LETTER • OPEN ACCESS You may also like - How just and just how? A systematic Recognizing the equity implications of restoration review of social equity in conservation research priority maps Rachel S Friedman, Elizabeth A Law, Nathan J Bennett et al. - Social equity in sustainability certification To cite this article: Bill Schultz et al 2022 Environ. Res. Lett. 17 114019 systems for the built environment: understanding concepts, value, and practice implications Adam Yeeles, Kim Sosalla-Bahr, Jennifer Ninete et al. View the article online for updates and enhancements. - Intensive rainfall recharges tropical groundwaters Scott Jasechko and Richard G Taylor This content was downloaded from IP address 158.109.174.204 on 16/05/2023 at 19:41 Environ. Res. Lett. 17 (2022) 114019 https://doi.org/10.1088/1748-9326/ac9918 LETTER Recognizing the equity implications of restoration priority maps OPEN ACCESS Bill Schultz1,∗, Dan Brockington2,3, Eric A Coleman4, Ida Djenontin5,6, Harry W Fischer7, RECEIVED Forrest Fleischman8, Prakash Kashwan9, Kristina Marquardt7, Marion Pfeifer10, Rose Pritchard11 2 April 2022 and Vijay Ramprasad12,13 REVISED 1 15 September 2022 Department of Political Science, Florida State University, Tallahassee, FL 32306, United States of America 2 ACCEPTED FOR PUBLICATION Institut de Ciència i Tecnologia Ambientals, Universitat Autònoma de Barcelona, Barcelona 08193, Spain 3 11 October 2022 Institució Catalana de Recerca i Estudis Avançats, Barcelona 08010, Spain 4 Department of Political Science, Florida State University, Tallahassee, FL 32306, United States of America PUBLISHED 5 25 October 2022 Grantham Research Institute on Climate Change and the Environment, The London School of Economics, WC2A 2AE London, United Kingdom 6 Department of Geography, Penn State University, State College, PA 16801, United States of America Original content from 7 Department of Urban and Rural Development, Swedish University of Agricultural Sciences, PO Box 7012, SE-750 07 Uppsala, Sweden this work may be used 8 under the terms of the Department of Forest Resources, University of Minnesota, St. Paul, MN 55108, United States of America 9 Creative Commons Department of Environmental Studies, Brandeis University, Waltham, MA 02453, United States of America Attribution 4.0 licence. 10 SNES, Newcastle University, NE1 7RU Newcastle Upon Tyne, United Kingdom 11 Any further distribution Global Development Institute, University of Manchester, M13 9PL Manchester, United Kingdom of this work must 12 Center for Ecology, Development and Research, Dehradun, Uttarakhand 248006, India maintain attribution to 13 Department of Environmental Studies, Ashoka University, Sonepat 131029, India the author(s) and the title ∗ of the work, journal Author to whom any correspondence should be addressed. citation and DOI. E-mail: wbs15@fsu.edu Keywords: restoration, mapping, priority-setting, climate change, social equity Supplementary material for this article is available online Abstract A growing number of studies seek to identify global priority areas for conservation and restoration. These studies often produce maps that highlight the benefits of concentrating such activity in the tropics. However, the potential equity implications of using these prioritization exercises to guide global policy are less often explored and articulated. We highlight those equity issues by examining a widely publicized restoration priority map as an illustrative case. This map is based on a prioritization analysis that sought to identify places where restoration of agricultural land might provide the greatest biodiversity and carbon sequestration benefits at the lowest cost. First, we calculate the proportion of agricultural land in countries around the world that the map classifies as a top 15% restoration priority. A regression analysis shows that this map prioritizes restoration in countries where displacing agriculture may be most detrimental to livelihoods: countries that are poorer, more populated, more economically unequal, less food secure, and that employ more people in agriculture. Second, we show through another regression analysis that a similar pattern appears sub-nationally within the tropics: 5 km × 5 km parcels of land in the tropics that are less economically developed or more populated are more likely to be top 15% restoration priorities. In other words, equity concerns persist at a subnational scale even after putting aside comparisons between the tropics and the Global North. Restorative activity may be beneficial or harmful to local livelihoods depending on its conceptualization, implementation, and management. Our findings underline a need for prioritization exercises to better attend to the risks of concentrating potentially negative livelihood impacts in vulnerable regions. We join other scholars calling for greater integration of social data into restoration science. © 2022 The Author(s). Published by IOP Publishing Ltd Environ. Res. Lett. 17 (2022) 114019 B Schultz et al 1. Introduction requires acknowledging and respecting diverse rights, values, and identities. Finally, contextual equity refers Ecosystem restoration seeks to recover degraded to the consideration of historical and structural socio- ecosystems through interventions like assisted natural cultural, economic, and policy contexts. regeneration, tree planting and species reintroduction Our study focuses most on distributional equity [1, 2]. The New York Declaration on Forests, the Bonn concerns, which have been at the forefront of envir- Challenge [3], and the UN Decade on Ecosystem onmental justice debates for decades, and are pos- Restoration [4] illustrate the high profile of restora- sibly easier to demonstrate quantitatively [36]. We tion in global environmental discourse. When imple- provide an illustrative analysis of a recent, well- mented effectively, restoration may provide a variety publicized priority mapping exercise [21] that exem- of benefits [5]. These benefits include carbon sequest- plifies the challenges faced by other studies in this ration, habitat protection, improving the availability literature. First, we show that an output map from of resources that support local livelihoods [1, 2, 6, 7], this study concentrates top priority restoration areas erosion control, flood reduction, and pollination in countries that are poorer, more densely popu- [8–12]. However, these benefits are not guaranteed lated, more reliant on agriculture, and more food [13–15], there may be spatial disparities in their dis- insecure. Though