Valuing Carbon Sequestration to Finance A ﬀ orestation Projects in China

: A ﬀ orestation programs have huge potential to store carbon, thereby contributing to mitigate climate change. However, the voluntary acceptance by landowners crucially depends on their economic outcome. We (i) assess the carbon sequestration potential of a ﬀ orestation projects in various Chinese regions by comparing di ﬀ erent tree species, project durations and regional particularities, (ii) analyze the costs and beneﬁts of tree species used for timber and fruit production as well as bamboo, and (iii) compare them with alternative crops under di ﬀ erent climatic and economic conditions. Finally, we calculate the minimum compensation required by farmers to convert cropland to forests assuming a joint production of timber or fruits and carbon sequestration. No compensation is needed when cropland with relatively low revenues is to be converted. In contrast, compensation payments must be high for converting land used for cash crops, such as sugarcane. For fruit trees, the amount of carbon sequestered is low, but selling fruits and nuts is quite proﬁtable. Consequently, in most cases there is no compensation needed. The minimum price per carbon credit decreases with increasing project duration because more carbon is stored per hectare, and in consequence, the required payments per credit are lower. This does not hold for fast-growing trees like eucalyptus and poplar, for which the minimum price increases with extended project duration. Bamboo shows a high carbon sequestration potential, and becomes economically more attractive by including carbon revenues. Carbon credits can often compensate the opportunity costs of alternative land uses, except for a ﬀ orestation on highly productive cropland. We demonstrate that the economic attractiveness of a ﬀ orestation projects is strongly context dependent and varies substantially across regions. The ﬁndings can contribute to the cost-e ﬃ cient design of carbon sequestration projects. The methodology can be applied to other regions in the developing world.


Introduction
Forests have great potential to store carbon, which contributes to reduce atmospheric carbon dioxide concentrations [1]. Globally, forestland has the capacity to sequester 2.4 ± 0.4 10 9 Mg of carbon emissions annually, equivalent to 25% of the global carbon emissions from fossil fuels used in 2018 [2,3]. A recent study has found an even higher potential and identified China as one of the most appropriate countries for afforestation efforts [4].
Economic incentives play an important role in the success of afforestation projects [5,6], especially in the beginning because it takes a number of years for newly planted forests to generate revenues may result in a significant variation of revenues from tree products. Finally, it is necessary to take the specific conditions in different regions into account. Accordingly, we address three research questions: (i) To what extent can carbon be sequestered both in tree biomass and soils through afforestation? (ii) Is it economically attractive to convert cropland into forest considering different tree species and crop types? (iii) What is the minimum price per ton of CO 2 to compensate the opportunity costs of farmers when switching from alternative land uses to forestry?
To answer these research questions, we first examine the carbon sequestration potential in biomass and soil of different plant species by afforestation in several Chinese regions. Then we compare costs and benefits of alternative land uses and calculate the minimum compensation required. Finally, we compare the economic attractiveness of afforestation projects under different conditions. Considering uncertainties about future economic conditions and developments, we conduct a sensitivity analysis to evaluate the changes of minimum prices per ton of CO 2 .

Study Area
According to Fang's classification [23], China is divided into six geographical regions: Northeast China, North China, East China, Central South China, Southwest China, and Northwest China ( Figure 1). The south and east of China mainly belong to the Tropics and Subtropics, while the north is a region with temperate climate. In order to cover different plant types and make the results applicable to different regions, we chose the most common species planted across five different regions except for Northeast China. The most common species used for afforestation in Northeast China is Larix, which is a deciduous softwood species. However, due to insufficient information on the soil organic carbon (SOC) stock changes after afforestation with deciduous softwood, we excluded Northeast China from this study. The analysis provides important information to classify carbon sequestration by trees based on different climatic and economic contexts.

123
The final dataset includes three ecological trees (eucalyptus, Chinese fir and poplar) and five 124 economic trees (apple trees in the north, orange and mango trees in the south and walnut and The final dataset includes three ecological trees (eucalyptus, Chinese fir and poplar) and five economic trees (apple trees in the north, orange and mango trees in the south and walnut and chestnut trees). Note that different species also represent different tree types. Poplar belongs to deciduous hardwood, eucalyptus is an evergreen hardwood, Chinese fir is an evergreen softwood, and economic trees are all deciduous hardwood. This classification not only allows us to further explore the effects of different values of tree products (timber or fruit) on the net income of afforestation projects but also to take the significant differences of carbon sequestration among various tree types into consideration.
To compare the income obtained from afforested land with income from cropland, we selected villages with converted land or cropland (most villages have both). The selected cropland should have similar characteristics as the converted land with respect to soil quality, water availability, slope, and other biotic and abiotic factors.

