Bringing the forest back: Restoration priorities in Colombia

Colombia has committed to ambitious forest restoration targets which include a 1 million ha Bonn Challenge commitment and 6.47–8.31 million ha (rehabilitation and restoration, respectively) under the National Restoration Plan. Determining where and how to implement programs to achieve these targets remains a significant challenge.


| INTRODUC TI ON
Nations have committed to increasingly ambitious forest restoration targets and multiple global initiatives have risen to support them to reverse biodiversity loss and climate change.For example, the United Nations (UN) has declared 2021-2030 the 'UN Decade on Ecosystem Restoration', the Bonn Challenge and the New York Declaration on Forests aim to restore 350 million ha worldwide by 2030, and Target 2 of the recently adopted Global Biodiversity Framework states: '… that by 2030 at least 30 per cent of areas of degraded terrestrial, inland water, and marine and coastal ecosystems are under effective restoration, in order to enhance biodiversity and ecosystem functions and services, ecological integrity and connectivity' (CBD, 2021;IUCN, 2020;United Nations, 2014, 2021).
However, determining where and how nations implement these ambitious forest restoration targets remains a significant challenge.
The complexity of navigating multiple motivations for implementing forest landscape restoration, and balancing these objectives with socio-economic goals, necessitates the use of decision support frameworks (Strassburg et al., 2019(Strassburg et al., , 2020)).Planning without systematic frameworks can lead to wasted resources, missed opportunities for gains in conservation or socio-economic benefits and project failure (López-Cubillos et al., 2022;Margules & Pressey, 2000).For instance, López-Cubillos et al. (2022) show that improved biodiversity and social outcomes in Colombia can be obtained through a systematic planning framework for the Payment for Ecosystem Services national policy, highlighting a more effective use of a limited national budget.Building upon previous approaches to optimise cost-effective restoration planning for carbon and biodiversity (Beyer et al., 2021;Strassburg et al., 2019Strassburg et al., , 2020)), we present a forest restoration assessment for Colombia.This is particularly important for Colombia as it is a megadiverse country, but has a history of complex landscape transformation and a rapidly expanding human footprint with an annual deforestation rate of 0.62% per year (Armenteras et al., 2013;Correa Ayram et al., 2020).This has left 25% of its ecosystems critically endangered (Etter et al., 2020).Restoration activities in Colombia are frequently underscored as an important strategy for reversing Colombia's biodiversity declines (Calle & Holl, 2019;Ocampo-Peñuela et al., 2022;Torres et al., 2022).Our restoration assessment takes into account potential conflicts with agriculture production, the capacity for natural forest regeneration, and aligns with Colombia's national targets around forest regeneration.Notably, we leverage the potential for assisted natural regeneration as it is gaining recognition as a compelling lower-cost approach for large-scale forest landscape restoration (Chazdon & Guariguata, 2016).
The complexity of navigating multiple motivations for implementing forest landscape restoration is particularly crucial for Colombia, where forest restoration is an important conservation and sustainable development objective of the government (Aguilar et al., 2015).
and establishment costs of restoration and maximising biodiversity conservation and climate change mitigation benefits.We explore four politically relevant restoration area-based targets (1, 6, 6.47 and 8.31 million ha) and identify minimum cost, and suites of maximum benefit and cost-effective solutions.

