Projecting global and regional outlooks for planted forests under the shared socio-economic pathways

There is rising global interest in growing more trees in order to meet growing population, climate change, and wood energy needs. Using recently published data on planted forests by country, we estimated relationships between per capita income and planted forest area that are useful for understanding prospective planted forest area futures through 2100 under various United Nations Intergovernmental Panel on Climate Change-inspired Shared Socio-economic Pathways (SSPs). Under all SSPs, projections indicate increasing global planted forest area trends for the next three to four decades and declining trends thereafter, commensurate with the quadratic functions employed. Our projections indicate somewhat less total future planted forest area than prior linear forecasts. Compared to 293 million ha (Mha) of planted forests globally in 2015, SSP5 (a vision of a wealthier world) projects the largest increase (to 334 Mha, a 14% gain) by 2055, followed by SSP2 (a continuation of historical socio-economic trends, to 327 Mha, or an 11% gain), and SSP3 (a vision of a poorer world, to 319 Mha, a 9% gain). The projected trends for major world regions differ from global trends, consistent with differing socio-economic development trajectories in those regions. Our projections based on empirical FAO data for the past 25 years, as well as those by other researchers, suggest that achieving the much more ambitious global planted forest targets proposed recently will require exceptional forest land and investment supply shifts.


Introduction
Planted forests provide many benefits including traditional wood production, employment and economic development, and have been identified as a key means to ameliorate climate change in the short to medium term. According to the recently published report by the Intergovernmental Panel on Climate Change (IPCC 2019), changes in land conditions, either from land-use or climate change, affect global and regional climate.
policymakers who seek to evaluate the scale of the possible consequences and the optimal design of programs aimed at expanding planted forest area.
In this study, we seek to provide insights into the prospective planted forest area futures in various countries, aggregated into major regions and the world, based on actual historical data and statistical analyses, under varying assumed economic and demographic futures. Projections providing such insights are done by, first, estimating the econometric relationship between global planted forest area and income per capita, along with a set of other socio-economic variables identified based on theory and recent empirical analysis. The estimated relationship with income per capita is then used to project future planted forest area under varying economic and demographic futures envisioned under different Shared Socio-economic Pathways (SSPs) (O'Neill et al. 2014, as discussed in the following sections.

Shared socio-economic pathways (SSPs)
This study provides topical and novel future projections of planted forest area under three distinct Shared Socio-economic Pathways (SSPs) representing varying world visions that were developed in conjunction with the IPCC Fifth Assessment. The SSPs describe different socio-economic, technological, environmental, and policy futures of the world, with varying degree of challenges for climate change mitigation and adaptation (Table 1). A recent study by Daigneault et al. (2019) builds upon existing SSP foundations by elaborating on the potential implications of SSP-related variables on forest resource management, forest product markets, wood-based bioenergy expansion, and other relevant trends in global forestry. We draw on some of those key assumptions related to plantation forest management under each SSP, as summarized in Table 1.
Compared to other SSPs, SSP1 and SSP5 represent visions of the wealthiest and most equal future worlds with the least population growth . While SSP5 envisions a world that is reliant on fossil-fuels, SSP1 characterizes a more sustainable outlook for the world, emphasizing the de-carbonization of society through low consumption growth and improved energy efficiency. SSP2 represents a world vision that more closely reflects a continuation of recent historical socio-economic trends. In contrast, SSP3 and SSP4, compared to other SSPs, envision poorer and unequal worlds, where SSP4 represents the most unequal world in terms of economic and technological development, investment in human capital and environmental emissions .
