web of Science (WOS) database for publications from 1992 to 2024, specifically addressing the application of remote sensing and GISs in terraced farmland research. After several rounds of meticulous testing, we established the following search strategy: Topic = (“Terrace” OR “Terraced fields” OR “Terraced landscapes” OR “Agricultural terrace”) AND (“remote sens*” OR “geographic information system*” OR “GIS”), yielding an initial collection of 538 records. The wildcard * allows for the inclusion of any word that starts with the preceding letter in the search. To ensure the relevance and quality of our results, we selected only those articles indexed in the Science Citation Index (SCI-E), excluding 3 non-English publications. In addition, we eliminated 18 non-research materials, including dissertation theses, conference papers, editorials, news articles, abstracts, data papers, review articles, and other non-research materials, as review articles lack original research contributions, and conference proceedings have minimal influence and acknowledgment within the field. We manually screened out 5 records as irrelevant to our research scope. This process resulted in 512 articles, with the final search and filtering concluded on 3 November 2024. Because of constraints on bulk downloads from the database (limited to 500 records per batch), we arranged the articles into two plain text files named “download_1” and “download_2”, corresponding to the order of downloads. After removing 4 duplicate records through CiteSpace’s deduplication function, a total of 508 publications were ultimately included for bibliometric analysis.

2.2. Data Analysis

We employed CiteSpace (version 6.2.R4) to analyze and visually depict the institutional collaboration, author collaboration, and keyword usage across all 508 downloaded papers. In the network maps generated, the interconnected lines illustrate the relationships between nodes, while the colors assigned to the nodes and lines correspond to different years. Due to the substantial volume of data and the wide range of publication years, we opted for a time slice of three years after several iterations, establishing a top 20 data threshold. This configuration excludes peripheral nodes, retains seminal works (≥10 citations), effectively identifies relevant literature, reflects research trends, and optimizes data processing within the software’s analytical framework. CiteSpace detects and examines the top 20 most cited or frequent items within each three-year time slice.

In building the network maps for institutional and author collaboration, as well as keyword co-occurrence, we applied the shortest path network method, a network pruning technique. This method computes the shortest paths between all network nodes, evaluates path similarities, and assigns weights to links to indicate their significance. This technique enhances the clarity and readability of the network map by eliminating less critical links while maintaining the essential ones. Based on the co-occurrence map, the keyword time zone map divided the years from 1992 to 2024 into 11 intervals according to the defined time slices, with each node assigned to the interval of its first occurrence.

The modularity value (Q) and silhouette value (S) are both critical measures used to evaluate modularity and clustering coherence in network analysis. These metrics effectively quantify the strength of the clustering structures within a network. Thus, we utilized CiteSpace to create keyword clustering maps using the Q and S values, which allowed us to assess the consistency of these clustering maps. The Q value measures the strength of the clustering structures by evaluating the extent to which the links between clusters stray from random links, generally ranging from 0 to 1. Values approaching 1 suggest a more pronounced clustering structure, with a Q value above 0.3 signifying significant clustering. Meanwhile, the S value assesses the density within clusters compared to the density between clusters, providing insights into the cluster homogeneity. This value ranges from −1 to 1, where values nearer to 1 indicate greater internal consistency within clusters, and those close to −1 suggest substantial inconsistency. An S value over 0.5 suggests a reasonable division of clusters. The Q and S values in our network maps all exceeded the established thresholds and approached the theoretical optimum of 1, confirming the analysis’ robustness and validity.

In addition, this study employed CiteSpace’s burst detection feature to generate a keyword burst map. After conducting multiple trials considering the broad temporal scope of the literature, we established a burst year length threshold of three years to identify both the short-term research focus and longer-term trends, thereby forming a robust basis for hotspot analysis. Furthermore, we collected statistical data on inter-referenced documents using the Web of Science (WOS) to gather relevant information regarding collaborations among countries, institutions, and authors. Subsequently, we employed CiteSpace (version 6.2.R4) and R (version 4.4.1) software to create visual representations of the collaborative connections among these entities.

3. Results

3.1. Publication and Citation Trends

The statistical summary of the annual publication volume and average citations per paper concerning the application of RS and GISs in the study of terraced farmland from 1992 to 2024 revealed a total of 508 published papers during this period. The results indicated a generally slow and fluctuating increase in the annual publication volume (Figure 1). Until 2007, the number of publications remained below ten per year, indicating that the research in this field was in its nascent stages. The consistently low output (<10 papers/year) mirrors broader limitations in geospatial technologies during this period, consistent with previous findings on early-stage adoption barriers in agricultural remote sensing [1,4]. After 2005, there was a marked upward trend in publications, particularly from 2010 onward, when the number of papers began to rise significantly. By 2016, the publication volume exceeded 20 for the first time, reaching 29 papers, and subsequently stabilized at a higher level (greater than 30 papers annually). The peak publication year was 2020, with a total of 58 papers. From 2016 to 2024, the cumulative publication volume (333 papers) accounted for approximately 65.55% of the total publications, highlighting the increased attention toward the application of remote sensing and GISs in terraced farmland research over the past nine years. The citation count is a crucial measure of a paper’s impact. The average citations per paper for the literature on the application of remote sensing and GISs in terrace research is approximately 23.83 citations per paper. The average citations fluctuated significantly between 1992 and 2015, peaking in 1998 and 2003 with 142 and 103.25 citations per paper, respectively. Despite these peaks, the overall trend in the average citations per paper has been downward, with a more pronounced decline after 2015. The high publication volume in the last nine years, coupled with the relatively low average citations per paper, may be attributed to the growing interest in this field, leading to a rapid influx of new articles, while citation counts require time to accumulate.