Results:
We identify solutions that simultaneously perform well across biodiversity and carbon objectives, despite trade-offs between these objectives.We find that cost-effective solutions can achieve on average 91.1%, 90.8%, 90.5% and 90.1% of maximum carbon benefit and 100% of the maximum biodiversity benefit while significantly reducing costs.On average, the cost-effective solutions reduce the cost by 87.5%, 56.8%, 59.6% and 46.2% compared to the maximum benefit solutions considering one, six, 6.47 and 8.31 million ha restoration targets, respectively.In a more direct approach, Colombia recently adopted an ambitious 20-year National Restoration Plan which recognises that many economic, social, political and even religious factors have led to unsustainable production, and the degradation and loss of natural ecosystems (Ospina Arango et al., 2015).The plan aims to ensure the sustainability of the natural environment, while balancing other socio-economic factors (Ospina Arango et al., 2015).The National Restoration Plan categorises areas across Colombia based on three approaches -to repair ecosystems, to restore cleared and rehabilitate degraded land, and to recuperate severely degraded areas -but the plan lacks realistic objectives and steps for implementation (Etter et al., 2020).The plan selected over 24 million ha as priorities for restoration, but restoring two thirds of the agricultural frontier of the country has been criticised as being unrealistic (Etter et al., 2020;Ospina Arango et al., 2015).

Main
To date, forest restoration in Colombia has occurred without explicit consideration of multiple benefits and cost, particularly those projects that were government-led (Murcia et al., 2016), leaving a crucial gap for strategic restoration planning to fill and help Colombia meet its ambitious restoration objectives (Etter et al., 2020).Previous analyses have identified 18 million ha of priorities for restoration based on endangered ecosystems using a multi-criteria analysis (Etter et al., 2020), and 10.3 million ha based on soil degradation, land suitability and land-use conflicts through an overlay analysis (Sylvester et al., 2020).The National Restoration Plan intersected natural areas from 2000 to 2002 with a disturbance map (agriculture, mining, low vegetation and urban) from 2005 to 2009 to explore transformation during that period (Ospina Arango et al., 2015).This plan highlights that there are 8.1 million ha for restoration (transition from agricultural use to forest), 8.4 million ha for rehabilitation (low and moderate degradation) and 6.8 million ha for recuperation (severe degradation) (Ospina Arango et al., 2015).Additionally, the Register of Ecosystems and Environmental Areas (REAA) is useful for creating maps of important ecosystems available to decision makers; however, all of these efforts fail to systematically assess the costs and benefits of different scenarios while assessing the trade-offs between biodiversity and carbon sequestration objectives (SIAC, 2016).
We address this gap by adopting a multi-objective optimisation framework to identify cost-effective solutions that leverage the potential for assisted forest natural regeneration benefits while accounting for restoration opportunity and establishment costs, maximise biodiversity conservation and climate change mitigation benefits, describe the trade-offs between objectives, and assess a suite of politically relevant scenarios.In doing so, we provide an evidence base to inform on proactive planning for environmentally sensible restoration policy and practice in Colombia.

| Study region
Over half of Colombia's territory was originally forested, thus the country has great potential for carbon sequestration to help meet global climate mitigation objectives through forest restoration (Phillips et al., 2016).To determine which areas were historically forested in Colombia, we used a potential ecosystem map developed by (Etter et al., 2017) (SM 1.2) and a recently released land cover map from the Institute of Hydrology, Meteorology and Environmental Studies (IDEAM) (IDEAM, 2018).Approximately 58.7% of the megadiverse country has been converted into anthropogenic land uses (IDEAM, 2018), and 258,231 km 2 were available for forest restoration in the country (Figure 1), which we partitioned into 3,094,040 planning units 500 × 500 m in size.

| Overview
We used spatial prioritisation, the process of using computational tools for the informed spatial allocation of actions or placement land uses, to achieve an objective of restoring forest to maximise biodiversity and carbon sequestration benefits within selected priorities, while considering establishment and opportunity cost.Tree planting and extensive site preparation are popular restoration strategies and can be effective, but implementation can be prohibitively expensive for some sites or at large scales.Where ecological conditions are such that forests can grow back on their own or with low-cost assistance, natural regeneration methods can be less costly (Williams et al., 2023).To leverage these potential costs, our establishment cost estimates account for the potential for natural regeneration by adjusting values relative to a spatially explicit random forest model (Williams et al., 2023).