Estimation of Carbon Sequestration
We estimated carbon sequestration in tree biomass (above-and below-ground) and changes in SOC stocks of afforestation projects on former cropland, focusing on ecological trees, economic trees and bamboo. The forest floor was not considered in this study, because due to management practices, there was little forest litter on the ground.

Carbon Sequestration by Ecological Trees
In the present study, we assessed the carbon sequestration capacity of three different ecological trees. We used the stand growth models developed in the literature to estimate carbon sequestration in forest biomass (Table 1). The models were developed for specific species and at regional scales, which allowed us to make respective comparisons. For carbon sequestration in soils, we considered the top 60 cm soil layer, which contains around 70%-80% of the first-meter SOC stocks, and is mostly affected by land-use change [25,26]. Several studies have shown the significant effects of tree types and stand ages on SOC sequestration capacities (e.g. [27,28]). However, these studies usually divide trees into hardwood and softwood, broadleaf and conifers or evergreen and deciduous [27,28], which may result in some overlap. For instance, species in genera such as eucalyptus could be viewed as hardwood and evergreen. In our study, we suggest a classification as (i) deciduous hardwood, (ii) evergreen hardwood and (iii) evergreen softwood. We used the results of our own meta-analysis based on world-wide data [29]. Table 2 shows how SOC sequestration changes after afforestation with deciduous hardwood, evergreen hardwood or evergreen softwood on cropland at different stand ages. SOC is continuously sequestered after afforestation with deciduous hardwood. Higher sequestration rates are observed after the first decade (Table 2). For evergreen plants, hard-as well as softwood experience a loss of SOC stock in the first two decades, while afterwards evergreen hardwood sequesters more SOC than the other two tree types. In general, they achieve the greatest SOC sequestration before the fourth decade, with some variation for deciduous hardwood.

Carbon Sequestration by Economic Trees
Wu et al. (2012) estimated the carbon sequestration capacity of apple trees at various ages in China's major apple regions and found the net carbon uptake is −7 (source until year 5), +19 (until year 18), and +17 Mg ha −1 (until year 22) [30]. They conclude that apple orchards older than 8 years can be considered as a carbon sink. At the age of about 18, they reach the peak of carbon sequestration capability, which then begins to decline. The relationship between tree age (X age ) and C sequestration (C seq ) can be expressed as Equation (1): For South China, Ren et al. (2014) estimated that the carbon accumulation in the biomass of orchards in Hainan was 21.8 Mg ha −1 at a stand age of 15, which is applied in the present study for tropical fruit trees such as mango [31]. For carbon sequestration of walnut trees in North China, we applied the findings from Shen et al. (2017), who calculated a biomass carbon sequestration of 10-year old walnut trees of 8.749 Mg ha −1 [32]. For chestnut trees, Peng and Wang (1998) determined the respective dry biomass for different trees densities [33]. In our study, we apply the results of lowest density since this is consistent with the density farmers usually plant: 11.6 (for tree age 1-4), 18.12 (for age 5-6), and 28.88 Mg ha −1 (for more than 7 years).
For SOC sequestration of these three tree types, we used the data of deciduous hardwood (compare Table 2), because economic trees normally belong to this type of wood.

Carbon Sequestration by Bamboo
Bamboo forest is one of the most important forest types in the world, and is mainly distributed in the tropical and subtropical regions [34]. In China, bamboo forests are mainly found in the south adding up to 6 million ha and accounting for 3% of the total forest area [35]. Moso bamboo (Phyllostachys pubescens Mazel ex Houz.) forests cover 4.43 million ha (74% of all bamboo forests). According to Zhang et al. (2014), the estimated annual carbon sequestration in the biomass of moso bamboo forests in southern China is 6.74 Mg ha −1 [34]. In our study, we apply this annual quantity of carbon accumulation because (i) the estimation was based on regional scale data from South China, which is the area under consideration in our study; (ii) the estimated overall carbon storage in the study area was 40.45 Mg ha −1 , which is consistent with previous studies [36]; and (iii) it reflects the regional management practices, where annually 1/6 of older culms from the moso bamboo forest are harvested by selective cutting.
For SOC sequestration of bamboo, we applied the results of evergreen hardwood species (compare Table 2), because bamboo (i) retains green leaves throughout the year; (ii) has flat leaves rather than needle-like leaves; and (iii) grows in high density.

