1. Introduction

Forest Transition theory has been used to explain the change from shrinking to expanding forest cover (Mather, 1992; Mather and Needle, 1998). Forest transition refers to the turning point from net forest area loss to net forest area gain. Forest transitions have been identified in countries of Europe and North America (Mather, 1992; Mather et al., 1998, 1999; Mather, 2004; Kauppi et al., 2006), and recently also in many developing countries (Mather, 2007; Rudel et al., 2005; Meyfroidt and Lambin, 2008, 2010).

Forest transition towards more forest cover is assumed to have the potential to improve environmental services (Mather, 2007; Rudel et al., 2005). In order to realize this potential, forest researchers have attempted to identify the causes of forest transition, as well as variables which may promote or speed up the process of transition (Xu et al., 2007; Culas, 2012; Mather, 2007; Rudel et al., 2005).

China passed through the turning point of its forest transition during the 1980s, when state policies played a central role (Mather, 2007; Rudel, 2009; Rudel et al., 2005). In our previous research, we found a paradox between a continuing decline in the area covered by natural forests and an increase in overall forest cover in tropical China (Xu, 2011; Ziegler et al., 2009; Zhai et al., 2014).

Rubber trees were introduced to tropical China more than fifty years ago, and were initially managed by state farms, the majority of which were established during the 1950s (Xu et al., 2014; Lardy, 1983). From the late 1980s onwards, the expansion of rubber plantations accelerated due to the involvement of local smallholders (Zhai et al., 2012; Xu et al., 2014). The area occupied by rubber plantations in Hainan has increased by 21.71% since the 1950s: from 0.54% of total land area in the 1950s to 14.16% in 1988, and to 22.25% in 2008 (Wang et al., 2012). In Xishuangbanna, the area occupied by rubber plantations has increased since the 1970s by 20.89%: from 1.25% of total land area in 1976 to 3.63% in 1988, to 11.30% in 2003, and to 22.14% in 2010 (Xu et al., 2014; Li et al., 2007).

During the same period, natural forest cover has decreased in both regions. In Xishuangbanna, the area covered by natural forest has decreased by more than 30% over the last forty years (Li et al., 2007; Jing and Ma, 2012). In Hainan, natural forest cover has decreased from 24.45% of total land area in the 1950s to 13.50% in 1980, and again to 8.30% in 1988. By 1995 however, the area covered by natural forest had increased to 12.00% (Zhang et al., 2000).

In our previous research in Xishuangbanna, we found that natural forest loss was mainly caused by the expansion of plantation forests (Zhai et al., 2015). We also found that in Xishuangbanna deforestation coexisted with plantation expansion in the overall process of forest transition. Based on our previous research findings and on our observations in the field, we hypothesized that it was mainly the expansion of plantation forests that has contributed to the increase of overall forest cover. We selected Hainan Island and Xishuangbanna prefecture as our study sites in order to investigate this hypothesis, and used data on rubber plantations to investigate the role of plantation forests. The current stage of forest transition in tropical China, with its interplay of forest cover increase, natural forest loss, and forest plantation expansion provides us with a unique opportunity to study current forest transition theory at a fine scale, and to demonstrate the importance of differentiating between forest types in forest transition.

In this research, we investigated the dynamics of changes in overall forest cover, natural forest cover, and rubber plantation cover in tropical China in order to answer the following questions: 1) When did the turning point of forest transition occur in tropical China? 2) What is the relationship between changes in natural forests and plantation forests in forest transition? and 3) What are the implications of this relationship for Reducing Emissions from Deforestation and Degradation and Enhancing Forest Carbon Stocks (REDDþ) projects?

2. Methods 2.1. Study sites

Hainan Island and Xishuangbanna are the two largest tropical regions in China and are both considered biodiversity hotspots (Francisco-Ortega et al., 2010; Zhu and Roos, 2004) (Fig. 1).