| Integration of outputs into policy and management processes
We explored four different restoration target levels that are politically relevant to Colombia.The first two targets were set by governmental bodies according to the rehabilitation and restoration goals from the National Restoration Plan (Ospina Arango et al., 2015), these were 8.31 and 6.47 million ha.The third target consists of Colombia's one million ha commitment under the Bonn Challenge (a global goal to restore 350 million ha of forest by 2030 (IUCN, 2020)).Finally, we used a six million ha target which was the area identified by a restoration prioritisation exercise that sought to ensure the inclusion of the Red List of Ecosystems in restoration planning (Etter et al., 2020).
For each area target level, we explored 22 scenarios representing a suite of solutions that correspond to different relative weights between carbon and biodiversity benefits, and that collectively describe a trade-off curve between these objectives.Specifically: 10 benefit-only scenarios, 10 cost-effectiveness scenarios, a cost-only reference scenario, and a random prioritisation reference scenario.
In the suite of scenarios that consider the potential for assisted natural regeneration, establishment costs are reduced (through multiplication) according to a ratio of 0.37.This ratio was derived from information on the se.plan model's global establishment cost layer, which was based on forest restoration project data (n = 234 observations) extracted from reports for 50 World Bank projects spanning 24 countries (SEPAL, 2023).The ratio describes the average ratio of establishment costs for assisted natural regeneration relative to tree planting based on available project data.

| Framing the decision support problem
We formulate the problem of where to restore forest to maximise benefits relative to costs as a linear programming problem, which we quantify in a range of ways among the scenarios.Specifically, the problem considers biodiversity conservation and climate change mitigation objectives, while accounting for establishment and opportunity costs: where x is a vector of decision variables representing the proportion of each planning unit i to restore; s is the expected change in carbon sequestration resulting from forest regeneration relative to current land cover conditions; b is the expected benefit to biodiversity conservation, summed across all species, following restoration (described in detail below); and c and e are the land opportunity costs and establishment costs associated with restoration respectively.The fractions within the summation signs are benefit-cost ratios.Carbon sequestration, biodiversity and cost metrics are quantified as rates per unit area of restoration.The relative contribution of climate change mitigation and biodiversity conservation objectives is determined by the weights w s and w b (which sum to 1).These weightings are required because the equivalence of objectives with different units is a subjective judgement to be made by planners or decision-makers.
The two components of the objective function represent the returns (benefits divided by costs) of forest restoration: climate change mitigation (s/(c + e); Mg ha −1 USD$ −1 500 m −2 ) and biodiversity conservation (b/(c + e); USD$ −1 500 m −2 -reduction in extinction risk divided by cost per planning unit) for each species j, where the total cost of forest restoration is the sum of the land opportunity costs (c; USD$ −1 500 m −2 ) and the establishment costs (e; USD$ −1 500 m −2 ).
N p is the total number of planning units and N s is the total number of species.The first constraint limits the total restoration area to target T, which restricts how many planning units can be included in a given solution.The second constraint limits the proportion of each planning unit that can be restored, where u (range: 0-1) is determined by the proportion of each planning unit containing cover types that are not available for restoration (e.g.water and urban areas).
We use this problem formulation to explore (i) a suite of costeffective solutions, (ii) a suite of maximum-benefit solutions, and (iii) a minimum cost solution.Equation 1is the problem formulation that identifies cost-effective reforestation planning solutions (i) where benefits are divided by costs.For comparison, a maximum-benefit formulation (ii) was also calculated in which the cost denominator is removed, as this provides insight into the locations that improve cost-effectiveness.A reference solution was also calculated that did not involve optimisation.The minimum cost solution (iii), which we included to act as a counterfactual scenario where costs are prioritised above all other objectives, restored forest in planning units in order of ascending cost until the target was achieved.See Figure 2 for the methodological flow.