Temporal Patterns of Carbon Sequestration
To obtain information on the changes of carbon sequestration with stand ages, we performed regression analyses to develop the functions expressing the temporal patterns of carbon sequestration for each species. All statistical analyses were performed using SPSS, ver. 25.0 (SPSS Inc., Chicago, IL, USA).

Economic Analysis
Our economic analysis focuses on the net income of afforestation activities and the opportunity costs of alternative land uses. We assess afforestation alternatives with different tree species, project durations and crop types. Then, we determine the costs of carbon sequestration for different scenarios.

Collection of Economic Data
Economic data were gathered through interviews with farmers in different regions, where we collected soil and biomass data in parallel. The interviews were conducted from July to August 2018 and April to May 2019. First, we selected a region in terms of (i) abundance of the tree or crop species we had selected, and (ii) physical accessibility. In the region, we used two approaches. The first was to randomly select villages, and within the villages randomly select interviewees to carry out door-to-door interviews. However, this procedure was not successful because of difficulties to find farmers who had planted our selected tree or crop species. The alternative approach was to ask the village committee to provide us with the phone numbers of farmers who had planted our selected trees or crop species, and we randomly select households from the list for face-to-face or phone interviews. We met and called more than 200 people, and 75 provided information that we considered reliable (five for each species).
We first asked farmers for the tree species they planted, the planting density and the stand age. The stands had to fulfill the following requirements: (i) planted as monocultures, (ii) managed to assure a survival rate of at least 70%, and (iii) no large-scale cutting. If these criteria were satisfied, we asked farmers, which costs and benefits they had when planting trees. The costs include (i) site preparation (preparation of the land and planting, seedlings or sapling); (ii) management (insecticides, herbicides, fertilizer, manual cleaning and weeding, pruning and thinning; replanting); (iii) harvest and (iv) tools and machinery. The benefits comprise (i) income from timber/wood use and/or (ii) sale of non-timber forest products (e.g. fruits, mushrooms, medicinal plants). To obtain an overview of the overall lifecycles, we collected data from farms with different stand ages. The information of economic costs and benefits of farmland were also collected through interviews with farmers who have participated in the afforestation program but still have cropland to cultivate as well as non-participants. These costs comprise site preparation, management, harvest and tools, while benefits stem from selling the harvested crops.
Local market prices are used to value traded forest and agricultural products. In case marketable products are not sold but consumed by the farmers, the market price is used for valuation, too, since directly consumed products increase the standard of living of farmers similar to the income generated by selling products. For each item, labor costs were considered. Since the labor price varied substantially among different regions, we used the local labor costs in each region.

Net Income Comparison of Forest and Cropland
A comparison of costs and benefits of the different land-use types was made to assess the attractiveness of afforestation investments. We applied the net present value (NPV) criterion shown in Equations (2) and (3).
Benefits (B) and costs (C) occur at different times (t) and are to be discounted (d = discount rate) in order to make them comparable: We assume that a profit-maximizing and risk-neutral landowner would switch from crop farming to afforestation if the NPV of the afforestation (NPV A ) is higher than or at least as high as the NPV of crops (NPV C ).
If the NPV A is lower than NPV C , we can assess whether the revenues of carbon sequestration (as a reward for the service provided by the landowner) could compensate for the net loss of switching to forestry [37].
PV CR is the present value of carbon revenues required to make forest an economically attractive alternative. These revenues occur, e.g., if Certified Emission Reduction units (CERs) are issued and sold on a market. As a result, the landowner might have an economic incentive to convert cropland to forest with the aim to jointly produce timber and carbon credits.
For discounting, we applied an interest rate of 4%, which is the average interest rate of 3-years Chinese national bonds in recent years (see https://www.chinabond.com.cn/). Because of uncertainties about future developments, we conducted a sensitivity analysis applying different interest rates, prices and rotation lengths.