Hainan Island (18°10'-20°10' N and 108°37'-111°03' E) is the largest tropical island in China with an area of 33, 920 km². The island's tropical rainforests are located at the northern margin of tropical Asia (Zhu and Roos, 2004) and are known for their high biodiversity. The northern part of Hainan Island is relatively flat with an overall elevation of 300 m above sea level. The southern part is hilly, with Wuzhishan in the center of the island being the highest mountain (1876 m) (Francisco-Ortega et al., 2010). Approximately 39% of Hainan Island is covered by mountains and hills. The island has a tropical monsoonal climate, with a rainy season from May to October and a dry season from November to April (Zhai et al., 2012). Its average annual rainfall is more than 1600 mm. The main forest types are tropical savannas, tropical monsoon forests with evergreen and deciduous trees, lowland or montane seasonal evergreen rainforests, and mangrove and tropical bamboo forests (Francisco-Ortega et al., 2010).

Xishuangbanna Prefecture (21°08'-22°36' N, 99°56'-101°50' E) in Yunnan Province covers 19, 150 km² and borders Laos to the south and Myanmar to the southwest. Its altitude varies from 475 m to 2430 m, and roughly 95% of the region is covered with mountains and hills. The climate in Xishuangbanna is influenced by warm-wet air masses moving in from the Indian Ocean in summer and by continental air masses in winter. Monsoonal changes in airflow result in a rainy season from May to October, and a dry season from November to April (Li et al., 2007). Annual rainfall ranges from 1100 mm to 2400 mm. Because of its geographical location and environmental conditions, Xishuangbanna is the region with the highest biodiversity in China (Zhang and Cao, 1995). There are five primary forest types: tropical seasonal rain forest, tropical mountain rain forest, evergreen broad-leaved forest, monsoon forest over limestone, and monsoon forest on river banks (Zhang and Cao, 1995).

2.2. Methodology

Forest and rubber plantation cover data in Hainan were obtained from the Hainan statistical yearbooks. Because data for natural forests was available only from 1986 to 1992, we engaged in another round of data collection by extracting forest cover data from published research papers. Relevant peer-reviewed publications were collected from the ISI Web of Science using a combination of search terms ("forest transition^*" OR deforestation OR degradation OR "forest regrowth" OR "land use change" OR "forest change" OR "forest cover", AND Hainan). Data quality, data consistency, and forest definition were checked according to natural forest data collected from the 1986-1992 Hainan statistical yearbooks. In the cross-checking process we first removed the data without reference to natural forests from the collected publications; we next removed data that were obviously fraught with errors (e.g. data in 1993 which reported natural forest cover to be twice that of 1992). Natural forest data were extracted from published research papers only for the period of 1993e1995. Comparing forest data after 1996 with those before 1996 was hampered by a range of difficulties including inconsistent definitions of forest and natural forest, and differences in methodology. Therefore, data for natural forest cover after 1996 were not included in the study.

Xishuangbanna data were collected using the same combination of search terms but different site search terms. From the search results, we found a series of publications with a consistent forest definition and a consistent methodology of obtaining both natural forests and rubber plantation data. Moreover, because in these publications forests were divided into different forest types, it was possible to separate forest cover data for natural forests and that for rubber plantations. Consistent data on both natural forest and rubber plantations in Xishuangbanna was available only from the publications of the Xishuangbanna Tropical Botanical Garden (XTBG).

The term deforestation is used in this research to denote the conversion of natural forests to another land use or the long-term reduction of the natural forest canopy cover to below a minimum threshold of 20 percent. Rubber plantations are the main forest plantation type in tropical China, covering more than one fifth of the total land area, and were therefore selected as a proxy for changes in plantation cover over time. The term turning point is used in the literature on forest transition to refer to the inflection point when overall forest cover changes from net loss to net gain. In this research, we apply the term turning point separately to changes in overall forest cover, changes in natural forest cover, and changes in plantation cover.

A Quadratic Non-linear Regression Model was conducted and visualized using SigmaPlot Version 11 to detect the relationships between time (year) and overall forest cover, natural forest cover, and rubber plantation cover.

3. Results 3.1. Time lag between overall forest and natural forest transitions

In Hainan, we found a time lag between the overall forest cover turning point and the natural forest cover turning point. While the turning point for overall forest transition was around 1978, the turning point for natural forests was in 1988 (Fig. 2). While overall forest cover increased significantly after its turning point in 1978, the loss of natural forest continued, though at a somewhat reduced rate (Fig. 2). Natural forest cover increased significantly after its turning point in 1988 (R²=0.87, p < 0.0001 vs. R²=0.93, p < 0.0001; Fig. 2).