3.2. Network Analysis of Cooperation

3.2.1. Analysis of Country Cooperation Network

Table 1. Top 20 countries with publications on the application of remote sensing and GISs in terrace studies between 1992 and 2024. The “Count” column represents the number of publications, and the “%” column displays the proportion of countries or authors relative to the overall total.

Country	Count	Citation	Average Citation	H-Index	Self-Citation Rate
CHINA	132	3544	29.05	30	0.0364
U.S.	108	3964	36.7	31	0.0025
ENGLAND	69	2646	38.35	24	0.0026
AUSTRALIA	64	1945	30.39	24	0.0041
GERMANY	44	1874	42.59	19	0.0005
ITALY	43	1852	43.07	22	0.0254
SPAIN	33	1525	45.91	20	0.0086
FRANCE	31	1478	47.68	16	0.0027
SCOTLAND	29	562	19.38	14	0.0036
INDIA	21	1209	57.57	9	0.0000
RUSSIA	18	1091	60.61	9	0.0046
CANADA	16	1479	92.44	9	0.0000
NETHERLANDS	16	376	23.5	10	0.0027
NEW ZEALAND	14	1188	84.86	8	0.0008
NORWAY	13	1251	96.23	10	0.0008
AUSTRIA	12	1306	108.83	9	0.0000
JAPAN	12	510	42.5	10	0.0078
IRELAND	10	1199	119.9	9	0.0008
SWITZERLAND	10	1130	113	7	0.0000
POLAND	8	139	17.38	5	0.0000

lists the top 20 countries ranked by publication volume, with European countries accounting for the majority at 12. China leads in publication volume with 132 papers, representing approximately 26% of the total output. The United States has the highest h-index at 31, followed closely by China with an h-index of 30. The U.S. also ranks first in total citations, with 3964 citations, followed by China with 3544 citations. Notably, the self-citation rate for the United States is 0.0025, whereas China has a significantly higher self-citation rate of 0.0364. Ireland has the highest average citations per paper at 119.9, followed by Austria (108.83 citations per paper) and Norway (96.23 citations per paper), all of which are European countries with relatively fewer publications.

Figure 2 shows a chord diagram depicting the collaborations among the top 20 countries by publication volume. The lengths of the arcs around the chord diagram represent the number of publications for each country, while the connecting lines between the arcs indicate collaboration between countries, with thicker lines signifying stronger collaboration. From Figure 2, it is evident that the United States is the country with the largest scale of international collaboration in this field, followed by Norway and Germany. Although China ranks second in terms of publication volume, its level of international collaboration is lower than that of Germany, indicating considerable potential for growth in international partnerships. European countries also exhibit extensive and close collaborations, which enhance their international influence. The predominance of international research collaboration among institutions in China, the United States, and Europe underscores the critical role of transnational knowledge exchange in addressing global sustainability challenges related to terraced agriculture [6].

3.2.2. Analysis of Institution Cooperation Network

Among the top 20 institutions ranked by publication volume, 5 Chinese institutions lead the count, followed by 4 U.S. institutions (

Table 2. Top 20 institutions with publications on the application of remote sensing and GISs in terrace studies between 1992 and 2024. The “Count” column represents the number of publications, and the “%” column displays the proportion of countries or authors relative to the overall total.

Institutions	Count	Acronym	Country	Betweenness Centrality	Year of First Publication
Chinese Academy of Sciences	53	CAS	China	0.23	2000
University of Adelaide	24	UoA	Australia	0.04	2000
Centre National de la Recherche Scientifique	21	CNRS	France	0.18	2009
University of Portsmouth	19	UoP	U.K.	0.03	1999
Consejo Superior de Investigaciones Científicas	14	CSIC	Spain	0.13	2003
Russian Academy of Sciences	14	RAS	Russia	0.04	2010
China Earthquake Administration	13	CEA	China	0.01	2020
Institut de recherche pour le développement	13	IRD	France	0	2013
Institute of Geographic Sciences and Natural Resources Research	13	IGSNRR	China	0.05	2005
Helmholtz Association of German Research Centres	11	HGF	Germany	0.12	2010
United States Department of the Interior	11	DOI	U.S.	0.15	1994
United States Geological Survey	11	USGS	U.S.	0.03	1999
University of Chinese Academy of Sciences	11	UCAS	China	0	2009
University of Padua	11	UNIPD	Italy	0	2019
Beijing Normal University	9	BNU	China	0	2013
Commonwealth Scientific and Industrial Research Organisation	9	CSIRO	Australia	0.04	2011
University of California System	9	UC	U.S.	0.22	2007
Arizona State University	8	ASU	U.S.	0	2012
Consiglio Nazionale delle Ricerche	8	CNR	Italy	0	2016
National Research Institute for Agriculture, Food and Environment	8	INRAE	France	0	2016

). The institution with the highest number of publications is the Chinese Academy of Sciences, with 53 publications, followed by the University of Adelaide in Austria (24 publications) and the French National Centre for Scientific Research (21 publications). The Chinese Academy of Sciences also holds the highest betweenness centrality (0.23), indicating its core position and strong influence in this field. Although the University of California system has fewer publications, it ranks second in terms of its betweenness centrality (0.22) and is accompanied by other influential institutions with centrality measures exceeding 0.10, such as the French National Centre for Scientific Research (0.18) and the U.S. Department of the Interior (0.15).