restoration and conservation inter- tribution, and there may be negative impacts of res- ventions may support local livelihoods [10], they toration as well [16, 17]. can also have negative impacts on the poor [38– A growing number of studies seek to identify res- 40]. Restoration can involve the loss of agricultural toration and conservation priorities spatially [18–24]. lands [21]. It may lead to increased competition for The maps these studies produce may be used by poli- remaining land [41, 42], the displacement of cus- cymakers to inform decision-making [25]. Such stud- tomary land uses [11, 43], or the displacement of ies can be valuable for improving the effectiveness of communities themselves [44]. It can also increase environmental policy [24, 26], identifying spaces with the risk of detrimental human-wildlife interactions climate mitigation potential [18, 20, 27], and identi- [45, 46]. A prioritization analysis that leads policy- fying synergies between biodiversity conservation and makers to concentrate restoration in lower-income carbon sequestration goals [28–30]. These studies can countries could have serious distributional con- also be useful for weighing the costs and benefits of sequences depending on subsequent decisions about land use change in different areas [19, 21]. design, implementation, and monitoring. However, these prioritization analyses and the Second, we show that similar distributional equity maps they produce have attracted criticism. For concerns persist at a subnational scale: areas within example, a study of global tree restoration potential tropical countries that are less developed or more [18] was criticized for projecting tree restoration populated are more likely to be top restoration prior- onto grasslands [31, 32]. Wyborn and Evans discuss ities. Even when cross-national comparisons between how mapping exercises might oversimplify complex the Global South and the Global North are set aside, decision-making problems or crowd out information using extant prioritization analyses to guide policy about local contexts [23]. Meyer and Pebesma point still risks concentrating any negative impacts of res- out that accuracy assessments of some global ecolo- toration on communities that are less prepared to gical maps are disputable, and that such maps may bear them. rely on inadequate reference data [33]. In contrast, Though our empirical analyses highlight distribu- others contend that despite such problems, the out- tional equity, we emphasize that priority setting ana- puts of these prioritization exercises provide broader lyses also raise potential procedural, recognitional, context for local decisions and help encourage global and contextual equity concerns. Proponents of res- co-ordination [25]. toration mapping may suggest that their outputs are To date, there are few empirical analyses of the not concrete proposals, but rather starting points for equity issues raised by the data and processes com- debate [25]. However, this argument sidesteps the monly used to produce global priority maps. We threat such analyses pose to fair decision-making. argue that a better understanding of equity in this area Priority-setting exercises may shape future policy is crucial, on both normative and feasibility grounds implementation, yet they often do not incorpor- [34]. We understand equity as an ethical requirement ate information on the values and interests of com- to take fair and just steps to support the sustainab- munities who would be most affected by restoration ility of socio-ecological systems [35]. Four dimen- [37]. These communities are often overlooked in sub- sions of equity are often underscored in environ- sequent steps of the restoration policymaking pro- mental planning [35–37]. First, procedural equity cess as well [47, 48]. This raises interrelated proced- requires that decision-making procedures are inclus- ural and recognitional equity concerns, which may ive and democratic, with effective participation of reinforce distributional consequences in the future. relevant voices. Second, distributional equity speaks Meanwhile, a failure to consider aspects of the local to how resources, costs, responsibilities, and benefits social context, such as land tenure security, risks are allocated and shared. Third, recognitional equity creating a misleading picture of global restoration 2 Environ. Res. Lett. 17 (2022) 114019 B Schultz et al priorities. This is problematic on contextual equity 2.2. Initial descriptive analysis grounds. In sum, our study reinforces calls for better We begin with some descriptive analysis of Strassburg integration of social data into restoration science. et al’s [21] restoration priority map. We follow an approach outlined by Strassburg et al to produce 2. Methods a map that identifies agricultural (‘converted’) land around the world at a 300 m resolution (see our sup- 2.1. Exploring a single map in-depth porting information for more detail). A cell is classi- As a case study into the equity issues underlying fied as being either converted cropland or pastureland global prioritization analyses, we explore Strassburg (1), or not (0). We then use the priority levels from et al’s restoration priority map [21]. We selected this the map in figure 1(e) of Strassburg et al’s [21] study prioritization analysis and one of its output maps for a to classify converted lands by their restoration prior- few reasons. First, it is a well-publicized global priorit- ity level and calculate the proportion of high prior- ization study13 . Moreover, it was published in a high- ity (global top 15%) agricultural lands that are in the impact journal, represents a clear methodological tropics. improvement, and exemplifies important tensions in Additionally, we estimate the proportion of agri- this literature. We emphasize that our intent is not to cultural land in each Food and Agriculture Organiz- critique this study in particular. We are responding to ation (FAO) member country that would need to be calls for deeper exploration of equity issues in the pri- restored if a global restoration initiative targeted the oritization literature raised by Strassburg et al them- planning units classified as a top 15% priority by this selves, as well as other recent studies [49]. map. Our estimates assume the complete restoration Strassburg et al [21] developed a multi-criteria of all agricultural land cells, as did the original prior- algorithm for optimally choosing areas for restoration itization analysis. while accounting for biome-specific characteristics of different parcels of land. They use this algorithm to 2.3. National-level regression analysis assign restoration priority levels to ‘converted lands’ We use ordinary least squares (OLS) linear regressions around the world (i.e. lands used for crops or pas- to explain variation in the proportion of agricultural ture). Specifically, they break down the world’s sur- land in a country that is classified as a top 15% res- face into 5 km × 5 km ‘planning units.’ Within each toration priority (the dependent variable; ‘top prior- planning unit, they map land cover in 300 m × 300 m ity land’). Our primary regression model includes the cells to both identify agricultural lands and estimate following explanatory variables: population, wealth the distribution of ecosystem types to which those (gross domestic product, GDP), economic inequal- agricultural lands should be restored. These authors ity (Gini index), governance (democracy and human focus on complete land use change, wherein agricul- rights scores), and a country’s land area (added as tural lands are entirely converted to a natural ecosys- a control)14 . We include population and wealth in tem, making opportunity costs easier to define. this model to explore whether more populated or less Strassburg et al [21] consider two benefits of wealthy countries tend to contain more top prior- restoring agricultural land: carbon sequestration ity land. We include a measure of economic inequal- and maintaining natural habitat. They estimate the ity given past research investigating the relationship opportunity costs of restoration based on projected between inequality and conservation policymaking profits from livestock or crop production, and they [48]. The Gini index is a well-established indicator of estimate implementation costs globally based on res- economic inequality that measures the dispersion of toration projects in Brazil’s Atlantic Forest [22]. Based income in a country, such that a low score indicates on these spatial estimates of costs and benefits, the a highly equal income distribution while a high score authors use simulations to classify planning units indicates a highly unequal income distribution. around the world into different restoration priority To see the logic for our inclusion of two gov- levels (top 5%, 6%–10%, 11%–15%, etc) under a ernance measures, consider that some countries may range of scenarios. These different scenarios assign pursue restoration and conservation in more coercive priorities to minimize costs, maximize carbon stor- ways than others, or may be more likely to infringe age, maximize biodiversity protection, or combina- on communities’ rights and implement environ- tions of the three. We focus on the restoration prior- mental policies without consideration of local live- ity map for one of their scenarios, visualized in figure lihoods. National-level measures of the incidence 1(e) of Strassburg et al [21], which identifies priority of conservation-induced displacement, for example, sites for restoration that maximize carbon sequestra- are not readily available, so we can only proxy for tion and habitat protection whilst minimizing costs. 14 Though land area is of less substantive interest, it may still be important to include in our models. Larger countries tend, for 13 For instance, it was referenced in an article of the New York instance, to have a greater GDP and a greater population. They may Times [83], and as of 8 February 2022, the study had an Altmet- also tend to contain more top priority land, simply by virtue of their ric score of 1867. size. 3 Environ. Res. Lett. 17 (2022) 114019 B Schultz et al these inclinations indirectly. Measures of the extent of Our first explanatory variable is a parcel’s level of liberal democracy—characterized by the protection development, based on nighttime lights data. Night- of individual liberties and limits on state authority time lights data can provide reasonable estimates of [50]—and the extent of human rights protection are economic development at local scales [51, 54], though some of the better proxies available at a cross-national it becomes especially difficult to link nighttime lights scale. A previous analysis of the interactions between to specific components of development (e.g. urban- inequality, democracy and the legal designation of ization, infrastructure growth, manufacturing sector protected areas in a global sample of 137 countries activity, etc). Nighttime lights are therefore more use- shows that, ceteris paribus, the relationship between ful in this analysis as a summary measure of over- inequality and protection varies depending on the all economic and infrastructure development [51]. strength of democracy: in relatively democratic coun- Additionally, note that nighttime lights data struggle tries inequality is associated with less land in protec- to measure the light emissions produced by poorer ted areas, whereas in relatively undemocratic coun- rural households in lower-income countries [59, 60]. tries the reverse is true [48]. It may be more appropriate to treat these data as Where appropriate, we log transform variables measures of development related to sectors of the eco- before including them in our analysis. See our sup- nomy aside from rural smallholder agriculture, rather porting information for more detail on our national- than presuming they provide a complete picture of level analysis, data sources (SI sections 1, 2, 4 and local variation in wealth. table S2), and for a full results table based on our Our second explanatory variable is the average primary regression model (table S3). After estimat- population density of a planning unit. To adjust for ing our primary model, we estimate four additional skew in the distribution of both explanatory vari- regression models that each replace GDP and the ables, as well as an established non-linear relation- Gini index with various alternative economic meas- ship between nighttime lights and economic devel- ures (and also omit the governance measures). The opment, we add a constant of 0.01 to both and then alternative economic measures are: (a) the percent log transform them before fitting our models [54, 61]. of farms smaller than 5 ha (as a proxy for the pre- Our supporting information provides more detailed valence of smallholder farming); (b) a measure of discussion of our parcel-level data, measurement and food availability (the percent of an average person’s transformation choices, and modeling choices (SI daily caloric needs that can be met by food availabil- appendix sections 1–4). It also includes parcel-level ity in a country); (c) the percent of a country’s citizens summary statistics (table S5). employed in agriculture; and (d) a country’s average We conduct this analysis within a multilevel mod- radiance at night (average ‘nighttime lights’), which eling framework, estimating random country-level is increasingly used as an indicator of national eco- intercepts. We also include additional control vari- nomic production [51], especially production linked ables for mean nighttime lights and total population to the manufacturing and service sectors [52–54]. at a country level (both logged and mean-centered). Again, see our supporting information for more detail For computational reasons, we approximate multi- on data sources, and for a full regression table includ- level logit models using a generalized additive mod- ing each of these additional models (table S4). eling (GAM) approach (see SI 4, tables S6 and S7). 