**1. Introduction**

In accordance with the International Union for Conversation of Nature (IUCN) [1,2], a protected area (PA) is defined as "a clearly defined geographical space, recognized, dedicated, and managed, through legal or other effective means, to achieve the long-term conservation of nature with associated ecosystem services and cultural values". In general, PAs are considered to be areas of land or sea, including national parks, national forests, natural reserves, conservation areas, wilderness areas, marine protected areas (MPAs), and wildlife refuges and sanctuaries, that are designated for the conservation of native biological diversity and natural and cultural heritage and significance [3]. Over the past century, the amount and coverage of both terrestrial and marine PAs have markedly increased [4]. As reported by the World Database on Protected Areas [5], as of July 2018, there were 238,563 designated PAs, covering about 14.9% and 7.3% of the Earth's land and ocean surface areas, respectively. PAs are central to nature conservation efforts with key environmental, social, cultural and economic functions throughout the world [3,6]. In addition, PAs play an important role in biodiversity conservation and ecosystem integrity [7–10].

Remote sensing refers to the art, science, and technology used for Earth system data acquisition through nonphysical contact sensors or sensor systems mounted on space-borne, airborne, and other types of platforms, data processing and interpretation from automated and visual analysis, information generation using computerized and conventional mapping facilities, and applications of the generated

data and information to benefit society and meet its needs. Remote sensing can provide comprehensive geospatial information to map and study PAs at di fferent spatial scales, e.g., high spatial resolution and large area coverage, di fferent temporal frequencies (e.g., daily, weekly, monthly, or annual observations), di fferent spectral properties (e.g., visible, near infrared, or microwave), and spatial contexts (e.g., the immediately adjacent areas of PAs vs. a broader background of land and water bases). Remote sensing is considered to be a cost-e ffective method to support the monitoring e fforts of PAs and has played a vital role in protecting natural resources, ecosystems, and biodiversity [11,12].

In terms of the large-scale observation ability of remote sensing, the technology is becoming a common practice for monitoring the characteristics and change of land surface properties of PAs [13]. For example, remote sensing has been applied to the assessment of night-time lighting within and surrounding global terrestrial PAs and wilderness areas [14], continuous monitoring of the landscape dynamics of national parks by Landsat-based approaches [15–19], the evaluation of forest dynamics within and around the Olympic National Park using time-series Landsat observations [20], and monitoring the wildlife habitat changes in Kejimkujik National Park and the National Historic Site in southern Nova Scotia of the Canadian Atlantic Coastal Uplands Natural Region [21]. One particular advantage that remote sensing can provide for the inventory and monitoring of protected areas is information to better understand the past and current status, the changes that occurred under di fferent impacting factors and managemen<sup>t</sup> practices, the trends of changes in comparison with those in the adjacent areas, and the implications of changes on ecosystem functions [22–24]. Remote sensing has unique advantages in monitoring frontier lands, which are always in remote and di fficult-to-reach locations and huge in their area coverage. Di fferent types of remote sensing data have been applied in the study of frontier lands—for example, using hyperspectral and radar data to monitor forests in the Amazon [25–30], in Africa [31], and in Siberia [32–35], and for hydrologic change detection in the lake-rich Arctic region [36,37], along the coastal zones [38–41], and in MPAs [42].

There have also been several reviews on PA monitoring using remote sensing. For example, Nagendra et al. [43] provided a review of remote sensing for conservation monitoring by assessing PAs, habitat extent, habitat condition, species diversity, and threats. Kachelriess et al. [44] reviewed the application of remote sensing for MPAs management. Gillespie et al. [45] reviewed advances in the spaceborne remote sensing of terrestrial PAs. Willis [11] provided a review of the remote sensing change detection methods employed for the ecological monitoring of United States PAs. The existing reviews have mainly been focused on a certain type of PAs or the monitoring method. There have been no bibliometric analyses of remote sensing applications in the monitoring of PAs.

Bibliometric analysis, introduced by Pritchard (1969), is a mathematical and statistical approach to analyze pertinent literature and understand the global research trends in a specific area [46,47]. Bibliometric analysis methods are frequently used to provide quantitative analyses of academic literature [48], and have been applied to environmental engineering and science, soil science, ecology, food safety, new energy utilization, and other areas. A bibliometric analysis helps identify research gaps and directions in one certain area [49]. In recent years, studies have applied this method to evaluate the research trends of remote sensing and its application in di fferent scientific fields [50–52]. For instance, Zhang et al. [53] combined the new index (geographical impact factor) and traditional bibliometric methods to study the global research trends in remote sensing studies. Viana et al. [54] performed a bibliometric analysis to appraise the publication, research trends, and characteristics regarding the application of remote sensing data in human health. Wang et al. used the bibliometric method to study the research status and trends in the remote sensing of crop growth monitoring in China [55]. However, no attempt has been made to evaluate the inventory and monitoring of PAs in the literature using bibliometric analysis methods. In recent years, the number of publications on the remote sensing monitoring of PAs has been increasing. Hence, it is necessary to summarize the current status and development trend in this field. With the help of bibliometric methods, researchers can better understand the current number of publications, what journals these documents are published

in, what the influential countries and institutions in this field are, how the research direction of this discipline is developing, etc.

Using a bibliometric approach, this article analyzes the relevant literature specialized in remote sensing applications in PAs. The aims of this work are to (1) summarize the variation in the characteristics of document types, total publication output, subject categories, and source journals; (2) analyze the publication output and international collaboration by countries, institutions, and authors; and (3) reveal the common research topics of PA monitoring research based on a keyword analysis. This research can help us understand the progress in this field and identify the relevant research and application directions.