Building on the original SSP narratives, Daigneault et al. (2019) make assumptions, for example, regarding technological development, stringency of regulation and rate of deforestation under each SSPs. They summarize that the best planted forest management practices are found in association with SSP1, consistent with high productivity growth, backed by strong regulations to avoid ecological tradeoffs in this pathway. Under SSP2, they imputed a medium rate of productivity growth coupled with a medium levels of regulation, with the net effect of a slow decline in deforestation. Under SSP3, very poor forest productivity growth is anticipated, and very little regulation is assumed to be in place. Under SSP4, the productivity growth and stringency of environmental regulations are unequal in different parts of the world, leading to deforestation in areas without regulation. Lastly, under SSP5, there is rapid productivity growth but, similar to SSP2, a slow decline in deforestation and relatively small increase in natural forest area. The authors do not, however, explicitly evaluate how the various SSPs may lead to distinct trends in planted forest area, which is a specific focus of our analyses. Based on this important question, Table 1 The summary of the Shared Socio-economic Pathway (SSP) narratives SSP Narrative  Description of SSPs (IPCC 2019, pp. 14-15) Anticipated effect on planted forest management (Daigenault et al. 2019) SSP1: Sustainability "taking the green road" (low challenges to mitigation and adaptation) Assumes a peak and decline in global population (~ 7 billion in 2100), high income and reduced inequalities, effective land-use regulation, less resource intensive consumption, free trade and environmentallyfriendly technologies and lifestyles. Relative to other pathways, SSP1 has low challenges to mitigation and low challenges to adaptation High growth in forest plantation productivity and forest management; rapid diffusion of best practices SSP2:Middle of the road (medium challenges to mitigation and adaptation) Assumes a medium global population growth (~ 9 billion in 2100), medium income, medium technological progress, production and consumption patterns that reflect continuation of past trends, and gradual reduction in inequality. Relative to other pathways, SSP2 has medium challenges to mitigation and medium challenges to adaptation Medium increase of productivity in managed forests and plantations SSP3: Regional rivalry "A rocky road" (high challenges to mitigation and adaptation) Assumes a high global population (~ 13 billion in 2100), low income and continued inequalities, materialintensive consumption and production, barriers to trade, and slow rates of technological change. Relative to other pathways, SSP3 has high challenges to mitigation and high challenges to adaptation Very low productivity development SSP4: Inequality "A road divided" (low challenges to mitigation, high challenges to adaptation) Assumes a medium global population growth (~ 9 billion in 2100), medium income but significant inequality within and across regions. Relative to other pathways, SSP4 has low challenges to mitigation, but high challenges to adaptation Productivity high in high income countries due to investment, low in low income countries SSP5: Fossil-fueled development "Taking the highway" (high challenges to mitigation, low challenges to adaptation) Assumes a peak and decline in population (~ 7 billion in 2100), high income, reduced inequalities, and free trade but emphasizes resource-intensive production, consumption and lifestyles. Relative to other pathways, SSP5 has high challenges to mitigation, but low challenges to adaptation Highly managed, resource-intensive; rapid increase in productivity 1 3 we aim to model the future planted forest area development under quantitative scenarios inspired by the SSP narratives , reflecting the developments also with recent SSP narratives created for forest sector (Daigneault et al. 2019).

Empirical approach
To characterize the relationship between economic development and planted forest area at the global level, we estimated an econometric model of planted forest area with the specification originally presented by Turner et al. (2006). They modeled annual changes in total forest area by country as functions of scale, technique, and composition effects represented by income per capita, rural population density, labor per unit of forest area, capital per unit of forest area, and by the country's openness to trade (see also Antweiler et al. 2001;Cole and Elliott 2003).