The analysis of the collaborative relationships among the institutions involved in research on remote sensing and GIS applications for terraces reveals three primary groups of closely collaborating organizations: one centered around the Chinese Academy of Sciences, another encompassing the University of Adelaide and the French National Centre for Scientific Research, and a third focused on the U.S. Department of the Interior and the U.S. Geological Survey (Figure 3).

3.2.3. Analysis of Author Cooperation Network

Among the top 20 authors ranked by publication volume, there are 12 from China and 3 from Italy (

Table 3. Top 20 authors and keywords of publications on the application of remote sensing and GISs in terrace studies between 1992 and 2024. The “Count” column denotes the number of publications, and the “%” column illustrates the percentage of keywords or institutions in relation to the overall total.

Author	Count	Affiliation	Country	Keywords
Tarolli, Paolo	9	University of Padua	Italy	Geosciences Multidisciplinary
Ostendorf, Bertram	6	University of Adelaide	Australia	Environmental Sciences and Ecology
Pijl, Anton	6	University of Padua	Italy	Environmental Sciences and Ecology
Wang, Jian	5	Huazhong Agricultural University	China	Environmental Sciences and Ecology
Yang, Shengtian	5	Beijing Normal University	China	Environmental Sciences and Ecology
Chen, Guan	4	Lanzhou University	China	Geology
Lin, Aiming	4	Kyoto University	Japan	Geochemistry and Geophysics
Liu, Xiaoyu	4	Inner Mongolia Agricultural University	China	Imaging Science and Photographic Technology
Na, Jiaming	4	Nanjing Forestry University	China	Remote Sensing
Tang, Guoan	4	Nanjing Normal University	China	Physical Geography
Wei, Wei	4	University of Chinese Academy of Sciences	China	Environmental Sciences and Ecology
Xu, Qiang	4	Chengdu University of Technology	China	Engineering
Yang, Xin	4	Nanjing Normal University	China	Geology
Zhang, Yi	4	Lanzhou University	China	Geology
Loreto Antón	3	Universidad Nacional de Educacion a Distancia	Spain	Geology
Bai, Juan	3	Beijing Normal University	China	Environmental Sciences and Ecology
Burton, Philip J.	3	University of Northern British Columbia	Canada	Environmental Sciences and Ecology
Casagli, Nicola	3	University of Florence	Italy	Geology
Chen, Die	3	Yangtze River Water Resources Protection Bureau	China	Environmental Sciences and Ecology
Cowley, Dave	3	Historic Environment Scotland	Scotland	Remote Sensing

). A network diagram illustrates the collaboration among authors in this field, highlighting those who have published two or more papers, as shown in Figure 4. The diagram reveals distinct clusters, with each cluster representing a collaborative team of authors. Notably, there is a lack of connections between most clusters, indicating a dispersed yet locally concentrated pattern of collaboration among researchers in this field. The color coding reflects the time periods of research activities, with the largest number of red clusters indicating a high level of research activity in the past three years. The three largest red clusters are centered around Paolo Tarolli (nine publications), Shengtian Yang (five publications), and Jiaming Na (four publications). Their research primarily centers on various applications of GISs and remote sensing, including the monitoring of land degradation in terraced areas, the assessment of gully erosion on the Loess Plateau and in lake environments, and the examination of the geomorphological evolution of terraces within the Loess Plateau region.

3.3. Analysis of Hot Research Topics and Trending Frontiers

3.3.1. Analysis of Hot Research Topics

The co-occurrence network of keywords illustrates the emerging trends and interconnections in the field of remote sensing and GIS applications for terrace studies (Figure 5). The size of each keyword node in the network corresponds to its frequency of occurrence. The largest nodes include remote sensing (55 occurrences), evolution (44 occurrences), erosion (37 occurrences), and climate change (30 occurrences), which occupy central positions in the network and represent the primary focus of the research in this area. The three keywords with the highest betweenness centralities are remote sensing (0.22), climate change (0.14), and landscape (0.13), all exceeding a centrality measure of 0.1, which thereby serve as the core nodes in the diagram. Interestingly, despite appearing only five times, vegetation possesses a high centrality score (0.11), indicating its role as a critical connector between different areas of research.

3.3.2. Analysis of Keyword Citation Bursts

The temporal distribution of the keywords illustrates the developmental trends in remote sensing and GIS applications for terrace research from 1992 to 2024, organized into three-year intervals (Figure 6). It is evident from the figure that the key topics emerged primarily between 1998 and 2006. The evolution of the studies in this domain can be roughly segmented into three phases: The Early Stage (1992–2006): during this period, the research focused on fundamental technologies and environmental characteristics, with keywords such as remote sensing, classification, and erosion predominating. The Mid Stage (2007–2015): the emphasis shifted to broader ecological factors and multi-scenario applications, including keywords like Loess Plateau, agriculture terraces, and deformation. The Recent Stage (2016–2024): more recent research has concentrated on localized practices and policy recommendations, with keywords such as China, region, and model gaining prominence.

The keyword clustering analysis was performed to identify hotspots in the research related to the application of remote sensing and GISs in terrace studies. The analysis yielded a Q value of 0.86, indicating a significant cluster structure, and an S value of 0.92, suggesting a coherent division of clusters (Figure 7). CiteSpace identified eight main clusters within this field, categorized as follows: #0 fluvial terraces, #1 Loess Plateau, #2 terraces, #3 dolines, #4 luminescence dating, #5 machine learning, #6 remote sensing, and #7 wildlife habitat. We extracted the three most typical keywords from each cluster in

Table 4. Characteristics of the cluster analysis for keywords in the study of remote sensing and GIS applications in terrace research.