Afforestation Scenarios
According to the UNFCCC (2013) rules, a CDM afforestation project can choose to prepare its first monitoring report (for the 'initial verification and certification') at any time [38]. It could be 10 years for a slow-growing forest, or e.g., five years for a fast-growing plantation. However, after the initial verification and certification of carbon stocks within the project boundary, verification and certification must be carried out every five years until the end of the crediting period. Therefore, forest projects starting in 2019 would have their carbon sequestration verified in 2024 or 2029, and the temporary CER (tCER) would expire in 2029 and 2034, respectively. Given these assumptions, tCER would always be valid for at least five years. Projects can have (i) a single fixed crediting period of up to thirty years; (ii) a renewable crediting period of up to twenty years, which can be renewed at most twice. Therefore, the maximum length of the crediting period of a project can be sixty years, if successfully renewed twice [38]. The costs of certification include project design and validation, carbon accounting and monitoring. Since no information was available about the exact certification costs in China, we neglected these costs in the main analysis and additionally conducted a sensitivity analysis to assess how certification costs can influence the minimum prices per tCER (see Section 3.3) Wang et al. (2007) analyzed two afforestation scenarios with different lengths of the evaluation period [16]. The first coincided with the national subsidy period for ecological trees in China (8 years); the second with the rotation period of 40 years for Larch, the tree species they studied. However, national subsidies will no longer play an important role in the determination of the evaluation period because (i) they are fading out from 2015 on (with a respective delay in areas where it started later) [21] and (ii) in the new period of the afforestation program, farmers can choose to plant ecological or economic trees with the same level of subsidies. In our study, a 30-year evaluation period was chosen because tree growth rates and SOC sequestration rates would decline afterwards (Tables 1 and 2). Thus, we assumed a project duration of 30 years with carbon credits verified and issued every five years. The PV CR of a 30-year project is determined by equation (7) [12,39]: (1 + d) 10 + · · · + ptCER 25 × C 25 (1 + d) 25 (7) ptCER t is the price for one tCER at time t, C t is the net amount of CO 2 stored at time t, and d is the discount rate. Carbon remaining in products made from harvested timber is not considered.
The minimum price per ton of CO 2 (ptCER) to make farmers switch from crop production to forestry can be calculated by the following equation [12]: In this study, we only conducted the economic evaluation for one rotation with different project durations because carbon sequestration data for successive rotations were not available. On the one hand, we determined the economic benefits of the same species across different regions since the carbon sequestration potential varies significantly among them. On the other hand, we compared conversion of cropland to various forest alternatives within the same region because the species commonly planted differ among regions. Furthermore, we compared the PV CR of a 30-year rotation with various rotation periods typically applied by farmers. For example, farmers who plant eucalyptus usually harvest it after five years, but this is too short to generate carbon credits. Therefore, we analyzed whether it would be more beneficial if timber is harvested after 10, 20, and 30 years, respectively.  ptCERt is the price for one tCER at time t, Ct is the net amount of CO2 stored at time t, and d is the 296 discount rate. Carbon remaining in products made from harvested timber is not considered.

297
The minimum price per ton of CO2 (ptCER) to make farmers switch from crop production to 298 forestry can be calculated by the following equation [12]: In this study, we only conducted the economic evaluation for one rotation with different project

314
To assess the timber production for a longer project duration, we applied the allometric growth 315 models of stand volumes from literature ( Table 3). Note that the price per m 3 of timber could 316 increase with diameter and stand age. Therefore, we asked farmers about the price of timber 317 stemming from bigger trees. However, farmers had hardly any information since they rarely harvest 318 and sell such timber. Thus, we had to refer to recent market prices reported at 319 http://www.chinatimber.org/. Note: V and A represent timber production (m 3 ha −1 ) and stand age (year), respectively. To assess the timber production for a longer project duration, we applied the allometric growth models of stand volumes from literature ( Table 3). Note that the price per m 3 of timber could increase with diameter and stand age. Therefore, we asked farmers about the price of timber stemming from bigger trees. However, farmers had hardly any information since they rarely harvest and sell such timber. Thus, we had to refer to recent market prices reported at http://www.chinatimber.org/. Table 3. Timber production (m 3 ha −1 ) by different tree species estimated by allometric growth models.

Species
Growth Note: V and A represent timber production (m 3 ha −1 ) and stand age (year), respectively.

Carbon Sequestration by Afforestation
The carbon sequestration potential differs among tree species and stand age as well as across regions (Table 4). In warmer and more humid climates, as in the southern and eastern regions, trees sequester carbon faster than in other regions due to the higher initial growth rate in a tropical climate. The temporal patterns of carbon sequestration including both tree biomass and soils by afforestation are shown in Figure S1 in Supplementary. Carbon sequestration rates of ecological trees achieve their peaks 25-30 years after planting, except poplar in E, S, and SW of China, where it accumulates most carbon before the age of 20. In contrast, fruit trees sequester less carbon because of the annual harvest of fruit products [30]. Fruit trees in the northern region are a net carbon source in the initial 8 years and after 30 years. For southern fruit trees and walnut trees, the rates of carbon sequestration peak at the age of 20, i.e., five years earlier than ecological trees. Bamboo shows a higher carbon sequestration potential than other species except eucalyptus (Table 4). It sequesters even more additional carbon in older stands ( Figure S1).