| Biodiversity conservation benefit
We estimated the biodiversity conservation benefit of restoring an area as the average reduction in local (national) extinction risk among all forest-associated species, based on the extinction risk model of Strassburg et al., (2019) and Thomas et al., (2004).This function is based on a re-working of the species-area relationship and operates at the level of individual species.The approach is imperfect, as it ignores the possibility of negative densitydependence at very low population sizes, and does not consider the time scale of resulting extinctions, which will vary with the life history and ecology of a species.However, unlike simpler formulations, it takes into account the non-linearity of the response of persistence to changes in population size, and has been used in several similar studies, many with restoration applications (Strassburg et al., 2012(Strassburg et al., , 2017;;Thomas et al., 2004).If the existing habitat area is small, there is a large benefit to increasing that area, but as the area of habitat increases, there is a diminishing benefit for the addition of more habitat area (Strassburg et al., 2019).The extinction risk (e) for each individual species as a function of habitat area was modelled as follows: where a is the current habitat area, or future projected habitat area following restoration, A 0 refers to the original habitat area, corresponding to pre-settlement conditions, and z determines how extinction risk increases as habitat is lost (here, as per previous studies Reptilia (n = 95) being included (Table 1).
The contribution of each species to the extinction risk reduction benefit is determined by a species weighting scheme that assigns equal total weights between species classes.Within each of those taxonomic groups the total weight is divided equally among all species in that group, and the weights sum to unity among all species.
Without a weighting scheme Aves and Magnoliopsida would have a disproportionately large influence on the solution due to their disproportionately high representation among the set of species we include.Through the use of the species extinction risk function, species that have a larger difference between their original extent of occurrence and their current area of occupancy have a higher extinction risk, and thus are prioritised.For this reason, there is no need to weight species based on their threat status.et al., 2006, 2004), but have not been validated by experts.Model overprediction was controlled by integrating spatial information into the fitting process of ecological niche models.Thus, the species distribution maps were refined overlaying the binary distribution map without spatial constraints on species occurrence records, and then selecting only the potential distribution areas supported by concrete evidence of the species' presence.This method operates under the assumption that suitable patches that overlap with species occurrences are more likely to be components of species distributions.This set of models was evaluated using AUC statistics, and only those models with an AUC > 0.7 were retained.AGB.For our study, we focused on areas suitable for restoration.

| Habitat maps
Pixels with more than 50% OGF-AGB were excluded, being less suitable for restoration.The remaining pixels, appropriate for restoration, were consolidated by summing the 100 m pixels into a coarser 500 m resolution.This approach enabled us to estimate the potential carbon uptake for each pixel, while also accounting for uncertainty.with all units in USD$ ha −1 .For extended cost methods see SM1.7.

| The potential for natural regeneration
We used a global model of the potential for natural forest regeneration (Williams et al., 2023) to determine the potential cost-savings that could be gained by considering the potential for assisted natural regeneration.This model (random forest) was built upon on an assessment of tropical and subtropical forest assisted natural regeneration between 2000 and 2016 to develop a model of the potential for assisted natural regeneration, which was applied to generate 30 m resolution predictions for tropical and subtropical regions globally (Fagan et al., 2022).The model is based on 42 biophysical variables including tree cover, climate, and soil variables.The model results are publicly available for download from the citation.

| Comparison of restoration priorities against Colombia's agricultural frontier
To explore potential challenges of implementing restoration priorities identified in this analysis, we overlaid them with the National Agricultural Frontier (Frontera Agricola Nacional) and Exclusion Zone (Zona de Exclusión).The National Agricultural Frontier (Frontera Agricola Nacional) is a national government-defined agricultural zone.
The aim of this zonation is to guide the formulation of public policy, For more details please see DataS1 -Extended Methods.