Although their specification was used to model annual changes in total forest area, we chose the same specification to model planted forest area because it captures the impact of investments designed to intensify forest management, including through tree planting. The inclusion of composition effects captures the situation where a country changes its derived forest product mixes as income rises (e.g., switching to alternative sources of energy from traditional fuelwood). Rural population density can represent the scale of forest use (including planted forests), whereas income per capita and its squared term models the technique effect. Finally, inclusion of a measure of a countries' openness to trade (or trade intensity), interacted with other variables in the model, is designed to represent the influence of the global market on domestic supply of forest products, i.e., the extent to which the relative comparative advantage of a country (e.g., Bonnefoi and Buongiorno 1990) is operational on the domestic forest sector. Following Turner et al. (2006), the planted forest area model estimated in this study is represented in Eq. 1: where PA it = planted forest area (ha); thus, any changes in planted forest area from period t − 1 to t are the result of net changes in afforestation during t − 1 to t (forest area established through planting and/or deliberate seeding); = a vector of parameters containing a constant ( 0 ), coefficients associated with the independent variables ( 1 to 7 ), and regional dummy variables ( 8 to 12 ); Y N it = per capita gross domestic product (GDP, 2010 constant US $) in country i at time t; R it = rural population density (person ha −1 ) in country i at time t, a proxy for scale of total forest use; L A it = the ratio of labor to total forest area in country i at time t; K A it = the ratio of capital to total forest area in country i at time t; I it = the ratio of the value of exports plus imports to GDP (trade intensity) in country i at time t; , suggesting that the marginal effect of trade intensity on planted forest area depends on the level of other variables; D 8 to D 12 = dummies for different regions in the following order: South America, Asia, Africa, Europe, and North America; it = a random error term that varies over country i at time t. (1) The dependent variable, planted forest area, conforms to the definition of planted forests by FAO (2015): (i) planted, seeded, or coppiced forests of native species; and (ii) plantations for timber or for the provision of services consisting of native or introduced species. Planted forest area is assumed to have a negative or positive relationship with rural population density (R). Although a negative relationship signifies that there is less area available for forest expansion with increasing rural population, the positive relationship may suggest that land use intensification by smallholders may free-up land for tree planting, despite no change or even increases in rural population densities (Lambin and Meyfroidt 2011). It is expected that countries with a higher share of labor per unit of total forest area and more capital per unit of total forest area (L/A, and K/A, respectively) can invest more in intensive forest management or new plantations, and therefore their signs should be positive. Trade intensity (I), the measure of a country's comparative advantage in producing and trading a forest product (the larger the value of I, the greater is the comparative advantage), should have a positive effect on planted forest area (Rudel et al. 2005;Turner et al. 2006). A positive sign for income per capita (Y/N) and a negative sign for its quadratic term, (Y/N) 2 , would suggest an inverse-U-shaped relationship of planted forest area with per capita income, whereas the reverse would indicate a U-shaped relationship.

Data
Data used in estimating Eq. 1 were comprised of a balanced panel of observations for 96 countries in five different time points (1990, 2000, 2005, 2010, and 2015), giving a total of 480 observations. The dataset consists of only those countries that regularly report their data on planted forest area. The planted forest area and total forest area data came from the Food and Agricultural Organization's Global Forest Resource Assessment Report (Food and Agriculture Organization 2015). Income (gross domestic product, GDP), population, and labor force data were obtained from the World Bank (2018). Data on capital stock and trade intensity were obtained from the Penn World Table database (Feenstra et al. 2015). Table 2 presents the description and the summary statistics of the variables. Specifically, the data indicate that for four of the six of the model variables, the standard deviation was many times greater than the mean. This fact, along with the relatively small sample size, suggests that findings of a high level of statistical significance will be difficult, which we do consider as the results unfold.

Estimating the econometric model of planted forest area
We first estimated three different panel data model specifications, each representing different assumptions about the distribution of the error term shown in Eq. 1: pooled ordinary least square (OLS), fixed effects (FE) and random effects (RE). The pooled OLS model, by assumption, does not capture any cross-sectional, time-invariant unobservable differences among countries, and the error term is assumed homoscedastic across all spatial-temporal units of observations (although generalized heteroscedasticity of the error term in Eq. 1 can be introduced). The FE model captures the variation of the independent variables in time but explicitly recognizes potential cross-sectional, time-invariant differences in the level of the dependent variable by allowing each country to have a separate constant term but a common coefficient for each explanatory variable. The variance of the error term in Eq. 1, however, is assumed homoscedastic (although, in estimation, generalized heteroscedasticity can be introduced). Time-invariant variables cannot be incorporated in the FE model and variance is assumed to be constant across the cross-sectional units.