Cluster Number	Size	Silhouette	Main Year	Main Keywords
0	37	0.812	2006	Evolution; river; rates
1	29	0.833	2010	Land use; soil erosion; management
2	27	0.897	1999	Erosion; catchment; patterns
3	19	0.952	2001	Landscape; Ebro Basin; volcano
4	17	0.965	2007	Landscape; evolution; deposits; shear wave seismics
5	17	0.838	2006	Climate change; model; artificial intelligence
6	16	0.951	2002	Imagery; deep learning; terrace extraction
7	16	0.998	2005	Biodiversity; vegetation classification; Peruvian Amazonia

, with the largest cluster (#0) containing 37 keywords, including evolution, river, and rates. The silhouette value for this cluster is 0.81, exceeding 0.5, which shows that this clustering is valid.

Our citation burst analysis of the top 20 emergent keywords (Figure 8) reveals three fundamental paradigm shifts in terrace research, contextualized within the broader evolution of geospatial agricultural studies. The delayed peak intensity (burst strength = 7.78) of remote sensing, despite its early emergence, mirrors documented patterns in precision agriculture [11], initially constrained by the technical limitations of pre-2000 satellite systems (e.g., Landsat 5–7’s 30 m resolution) before transformative post-2015 advances in UAV photogrammetry (<5 cm resolution) and hyperspectral sensors (PRISMA’s 30 m/30 bands) enabled unprecedented terrace feature extraction, as demonstrated in Andean systems [10]. The sustained prominence of land use (2013–2024) reflects both theoretical advances in terrace-specific land change frameworks and policy-driven research expansion following the FAO’s call for “Globally Important Agricultural Heritage Systems” monitoring [25]. Most notably, the recent emergence of soil erosion and Loess Plateau keywords (2019–2024) correlates with mounting empirical evidence of terrace-mediated environmental benefits, including documented erosion reduction and soil organic carbon accumulation, underscoring the growing recognition of terraces as nature-based solutions for climate adaptation in vulnerable agroecosystems [29].

4. Discussion

Our bibliometric analysis reveals both the remarkable progress and inherent challenges in applying remote sensing and GISs to terrace research, as evidenced by the publication trends in Figure 1. While the field has demonstrated exponential growth since 2010 (333 papers, 65.55% of the total output), with peak productivity in 2020 (58 papers), this expansion must be interpreted cautiously given several methodological constraints. The early period (1992–2007, <10 papers/year) reflects not only the technological limitations of contemporary geospatial systems [1,4] but also potential publication bias favoring successful applications over null results in nascent fields. The post-2015 surge coincides with democratized data access (e.g., the Landsat open archive, Sentinel-2 launch), yet the concurrent decline in average citations (Figure 1) suggests either the dilution of impactful research or a temporal lag in citation accumulation—a phenomenon well documented in rapidly evolving disciplines. These patterns mirror broader concerns in geospatial agriculture research, emphasizing the need for more balanced, validation-focused studies despite the field’s impressive quantitative expansion.

The bibliometric analysis also highlights a geographically concentrated pattern of research activity. For instance, studies from regions such as China, Southeast Asia, and South America are notably active (Figure 2, Figure 3 and Figure 4), likely due to the extensive distribution and significant ecological, agricultural, and cultural value of terraces in these areas. However, in regions where terraces are less prevalent, related research remains relatively sparse. This disparity may be attributed to the distribution characteristics of terraces and the allocation of research resources. Future studies should focus on addressing the technical applicability and development potential in under-resourced regions, aiming to bridge these gaps.

Based on the co-occurrence map of keywords, this study conducted a clustering analysis by examining the keyword frequencies and interconnections (Figure 5 and Figure 6). The results indicate that research on remote sensing and GISs in terrace-related studies from 1992 to 2024 can be categorized into eight distinct clusters (Figure 7). An in-depth discussion of the keyword clusters and their temporal trends helps uncover the knowledge structure of this field, identify key research hotspots and subfields, and track the evolution of these hotspots over time. This offers deeper insights into the application of remote sensing and GISs in terrace research and provides a roadmap for future advancements.

4.1. Fluvial Terraces

Our cluster analysis reveals fluvial terraces (Cluster #0) as the dominant research focus (37 keywords, silhouette = 0.812), not only reflecting their agricultural significance but also exposing several methodological limitations. The presence of fluvial terraces provides favorable conditions for agriculture because the suitable topography can reduce soil erosion and improve water resource utilization, making fluvial terraces an ideal foundation for terrace construction. In recent years, rapid advancements in remote sensing and GIS technologies have led to substantial breakthroughs in fluvial terrace research, particularly in enhancing the accuracy and efficiency of data acquisition and processing. High-resolution remote sensing imagery (e.g., Landsat 8, Sentinel-2, and drone-based imagery) and advanced data sources such as LiDAR have enabled researchers to precisely capture the spatial distribution, topographic features, and microgeomorphic details of fluvial terraces. This overcomes the limitations of traditional ground-based measurements, which often suffer from restricted spatial coverage.

Simultaneously, progress in GIS technologies has provided robust tools for the integration and analysis of multi-source data. For example, the automated extraction and analysis of digital elevation models (DEMs) have significantly improved the efficiency of fluvial terrace delineation. Spatial statistical analysis and modeling techniques have further facilitated the quantification of the spatiotemporal evolution of fluvial terraces and their environmental impacts. Additionally, the adoption of big data and cloud computing platforms for remote sensing data processing—such as Google Earth Engine—has accelerated the processing of large-scale, multi-temporal datasets. This has enabled efficient and precise investigations into the dynamics of fluvial terrace terraces, their formation and evolution, and their associations with climate change.

