**3. Results**

#### *3.1. General Characteristics and Trends of Publication Outputs*

The trend for publications from 1991 to 2018 is illustrated in Figure 1. In general, the number of publications has shown an increasing trend over the years, with small fluctuations between individual years. According to the dates, the evolution of the published article output can be divided into three stages. The first stage extends from 1991 to 2003, with a relatively slow growth period. The second stage features a steady growth period from 2004 to 2011. The third stage is a fast growth period from 2012 to 2018.

The sample documents covered a total of 108 subject categories. The research domain covered a wide variety of themes and disciplines. The top 10 subject categories with more than 200 documents are displayed in Figure 2. The results indicate that environmental sciences ranked first with 1524 publications, followed by remote sensing with 1062 publications, ecology with 946 publications, and imaging science and photographic technology with 652 publications. Multidisciplinary geosciences, physical geography, forestry, biodiversity conservation, water resources, and meteorological and atmospheric sciences were also relevant subject categories.

**Figure 2.** Top 10 subject categories in the field of the remote sensing monitoring of protected areas (PAs).

For the source journals, 739 di fferent journals published papers related to remote sensing for PA monitoring. Table 1 shows the top 20 journals in terms of total relevant publications. Remote Sensing of Environment ranked first, with 256 articles covering 5.63% of the total publications. Remote Sensing ranked second, with 174 articles accounting for 3.83%, while the International Journal of Remote Sensing ranked third, with 153 articles accounting for 3.37%. The ISPRS Journal of Photogrammetry and Remote Sensing, Forest Ecology and Management, and the International Journal of Applied Earth Observation and Geoinformation (ranked 4th, 5th, and 6th places) accounted for 2.38%, 2.11%, and 2.02%, respectively.


**Table 1.** Top 20 main source journals in the research field.

#### *3.2. Countries, Institutions, and International Collaboration*

According to the retrieved results, the papers covered a total of 153 di fferent countries (or territories, hereafter referred to as "countries" for simplification). The geographical distribution of the top 20 productive countries for the overall study period is shown in Figure 3. The USA ranked first with a dominant output of 1655 papers or a share of 36.41%. China had 619 papers (13.62%) and UK had 479 (10.54%), ranking second and third, respectively. Other top ranked countries are Germany (7.92%), India (7.90%), Australia (7.11%), Canada (6.64%), and Italy (5.65%).

The co-authorship analysis studied a network of the main countries, which is plotted in Figure 4. These countries published more than 60 papers. There were four main clusters formed in the network (Table 2). The USA showed 62,644 citations and a link strength of 634, the UK showed 14,335 citations with a link strength of 241, and China showed 12,906 citations with a link strength of 265, which surpassed all the other clusters. The strongest link strength was evidenced by the USA and China, with a 151.93 link strength, followed by the USA and Canada with a 64.89 link strength, the USA and the UK with a 58.69 link strength, the USA and Germany with a 49.93 link strength, the USA and Australia with a 46.48 link strength, and the USA and Brazil with a 43.59 link strength.

According to the results, 4451 institutions contributed to the analyzed publications. The top 15 research institutions with the largest number of documents are listed in Table 3. By far the most productive institution was the Chinese Academy of Sciences in China, with 296 publications. The University of Maryland was in second place with 118 publications. The Chinese Academy of Sciences also ranked first in number of citations, followed by NASA, University of Maryland, and the U.S. Forest Service.

**Figure 3.** The geographic distribution of the top 20 productive countries.

**Figure 4.** Co-authorship cooperation between productive countries. Each node represents a country. The size of the nodes reveals the citations of the countries, while the thickness of the lines between them shows the strength of collaboration.


**Table 2.** 5 main clusters for country collaboration.

**Table 3.** Top 15 institutions based on total publications.


An institutional cooperation network based on the VOSviewer software for the construction of scientific maps is presented in Figure 5. This figure presents the four clusters of collaboration among the prolific institutions with 35 or more publications. The largest cluster (red) contains nine institutions. All the institutions in the red cluster belong to the USA. The green and blue clusters both contain five institutions. Two of the institutions in the green cluster belong to the Netherlands, and the remaining three are from Australia, the UK and the USA. The blue cluster is composed of three Chinese institutions and two American institutions. The fourth cluster (yellow) includes three institutions from Canada. It can be seen that the cooperation between institutions is mainly focused within the same country or neighboring countries.

**Figure 5.** Co-authorship cooperation between productive institutions. The colors represent clusters of institutions, the size of frames represents the number of articles published by these institutions, and the lines represent the strength of cooperation among institutions.

#### *3.3. Common Interests in Research Topics*

Keywords, a core element of papers, offer a highly summarized form of a paper's contents. In order to understand the focus areas and development trends of one field, it is necessary to systematically analyze the selection of keywords in relevant studies [64]. Table 4 shows the 20 most frequently used author keywords from 1991 to 2018, including "remote sensing", "GIS", "Landsat", "deforestation", "LiDAR", "conservation", and "biodiversity", for research on PA monitoring that is concentrated on deforestation and biodiversity conservation.