The RE model allows for cross-sectional (cross-country) variation in error variance but no time-invariant differences in the level of the dependent variable across countries. To partially relax the assumption of no time-invariant cross-sectional differences in the level of the dependent variable, following previous research suggesting that planted forest area changes vary systematically across regions (Payn et al. 2015), we included region-specific dummies (for South America, Asia, Oceania, Europe, and North America).
Each of these three model specifications was estimated using natural logarithmic transformations of the continuous variables, so these variables' estimated coefficients represent elasticity estimates. The models were estimated using Stata IC/14 using Normal, robust and clustered (by country) standard errors to control for the effects of heteroscedasticity in the model. Furthermore, acknowledging the possibility of nonstationarity of regression variables (for which, due to a paucity of time series observations per country, were not amenable to unit root tests), we estimated the model using the first-differences of the natural logarithms of the continuous variables, with robust standard errors to mitigate problems of omitted variables biases and potential spurious regression. The statistically most superior model was selected based on a pooling test and overall model goodness of fit.
Rejection of the null hypothesis in the pooling test (Table 3) suggested a better fit for the RE model compared to the OLS model. The analysis further revealed several statistically non-significant coefficients with unexpected signs found for FE model. Because the empirical fit proved better, we decided to proceed with the RE model with regional dummy variables. Although a FE-model is perceived as more suitable for country level analysis (Bell et al. 2019), our results suggest that there are some location-specific attributes, other variables, or data limitations that the FE model was unable to account for in our study context.

Projecting global planted forest area under different SSPs
The estimated parameters of Eq. 1 were used to project planted forest area by country to 2100, under various scenarios of economic and demographic changes represented in various SSPs. In this study, we report the projected planted forest area only for three out of the five SSPs (SSP2, SSP3, and SSP5) because of the similarity of results for SSP1 and SSP5 and the similarity of results for SSP3 and SSP4.
The projection of total forest area was made for 152 countries under each SSP for which projected data on the explanatory variables were available. The available future data needed for projecting planted forest area were income per capita, Y/N, its squared term, (Y/N) 2 , rural population per unit of forest area, R (a proxy for rural population density), and labor per unit of total forest area, L/A. Data on future GDP (Y) and population (N) for each country in each SSP were obtained from the International Institute for Applied Systems Analysis (IIASA) projection database (IIASA 2018). The projected L/A data were generated by forcing the 2015 L/A value to grow at the same rate as the projected population in each SPP. Similarly, projected values for R for each country in each SSP were estimated by dividing rural population (IIASA 2018) by rural land area (Jiang and O'Neill 2017) (detailed calculations available by request). The projected average values of Y/N, R, and L/A for different geographical regions are shown in Figs. 2, 3 and 4. The projected trends on these variables for each region were generally similar to the respective trends for individual countries within each region (discussed further in results section), which largely determined projected trends of planted forest area across the SSPs. The quantitative Table 2 Description and summary statistics of the variables used in the econometric equation for planted forest area Data covering 1990Data covering , 2000Data covering , 2005Data covering , 2010Data covering , and 2015. Includes 96 countries from Africa (38), Asia (18), Europe (19), North America (10) projections of the remaining explanatory variables for the SSPs were not available, so their effects on forest area were held constant when projecting planted forest area under each SSP. Table 3 summarizes the coefficients (which are interpreted as elasticities), standard errors, and the statistical significances of the estimated RE model specification (results for other specifications available by request). While the per capita income term (Y/N) was estimated to have a positive and significant relationship with planted forest area, its squared term (Y/N 2 ) was negative and non-significant at traditional thresholds (p value = 0.127). However, to avoid potential Type 2 errors, we decided to keep the (Y/N) 2 term in the model in projecting planted forest area. In addition, the independent variables in this final model were usually significant with the expected signs, indicating the merits of the model selected.