For instance, in the mountainous regions of northern Guangdong, terraces are a critical agricultural resource. However, with the advancement of urbanization, the phenomenon of terrace abandonment has become increasingly pronounced. Wu et al. (2025) conducted an in-depth study on the spatial differentiation characteristics and driving mechanisms of terrace abandonment in this region [22]. Their findings revealed that factors such as the topographic conditions, farming distance, and shape index were positively correlated with terrace abandonment, while the terrace quality and plot contiguity were negatively correlated. These results suggest that fluvial terraces with favorable topographic conditions are more suitable for terrace construction, helping to reduce abandonment and improve the land use efficiency. Therefore, when planning terrace construction, local governments and management agencies should prioritize fluvial terraces, leveraging their topographical advantages to promote sustainable agricultural development.

4.2. Terraces on the Loess Plateau

The cluster analysis reveals the Loess Plateau (Cluster #1) as the second dominant research focus (silhouette = 0.833, Table 4 and Figure 7), reflecting its global significance as a model system for studying soil erosion and terrace sustainability. The Loess Plateau is characterized by its thick loess deposits and relatively steep terrain. This region frequently experiences severe soil erosion, particularly during periods of heavy rainfall, leading to significant soil loss. Terraces, as a key soil and water conservation measure, have proven highly effective at reducing runoff, controlling slope erosion, improving the land quality, and providing multiple ecological and agricultural benefits. Additionally, terraces significantly influence carbon cycling processes by promoting soil organic carbon (SOC) accumulation [29]. Since the 1950s, the Loess Plateau has been a central region for slope-to-terrace conversion projects. By 2017, terraces accounted for 60% of the total cultivated land area in the region [30]. Scientifically designed, properly constructed, and well-managed terraces can alter the microtopography to reduce runoff erosivity, enhance local rainfall infiltration, and decrease runoff and sediment production. At the same time, terraces improve soil physicochemical properties, benefiting crop growth and nutrient accumulation [31].

Although terraces are widely recognized as effective measures for soil and water conservation, their construction and maintenance can disrupt carbon fluxes between the atmosphere and soil by altering land use types and terrain, potentially impacting the SOC dynamics [30]. The existing studies investigated changes in the SOC levels under different terrace types, revealing both positive and negative effects of terraces on the SOC [4]. The current research on the carbon sequestration potential of terraces is limited by the complexity of conservation practices, the scale of implementation, and the underlying mechanisms. Additionally, discrepancies in SOC estimates arise due to differences in estimation methods, sample sizes, and spatial distributions [32]. Specifically, the estimation of the soil carbon pool in the terraces of the Loess Plateau remains uncertain, with most studies focusing on small-scale mechanisms. Large-scale assessments are still lacking, hindering the scientific evaluation of carbon sequestration in soil and water conservation practices [33].

The existing studies on terraces in the Loess Plateau predominantly emphasize the impact of terrace construction on the SOC content [34]. Research on the mechanisms of SOC sequestration and the accounting of carbon sink effects remains relatively underdeveloped. This is especially true for long-term carbon sink effects, which are critical for regional ecological security and achieving carbon neutrality targets. Due to incomplete statistics on the number of terraces and the lack of spatiotemporal data on terrace use, studies on the accounting and potential assessment of terrace soil carbon sinks are scarce. This gap limits our understanding of the long-term carbon sink effects of terraces. Therefore, it is imperative to conduct research on the mechanisms driving soil carbon sequestration and the potential evaluation of carbon sinks under different terrace utilization methods in typical regions of the Loess Plateau. Such studies are essential to provide scientific evidence for achieving high-quality development in soil and water conservation and advancing sustainable agricultural practices.

4.3. Terracing Effects on Water Resources, Hydrological Processes, and Soil Erosion

The cluster analysis reveals that Cluster #2’s focus is on the relationship between terraces, water resources, hydrological processes, and soil erosion (silhouette = 0.897, Table 4 and Figure 7). In mountainous and hilly regions, steep terrain and intense rainfall make water resource shortage and soil erosion a significant challenge. Studies have shown that terraces effectively mitigate runoff and soil erosion by reducing the slope gradient and increasing the soil water retention capacity. High-resolution imagery obtained through remote sensing enables the identification and quantification of the soil erosion severity, while the GIS facilitates spatial analyses, allowing researchers to evaluate the impact of terraces on soil erosion reduction and to assess the feedback effects of soil erosion on the terrace evolution.

As of 2018, the total area of level terraces constructed in the Loess Plateau had reached 3.69 million hectares [35], accompanied by a significant increase in vegetation coverage [36]. The primary objective of terrace construction in this region is to comprehensively control soil and water losses and improve the ecological environment of the middle reaches of the Yellow River. Consequently, the impact of terraces on watershed soil erosion and sediment transport has become a key research focus in studies on soil erosion and sediment reduction in the Yellow River Basin [37]. For instance, in the He-Long section of the Yellow River’s middle reaches, results showed that human activities contributed 77% and 76% of the reductions in the runoff and sediment load, respectively [38]. Similarly, another study found that human activities reduced the runoff by 67.13% and the sediment yield by 80.10% in the Yan River Basin between 1952 and 2011 [39]. Using the SWAT model, simulated hydrological processes showed that in the Yan River Basin, check dams intercepted 14.3% of the runoff and 85.5% of the sediment yield [40]. While variations exist among these methods, the consensus is clear: terraces play a crucial role in controlling soil erosion in the Loess Plateau and reducing sediment transport in the Yellow River.