2.4. Subnational-level regression analysis 3. Results Finally, we analyze approximately 820 000 planning units (parcels) from FAO member countries with at 3.1. Targeting the tropics for restoration least 20% of their land area in the tropics. Parcels were Figure 1 shows estimated converted (agricultural) only included in our sample if they also had a res- lands around the world, classified by their restoration toration priority level assigned and if they contained priority level [21]. As explained above, these priority at least 10% agricultural land. We first analyze the levels were assigned based on an efficiency criterion entire sample of parcels that meet our criteria. Then, (maximizing environmental benefits while minimiz- to help address possible spatial dependence concerns ing economic costs). Of all the 5 km × 5 km plan- [55], we repeat our analysis on a randomly chosen ning units around the world classified as a top 15% 5% of parcels, decreasing the likelihood that remain- priority, we find that approximately 85% are located ing parcels will be close together. A similar strategy in the tropics. Importantly, this prioritization ana- has been applied, for instance, in work examining the lysis is not unique in that regard—due to its high impacts of protected areas [56–58]. concentrations of biodiversity and carbon sequestra- Our dependent variable here is binary: whether a tion potential [20, 28], other analysts also identify the parcel is classified as a top 15% priority (Y = 1) or not humid tropics as a place where climate change mitiga- (Y = 0). We explain variation in this measure using tion strategies could be most efficiently concentrated both logit regression models and linear probability [18, 27, 62]. We return to this issue below. models (LPMs), though we focus on the linear prob- Next, for every FAO member state, we calculated ability model results for ease of interpretation. the proportion of its agricultural land classified as 4 Environ. Res. Lett. 17 (2022) 114019 B Schultz et al Figure 1. Estimated agricultural lands by their restoration priority. We mapped agricultural (i.e. converted) land in each FAO member country following a procedure described by Strassburg et al [21]. We then used their map of priority scores to classify these lands based on their restoration priority percentage and aggregated these percentages into five categories (see the legend). Figure 1e in their study presents similar information. a top 15% restoration priority (i.e. that would be Table 1. Agricultural land to restore under a global 15% plan following Strassburg et al [21]. restored under a hypothetical global 15% initiative guided by this map). We provide a sample of these Proportion of agricultural estimates in table 1. Our supplementary information Country land to restore includes an extended version of this table (table S1) Equatorial Guinea 99.6% and a color-coded map that visualizes these estim- Philippines 96.7% ates for every FAO member state (figure S1). Indeed, Nicaragua 87.2% many tropical countries would need to restore sub- Nepal 83.9% stantial portions of their agricultural land to natural Indonesia 82.5% ecosystems, whereas most countries outside the trop- El Salvador 62.4% Uganda 34.1% ics would not. Tanzania 31.5% Brazil 16.7% 3.2. The national implications of a prominent India 12.6% restoration map Malawi 10.6% Results from our primary national-level regression New Zealand 1.4% model (table S3), visualized in panel (a) of figure United States 0.1% 2, show that countries have less of their agricultural Ukraine 0.1% land classified as a top 15% restoration priority (‘top Germany <0.1% priority land’) if they are wealthier, less populated, If Strassburg et al [21] map from figure 1(e) of their study was and have less economic inequality. Panel (b) visual- followed to implement a 15% global restoration target, the izes the additional results for our regression models estimates in this table represent the percent of agricultural land that 15 select countries would need to restore to a natural state. that use alternative economic measures (table S4). We selected countries for this table based on the expertise of It shows that countries have more top priority res- members of our research team and with an eye towards toration land if they have higher employment in the representing different continents. agricultural sector and if farms are smaller (less than 5 ha). Countries have less top priority land if they have more food availability (% daily dietary energy to be top 15% restoration priorities. For instance, available), or if they are brighter at night (a proxy for per the linear approximations in Model 1 of table 2, economic development). when (logged) parcel lights increase by one standard deviation (+1.65), the probability that a parcel is a 3.3. Subnational implications within the tropics top 15% restoration priority declines by 3.47%. Fur- On average, 5 km × 5 km parcels of land in a trop- ther, the average within-country difference between ical country that are darker at night are more likely the brightest and dimmest (logged) parcel light value 5 Environ. Res. Lett. 17 (2022) 114019 B Schultz et al Table 2. Parcel level regression results (tropical countries). Dependent variable Top 15% priority planning unit? (0/1) LPM LPM Logit Logit (1) (2) (3) (4) Parcel lights (log) −0.021∗∗∗ −0.020∗∗∗ −0.139∗∗∗ −0.133∗∗∗ (0.0004) (0.002) (0.003) (0.012) Parcel population (log) 0.012∗∗∗ 0.011∗∗∗ 0.085∗∗∗ 0.073∗∗∗ (0.0003) (0.001) (0.002) (0.010) Country lights (log) 0.051∗∗∗ 0.048∗∗ 0.411∗∗ 0.359∗∗ (0.016) (0.019) (0.164) (0.154) Country population (log) −0.075∗∗∗ −0.067∗∗∗ −0.670∗∗∗ −0.415∗∗ (0.016) (0.021) (0.176) (0.175) Constant 0.240∗∗∗ 0.256∗∗∗ −2.093∗∗∗ −1.673∗∗∗ (0.052) (0.057) (0.518) (0.455) Random sample (5%) No Yes No Yes GAM No No Yes Yes sd (constant) 0.318 0.327 3.117 2.599 Countries 97 89 97 89 Observations 837 468 41 860 837 468 41 860 ∗∗ ∗∗∗ Note: p < 0.05 p < 0.01. This table presents coefficients from two multilevel linear probability models and two multilevel logit models (estimated using a generalized additive modeling approach). All models include a country-level random intercept and country-level controls for development and population (logged and mean-centered). Models 1 and 3 are fit for all parcels that satisfy our inclusion criteria (in a tropical country, at least 10% agricultural land, and a non-missing restoration priority score), while Models 2 and 4 are fit on a dataset produced by randomly sampling 5% of these parcels. Coefficients for the linear probability models represent average changes in Pr (Y = 1) associated with X + 1. Coefficients for the logit models instead represent changes in log odds. Figure 2. Country-level correlates of expected restoration on agricultural land. Results in panel (a) are based on coefficient estimates from a country-level regression model that also controls for a country’s land area. Results in panel (b) are based on coefficients from four separate linear regression models, each controlling for a country’s population and land area. For explanatory variables included in these regressions on a log scale, we use a 50% increase as a benchmark, and 1 s.d. increase otherwise. We present normal-theory 95% confidence intervals for these estimates based on HC2 robust errors. Sample size is based on data availability. Units are FAO member countries (out of 194 total). 6 Environ. Res. Lett. 17 (2022) 114019 B Schultz et al is 7.03. Model 1 therefore suggests that the brightest top restoration priority are also those whose popula- parcels in an average tropical country are approxim- tions are most dependent on agriculture. While the ately 14.7% less likely to be top restoration priorit- restoration of nature can facilitate human well-being, ies than the dimmest parcels. Moving to the result for these benefits may not outweigh the costs incurred population, our analysis shows that more populated where activities like farming are the main source of parcels are more likely to be top restoration priorities. food security and income. Plans that do not explicitly A one standard deviation increase in logged popula- call for restoring agricultural land could still threaten tion (+2.55) is associated with a 3.06% increase in the communities in highly populated, agriculturally reli- probability of being a top restoration priority, while ant countries due to increased competition for land the average within-country range of logged popula- or land-use displacement [11, 41, 42]. tion (8.46) implies a 10.15% increase. This finding may seem obvious, in that the eco- nomies of tropical countries are known to be relat- 4. Discussion ively more dependent on agriculture. However, that obviousness raises a troubling question: why have the We provide an illustrative case study of a prioritiz- possible negative impacts of concentrating restorative ation analysis that sought to locate the restoration activity in lower-income countries been an infrequent of agricultural land in areas where it would provide consideration in priority setting analyses? the greatest ecological benefits (in terms of habitat Second, our subnational regression analyses sug- protection and carbon storage) at the lowest cost gest that areas of the tropics with relatively fewer live- [21]. We show that the map produced by this ana- lihood opportunities besides agriculture (i.e. with less lysis concentrates top priority restoration areas in the light-emitting economic activity) are also more likely global tropics. Moreover, we show that many trop- to be targeted for restoration. Again, it could also be ical countries have substantial portions of their agri- argued that this finding is not surprising. Unpopu- cultural land classified as a top restoration priority, lated parcels, or those with a low population, may rep- while countries outside the tropics largely do not. It is resent existing natural ecosystems that need less res- worth emphasizing that countries at the top of table 1, torative activity. Meanwhile, parcels that are brighter for example, would not be able to restore so much at night may represent areas where expensive infra- agricultural land—food needs, employment needs, structure is more concentrated, leading to higher res- and the cultural value of agricultural lands would toration opportunity costs. However, again, this sup- clearly prohibit it. Moreover, earning the cooperation posed obviousness is part of our point. Prioritization of decision-makers in many countries with a restora- analyses may often locate restoration and conserva- tion initiative that involves substantial loss of agricul- tion priorities in places where these activities raise tural land would require a complicated political pro- greater social equity concerns. Seemingly rational pri- cess, as would developing a program to compensate oritization criteria may obscure equity issues that communities whose livelihoods are displaced [24]. need to be discussed more openly. Similar compensation programs have struggled in the More broadly, we reiterate that restoration inter- past [63–65], and it is not clear how an initiative to ventions have the potential to benefit or harm substantially increase restoration in the tropics would local livelihoods depending on their consideration overcome this hurdle. of social equity [35–37, 66]. This includes the dis- Of course, many other analysts also point out tributional concerns that we highlight above, but the possible benefits of concentrating restoration also inter-related procedural and recognitional con- and conservation interventions in the tropics. For cerns regarding whose interests are acknowledged and instance, Bastin et al argue that loss of forest habitat incorporated into restoration decision-making. Addi- in the tropics presents the greatest threat to global tionally, it includes contextual concerns regarding potential canopy cover [18]. Similarly, Walker et al how interventions are adapted to their circumstances. seek to map the global distribution of carbon stor- In turn, whether it is problematic to allow optim- age opportunities, and in particular the distribution ization exercises to guide global restoration prior- of unrealized opportunities for carbon storage [27]. ities also depends on their engagement with these These authors argue that a majority of unrealized car- issues. For example, it depends on analysts’ decisions bon storage potential is found in tropical regions. about what constitutes ‘restoration.’ Mapping exer- Lewis et al provide a commentary which argues that cises often estimate the benefits of restoration by expanding natural forest restoration in the Tropics assuming complete land use change [19, 21, 22]; and Subtropics is essential to fighting climate change, e.g. restoring a pasture entirely to natural forest. How- given the quick pace at which trees grow and sequester ever, it is also possible to integrate restoration into carbon, and this region’s cheaper land prices [62]. existing agrarian landscapes, which requires less dis- Our regression analyses help better illustrate the placement of local land uses [10, 67, 68]. Making distributional equity issues that such arguments raise. an informed choice between complete and partial First, our national level regression results imply that land use change requires weighing the different social countries with more agricultural land classified as a impacts of these strategies, as well as their relative 7 Environ. Res. Lett. 17 (2022) 114019 B Schultz et al contributions to carbon storage and biodiversity pro- of prioritization maps. Recent defenses of restora- tection. Extant prioritization analyses often do not tion mapping move in the right direction by laying provide sufficient information on these tradeoffs. As out some possible benefits of these exercises more another example, prioritization analysis also requires precisely [25]. But there is also a need for more discus- making decisions about how the human population sion about whether these benefits outweigh the pos- of an area should influence its status as a restoration sible consequences of prioritization exercises: first, or conservation priority. While Strassburg et al [21] a risk of concentrating restoration in regions where focus on restoring agricultural land, other studies rule possible negative livelihood impacts would be the out crop lands, grazing lands, and areas of human most severe [39, 43, 44, 81, 82]; and second, a risk habitation ex ante on the grounds of equity and of crowding-out other strategies for identifying more feasibility [27, 69]. just and pragmatic policy goals [23] or undermining Relatedly, it is important to critically evaluate the the ability of local stakeholders to participate in land data inputs to global environmental assessments [70]. use decisions [47]. Both consequences point to a need The various decisions we highlight above should, on for complementary, bottom-up approaches to setting equity grounds, consider the interests and welfare of restoration priorities. affected communities [35, 71], but omitting relevant social data makes doing so impossible. Moreover, res- Data availability statement toration and conservation prioritization analyses may not properly account for the spatial distribution of The data that support the findings of this study are various costs [72–74], the extent of feasible land area available upon reasonable request from the authors. [69], or important spatial variation in the probabil- ity of success [75]. Information on land tenure, local Acknowledgements socio-economic conditions, and the priorities of gov- ernment officials may be important to understand- The authors are grateful for generous support from ing restoration’s true costs in a given locale, or the Formas, a Swedish government research council probability that its intended benefits will materialize focused on sustainable development (grant numbers [14, 76]. But prioritization studies do not commonly 2020-02781 and 2020-02872). include this information in their analyses. As we point out above, Strassburg et al call atten- Author contributions tion to several of the issues we discuss, including the large proportion of agricultural land in some E C, F F, and P K conceived the study. B S led the writ- countries that is classified as a top 15% restora- ing process and conducted the analyses. All authors tion priority [21]. They also discuss the need for contributed to writing and the development of our deeper consideration of equity issues. Our explora- arguments. tion of one of their output maps takes this call fur- ther. Incorporating more social data into prioritiz- Conflict of interest ation analyses will help re-orient those who create such maps towards a broader conceptualization of We have no competing interests to disclose. sustainability [35, 49] and aid in the search for com- promise solutions between opponents and support- ORCID iDs ers of global mapping exercises. Ideally, prioritization analyses should provide opportunities for affected Bill Schultz  https://orcid.org/0000-0002-6180- communities to participate in the mapping process, 6679 helping these analyses to better consider local know- Eric A Coleman  https://orcid.org/0000-0002- ledge, experiences, and aspirations. Examples include 5315-7111 mapping done through the Restoration Opportun- Prakash Kashwan  https://orcid.org/0000-0001- ities Assessment Methodology [77, 78], as well as 5540-5994 a recently developed systems approach to evaluat- Rose Pritchard  https://orcid.org/0000-0002-6174- ing tree-restoration interventions which incorporates 3460 indicators of social and wellbeing outcomes along- side indicators of biodiversity and climate change References mitigation [79]. Participatory mapping processes can [1] Gann G D et al 2019 International principles and standards be further strengthened with land tenure diagnostic for the practice of ecological restoration Restor. Ecol. tools [80]. Assessing the tradeoffs that restoration 27 S1–S46 may imply under varying governance and land use [2] Besseau P, Graham S and Christophersen T 2018 Restoring arrangements should become a more foundational Forests and Landscapes: The Key to a Sustainable Future Global Partnership on Forest and Landscape Restoration aspect of restoration planning [79]. [3] Fagan M E, Reid J L, Holland M B, Drew J G and Zahawi R A In conclusion, we underscore the importance 2020 How feasible are global forest restoration of engaging in critical discussions about the goals commitments? Conserv. Lett. 13 e12700 8 Environ. Res. Lett. 17 (2022) 114019 B Schultz et al [4] Aronson J, Goodwin N, Orlando L, Eisenberg C and [28] Soto-Navarro