The turning point for overall forest transition in Xishuangbanna, on the other hand, occurred in 1988, i.e. 10 years later than in Hainan (Fig. 3). Natural forest cover in Xishuangbanna has continued to decrease significantly with time (R²=0.97, p < 0.0001; Fig. 3).

3.2. Changes in rubber plantation and natural forest cover

In Hainan, before the natural forest turning point, natural forests cover greatly decreased while the rubber plantations increased (Fig. 2). There was a surge in the expansion of rubber plantations in Hainan after 1988, followed by a period of when both rubber and natural forest cover steadily increased (Fig. 2). In Hainan, from 2002 to 2010 there was an upsurge in overall forest cover (Fig. 2a) with rubber plantation area continuing to increase.

After the turning point for forest transition in 1988, overall forest cover in Xishuangbanna entered a relatively stable state (Fig. 3). From 1992 to 2003, rubber plantations in Xishuangbanna showed a linear increase, while at the same time natural forests showed a linear decrease. From 1999 to 2010, both the expansion of rubber plantations and the decline of natural forests accelerated dramatically. Rubber plantation cover in particular increased significantly over time (R²=0.8782, p < 0.0006; Fig. 3). However, from 2002 to 2010, the pattern of change in overall forest cover in Xishuangbanna differed from that of Hainan. Overall forest cover in Xishuangbanna increased slightly from 2001 to 2005, and decreased slightly after 2005.

4. Discussion

We observed that in tropical China, overall forest transition coexists with continuing loss of natural forests. It is important to note that under the UN Food and Agriculture Organization (FAO) definition of forests (http://www.fao.org/docrep/008/a0400e/a0400e00.htm), which is widely used in China (both in government reports and policy documents), the loss of natural forest cover cannot be recognized. This is because the term "forest" according to the FAO definition is applied to both natural forests and planted forests. Indiscriminate use of the term "forest" disguises the fact that the transition which has occurred in tropical China was a transition from natural forests to plantation forests (Zhai et al., 2015) rather than simply a forest transition from less forest cover to more forest cover (Figs. 2 and 3). Separating the dynamics of natural forest cover and plantation cover from those of overall forest cover will not only reveal these changes, but may also help identify the driving forces behind natural forest loss and displacement by monoculture plantations. Previous research in China and abroad has demonstrated that forest cover dynamics can be misunderstood when overly general definitions of 'forest' are applied (Li et al., 2007; Zhai et al., 2012; Gibbs et al., 2010; Aziz et al., 2010). Simultaneous occurrence of natural forest loss and plantation expansion has also been observed in Chile and Ecuador, where researchers found that the increase in overall forest area cover had negative effects on carbon storage (Hall et al., 2012) because it consisted to a large extent in the conversion of natural forests to plantations. However, despite growing awareness of the shortcomings of current forest definitions (Lambin and Meyfroidt, 2010), research within the context of forest transition theory has so far looked only at overall forest cover and has not differentiated between different forest types such as natural forests and forest plantations.

By classifying natural forests and plantations separately in this research, we were able not only to differentiate the divergent dynamics of these two categories, but also to demonstrate a potential means of identifying conflicts between natural forest and rubber plantations. In previous researches in tropical China which distinguished between forest types, rubber plantations were found to be a major driving force of natural forest loss (Li et al., 2007; Zhai et al., 2012, 2015). We therefore recommend that research on forest transition applies a forest definition that emphasizes the differences between forest types, and at least distinguishes between natural forest and forest plantations. As different forest types also have very different capabilities in regards to delivering ecosystem services (Sasaki and Putz, 2009), such definitional issues have important implications for future reforestation/afforestation and Payments for Ecosystem Services (PES) projects.

Our own findings show that trends in forest cover after turning points are not always consistently upward or downward. While the data used in this paper indicate an uninterrupted net gain in the natural forest cover in Hainan since 1988, other research has shown that natural forests may be at risk of net losses in the future. Natural forest cover on Hainan Island has declined by 5.89% between 1998 and 2008 (Wang et al., 2012). Other research found that from 1991 to 2008, natural forest area on Hainan Island had decreased by 4.38% from 1991 to 2008 (Zhang et al., 2010). In our previous smaller-scale research project in Hainan, carried out in the upper reaches of the Changhua Watershed, we have also observed a trend of declining natural forest cover from 1995 to 2005 (Zhai et al., 2012, 2014). There is therefore a distinct possibility that the trend of increasing natural forest cover has reversed in Hainan after 1995. This, however, needs to be supported by more time-series data on natural forests.