Recent studies have noted a shift in the runoff generation modes in the Loess Plateau, characterized by declining infiltration–excess runoff and an increasing contribution of saturation excess and mixed runoff. This transformation in the runoff mechanisms is likely associated with the terrace construction [41]. To advance research on the impact of terraces on watershed hydrology and sediment transport, it is essential to develop a mechanistic understanding of the underlying surface processes and create hydrological models that incorporate shifts in the runoff generation mechanisms. These efforts are pivotal for accurately quantifying the effects of the terrace construction on the runoff and sediment reduction and for devising effective watershed management strategies. Terraces are often situated within catchment areas, and proper watershed management can enhance the water resource utilization efficiency and support crop growth. Remote sensing technologies can monitor rainfall runoff and water accumulation, while GISs enable spatial analyses of catchment areas, helping to identify optimal water management practices and drainage systems to maximize the water use efficiency in terraces. By analyzing land use changes using remote sensing and GISs, researchers can map the spatial distribution of different terrace systems and assess their relationships with environmental factors. This information provides a scientific foundation for optimizing terrace layouts and promoting sustainable development.

Future research should focus on leveraging advanced technologies, such as UAV-based remote sensing and LiDAR, to enhance the monitoring of terraces and their surrounding environments. It is also essential to explore the integration of diverse data sources (e.g., climate, soil, and vegetation) to improve the accuracy and applicability of hydrological models in catchment areas. Based on the results of data analyses, policy recommendations can be proposed to support sustainable terrace management, including strategies for scientific land use planning, ecological conservation, and restoration measures to mitigate soil erosion risks.

4.4. The Geographic Distribution of Terraces and Dolines

The results from the cluster analysis shows that Cluster #3 is about the spatial arrangement of terraces and the geographic distribution of dolines (depressions or sinkholes, Table 4 and Figure 7). Naturally occurring dolines are closely linked to soil and water conservation, land use, and ecosystem services, as they influence water flow, soil accumulation, and biodiversity [42]. For instance, a study in the Torrentejo region of the Ebro Basin, an area characterized by extensive karst landscapes, highlighted three significant phases in the evolution of the terraced agricultural landscape: (1) the establishment of clustered settlements during the Early Middle Ages, characterized by the development of terraced farming systems and their subsequent management under diverse feudal authorities during the High Middle Ages; (2) the extensive restructuring and enlargement of regional terrace systems during the 18th and 19th centuries, influenced by the shift towards agricultural specialization and market-oriented production; and (3) the modern transformation of terrace landscapes driven by advancements in mechanization and the increasing dominance of capital-intensive farming methods. These findings underscore the profound interplay between terraces and local doline landscapes [43].

Understanding the spatial arrangement of terraces and their relationship with doline distribution can offer critical insights into their connection with local agricultural practices. Dolines play a pivotal role in the hydrological cycle by influencing rainwater collection, groundwater recharge, and surface water dynamics [44]. Research on terraces and dolines can illuminate their combined effects on water resource management. Furthermore, dolines serve as vital habitats for biodiversity and significantly impact terraced agroecosystems [45]. Conversely, terrace construction can also alter the formation and evolution of dolines [43]. Analyzing the interactions between dolines and their surrounding landforms can deepen our understanding of their effects on microclimates and agricultural productivity.

To advance this field, it is essential to leverage remote sensing and GIS tools to study the spatial characteristics and dynamics of dolines, integrating these findings into terrace planning, design, and management. Optimizing dolines’ advantages—such as improving soil moisture management—could enhance crop production. Simultaneously, remote sensing and GIS technologies should be employed to monitor and assess how terrace construction influences dolines, ensuring the sustainable development of terrace ecosystems. Ultimately, such research will inform the development of sustainable terrace management practices that protect and harness dolines and their ecological functions.

4.5. Application of Luminescence Dating in Terrace Research

The cluster analysis of the keywords shows that Cluster #4’s focus is on luminescence dating (Table 4 and Figure 7), a technique that determines the last exposure of mineral grains in soils and sediments to light [46]. Luminescence dating offers a precise timeline for land use changes and terrace formation processes. By leveraging luminescence dating, researchers can reconstruct the evolutionary history of terraces and their surrounding landscapes, shedding light on historical farming practices and the impacts of environmental changes across different periods [47]. When combined with high-resolution topographic and soil data obtained via remote sensing, luminescence dating provides critical insights into the morphological evolution of terraces over varying timescales. Using GIS tools, researchers can compare land use patterns across different periods, generate historical landscape evolution models, and assess the influence of human activities on the terrace morphology and ecosystem dynamics [48].

The luminescence dating of sediment samples further unravels the soil formation processes in terrace regions and their interactions with surrounding environments, such as riverine deposition and weathering processes [49]. When integrated with remote sensing data on topographic changes, researchers can analyze the spatial and temporal distribution of sediment layers, revealing correlations between historical natural disasters (e.g., floods, landslides) and terrace development [48]. The GIS mapping of geographic information and historical changes, enhanced by luminescence dating, facilitates the analysis of terrace evolution patterns and identifies the roles of natural environmental factors and socio-economic drivers in terrace development [50]. These insights can guide the formulation of sustainable terrace management strategies, balancing the preservation of traditional agricultural landscapes with the need to adapt to climate change and evolving human activities. Moreover, when coupled with shear wave seismic techniques, luminescence dating can assess the impacts of seismic activity on terraces and surrounding landscapes, offering valuable insights into the soil physical properties, sediment stability, and potential risks to terrace structures [51]. This integration enables a deeper understanding of how specific events influence the geomorphic evolution of terraces over time.