| RE SULTS
We found that across all target levels there is still strong potential for achieving both biodiversity and carbon benefits simultaneously.
Interestingly, at higher restoration target levels, trade-offs between carbon and biodiversity decrease, and thus there is more opportunity for improving one objective without sacrificing the other.This is shown by the sharper upper-right portion of the curve (Figure 3b-d).
The curved trade-off can be more clearly seen at the 1 million ha (Figure 3a) restoration target level.Our results show that the costeffective solutions were on average 28.2% of the cost of the maximum benefit solutions (that ignore costs, purple points in Figure 3), and were on average only 9.36% and 0.35% lower in carbon and biodiversity benefit, respectively.In some cases, the cost-effective solutions yield higher biodiversity benefits than the maximum benefit solutions (Figure 3).This is because the species extinction risk function is a non-linear problem which is being solved linearly in incremental steps (and values are therefore an approximation).The slight differences in values are within the tolerance of the algorithm.
Our analysis shows there are many rare and endemic species with original ranges that are small relative to the size of Colombia and concentrated in specific areas (Figure 4) -53.3% of species have original ranges <10% of the country range.Species with the smallest original ranges are also the species with highest extinction risks.There are 90 species with an extinction risk >60%, and their original ranges span 0.43% of the country on average (range < 0.01% to 5.99%).We found that 57 of 2563 species have extinction risk values of 1, indicating that there is no remaining habitat for these species that can be quantified using the datasets available.Species with the highest extinction risks benefit disproportionately from habitat restoration (because extinction risk changes non-linearly as a function of the ratio of current: original range area).There is a 10 order of magnitude difference between the species with the highest and lowest delta extinction risk (the measure of benefit to the species from habitat restoration, accounting for species weights) because of these 57 extreme value species.The restoration allocation occurs in incremental steps, with the biodiversity benefits recalculated after each step.For the scenarios that include biodiversity benefit, much of what occurs in the early steps is driven by these extreme species (see SM2.1 for extended exploration of biodiversity data).This highly skewed biodiversity data has a large influence on the results (see SM2.2 -Sensitivity Analysis).
Despite the highly skewed biodiversity data, there is still strong potential for good compromise solutions (Figures 3 and 5) that simultaneously achieve biodiversity and carbon benefits (see Data S2).For example, compromise solutions (that balance biodiversity, carbon and cost) achieve on average 85.7%, 87.7%, 87.8% and 87.5% of maximum carbon benefit and on average 95.1%, 99.5%, 99.9% and 99.9% of the maximum biodiversity benefit under the one, six, 6.47 and 8.31 million ha restoration targets, respectively (Table 2).
For the scenarios that are weighted towards maximising biodiversity benefit, the spatial priorities are more concentrated across the Andes, Amazonian foothills and the departments of Antioquia and Santander, when compared to the scenarios that are weighted towards maximising carbon sequestration benefit.Priority areas weighted towards carbon sequestration benefits are more concentrated in the department of Caquetá, Cauca river valley and north of Antioquia.In the cost-effective solutions the northern part of the country is more frequently selected (Figure 5).
For the cost effective, maximum benefit and minimum cost scenarios, around 70 to 80% of the area across all targets is within the agricultural frontier whereas close to 10% is within the exclusion zone (Figure 6).Priorities within the exclusion zone may be more amenable to natural regeneration restoration activities compared to those within the agricultural zone which may require more active restoration techniques coupled with incentive-based programs such as payments for ecosystem services.