C et al 2020 Mapping co-benefits for carbon Cross A T 2020 A world of possibilities: six restoration storage and biodiversity to inform conservation policy and strategies to support the United Nation’s Decade on action Phil. Trans. R. Soc. B 375 20190128 Ecosystem Restoration Restor. Ecol. 28 730–6 [29] Possingham H P, Bode M and Klein C J 2015 Optimal [5] Meyfroidt P et al 2022 Ten facts about land systems for conservation outcomes require both restoration and sustainability Proc. Natl Acad. Sci. 119 e2109217118 protection PLoS Biol. 13 e1002052 [6] Griscom B W et al 2017 Natural climate solutions Proc. Natl [30] Dinerstein E et al 2020 A “Global Safety Net” to reverse Acad. Sci. 114 11645–50 biodiversity loss and stabilize Earth’s climate Sci. Adv. [7] Lamb D, Erskine P D and Parrotta J A 2005 Restoration of 6 eabb2824 degraded tropical forest landscapes Science 310 1628–32 [31] Grainger A et al 2019 Comment on “The global tree [8] Aradottir A L and Hagen D 2013 Ecological restoration: restoration potential” Science 366 eaay7976 approaches and impacts on vegetation, soils and society Adv. [32] Veldman J W, Aleman J C, Alvarado S T, Anderson T M, Agron. 120 173–222 Archibald S, Bond W J 2019 Comment on “The global tree [9] Das S 2017 Ecological restoration and livelihood: restoration potential” Science 366 contribution of planted mangroves as nursery and habitat [33] Meyer H and Pebesma E 2022 Machine learning-based for artisanal and commercial fishery World Dev. global maps of ecological variables and the challenge of 94 492–502 assessing them Nat. Commun. 13 1–4 [10] Erbaugh J T, Pradhan N, Adams J, Oldekop J A, Agrawal A, [34] Kashwan P, Biermann F, Gupta A and Okereke C 2020 Brockington D, Pritchard R and Chhatre A 2020 Global Planetary justice: prioritizing the poor in earth system forest restoration and the importance of prioritizing local governance Earth Syst. Gov. 6 100075 communities Nat. Ecol. Evol. 4 1472–6 [35] Leach M, Reyers B, Bai X, Brondizio E S, Cook C, Díaz S, [11] Malkamäki A, D’Amato D, Hogarth N J, Kanninen M, Espindola G, Scobie M, Stafford-Smith M and Pirard R, Toppinen A and Zhou W 2018 A systematic review Subramanian S M 2018 Equity and sustainability in the of the socio-economic impacts of large-scale tree anthropocene: a social–ecological systems perspective on plantations, worldwide Glob. Environ. Change 53 90–103 their intertwined futures Glob. Sustain. 1 1–13 [12] Adams C, Rodrigues S T, Calmon M and Kumar C 2016 [36] Friedman R S, Law E A, Bennett N J, Ives C D, Thorn J P R Impacts of large-scale forest restoration on socioeconomic and Wilson K A H J 2018 How just and just how? A status and local livelihoods: what we know and do not know systematic review of social equity in conservation research Biotropica 48 731–44 Environ. Res. Lett. 13 53001 [13] Coleman E A et al 2021 Limited effects of tree planting on [37] Pritchard R, Sauls L A, Oldekop J, Kiwango W and forest canopy cover and rural livelihoods in Northern India Brockington D 2022 Data justice and biodiversity Nat. Sustain. 4 1–8 conservation Conserv. Biol. 36 1–10 [14] Fleischman F et al 2020 Pitfalls of tree planting show why we [38] Agrawal A and Redford K 2009 Conservation and need people-centered natural climate solutions Bioscience displacement: an overview Conserv. Soc. 7 1–10 70 947–50 [39] Gerber J-F 2011 Conflicts over industrial tree plantations in [15] Holl K D and Brancalion P H S 2020 Tree planting is not a the South: who, how and why? Glob. Environ. Change simple solution Science 368 580–1 21 165–76 [16] Wilson K A, Davis K J, Matzek V and Kragt M E 2019 [40] Andersson K, Lawrence D, Zavaleta J and Guariguata M R Concern about threatened species and ecosystem disservices 2016 More trees, more poverty? The socioeconomic effects underpin public willingness to pay for ecological restoration of tree plantations in Chile, 2001–2011 Environ. Manage. Restor. Ecol. 27 513–9 57 123–36 [17] Shackleton C M, Ruwanza S, Sanni G K S, Bennett S, De [41] Latawiec A E, Strassburg B B N, Brancalion P H S, Lacy P, Modipa R, Mtati N, Sachikonye M and Thondhlana G Rodrigues R R and Gardner T 2015 Creating space for 2016 Unpacking Pandora’s box: understanding and large-scale restoration in tropical agricultural landscapes categorising ecosystem disservices for environmental Front. Ecol. Environ. 13 211–8 management and human wellbeing Ecosystems 19 587–600 [42] Melick D, Yang X and Xu J 2006 Seeing the wood for the [18] Bastin J-F, Finegold Y, Garcia C, Mollicone D, Rezende M, trees: how conservation policies can place greater pressure Routh D, Zohner C M and Crowther T W 2019 The global on village forests in southwest China Plant Conservation and tree restoration potential Science 365 76–79 Biodiversity (Berlin: Springer) pp 385–97 [19] Brancalion P H S et al 2019 Global restoration opportunities [43] Ramprasad V, Joglekar A and Fleischman F 2020 Plantations in tropical rainforest landscapes Sci. Adv. 5 eaav3223 and pastoralists: afforestation activities make pastoralists in [20] Cook-Patton S C et al 2020 Mapping carbon accumulation the Indian Himalaya vulnerable Ecol. Soc. 25 potential from global natural forest regrowth Nature [44] Henrique K P and Tschakert P 2019 Contested grounds: 585 545–50 adaptation to flooding and the politics of (in) visibility in [21] Strassburg B B N et al 2020 Global priority areas for São Paulo’s eastern periphery Geoforum 104 181–92 ecosystem restoration Nature 586 724–9 [45] Sharma P et al 2020 Mapping human–wildlife conflict [22] Strassburg B B N et al 2019 Strategic approaches to restoring hotspots in a transboundary landscape, Eastern Himalaya ecosystems can triple conservation gains and halve costs Nat. Glob. Ecol. Conserv. 24 e01284 Ecol. Evol. 3 62–70 [46] Ghoddousi A, Buchholtz E K, Dietsch A M, Williamson M A, [23] Wyborn C and Evans M C 2021 Conservation needs to break Sharma S, Balkenhol N, Kuemmerle T and Dutta T 2021 free from global priority mapping Nat. Ecol. Evol. 5 1322–4 Anthropogenic resistance: accounting for human behavior in [24] Jung M et al 2021 Areas of global importance for conserving wildlife connectivity planning One Earth 4 39–48 terrestrial biodiversity, carbon and water Nat. Ecol. Evol. 5 [47] McCarthy J and Thatcher J 2019 Visualizing new political 1–11 ecologies: a critical data studies analysis of the World Bank’s [25] Chaplin-Kramer R et al 2022 Conservation needs to renewable energy resource mapping initiative Geoforum integrate knowledge across scales Nat. Ecol. Evol. 6 118–9 102 242–54 [26] Schmidt-Traub G 2021 National climate and biodiversity [48] Kashwan P 2017 Inequality, democracy, and the strategies are hamstrung by a lack of maps Nat. Ecol. Evol. environment: a cross-national analysis Ecol. Econ. 1–3 131 139–51 [27] Walker W S et al 2022 The global potential for increased [49] Allan J R et al 2022 The minimum land area requiring storage of carbon on land Proc. Natl Acad. Sci. conservation attention to safeguard biodiversity Science 119 e2111312119 376 1094–101 9 Environ. Res. Lett. 17 (2022) 114019 B Schultz et al [50] Lindberg S I, Coppedge M, Gerring J and Teorell J 2014 Embodiments in sub-Saharan Africa and implications for V-Dem: a new way to measure democracy J. Democr. future design Environ. Manage. 