#### **2. Methodology and Data Collection**

The bibliometric indicators analyzed in this research include a number of publications, subject categories, source journals, countries, and institutions, which were all obtained directly from the Web of Science. The Web of Science database can o ffer various statistics on retrieved papers, including the author, series name, conference name, country/region, document type, editor, fund funding institution, authorization number, group author, language, institution, publication year, research direction, source publication name, and the Web of Science category. Another function of web of science is to "Create a Citation Report", which can directly generate the total quoted frequency of the retrieved documents, the total quoted frequency of the removed self-cited documents, the quoted documents, the quoted documents of the removed self-cited documents, the average times a document has been cited, and the H-index of each item.

Co-authorship among countries and institutions was also analyzed in this research. Co-authorship mainly analyzes the co-signature of authors in the published paper. If the authors sign their names together in the paper, they are considered to have a cooperative relationship. At present, co-authorship analysis not only focuses on an analysis of researchers, but also includes an analysis of the cooperation between countries and institutions. In the case of the co-authorship analysis, the link strength between countries and institutions indicates the number of publications that two a ffiliated countries and institutions have co-authored, whereas the total link strength indicates the total strength of the co-authorship links of a given country or institution with other countries and institutions. Similarly, in the case of the co-occurrence analysis, the link strength between the author keywords indicates the number of publications in which two keywords occur together. In order to investigate the development of remote sensing in the field of protection area monitoring, we determined the keywords related to satellite, sensor and remote sensing monitoring method from the results of the co-occurrence keywords.

In this study, the VOSviewer software was utilized to visualize the co-authorship collaboration networks of countries and institutions and produce a keywords co-occurrence analysis. Invented by Van Eck and Waltman (Leiden University) in the Netherlands, VOSviewer is freely available -text mining software for generating bibliometric maps and analyzing trends in the scientific literature [56]. The outstanding feature of VOSviewer is its strong graphic display ability, which is especially suitable for analyzing large-scale sample data. This visualization e ffect is better than that of other similar analysis software, and the analysis function is more comprehensive. VOSviewer is a robust tool that uses clustering algorithms and functionalities based on the strength of the connections among items to facilitate the analyses of the network. [57]. The VOSviewer software uses a circle and label to represent an element, in which the circle size represents the importance, and circles with the same color belong to the same cluster.

Bibliometric maps are created by VOSviewer. These maps include items such as countries, institutions, and author keywords in the present study. The connection or relation between two items is named a link. The strength of a link indicates the number of publications that two countries or institutions have co-authored in the case of co-authorship links, or the number of publications in which two author keywords occur together in the case of co-occurrence links [46,58]. In the VOSviewer, there are two methods used to calculate link strength: full counting and fractional counting. Full counting means that a co-authored publication is counted with a full weight of one for each co-author, which implies that the overall weight of a publication is equal to the number of authors of the publication. Fractional counting means that a co-authored publication is assigned fractionally to each of the coauthors, with the overall weight of the publication being equal to one. As analyzed by Perianes-Rodriguez et al. (2016), a fractional counting approach is preferable to full counting [59]. Therefore, we chose fractional counting to calculate the link strength.

VOSviewer uses a clustering algorithm to cluster the literature network, which is similar to the network clustering method of Modularity, specifically the maximization formula:

$$V(c\_1, \dots, c\_n) = \frac{1}{2m} \sum\_{i$$

$$w\_{i\bar{j}} = \frac{2m}{c\_i}c\_{\bar{j}}\tag{2}$$

where *ci* is the cluster of element *I*, and γ is the resolution of clustering. By adjusting its size, different resolution clustering can be obtained. The larger γ is, the more clustering will be obtained, and the finer the classification will be.

In VOSviewer, the number of clusters is determined by the option "choose threshold". In the case of a co-authorship analysis, the threshold is the minimum number of documents of a country or an institution. In the case of co-occurrence analysis, the threshold is the minimum number of occurrences of a keyword. We can choose the threshold according to our own needs.

The VOSviewer software has been widely used in bibliometric analyses in many fields. For example, Santos et al. used VOSviewer to map knowledge networks on female entrepreneurship [60]. Sainaghi et al. mapped the co-citation network of journals and authors on the foundations of hospitality performance measurement research using VOSviewer [61]. Sweileh et al. used VOSviewer to visualize map-based bibliometric indicators for the global research output on antimicrobial resistance among uropathogens [62].

The relevant documents were retrieved from the Science Citation Index Expanded (SCI-Expanded) and Social Science Citation Index (SSCI) of the Web of Science database, which is a multidisciplinary database of Thomson Reuters [63]. The following keywords were used to search all archived documents: TS (Topic) = "protected area\*" or "natural reserve\*" or "conservation area\*" or "national park\*" or "national forest\*" or "marine protected area\*" or "wilderness area\*" or "frontier land\*" or "natural monument\*" or "biodiversity conservation" and "remote sensing". The publications that contained any of those keywords or variants (with\*) in their titles, abstracts, and keyword lists were included [48]. The information on title, authors, institution, abstract, keywords, and cited references was downloaded. We set the starting time of this study as 1991, considering that the number of publications under the subject of remote sensing applications in PAs and relevant studies increased significantly after 1991 in professional journals and publications. This is illustrated in Figure 1. The data collection was conducted on 16 November 2019. Until 2018, a total of 4546 records were retrieved as the data for this analysis. Among these records, 3994 papers were focused on the remote sensing monitoring of terrestrial PAs, while the other 552 papers were on MPAs.

**Figure 1.** Annual publication and cumulative number, 1991–2018.