Economic Evaluation of Converting Cropland to Forest
We compare the economic results of afforestation alternatives and cropland. Table 5 shows the NPV changes of various tree species and crop types with different project durations. In many cases, afforestation generates a higher net income than cropland. However, in most situations two-harvest corn or one-harvest sugarcane generate a higher NPV than afforestation alternatives (Table 5). Two-harvest corn grown in the tropical area generates an NPV (ranging from 6731 to 26,146 USD ha −1 for five years and 30 years, respectively) more than two times higher than that of one-harvest corn (ranging from 1093 to 3661 USD ha −1 for five years and 30 years, respectively), due to higher yields and selling prices. In China, sugarcane is a cash crop with high economic returns. This leads to high opportunity costs of converting cash crops to forests. The NPV A of ecological trees and bamboo are mainly based on the income from timber products, depending on the growth rate of timber volume. In the first decades, eucalyptus, as a fast-growing tree species, generates a NPV A (7365 USD ha −1 ) higher than crops, such as one annual harvest of wheat (2956 USD ha −1 ) or corn (1991 USD ha −1 ), or two harvests of rice (2606 USD ha −1 ). After 20 years, growth rates decline, which results in a decreasing present value of net revenues from timber. In contrast, bamboo grows fast during the whole study period because it is self-regenerating. Therefore, afforestation with bamboo produces a higher NPV than that of one-harvest corn after the first decade (Table 5). However, the lower price of bamboo makes it economically less attractive compared to other afforestation alternatives. For economic trees, the NPV is often higher than that of crops because selling fruits is much more profitable than selling grains. The result holds even when taking into account the relatively low revenues of fruit trees during the first 5-10 years. This is due to the fact that fruit trees normally need 7-8 years to achieve their full fruit-producing potential. In our study, we assume a constant fruit yield once the trees have achieved this potential. This might lead to a possible overestimation of the revenues, if fruit trees tend to produce less fruits after a few decades [30]. Notes: normal means the common rotation period applied by farmers. One-or two-harvest means the crops are harvested once or twice annually, respectively. * We assume 1/6 selective harvest after the first 3 years, clear-cut at the end of each rotation period. 1 RMB = 0.15 USD. Shaded areas indicate opportunity costs are higher than net income of afforestation.
It should be noted that crop production in China is often realized by small-scale farming. Large-scale farming requires bigger machines and investments. In addition, Chinese small-scale farmers can sell their harvest on the market at higher prices without using middlemen. Taking sugarcane as an example, the price when selling to individuals is three times higher than the selling price to a sugar company. Table 6 shows the amount of tCERs generated by each alternative. The number of tCERs is determined by the specific tree species, regions and project durations. Eucalyptus produces the highest amount of tCERs during the whole study period with 1519.8 Mg −1 ha CO 2 sequestered (Table 6). This is because it is fast-growing and it is mostly planted in tropical areas, resulting in higher tree biomass. However, it should be noted that there are no tCERs generated, if the project duration is too short. This is because carbon credits could only be verified for at least five years and they will expire if the trees are harvested during the crediting period, in this case within the 10 years since the trees have been planted [38]. In other words, if farmers plant eucalyptus only for five years and then harvest it, no tCER will be issued. Therefore, the minimum project duration to get tCERs is 10 years. For Chinese fir, carbon accumulation is lower than that of eucalyptus due to its lower growth rate. Poplar produces significantly more tCERs in E, S, and SW regions compared with N, NE, and NW regions because of the favorable climatic conditions in eastern and southern regions ( Table 6). For economic trees, carbon accumulation and consequently the number of tCERs is lower. The regional impact becomes obvious with fruit trees in the southern region producing more tCERs than in the northern parts. Nut trees generate more tCERs than fruit trees because of their higher tree biomass (Table 6). Bamboo produces the third highest number of tCERs, which indicates a high potential to sequester carbon.
The minimum present value of carbon revenues (PV CR ) required to convince farmers to plant trees as calculated according to Equation 6 are given in Table 6. The blank area indicates that no compensation is needed for forest alternatives because of their higher NPV A ( Table 5). The total compensation to farmers strongly depends on the opportunity costs of cropland. For two-harvest corn and one-harvest sugarcane, the payments would be substantial because they are more profitable than forests (Table 5). However, for farmers cultivating other crop types such as one-harvest wheat, one-harvest corn or two-harvest rice, it could be more attractive to plant trees.
We also calculated the minimum price per tCER for different afforestation alternatives based on various crop types according to Equation 8 (Table 6). For ecological trees, the minimum price per tCER for converting crops with high NPVs is highest with 899 USD Mg −1 CO 2 , e.g., for converting sugarcane crops to forest with Chinese fir of 10-year rotation period. The minimum price per tCER is highly affected by the amount of tCER generated by forest alternatives. This price normally decreases with longer project duration because of the higher amount of tCERs generated per hectare. However, for afforestation with eucalyptus on land used for one-harvest wheat, one-harvest corn, and two-harvest rice cropland, the minimum price per tCER of projects with a longer duration (30 years) increases from 0 to 2.76, 1.04, and 2.28 USD Mg −1 CO 2 compared to projects with a duration of 10 years, respectively. This is due to the variation of discounted timber revenues, which are highest with 7365 USD ha −1 for 10 years, and then decrease steeply to 2052 USD ha −1 until year 30 (Table 5) based on decreasing growth rates and discounting effects. The highest minimum prices per tCER are reported for afforestation on cropland with two-harvest corn and one-harvest sugarcane ( Table 6). This is a consequence of the high revenue loss when switching from highly profitable crops to forestry. However, since cash crops such as sugarcane are very labor intensive, households with little labor available may prefer to grow trees even if they earn less.
Note that the analysis is based on current timber prices and that we assume farmers can decide when and how much to harvest. However, in case harvesting is not permitted or severely curtailed due to strict cutting quota, farmers can hardly get revenues from selling timber after the end of national subsidies in the afforestation program. The results show the NPV A of eucalyptus in the first 10 to 20 years is sufficiently high (ranging from 4934 to 7365 USD ha −1 ) to cover the opportunity costs of crop types with a relatively low NPV (ranging from 1991 to 4367 USD ha −1 ) ( Table 5). For longer periods, especially after 20 years, the required compensation to keep the trees increases ( Table 6). The minimum ptCER reaches 2.76 USD Mg −1 CO 2 to convince farmers not to harvest the trees until 30 years.
Generally, it is more profitable to plant economic trees, especially apple and nut trees, and there is no need for compensation for any of these forest alternatives (Table 6). Only farmers who plant orange trees on former sugarcane land need payments. To compensate the opportunity costs of changing the land use, for the first 10 years such farmers need about 767 USD Mg −1 CO 2 because of (i) the low amount of tCERs generated; and (ii) the low yield and revenues of fruits trees in the initial years. The minimum price per tCER for afforestation with mango trees varies considerably, from 94 to 2680 USD Mg −1 CO 2 , representing the significant effects of alternative crop types and the number of tCERs generated.
Similarly, the minimum price per tCER of bamboo varies from 0 to 927 USD Mg −1 CO 2 depending on the project duration (Table 6), but there is an obvious trend that the compensation can be reduced with longer rotation periods. This is due to the rapid increase of tCERs based on the increasing growth rate of bamboo clumps during the overall study period ( Figure S1).
The results reveal the significant impact of the project duration and tree species on the minimum price per tCER. In general, the minimum price decreases with the increase of the project duration because more tCERs are generated in later years, which is consistent with the findings of De Koning et al. (2005) [12]. However, the situation changes for fast-growing tree species like eucalyptus and poplar, with an minimum price per tCER even higher for older stands. This can be explained by the fast growth rate during the first 10 to 20 years. The present value of timber revenues per hectare for 10-year eucalyptus and 20-year poplar is significantly higher than that of shorter and longer periods.
The minimum price per tCER varies hugely from 0 to 2680 USD Mg −1 CO 2 depending on the afforestation alternatives (Table 6). Our results show a greater variation than that found by Phan et al. (2014) [19], most likely because of the different opportunity costs of cropland alternatives considered in our study. There are three kinds of costs of the afforestation projects, namely opportunity, implementation, and transaction costs. Opportunity costs are considered the most important factor, which on average can be up to three times higher than the transaction and implementation costs [43]. The average carbon price in China's national carbon market is expected to be 8 USD Mg −1 CO 2 in 2020 and 15 USD Mg −1 CO 2 in 2025 [44], which can cover most of the costs per tCER for afforestation alternatives, except afforesting crop land that generates high grain revenues (two-harvest corn) or cash crops (one-harvest sugarcane).
This study determined the minimum price per tCER to make ecological and economic trees as well as bamboo forest economically attractive. Although economic trees are not officially included in the CDM afforestation framework, we assumed that afforestation with economic trees can also generate tCERs because (i) they indeed can contribute to carbon sequestration [30][31][32][33]; and (ii) afforestation projects in some countries like China include economic trees [21]. Therefore, our study followed the tCER rules to provide new insights and give policy recommendations under the condition that economic trees fulfil the respective requirements. We found fruit orchards to be a cost-efficient way to sequester carbon, especially nut trees since a considerable amount of carbon is stored, and there is no need for compensating the farmers. Bamboo also has a high potential to sequester carbon, even higher than many ecological tree species. However, the initial payments required for carbon sequestration can be high (varying from 2.31 to 13.62 USD Mg −1 CO 2 for replacing even economically less attractive crops) due to the low revenues from bamboo harvesting ( Table 6). Switching to a longer rotation period of about 30 years reduced the compensation requirements to e.g., 0.54 USD Mg −1 CO 2 for replacing two-harvest rice land use (Table 6). Table 6. Amount of temporary Certified Emission Reductions (tCER) (Mg ha −1 ) generated, present value of carbon revenues (PV CR ) (USD ha −1 ) required to make afforestation economically attractive, and minimum price per tCER (USD Mg −1 CO 2 ) for different afforestation alternatives.   Note: C, E, N, NE, NW, S, and SW are abbreviations for China, and the eastern, northern, northeastern, northwestern, southern, and southwestern region; normal means the common rotation period applied by farmers. One-or two-harvest means the crops are harvested once or twice annually, respectively. Blank area indicates the forest alternative is feasible without compensation needed for carbon sequestration. NA means that kind of crop is normally not cultivated in that region. *We assume annual 1/6 selective harvest after first 3 years, clear-cut at the end of each rotation period. 1 RMB = 0.15 USD.