| DISCUSS ION
When using the results from this analysis to guide restoration, a key question that decision makers must address is the relative importance of each objective and the degree of concession between objectives they are willing to accept.In our case, across all target levels, clear trade-offs between biodiversity conservation and carbon sequestration restoration objectives were discovered.Despite these trade-offs, we identified national-level compromise solutions that balance biodiversity and carbon and provide good outcomes for each objective.This finding positions Colombia as a country with high potential for carbon capture through forest restoration that also brings high benefits to biodiversity.
To ensure that monetary investments in restoration programs are efficient and strategically planned, we accounted for opportunity and establishment costs.Our establishment cost estimates ranged between USD 1836 and USD 2439 per hectare, depending on the type of management used after the implementation (e.g.irrigation, fertilisation, pruning and exotic species control).These values are similar to other active restoration cost values found in Colombia.For instance, (Vogl, 2018) showed that restoring regions close to large cities in Colombia could cost between USD 1638 and USD 2040 per hectare.
Our results show there is a high potential for cost-effective restoration options in Colombia, for example through assisted natural regeneration, which we found were on average only 28.2% the cost of maximum benefit scenarios.Many restoration strategies focus on planting trees, which though more rapid can be more costly and result in fewer biodiversity benefits compared to approaches based Performance of a range of forest restoration prioritisation scenarios (points) with respect to climate change mitigation benefit (x axis), measured as above-ground carbon sequestration, and biodiversity conservation benefit (y axis), measured as the mean percent reduction in extinction risk among all species.The points the top left of the plot represent solutions that perform best for biodiversity conservation benefit, while the points at the bottom right represent solutions that perform best for carbon sequestration -these correspond to the relative weights assigned to the solutions during the optimisation analysis.Cost-effective scenarios (solid line) maximise return on-investment (benefit in physical terms, divided by cost in monetary terms) and are likely to represent the most feasible restoration planning options.For reference, maximum-benefit scenarios (dashed line) that maximise returns irrespective of cost are also presented.The point circled in black represents a potential good compromise solution.A further reference scenario is also presented in which restoration is allocated to minimise costs irrespective of benefit (square).For all scenarios, the colour represents the total cost of the solution in USD (see legend).
on natural regeneration (Crouzeilles et al., 2017).Assisted natural regeneration is an under-utilised tool for broad scale forest restoration.When well designed, managed and monitored, it has great potential to achieve cost-effective restoration, especially in tropical regions (Chazdon & Guariguata, 2016;Crouzeilles et al., 2020).
With approximately 40% of the country considered degraded, and deforestation continuing, landscape scale forest restoration in Colombia is a means of generating natural and social capital while leveraging change across social and political spectra in an integrative way (Aguilar et al., 2015;Isaacs-Cubides et al., 2020;Cabrera Montenegro et al., 2011;Etter et al., 2008).Considering a large proportion of priority areas are found within the agricultural frontier (Figure 6), our restoration solutions should only be implemented if it facilitates social equity.Many strategies have been proposed for achieving social equity including the Colombian government's payment for ecosystem service scheme (López-Cubillos et al., 2022), the inclusion of all stakeholders within decision making processes, and the inclusion of equity and justice in the centre of all restoration programme design (Löfqvist et al., 2023).Many local socio-economic and ecological factors that are not considered in our analysis must be key considerations by on-the-ground decision makers.
The next important step for restoration in Colombia is the generation and implementation of plans and solutions.An underutilised mechanism for broad-scale restoration in Colombia is leveraging market-based opportunities, which can speed up the execution of large-scale restoration initiatives (Brancalion et al., 2017).For example, carbon removal markets might be able to cover project costs and even While the data used in our analysis represent the best available for Colombia, there are still potential limitations.From the original 5679 species distribution models at our disposal, only 2844 had habitats associated with them and were therefore included in the analysis.A further 65 species were omitted because their potential area of occurrence was zero (i.e. they were a forest occurring species but did not have any of their potential distribution within forested habitats).Defining associated habitats for these species may help address the skewed biodiversity data.Further, all species with available data were included in the analysis, but they may not fully represent the unique suite of species that could benefit from forest restoration in Colombia.An important research need is to develop a comprehensive dataset of species that require forest restoration in Colombia.This planning framework can be revised and repeated over time to reflect new information as it becomes available.
While the majority of Colombia is forested (Arbeláez-Cortés, 2013), the country hosts a plethora of important species within its unique savanna, mangrove, and wetland ecosystem types -which are also being degraded (Etter et al., 2017).An important future research need is to expand our analysis to account for non-forest ecosystem types, to similarly inform on their restoration.Further, our optimisation framework considers only two motivations for forest restoration; namely, biodiversity conservation and carbon sequestration.However, there are many other ecosystem services that forest provide in Colombia, and thus, many other motivations for their restoration.Some of them include improving community wellbeing, social equity, and the need for freshwater (López-Cubillos et al., 2022;Norden et al., 2020).For example, a recent study found that when prioritising social equity within Colombia's national Payment for Ecosystem Services Scheme, of which restoration is a core objective, the municipality of Tolima was identified as the highest priority (López-Cubillos et al., 2022).The plots show that there many species with original ranges that are small relative to the size of Colombia and concentrated in specific areas, which are also the species with highest extinction risks.
our analysis.Future work might expand on our framework by accounting for threats to restoration success, such as through a deforestation risk model (Negret et al., 2019), or identifying actions to reduce the drivers that reduce restoration success (Armenteras et al., 2013).
Given our restoration planning problem focusses on maximising restoration outcomes for species and carbon benefits, our identified priorities would differ if we had taken and ecosystem-based approach.For example, other assessments identify the threatened dry forests as high restoration priorities (Etter et al., 2020;Ospina Arango et al., 2015;SIAC, 2016), and while some of our identified priority areas fell within dry forests, this was to a lesser extent.This is due to the relatively higher species and potential future carbon sequestration benefits that can be gained by restoring other forested areas such as the inter-Andean, Andean and high Andean valleys.While large areas of tropical rainforests were identified as priorities in many solutions (such as in the department of Caquetá), priorities spanned a variety of F I G U R E 5 Trade off curve for a million ha restoration target (Figure 3b) with examples of the spatially explicit ranked solutions for scenarios with a weighting of 1:0 and 0:1 for biodiversity and carbon (both of which ignore cost) and a good compromise solution (point circled in black) that balances biodiversity, carbon, and cost.The maps depict the order in which planning units were selected by the prioritisation analysis (objective function) (shown by the gradient from yellow (selected first) to red (selected last)).