Results
The relationship between planted forest area and Y/N demonstrated near unitary elasticity, with a 1% increase in Y/N being associated with a 1.07% increase in planted forest area. All other variables had their expected signs, although the signs on the coefficients of interaction variables and regional dummies were unknown beforehand. The estimated Table 3 Estimated coefficients, standard errors, and significance levels of the random effects version of the Eq. 1 of global planted forest area Numbers in parentheses are standard errors ***Statistically significant at 1%, **at 5%, *at 10%; see Table 2 for variable explanations a The specified model was a log-log model; therefore, the estimated coefficients are also of elasticity estimates coefficient of R was negative and statistically significant, with an elastic relationship with planted forest area, suggesting that a 1% increase in rural population density would decrease planted forest area by 1.31%. The L/A coefficient demonstrated a positive and highly elastic and statistically significant relationship with planted forest area, indicating that a 1% increase in labor force per unit of forest area is associated with a 1.98% increase in planted forest area. The K/A coefficient was not statistically significant, although it had the expected positive sign. The interaction term of L/A and K/A was statistically significant and negative. The negative sign was contrary to expectation, since a positive sign was expected for both L/A and K/A. However, the estimated coefficient was small (− 0.05), indicating that a 1% increase in both L/A and K/A is associated with a 0.05% decrease in planted forest area. Trade intensity (I) was not statistically significant, although its sign was positive, as expected.
Finally, the coefficient of the interaction term of I, L/A, and K/A was statistically significant and negative. The negative relationship with planted forest area suggests that the degree of a country's comparative advantage in the trade of forest products in global markets depends on the interaction of labor per unit forest area and capital per unit of forest area. Stated differently, increasing trade openness is not helpful in expanding planted forest area when a country has a large labor force and a large capital stock per unit of forest area available.

Projected global planted forest area
For each SSP, the projected global planted forest area using the estimated model parameters showed increasing trends for the next three to four decades and declining trends thereafter. The projected planted forest area in each SSP over time is consistent with the estimated Eq. 1 (Table 3) and the projected trends for the explanatory variables (Figs. 2, 3, 4). Compared to initial planted areas totaling 293 Mha in 2015, SSP5 showed the largest increase by 2055 (to 334 Mha, a 14% increase), followed by SSP2 (to 327 Mha or an 11% increase). For SSP3, the projected increase in planted forest area was highest in 2045 (to 319 Mha, a rise of 9% from the 2015 level). By 2100, although SSP2 and SSP3 showed increases of only 2% and 3%, respectively, SSP5 did not show any increase in global planted forest area, relative to the 2015 level (Fig. 1).
The estimated model indicates a positive impact of per capita income and labor per unit forest area but a negative impact of rural population density on planted forest area, the net effect on planted forest area depending on the magnitude of changes in those explanatory variables. For instance, the largest per capita income (Fig. 2a-f) growth, coupled with a lower population density (Fig. 3a-f) projected for most countries in SSP5, contributed to the largest global increase in planted forest area before 2055, although a lower rate of projected growth in labor per unit forest area (Fig. 4a-f) in this scenario dampened the projected net increase in planted forest area. By 2100, the higher magnitudes of growth in labor per unit forest area in SSP3 and in SSP2 compared to SSP5 led to lower declines in planted forest areas in these two scenarios compared to SSP5.
Projected trends in planted forest area varied substantially by world region, a result linked to their differing trajectories in explanatory variables. For instance, Africa showed the largest percentage increases in planted forest area in all SSPs, among all regions (Fig. 5a), resulting in increases by 2100 of 133%, 122%, and 72% for SSP3, SSP2, and SSP5, respectively, compared to the observed 2015 level (16 Mha). While SSP2 and SSP3 showed almost linear trends in planted forest area growth in Africa, it declined after 2075 for SSP5. The largest increase in planted forest area in Africa in SSP3 is consistent with the largest projected increase in labor force growth in this scenario, which more than doubles by 2100 compared to the 2015 level (Fig. 4a). However, lower income per capita (Fig. 2a) and higher rural population density (Fig. 3a) growth dampened the projected overall increase in planted forest area in SSP3 for Africa.