Sensitivity Analysis
Because of uncertainties concerning future economic conditions and developments, we conducted a sensitivity analysis to assess the changes of minimum prices per tCER using different timber, fruit and grain prices, interest rates, and certification costs (Table 7). Further, we compared the minimum prices per tCER with and without considering soil organic carbon in tCER generation, given that SOC is often not properly included in CDM afforestation projects [9]. We selected one tree species from each tree group to represent the respective changes. We chose poplar (ecological tree), chestnut trees (economic tree), and bamboo since the minimum prices per tCER of these tree species reflect different aspects of relevant changes. For comparison, we selected one-harvest wheat and two-harvest corn as crop alternatives because (i) their revenues differ significantly, and (ii) they grow in different regions.
The minimum compensation required to make tree planting profitable declines with rising timber prices. With timber prices increasing by 25%, poplar afforestation becomes economically attractive in the northern regions even without compensation (Table 7). In contrast, the required compensation rises with increasing prices of crop grains. Note that huge afforestation activities in a region might lead to (i) reduced timber prices in the longer run due to an excess of timber supply, and (ii) increased crop prices due to reduced grain production. Both effects would reinforce each other and lead to higher minimum prices per tCER.
The interest rate level also has a significant impact on the minimum tCER prices (Table 7). In general, higher interest rates lead to lower NPV from timber, mainly because revenues occur only at the end of the project duration, and are therefore strongly affected by discounting. However, the situation for bamboo is different because we assume bamboo to be harvested by selective cutting after year 3 (1/6 annually) and clear cut at the end of the project. This means, revenues are generated continuously and relatively early and, are thus, less affected by discounting.
The afforestation alternatives with a minimum ptCER higher than 10 USD Mg −1 CO 2 are less affected by the change of certification costs, because these costs are relatively low compared with the generated carbon revenues. In contrast, for minimum ptCER lower than 10 USD Mg −1 CO 2 , the certification costs can sum up to about 85% of the carbon price. Concerning the influence of SOC sequestration, our results show that the minimum ptCER increases up to three times without considering tCERs generated in soils. Note: C, E, N, NE, NW, S, and SW are abbreviations for China, and the eastern, northern, northeastern, northwestern, southern, and southwestern region. One-or two-harvest means the crops are harvested once or twice annually, respectively. Numbers in bracket indicate negative value, which means no compensation is needed for forest alternatives. Bold figures indicate the results of Table 6. 1 RMB = 0.15 USD.

