Global protected areas boost the carbon sequestration capacity: Evidences from econometric causal analysis

Highlights

• The CDI can be improved by 0.39% by construction global PAs.

• Regional differences with CDI improvement globally remain.

• Upgrading PAs in Asia and South America can contribute to an increase of CDI.

• Strictly implementing planning is conducive to improving CDI.

• Enhancing the power of local governments is conducive to improving CDI.

Abstract

Carbon sequestration capacity is the key factor in maintaining biodiversity and ecosystem services. However, further research is required on how to evaluate the impact of protected areas on carbon sequestration capacity from a global scale. To date, we propose a carbon density index of global protected areas (>10 km², 32,756 samples) by the Integrated Valuation of Ecosystem Services and Trad'eoffs carbon model over the period 1994–2015. Then, we use the propensity score matching and difference-in-difference methods to separate the time effect and policy effect of the construction of protected areas on carbon sequestration capacity. Our analysis reveals that the carbon sequestration capacity can be improved by 0.39% by constructing global protected areas. There are regional differences with carbon sequestration capacity improvement globally. Africa has the largest value of increased carbon sequestration capacity, followed by Asia, Oceania and Europe. Upgrading protected areas (0.05%), strictly implementing planning (0.18%) and enhancing the power of local governments (0.08%) are conducive to improving carbon sequestration capacity. The assessment of the carbon sequestration capacity dynamic with protected areas is of great significance to meet the Convention on Biological Diversity.

Introduction

Wild inhabitants and species are disappearing rapidly, exacerbating the sixth mass extinction of species worldwide (Barnosky et al., 2011; Pimm et al., 2014). To curb this trend, legally protected areas (PAs) have been set up around the world (Watson et al., 2014). Although the area of existing global PAs has more than tripled in the last 40 years, accounting for 15.4% of the world's terrestrial area, it still falls behind the target of the Convention on Biological Diversity which requires at least 17% by 2020 (Juffe-Bignoli et al., 2014). In addition to expanding the area and increasing the number of PAs, human beings also need to pay attention to the quality of protection of ecosystem services and biodiversity in PAs. Stocking ecosystem carbon is one of the ecosystem services of PAs (Ouyang et al., 2016). Ecosystem carbon stocks in terrestrial PAs account for approximately 20% of all terrestrial ecosystem carbon stocks (Melillo et al., 2015). Small changes in carbon stocks can affect the concentration of CO₂ in the atmosphere, thereby affecting the distribution, composition, and function of terrestrial ecosystems (Schimel et al., 2000). To determine whether the protection of biodiversity and the ecosystem service of PAs have achieved the desired purpose, it is necessary to study PAs carbon sequestration capacity (Andam et al., 2008; Hill et al., 2015).

The conclusion that PAs improve carbon sequestration is controversial due to the size of the samples and methods (Table 1). Some studies showed that although there is less carbon loss inside PAs than that outside PAs, carbon stocks of PAs continue to decrease such as in Humid Tropical Forests (Scharlemann et al., 2010), Brazilian Amazonia (Nogueira et al., 2018), and Czech Republic (Vackar et al., 2016). Other researchers showed that PAs are helpful to increase carbon stocks, such as in Uganda (Gizachew et al., 2018), Brazilian Amazon (Nolte et al., 2013), and United States (Lu et al., 2018). Although sometimes the same samples are selected, the results are difference due to different research methods. For example, in the Brazilian Amazon region, Nogueira et al. (2018) suggested that carbon lost in PAs is 0.787 (2.3%) Pg C in 2014 by using the comparison method inside and outside of PAs. In contrast, Nolte et al. (2013) believed that PAs can avoid deforestation and carbon loss by propensity score matching (PSM). In the humid tropical region, Scharlemann et al. (2010) confirmed that the loss of carbon inside PAs is about half of unprotected areas. Collins and Mitchard (2017) also believed that the PAs are losing forest at a mean rate of 0.18% year⁻¹ in 2018 PAs in tropical countries by remote sensing. But Bebber and Butt (2017) suggested that tropical PAs reduced deforestation carbon emissions by around 29% by PSM model.

Although some studies have evaluated the carbon sequestration capacity of PAs, gaps in the literature remain. First, there is a general lack of effective methods to separate the “time effect” and “policy effect” because the impact of construction of PAs on the carbon sequestration capacity is two-fold. One is the so-called “time effect”, which is generated by natural growth or change, and the other is the “policy effect”, which is mainly affected by the construction of PAs (Blundell and Dias, 2000). From the existing literature, the main methods to estimate the carbon sequestration capacity of PAs are the comparison method inside and outside of PAs, remote sensing, and PSM. It is difficult to solve the sample selection deviation with the comparison method inside and outside of PAs because there are geographical and climatic differences inside and outside of PAs (Bebber and Butt, 2017; Bowker et al., 2017). PSM method can eliminate the sample selection deviation by using a series of control variables to keep the same trend and characteristics of the treated group and the control group, without losing a large number of observations (Melissa et al., 2014; Nolte et al., 2013). However, the disadvantage of PSM is that it only considers the observational control variables and ignores the hidden bias of the potential variables. In addition, PSM and remote sensing models cannot solve the endogenous problems of the samples themselves, such as the separation of “time effect” and “policy effect” (Weisglaskuperus et al., 2009). The DID model can solve the endogenous problem of samples by using the twice differentials to obtain the “policy effect”. Therefore, this paper adopted PSM-DID model to evaluate the impact of the construction of global PAs on carbon sequestration capacity. Second, it still remains the lack of assessing the carbon sequestration capacity of PAs from a global scale. One of the main reasons is that there are only few remote sensing products of terrestrial ecosystem carbon stocks at the global scale, as it is difficult to estimate the cost of global carbon stocks through site measurements (Table S1) (Liu et al., 2015). In addition, the existing data of global carbon stocks are generally at a resolution of 0.25° by remote sensing, which is much larger than that required by the PAs. Therefore, the carbon sequestration efficiency of global PAs has barely been assessed due to the low resolution of carbon storage (Juffe-Bignoli et al., 2014). In recent years, the Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) has become a popular method to simulate directly the impact of land use changes on terrestrial ecosystem carbon stocks (Tallis et al., 2013). To address the low resolution of global carbon stocks products, we selected InVEST model in this paper as the tool to calculate the changes in terrestrial ecosystem carbon stocks caused by land use in different periods with a spatial resolution of 1 km.

The innovation of this paper is twofold. First, we adopted InVEST model in this paper to estimate the carbon stocks of global PAs with a spatial resolution of 1 km, which solved the problem of low resolution of global carbon stocks and made it possible to assess the carbon sequestration capacity of global PAs. Second, this paper used the PSM-DID model to solve the sample selection deviation and endogenous problem, and to separate the “policy effect” of PAs. To conduct this study, after this introduction, we structured this paper as follows. Section 2 presents the methodology and data and defines the related variables. In 3 Results, 4 Discussion, 5 Conclusion and implications, the empirical results, discussion, and conclusion and implications are presented, respectively.