Asia showed contrasting trends in projected planted forest areas to 2100 (Fig. 5b), compared to its observed 2015 level. It increased through 2035 in all scenarios but then decreased in both SSP2 and SSP5, owing largely to higher rural population densities (Fig. 3b), coupled with a smaller labor force (Fig. 4b) projected in these scenarios in the later projection years, relative to SSP3. For instance, projected planted forest area in SSP2 and SSP5 declined by 14.9% and 29.1%, respectively, but SSP3 showed a slight increase (4.3%) by 2100, compared to the observed 2015 area of planted forests.
For Europe, planted forest area in 2100 was projected to decline in SSP2 and SSP3 by 4.5% and 18.1%, respectively, but was projected to increase in SSP5 by 6.3%, relative to the 2015 level, although the trends were increasing until 2070 for SSP5 and until 2055 for SSP2 (Fig. 5c). The largest projected decline in planted forest area in Europe (SSP3) can be linked to the slowest growth in per capita income (Fig. 2c) and labor per unit forest area (Fig. 4c) compared to other world regions, although the low rural population density (Fig. 3c) in this scenario helped increase Europe's planted forest area. The largest projected increase in planted forest area in Europe (SSP5) can be linked to relatively higher growth in per capita income (Fig. 2c) and labor per unit forest area (Fig. 4c), although projected higher growth in rural population density (mostly before 2075) helped to decrease planted forest area development in this scenario (Fig. 3c).
Unlike other regions, North America showed the largest variation in projected planted forest area across SSPs, showing a decline of 16.7% in SSP3, an increase of 19.3% in SSP2, and an increase of 49.9% in SSP5, by 2100, compared to its 2015 area of planted forests (Fig. 5d). The decline in planted forest area in SSP3 is associated with the lowest growth in average per capita income (Fig. 2d) coupled with the highest growth in rural population density (Fig. 3d). This planted forest area decline was despite a higher projected labor force in this scenario (Fig. 4d), which, because of its positive relationship with planted forest area, would tend to put a damper on losses. Similarly, the largest projected increase in planted forest area, found in SSP5, is explained by the higher projected income per capita growth and lower projected growth in rural population density, despite lower projected growth in labor per unit of forest area (Figs. 2d,3d,4d).
The prospective planted forest area futures for Oceania (Fig. 5e) are similar to those projected for North America, showing large variation in projected planted forest area changes across the SSPs, with SSP2, SSP3, and SSP5 showing a 26.7% increase, a 6.0% decline, and a 53.3% increase, respectively, by 2100, compared to 2015 areas. Similar to North America, the projected planted forest area for Oceania can also be attributed to projected growth in each of income per capita (Fig. 2e), rural population density (Fig. 3e), and labor per unit of forest area (Fig. 4e).
Finally, the projected planted forest area outlook for South America (Fig. 5f) varied from a decline of 23.2% in SSP5 to an increase of 34.2% in SSP3 by 2100, compared to its 2015 level, with SSP2 showing an increase of just 1.3%. The projected decline in planted forest area in South America in SSP5 is linked to low growth in labor per unit forest area (Fig. 4f), although the projected highest growth in per capita income (Fig. 2f) and the lowest rural population density (Fig. 3f) tended to increase planted forest area in this scenario.
The largest increase in planted forest area in South America, under SSP3, is mainly attributable to the projected high labor per unit of forest area compared to SSP2 and SSP5.

Discussion and conclusions
Recently published research modeled planted forest area change as linear function of income per capita change (Korhonen et al. 2014;Nepal et al. 2019c), and our study bolsters that modeling approach. In addition, our study offers modest evidence of a negative quadratic per capita income term at a 13% significance level, which is reasonable given the quite sparse and variable planted forest panel data available. This quadratic term does project that planted forest area changes may begin to decline at higher per capita income levels. This outcome is consistent with the view that rising incomes are associated with progressively greater attention to the adverse environmental or social impacts of planted forests (Malkamäki et al. 2018). Thus, the nature of the relationship between economic growth and sustainability of planted forest area growth merit additional scrutiny, focused on understanding the role of planted forests in achieving sustainability goals (SDG 12, UN 2017).