In summary, luminescence dating provides a robust framework for studying terraces across diverse regions and historical contexts. When combined with remote sensing and GIS technologies, it facilitates the creation of terrace time-series datasets, enabling analyses of terrace responses under different climatic conditions. By examining the interplay of natural and human factors, this approach highlights the role of socio-historical contexts, policy changes, and traditional knowledge in terrace evolution. The findings from luminescence dating not only deepen our understanding of terrace dynamics but also provide scientific evidence to inform modern land management and sustainable agricultural practices in the face of future challenges.

4.6. Machine Learning and Deep Learning in Terrace Research

The cluster analysis of keywords shows that Clusters #5 and #6 focus on the application of machine learning, deep learning, remote sensing, and GISs for terraced land research (Table 4 and Figure 7). Machine learning algorithms are capable of processing and analyzing vast amounts of remote sensing data and GIS information, identifying patterns and trends within them [52]. This is particularly crucial for terraced landscapes, which often exhibit complex interactions between the topography, climate, and human activities. Through the use of machine learning, predictive models can be constructed to simulate future climate changes and their potential impacts on terraced agriculture [53]. Such models assist researchers and policymakers in formulating more scientifically grounded management strategies, ensuring the sustainable use of terraced lands. Deep learning algorithms can be employed to analyze historical climate data in relation to crop yields [21]. Furthermore, image classification techniques based on machine learning can automate the detection of land use changes in terraced areas and their surroundings. This capability aids in evaluating the environmental impacts of various land use patterns and contributes to effective land planning.

Machine learning models can also be used to analyze the effectiveness of different terraced land management strategies, facilitating the identification of optimal farming practices and fertilization schemes. Such intelligent decision support systems can enhance yields, reduce resource wastage, and improve ecological safety. By combining remote sensing data with machine learning, the real-time monitoring of the soil moisture, nutrient levels, and vegetation health can provide insights into the ecological status of terraced land, enabling timely interventions to mitigate risks associated with climate change [54].

A variety of machine learning algorithms (such as decision trees, random forests, support vector machines, and deep learning) have been used in terraced land research, and each method has its specific advantages and applicable scenarios [55]. It is essential to select the appropriate algorithm tailored to specific research questions. Additionally, interdisciplinary collaboration encompassing climate science, agricultural science, and computer science should be prioritized to enhance the depth and breadth of terraced land studies. Future research should focus on expanding access to high-quality remote sensing and geographic data, including open data sources and long-term time-series data, as well as addressing the issues of data imbalance and missing values to improve model accuracy and reliability. It is crucial to consider how advancements in new technologies, such as drone-based remote sensing, can further propel terraced land research. Moreover, translating the findings from machine learning-based models into practice and policy will strengthen cooperation between local governments and farmers, guiding sustainable land management practices.

4.7. Terraces and Wildlife Biodiversity

Our cluster analysis reveals that Cluster #7 focuses on the relationship between terraces and biodiversity, especially in some specific ecosystems, such as the Amazon Basin in Peru (Table 4 and Figure 7). Terraced farming creates diverse microenvironments that support the survival of various plant and animal species [20]. For instance, different hydrological conditions and soil types within terraces can promote the reproduction and survival of specific species, thereby enhancing regional biodiversity [56]. In terraced systems, traditional agricultural practices interact with natural ecosystems. Many terraced farmlands still maintain traditional farming methods, which help preserve the habitats of local species to some extent [57]. Remote sensing technologies can be employed to extract and classify the vegetation types in terraced areas from satellite or aerial imagery. By using high-resolution remote sensing data alongside machine learning algorithms, researchers can effectively identify and analyze the various vegetation types and their distributions within terraced landscapes, providing essential baseline data for biodiversity studies. GISs can integrate time-series remote sensing data to monitor changes in terraced farmlands and their surrounding ecosystems, analyze how land use patterns impact biodiversity, and evaluate the effectiveness of ecological restoration efforts.

For example, the Amazon region of Peru is a biodiversity-rich ecosystem where terraced agriculture plays a crucial role in maintaining this biodiversity [19]. Research on terraced farming can help elucidate the unique ecological functions and agricultural practices in this area. Through the combination of remote sensing and GISs, researchers can analyze the impact of terraced agriculture on adjacent wildlife habitats and explore how human activities alter the biodiversity. A comprehensive analysis of terraced lands and their surrounding environments can provide a scientific basis for regional conservation efforts, allowing for the prediction of potential habitat loss and species extinction risks through GIS modeling, which can inform the development of suitable conservation measures.

4.8. Applications of Remote Sensing and GISs in Terrace Hydrology Management

The integration of remote sensing and GIS technologies has revolutionized our capacity to monitor and manage terrace water resources under climate change pressures. Recent advances in multi-spectral and hyper-spectral sensors (e.g., Sentinel and Landsat) enable unprecedented spatial–temporal monitoring of soil moisture dynamics, crop water stress, and terrace-induced runoff patterns at various scales [58]. Particularly noteworthy is the emergence of Synthetic Aperture Radar (SAR) systems, which overcome cloud-cover limitations in tropical terrace regions while providing millimeter-scale deformation measurements crucial for assessing the terrace structural integrity under extreme rainfall events [59].