A statistical analysis of the changes in the author keywords between different stages is beneficial for a comparative analysis of the changes in common research subjects and the development process of PA monitoring studies [19,65,66]. Table 4 separates the development of PA monitoring research into three stages, namely 1991–2003, 2004–2011, and 2012–2018. "Remote sensing" and "Landsat" were the most frequently used author keywords and appeared in the top 20 in all three periods. The "MODIS" and "LiDAR" keywords increased in frequency of appearance from 1991 to 2011 and increased further in 2012–2018, which indicates that the platform played a significant important role in PA monitoring. Comparing the three different stages, the keywords rankings changed considerably. The keyword "climate change" began to appear in the top 10 during 2012–2018, which suggests that more attention was being given to climate change on PA research. The research focus of each stage is as follows. The early stage of research focuses on landscape ecological change and human disturbance. The middle stage focuses on the change detection of land cover and land use caused by deforestation. The late stage focuses on the impact of climate change on PAs.

In order to trace the trend of the remote sensing data used in PAs research, the most frequently selected keywords related to satellites and sensors were counted. The top ten are Landsat, MODIS, LiDAR, SPOT, AVHRR, ASTER, IKONOS, PALSAR, Sentinel (Sentinel-1 and Sentinel-2), and WorldView, with low, moderate, or high-resolution sensors. The annual publications of the top ten satellites and sensors are shown in Figure 6. In terms of quantity, Landsat was the most frequently used satellites and sensors type, with 1078 papers, followed by MODIS with 439 papers and LiDAR, with 370 papers. In addition, with the continuous development of remote sensing technology, some new platforms and satellites have emerged and have been applied to monitor PAs in recent years. For example, there were 35 papers on the UAV monitoring of PA, and 26 papers on small satellites from 2001 to 2008.


**Table 4.** Top 20 author keywords in different stages, 1991–2018,1991–2003, 2004–2011, and 2012–2018. F(%)—frequency of author keywords and their percentage of

**Figure 6.** Annual publications of the main satellites and sensors in the research field.

Based on the co-occurrence analysis, the remote sensing monitoring methods are also counted in Table 5. The remote sensing monitoring methods mainly include classification, time-series analysis, model methods, object-oriented method, visual analysis, direct comparisons, and hybrid methods [67,68]. The classification method holds the first position with 526 papers and 11.57% of the total publications, followed by time-series analysis (288, 6.34%) and model method (159, 3.50%).


**Table 5.** The main remote sensing monitoring methods used for protected areas (PAs).

Figure 7 shows a co-occurrence network analysis of the keywords, which can be used to identify the research front in terms of topical trends for PA monitoring. In this analysis, the minimum number of occurrences of a keyword is 30 times for titles and abstracts in all publications. The research theme of PA monitoring has been categorized into six colored clusters, which were analyzed as follows. The red cluster with the highest number of keywords (12) is led by "land cover"; In addition, "land use", "monitoring", "mapping", "hyperspectral", and "classification" are also the main keywords of this cluster. Most keywords in this cluster are associated with studies on land use and land cover classification using hyperspectral remote sensing data. The blue cluster, with 11 keywords, has "Landsat", "MODIS", "NDVI", "climate change", "change detection", and "wetland" as its main related keywords, which appear in the relevant research on the habitat mapping and change detection of PAs, as well as the impact of climate change. The green cluster (11 keywords) focuses on the keywords: "deforestation", "LiDAR", "REDD", "biomass", "forest inventory", "tropical forest", "forest management", and "carbon". The keywords of this cluster are closely related to estimating forest biomass and carbon storage in PAs using LiDAR data. The yellow cluster has 10 keywords; the most frequently used is "remote sensing" followed by "conservation", while "biodiversity", "protected areas", and "fragmentation" are ranked 3rd–5th, respectively. Most keywords in this cluster relate to

the use of remote sensing to support biodiversity conservation in PAs. The number of keywords in the purple cluster is four, including "land-use change", "land-cover change", "ecosystem service", and "landscape metrics". This cluster is related to the analysis of land-use/land-cover change and ecosystem service evaluation by remote sensing and landscape metrics. The orange cluster includes only three keywords. The keyword "GIS" appears most frequently, with a total of 387 occurrences. The other two keywords are "soil erosion" and "RUSLE (The Revised Soil Loss Equation)". This cluster has connections with keywords related to the study of soil erosion and its spatial distribution in PAs using the GIS analysis method.

**Figure 7.** Keywords co-occurrence network. Each node represents a keyword, the size of the node indicates the number of occurrences of the keyword, and the line thickness of the two nodes represents the degree of connection.