In addition to our finding that planted forest increases may be tempered by higher income levels, this article provides insights into the potential future of planted forest area under varying economic and demographic trajectories envisioned in different SSPs. Our results show that different SSPs are likely to result in different outlooks for, and distributions of, planted forest area among the continents. SSP5, which anticipates resource intensive production and consumption, brings with it a prediction of the greatest increase in areas of planted forests in Africa, Europe, North America and Oceania; in Asia and South America, in contrast, the planted forest growth is the most modest. Respectively, the planted forest area in Asia and South America grows the most under SSP3, which emphasizes "regional rivalry". The sustainability consequences of these projected planted forest area changes depend on multiple interlinked drivers that we could not capture with the data and statistics available for our modeling. For example, based on our analyses, we cannot conclude whether planted forests will substitute for natural forests or even lead to losses of other ecologically prestigious habitats. Nonetheless, we can suggest some risks related to the different SSPs and planted forests. In cases where we anticipate high population growth and consumption growth without a stringent regulation (SSP3), Asia or South America might be particularly under at risk of unsustainable forest plantation levels. In the other regions, greater planted forest development is found with SPP5, which could be attributable to the highly resourceintensive and fossil-fuel-dependent future , coupled with a positive assumption of high productivity growth and diffusion of best practices in planted forest management (Daigneault et al. 2019). We compared our projections with our previous planted forest area outlook reported by Nepal et al. (2019c), which provided outlooks of planted forest area development through 2070 for each SSP and which were linearly driven by GDP per capita and roundwood production quantities projected in each country in each SSP. Our model projects smaller increases in global planted area than the increases reported by Nepal et al. (2019c) for the same scenario. For instance, while we projected a 10% increase in global planted forest area by 2070 relative to 2015 level for SSP2 ("the middle road"), Nepal et al. (2019c) projected a 46% increase for the same scenario during the same period. The differences in two model projections are due to evident differences in model specification between mostly  Jiang and O'Neill (2017). Rural land area was assumed to change in the future in proportion to the projected trends in rural population linear versus quadratic independent variable formulations and the more extensive planted forest area drivers included in our study.
Our model projects larger increases in planted forests for Africa (increasing by 99% from 2015 to 2070), North America (+ 20%) and Oceania (+ 28%) under SSP2, compared to Nepal et al. (2019c), which showed increases in those regions by 65%, 16%, and 14%, respectively, during the same time span and scenario. More differences in projections were observed for Asia and Europe, where our model showed almost no growth in planted forest area, and also for South America, where our model showed a 13% increase in SSP2, by 2070, relative to 2015 levels. In contrast, Nepal et al. (2019c) showed increases of 46%, 23%, and 38% in those regions, respectively, during the same period and same scenario.  (2015). The base year labor per forest area was assumed to follow the population growth trends projected in the various shared socio-economic pathways Both of our approaches here and in research reported in Nepal et al. (2019c) still project less future plantation area than the largely linear projections based on sigmoid projections of past data (e.g., Carle and Holmgren (2008), d 'Annunzio et al. (2015)). Our study also complements the study by Daigneault et al. (2019) by associating the SSPs narratives created for forest sector with the quantitative sector scenarios that are inspired by SSPs imagining more alternative socio-economic pathways .
It bears stating that projections to the next century of planted forests are made under the assumption that model parameters do not change. Hence, broad conclusions about absolute levels of planted forests, given assumed future incomes per capita, should be made with caution. Such caution would also carry over to evaluation of the effects of policies, including those intended to counteract potential reductions (assuming a negative sign on quadratic income per capita) in planting at higher income levels. This study provides potentially valuable background information for private forest investors and for public policy makers. The earlier years of our projections are well within the investment planning horizons for investors and policy makers considering planted forests in the context of timber markets, country competitiveness, and environmental benefits such as carbon storage.