GIS-based hydrological modeling, when coupled with remote sensing-derived parameters, has demonstrated particular efficacy in three key areas: (1) quantifying the terrace impacts on the watershed-scale water retention through spatially explicit SWAT models [60], (2) optimizing terrace design configurations using terrain analysis and runoff simulation tools [61], and (3) predicting climate change vulnerabilities through scenario-based water balance modeling [62]. The fusion of UAV-based photogrammetry with machine learning algorithms has further enhanced our ability to map microtopographic features governing terrace water distribution—a critical advancement given that traditional surveys often miss sub-meter variations responsible for localized waterlogging or drought stress [63].

However, significant challenges persist in scaling these technologies. Discrepancies arise when reconciling coarse-resolution satellite data (e.g., 10–30 m pixels) with terrace systems characterized by intricate spatial heterogeneity, particularly in ancient or irregularly constructed terraces [64]. Emerging solutions include data fusion techniques blending microwave, thermal, and optical datasets; object-based image analysis for terrace unit delineation; and edge computing approaches for real-time terrace water monitoring. Future research directions should prioritize the development of standardized protocols for remote sensing and GIS-based terrace water assessments, particularly in data-scarce regions where smallholder terraces dominate. Furthermore, the integration of IoT sensor networks with satellite-derived data presents a promising avenue for creating adaptive terrace water management systems resilient to climate variability.

4.9. Limitations and Perspectives

Despite significant advances in the application of remote sensing and GIS technologies for terraced agriculture research, several limitations remain. The current research often focuses on specific regions or relies on single data sources, leading to a lack of large-scale, cross-regional comprehensive studies. This narrow focus hinders our understanding of the broader implications of terraced agriculture on biodiversity and ecosystem dynamics.

When it comes to monitoring the dynamic changes of terraced farmland over time, the existing studies have not fully harnessed the potential of time-series remote sensing data. The ability to track changes across multiple time points is crucial for understanding the impacts of climate change, land use practices, and other anthropogenic factors on terraced systems.

Our current search strategy focused exclusively on publications indexed in the Web of Science, which may have introduced selection bias and limited the breadth of our analysis. While this approach ensures the inclusion of peer-reviewed literature of high academic standards, it potentially overlooks significant contributions found in other databases, such as Scopus and Google Scholar. Consequently, important insights and emerging trends in terracing and soil hydrology might remain undiscovered.

Another potential limitation of this study is its reliance on keyword analysis for the bibliometric mapping. While keyword analysis can effectively highlight prevailing research themes, it inherently depends on the selection and accuracy of the chosen keywords. The process of selecting keywords may introduce biases, as it may overlook relevant studies that use different terminologies or synonyms. Additionally, the dynamic nature of research fields often leads to the emergence of new keywords over time, which may not be fully captured in our analysis. Acknowledging these limitations is crucial for understanding the context and representativeness of our findings within the broader literature on terracing effects and water resources under climate change.

In the future, considering a more inclusive bibliometric analysis that integrates diverse data sources to provide a holistic understanding of this evolving field, exploring the integration of drone technology, ground-truthing data, and cloud computing for multi-source fusion analysis presents a promising direction for research. Drones can provide high-resolution imagery and data collection capabilities, while ground-truthing can help validate and enhance the accuracy of remote sensing findings. Cloud computing will facilitate the processing of large datasets, making it easier to analyze complex spatial and temporal patterns.

Enhancing the interpretability of deep learning algorithms is a critical step forward for this field. Increasing the transparency and comprehensibility of these models is essential to foster trust in their outputs, particularly when the results will inform policy decisions and conservation strategies. By addressing these challenges, future research can better contribute to our understanding of terraced agriculture and its multifaceted role in regional ecosystems and biodiversity preservation.

5. Conclusions

This comprehensive bibliometric analysis elucidates three decades (1992–2024) of evolving research trends in remote sensing and GIS applications for terraced farmland, offering actionable insights for both policymakers and researchers. The observed surge in publications, particularly post-2019, underscores the global urgency to address terraced agriculture’s role in mitigating climate-induced challenges such as soil degradation and water scarcity. The dominance of collaborative networks led by institutions in China, the United States, and Europe highlights the necessity of cross-border knowledge transfer to tackle shared sustainability goals. The temporal keyword analysis reveals a critical shift from basic remote sensing applications (pre-2005) to advanced machine learning and deep learning modeling in recent years. This transition advocates for upscaling AI-driven tools (e.g., convolutional neural networks) to predict terrace-specific hydrological responses under climate extremes and for integrating IoT-based sensor networks with satellite data for real-time terrace monitoring in data-scarce regions. The persistent focus on the Loess Plateau (2005–2015) and emerging hotspots (e.g., Andean terraces) calls for region-specific guidelines leveraging remote sensing and GISs to optimize terrace design for water conservation, and for incentivizing farmer participation in crowdsourced geodata collection to bridge ground-truth gaps.

To advance the field, we propose four concrete directions: (1) Standardized benchmarking: the development of unified metrics to evaluate the remote sensing and GIS accuracy across diverse terrace types (e.g., bench vs. contour) and climatic zones. (2) Hybrid modeling: combining process-based hydrological models with machine learning algorithms to enhance the predictive capability for terrace water management. (3) Equitable technology access: addressing disparities in remote sensing and GIS adoption by creating low-cost, open-source tools tailored to smallholder terraces. (4) Long-term impact studies: quantifying the socioecological trade-offs of terrace rehabilitation using longitudinal satellite data (e.g., Landsat archives).

By translating these findings into practice, the scientific community can empower stakeholders to transform terraced landscapes into resilient, productive systems. This study not only maps the past and present of terrace research but also charts a pathway toward achieving SDG 2 (Zero Hunger) and SDG 15 (Life on Land) through geospatial innovation.