Global forest change is a public concern [1
] because humans are heavily reliant on forests. On one hand, forests provide forest products [7
] including food, medicine, fodder for livestock, fuel, and shelter [8
]. The livelihoods of 1.6 billion rural people are estimated to depend on forests [9
]. On the other hand, the ecological services provided by forests can solve some environmental problems [10
]. For example, forests can reduce soil erosion and improve water quality by regulating the hydrological cycle [12
], ameliorate climate change by increasing carbon sequestration [14
], and protect biodiversity [15
]. Forest recovery after severe deforestation first occurred in European countries in the 18th century [16
] and later in North America. In recent years, forest cover has begun to increase in many developing countries [18
]. This increasing trend in forest cover led Mather to coin the term “forest transition” (FT), which refers to the process of moving from a reduction to an expansion in forest area within a certain country or region [17
Since the introduction of the term FT [17
], many researchers have studied this process and its underlying causes. In European countries, economic development and forest scarcity are regarded as two pathways of FT. Economic development and urbanization reduced the conversion of forest lands, and forest loss during agricultural expansion caused a countervailing trend [10
]. Later, several theories supplemented the theoretical framework and proposed other FT pathways, such as government policies [20
], globalization [22
], and smallholder tree farming based on land use identification [25
]. Many studies on FT have been carried out on subnational [26
], national [27
] and multinational [19
] scales; however, global-scale studies on FT are lacking.
To obtain an understanding of global FT conditions, consistent data of annual forest cover are needed. Many studies have used forest resource assessment (FRA) data from the Food and Agriculture Organization (FAO) to investigate FT [10
], since these data provide the annual forest area. The quality of FRA2015 data has improved as more countries have added remote sensing data to supplement the standard data sources [10
]. However, FRA data still suffer from some problems, including uneven data quality and differences in forest definitions among countries [32
]. While other recent forest cover datasets have high spatial resolution, they only provide forest cover data for certain years. For example, the GlobeLand30 land cover dataset [33
] has a spatial resolution of 30 m and contains two baseline years (2000 and 2010). The Global Land Cover Facility (GLCF) [35
], which has a spatial resolution of 30 m, covers 1990, 2000, and 2005. The Global Forest Watch (GFW) dataset developed by Hansen [36
] has a resolution of 30 m and includes only annual gross loss of forest cover for 2000–2012. In 2017, the European Space Agency (ESA) released global annual land cover data from 1992–2015, including data for the entire surface of the Earth at a spatial resolution of 300 m [37
]. This dataset provides continuous, consistent, and long-term series data on global land cover during the last two decades, presenting a unique opportunity for exploring FT.
The primary objective of this study was to use the ESA dataset to provide a global view of forest cover change, FT status, and their drivers. Because the ESA dataset only covers 1992–2015, developed countries were excluded from the analysis because these countries experienced FT much earlier [10
The objectives are summarized as follows:
Analyze the spatial and temporal variations in forest cover between 1992 and 2015 in developing countries on the global, continental, and country scales; and
Determine the driving factors for FT occurrence based on the binary logistic model.
Based on the latest annual data published by ESA, we analyzed the spatiotemporal characteristics of forest cover and determined the FT statuses in developing countries from 1992–2015 on the global, continental, and country scales. Finally, we attempted to identify the driving forces of FT in developing countries and different continents. The results are expected to be useful for understanding FT and provide implications for further theoretical explorations of FT. The conclusions are summarized as follows.
(1) The forest area in developing countries decreased from 1992–2015, although the rate of decrease slowed after 2004. The areas of forest reduction were mainly distributed in South America, which had the largest area of forest loss (505,100 km2) and accounted for approximately 85% of total forest loss.
(2) On a national scale, the countries with the largest decreases in forest area between 1992 and 2015 were Brazil, Argentina, Paraguay, and Indonesia, three of which are in South America. The two countries with the largest increases in forest area were African countries, South Sudan and Ethiopia, where forest coverage continued to grow over the study period.
(3) Over 80% of African countries had experienced FT by 2015; in Asia and North America, the percentages exceeded 60.0%. In South America, FT had only occurred in half of the countries by 2015.
(4) The factors influencing FT are different at different scales, and the same variable may have opposite effects in different continents. At a global scale, forest coverage and trade ratio had negative effects on the probability of FT occurrence, whereas the effect of urbanization level was positive.