Uncertainty Analysis
This study comprehensively determined the extent of carbon sequestration by afforestation with different tree species and assessed the costs and benefits of afforestation projects from the farmers' perspective in China. Several uncertainties and bias could have affected the estimation of the carbon sequestration potential and the collection of economic data from the farmers. We controlled uncertainties in two ways. First, we applied the models and results of previously published studies to determine the carbon sequestration potential after afforestation. These findings are based on large sample sizes and therefore the sampling error is minimized in the estimation of tCERs. Second, for the economic data collection, we carried out at least five interviews for each species: if the values given by the interviewees varied by more than 10% we interviewed more people until we obtained a variation lower than 10%. Then we calculated the average of these answers. Most farmers gave very similar answers, so we are confident that our results provide well-founded and realistic average data. However, given the relatively small sample size, the representativity of our results for the diverse overall area might be limited. Further, it should be noted that we estimated the carbon sequestration potential based on average regional site qualities. In consequence, the timber revenues and tCER generated can be higher at sites with better soil quality and overestimated for poorer soil quality.
Besides the analyzed economic variables, some additional factors might play a role in the decision-making of farmers. Possible reasons why landowners do not switch to more profitable land uses are (i) lacking planting skills for trees or the most profitable crops; (ii) limited market access for particular crops or tree products (e.g., transportation network for timber); (iii) insufficient household labor to grow more labor intensive and profitable crops (e.g. sugarcane); (iv) maintaining the tradition of growing certain crops (e.g., for self-consumption).