68 619–41 25 159–69 [68] Chazdon R L, Cullen J L, Padua S M and Padua C V 2020 [51] Asher S, Lunt T, Matsuura R and Novosad P 2021 People, primates and predators in the Pontal: from Development Research at High Geographic Resolution: an endangered species conservation to forest and landscape analysis of night-lights, firms, and poverty in India using the restoration in Brazil’s Atlantic Forest R. Soc. Open Sci. SHRUG open data platformevelopment research at high 7 200939 geographic resolution World Bank Econ. Rev. 35 845–71 [69] Gopalakrishna T, Lomax G, Aguirre-Gutiérrez J, Bauman D, [52] Ghosh T, L Powell R, D Elvidge C, E Baugh K, C Sutton P Roy P S, Joshi P K and Malhi Y 2022 Existing land uses and Anderson S 2010 Shedding light on the global constrain climate change mitigation potential of forest distribution of economic activity Open Geogr. J. 3 147–60 restoration in India Conserv. Lett. 15 e12867 [53] Smith B and Wills S 2018 Left in the dark? Oil and rural [70] Scoones I 2022 Livestock, methane, and climate change: the poverty J. Assoc. Environ. Resour. Econ. 5 865–904 politics of global assessments Wiley Interdiscip. Rev. Clim. [54] Weidmann N B and Schutte S 2017 Using night light Change e790 emissions for the prediction of local wealth J. Peace Res. [71] Pritchard R 2021 Politics, power and planting trees Nat. 54 125–40 Sustain. 4 932 [55] Mets K D, Armenteras D and Dávalos L M 2017 Spatial [72] Iacona G D et al 2018 Standardized reporting of the costs of autocorrelation reduces model precision and predictive management interventions for biodiversity conservation power in deforestation analyses Ecosphere 8 e01824 Conserv. Biol. 32 979–88 [56] Nolte C, Agrawal A, Silvius K M and Soares-Filho B S 2013 [73] Iftekhar M S, Polyakov M, Ansell D, Gibson F and Kay G M Governance regime and location influence avoided 2017 How economics can further the success of ecological deforestation success of protected areas in the Brazilian restoration Conserv. Biol. 31 261–8 Amazon Proc. Natl Acad. Sci. 110 4956–61 [74] Nolte C 2020 High-resolution land value maps reveal [57] Ferraro P J, Hanauer M M, Miteva D A, underestimation of conservation costs in the United States Canavire-Bacarreza G J, Pattanayak S K and Sims K R E 2013 Proc. Natl Acad. Sci. 117 29577–83 More strictly protected areas are not necessarily more [75] Wilson K A, Carwardine J and Possingham H P 2009 protective: evidence from Bolivia, Costa Rica, Indonesia, and Setting conservation priorities Ann. New York Acad. Sci. Thailand Environ. Res. Lett. 8 025011 1162 237–64 [58] Zhao H et al 2019 Individual-level performance of nature [76] Fleischman F D 2014 Why do foresters plant trees? Testing reserves in forest protection and the effects of management theories of bureaucratic decision-making in central India level and establishment age Biol. Conserv. 233 23–30 World Dev. 62 62–74 [59] Gibson J, Olivia S and Boe-Gibson G 2020 Night lights in [77] Maginnis S, Laestadius L, Verdone M, DeWitt S, economics: sources and uses 1 J. Econ. Surv. 34 955–80 Saint-Laurent C, Rietbergen-McCracken J and Shaw D M [60] Keola S, Andersson M and Hall O 2015 Monitoring 2014 A guide to the Restoration Opportunities Assessment economic development from space: using nighttime light Methodology (ROAM): Assessing forest landscape restoration and land cover data to measure economic growth World Dev. opportunities at the national or sub-national level : working 66 322–34 paper (road test edition) Int. Union Conserv. Nat. [61] Bruederle A and Hodler R 2018 Nighttime lights as a proxy [78] Hanson C, Buckingham K, DeWitt S and Laestadius L 2015 for human development at the local level PLoS One The restoration diagnostic: a method for developing forest 13 e0202231 landscape restoration strategies by rapidly assessing the status [62] Lewis S L, Wheeler C E, Mitchard E T A and Koch A 2019 of key success factors World Resources Institute Restoring natural forests is the best way to remove [79] Pfeifer M et al 2021 A systems approach framework for atmospheric carbon Nature 568 25–28 evaluating tree restoration interventions for wellbeing and [63] Pelletier J, Horning N, Laporte N, Samndong R A and ecological outcomes in rural tropical landscapes Phil. Trans. Goetz S 2018 Anticipating social equity impacts in REDD+ R. Soc. B (https://doi.org/10.1098/rstb.2021.0111) policy design: an example from the Democratic Republic of [80] McLain R, Lawry S, Guariguata M R and Reed J 2021 Toward Congo Land Use Policy 75 102–15 a tenure-responsive approach to forest landscape restoration: [64] Poudel M, Thwaites R, Race D and Dahal G R 2015 Social a proposed tenure diagnostic for assessing restoration equity and livelihood implications of REDD+ in rural opportunities Land Use Policy 104 103748 communities—a case study from Nepal Int. J. Commons [81] McElwee P 2009 Reforesting “bare hills” in Vietnam: social 9 177 and environmental consequences of the 5 million hectare [65] Duchelle A E, De Sassi C, Jagger P, Cromberg M, Larson A M, reforestation program Ambio 38 325–33 Sunderlin W D, Atmadja S S, Resosudarmo I A P and [82] Andersson K P, Gibson C C and Lehoucq F 2006 Municipal Pratama C D 2017 Balancing carrots and sticks in REDD+ politics and forest governance: comparative analysis of implications for social safeguards Ecol. Soc. 22 decentralization in Bolivia and Guatemala World Dev. [66] Wells H B M, Kirobi E H, Chen C L, Winowiecki L A, 34 576–95 Vågen T-G, Ahmad M N, Stringer L C and Dougill A J 2021 [83] Einhorn C 2020 Restoring farmland could drastically slow Equity in ecosystem restoration Restor. Ecol. 29 e13385 extinctions, fight climate change (The New York Times) [67] Djenontin I N S, Zulu L C and Etongo D 2020 Ultimately, (available at: www.nytimes.com/2020/10/14/climate/ what is forest landscape restoration in practice? biodiversity-farmland-extinction.html) 10