Our robust analyses and close linkage to SSPs provide a relatively narrow range of total planted forest projections of between 319 and 334 Mha by 2100, depending on the SSP pathway. This projected amount is still less than the sum of individual country or global expectations stated in the last few years, as reviewed in our introduction (e.g., IUCN 2015; WWF 2012). For example, Bastin et al. (2019) estimated global technical potential for woodlot creation to meet global demand for forest product markets and for climate change mitigation to be as high as 900 million ha. The World Economic Forum (WEF 2020) goal of 1 trillion trees (1t.org 2020) would represent roughly 1 billion ha of forests at 1000 trees per ha, or three times as much area as we project. Furthermore, at an establishment costs of about $1500 per ha without overhead costs (Cubbage et al. 2007(Cubbage et al. , 2020, an additional billion ha of forests would require about $1.5 trillion dollars of new funds, an amount many times greater than all prior global forest planting investments in total. The differences between our projected planted forest area and upper level global aspirations indicate that historical barriers to planted forest expansion would need to be addressed, including land availability, complex land tenure considerations, trade-offs with natural forests and associated environmental impacts, and identification of novel sources of funding. Issues such as these have prompted criticism of the 1 trillion goal as well (e.g., Calma 2020), so our econometric analyses of historical trends suggest that lesser levels of planted forests are more plausible, providing a useful reality check to policy goals. Projecting futures inherently involves a significant level of uncertainty, and projecting planted forests at a global scale requires a balance between a theoretical structure that is general enough to cover the variability of economic situations in different countries and the availability of international data (Turner et al. 2006).
Our analysis indicates that projecting the continuation of historical socio-economic trends under SSP2, planted forest areas are apt to increase at lower rates than found in previous studies. Of course, this analysis does not provide information on the planted forest area and natural forest area interplay, or possible adverse impacts of planted forest area expansion on, e.g., local communities (Malkamäki et al. 2018), which will largely define whether the planted forest expansion is sustainable. Nonetheless, the merits of planted forests for fiber and especially for climate change reduction in tropical countries, such as suggested by Busch et al. (2019), should be considered by policy makers when they consider policies such as payments for net carbon emissions reductions, or indeed for many other benefits of planted forests such as water quality, biodiversity, or recreational values. These broad benefits would augur for concerted global efforts to maintain higher planted forest increases than we projected here with this more comprehensive model than employed in previous research.
Our findings here of somewhat lesser projected areas of planted forests in the future do have significant policy implications for both advocates and critics of plantations. Our findings suggest that wood production from planted forests will stabilize, unless higher productivity and growth rates offset more modest area expansions (see discussion in Daigneault et al. (2019) about anticipated effects of SSPs on the planted forest management). Our findings also indicate that in order to achieve the desired effects in reducing climate change, greater policy interventions will be required, such as those supporting a higher price for carbon storage (e.g., Busch et al. 2019). For instance, based on the estimated marginal abatement cost curve for 90 tropical and subtropical countries, Busch et al. (2019) found that carbon prices of USD 20 and 50 per ton would incentivize tropical land users to increase reforestation, 2020-2050, by 8% and 22%, respectively, relative to their projected reforestation under their business-as-usual scenario with a zero carbon price.
On the other hand, the lower projected area estimates suggest that fears of adverse effects from planted forests are not as likely given the trends we projected here. Regardless, more systematic studies on the diverse socio-economic drivers and projected trends in planted forests, and their interlinkages, e.g. with a carbon price (Busch et al. 2019), carbon storage (Pihlainen et al. 2014;Assmuth and Tahvonen 2018), biodiversity policies, or tree species selection (Potter et al. 2017a, b, Matthies andValsta 2016), at regional and global levels, could improve information and discussion and better guide private investors and public decision makers.