While in Xishuangbanna, overall forest cover has already entered its transition stage (Xu et al., 2007; Zhai et al., 2015), loss of natural forests continues (Zhai et al., 2015) and it is unclear when this trend will reverse. It is important to note that our figures for overall forest cover after the turning points may underestimate the true amount, as forest plantations types other than rubber were not included in our calculations. The area covered by pulp plantations, for example, has increased by 5% from 1998 to 2008 on Hainan (Wang et al., 2012). Plantations other than rubber were not included due to a lack of data and because of the fact that these plantations account for only a small proportion of the total land area (Li et al., 2009).

We found that during the Sloping Land Conversion Program (SLCP) period (2002-2010) overall forest cover increased, while both the loss of natural forest and the expansion of rubber plantations accelerated (Figs. 2 and 3). The rapid expansion of rubber plantations can be explained to some extent by policy interventions. The SLCP was initially launched in 1999 by the Chinese government as a PES project in response to the 1998 floods in the Yangtze River watershed. The Program aimed at restoring natural ecosystems and redressing environmental degradation caused by cropland expansion and deforestation on steep slopes (Liu et al., 2008). It encouraged farmers to convert croplands on steep slopes to forest and grassland by providing them with grain and cash subsidies, and farmers in tropical China planted rubber trees that were considered as forests. The coincidence of deforestation and rubber plantation expansion was also found in our previous work (Zhai et al., 2015), while different patterns of change in overall forest cover during the SLCP period were detected for the first time. The different patterns of change in overall forest cover during the SLCP period, and especially the decrease of both natural forests and overall forest cover is ecologically significant. However, the explanation of this issue is beyond the scope of this research. The question of whether or not the SLCP accelerated deforestation and the expansion of rubber plantations in Xishuangbanna still requires further research. It is also unclear whether the coincidence of deforestation and the expansion of plantations in forest transition has occurred in other regions of the world.

Gaining a better understanding of these issues is particularly important when implementing large-scale PES projects and REDD₊ projects. For example, the design of REDD₊ projects which aim to increase carbon storage by afforestation and reforestation should take into account the dynamics of natural and plantation forest transition, and not simply changes in overall forest cover. Similarly, PES projects are unlikely to successfully safeguard or boost the provision of ecosystem services if natural forests are replaced by plantations or displacing ecosystem services in other regions. The displacement could facilitate a region or country forest transition through a displacement of land use in another region or country. While leakage, a form of land use displacement, refers to improving land use sustainability or ecosystem services (e.g. REDD₊) in a region or country through a displacement of the environmental or ecosystem in another region or country (Meyfroidt and Lambin, 2008; Meyfroidt et al., 2013). Future studies of forest transitions, therefore, should assess both overall forest cover, natural forest cover, and other forest types including plantations. Future research into forest transition should also closely examine ecosystem services and the potential for leakage and displacement effects during forest transition.

5. Conclusions

We found that loss of natural forest cover continues after the turning point of forest transition in tropical China, and that this loss is obscured by use of indiscriminate definitions of forest in forest transition theory. However, in order to prove that current concepts of forest transition fail to adequately distinguish between forest types, we need more time-series data of different forest types collected under common methodologies and definitions. Our research shows that forest transition in tropical China has mainly been caused by the rapid and large-scale expansion of plantations. This is of relevance for future reforestation and afforestation programs such as PES and REDD ₊ projects; we therefore recommend that the design of such projects take into account the differences in transition dynamics between forest types in order to avoid increasing tree cover while losing natural forests.

Acknowledgements

This work is funded by the National Natural Science Foundation of China (Grant 31300403) and the China Postdoctoral Science Foundation (Grant 2013M540722). This research is part of the CGIAR Research Program 6: Forests, Trees, and Agroforestry. The research is also part of the BMZ/GIZ "Green Rubber" (Project No. 13.1432.7-001.00) and 'SURUMER' (Project No. 01LL0919A) funded by the German Federal Ministry of Education and Research (BMBF) under Grant number FKZ 01LL0919. The authors are very grateful to Dr. Chen Si-Chong from the University of New South Wales for providing comments and suggestions on the methods.