Conclusions
Afforestation programs have a great potential to sequester carbon. The revenues of selling carbon credits can in turn contribute to finance the afforestation costs. Our study determined the carbon sequestration potential of afforestation programs considering both tree biomass and soils and compared the costs and benefits of carbon sequestration by different tree species.
Economic trees sequester less carbon than ecological trees and bamboo. Among economic trees, nut trees sequester more carbon than fruit trees. The regional context significantly influences the carbon sequestration potential with more carbon sequestered in southern and eastern regions than in northern parts. Bamboo also shows a remarkable carbon sequestration potential, which is even higher than Chinese fir and poplar in northern regions. Older bamboo stands (>30 years) could even be stronger carbon accumulators than eucalyptus, although the latter shows the greatest average carbon sequestration capacity during the study period.
Generally, it is less attractive to afforest highly productive cropland because of the high opportunity costs. The minimum price per tCER decreases with increasing project duration, except for some alternatives with fast-growing trees like eucalyptus and poplar, for which the minimum price per tCER is higher with a 30-year duration compared to a 20-year duration.
Tree product revenues differ depending on the tree species, which results in a significant variation of economic net benefits. However, our results also show that for most afforestation alternatives no compensation is needed to convert cropland with relatively low revenues such as one-harvest wheat, one-harvest corn, and two-harvest rice. For ecological trees, afforestation is profitable if farmers harvest the timber after the normal rotation period of each tree species. Furthermore, with the recent increase of timber demand in China, the timber price has increased substantially, which makes afforestation projects even more attractive. For economic trees, the amount of carbon sequestered is low, but selling fruits and nuts is quite profitable. Therefore, in most cases, there is no compensation needed for afforestation efforts. Further, our study suggests that bamboo afforestation can become economically more attractive, when carbon revenues are included. The same holds for other afforestation alternatives which require compensations. If the carbon credits generated would be traded, e.g., in a Chinese carbon market, the opportunity costs of alternative land uses could be fully compensated, except for afforestation on highly attractive cropland. However, considering the uncertain development of future China's carbon credit demand and supply [45], it could be advantageous to trade CER generated by China's afforestation projects in an international emission trading system.
Our study provides a comprehensive economic valuation of carbon sequestration by afforestation projects in China considering ecological trees, economic trees and bamboo. It demonstrates that the attractiveness of different afforestation alternatives is strongly context dependent and varies substantially across regions. The comparison of costs and benefits can help to determine efficient ways to sequester carbon by afforestation projects. The findings can contribute to the design of carbon sequestration projects in China [45]. The methodology and results can be adapted and transferred to other regions in the developing world.
Supplementary Materials: The following are available online at http://www.mdpi.com/1999-4907/10/9/754/s1. Figure S1. Temporal patterns of the amount of carbon sequestration (including biomass and soil carbon) after afforestation with various tree types on cropland.