TA B L E 2
Average values of solutions across all scenarios (excluding minimum cost scenario) at each restoration target.forest ecosystems, and thus the strategies which should be used for implementation will depend on each ecosystems ecological condition, its intactness, threat status, social context, according to internationally accepted guidelines (Gann et al., 2019).For threatened forested ecosystems in rural areas current restoration programs are using landscape management tools that consider economic and social viability considerations alongside conservation objectives (Lozano-Zambrano, 2009).
Social viability for restoration projects is fundamental as there are many restoration programs in Colombia that have failed due to communities lack of participation and understanding of their benefits (Ministerio de Ambiente y Desarrollo Sostenible, 2023).Our analysis is potentially limited by the differences in mapping methodologies between the potential ecosystems (Etter et al., 2017)  Another limitation is that our analysis only considers forestland converted to agriculture as being available for restoration.There is significant scope to include recognised degraded forestland as being available (Sylvester et al., 2020), which would involve the quantification of partial benefits towards meeting objectives when some natural ecosystems remain.Some sources of degradation other than agriculture include areas of crops for illicit use that promote land degradation by depleting organic matter (Rodríguez et al., 2020), and invasive species (Baptiste et al., 2020).A recent study found that at least 42 invasive species negatively impact restoration activities (Cárdenas-López et al., 2017).Therefore, a comprehensive map of forest degradation is a crucial future research need and could be incorporated into future analyses.Prioritising restoration in degraded forest areas that retain 50-80% of their potential biomass could deliver rapid biodiversity and climate mitigation benefits, relative to restoring forest on cleared land (Rayden et al., 2023).

CO N FLI C T O F I NTER E S T S TATEM ENT
Authors have no conflicts to declare.

PEER R E V I E W
The peer review history for this article is available at https:// www.
Conclusions: Colombia has committed to bold restoration and conservation targets, such as those under the new 2030 Convention on Biological Diversity Global Biodiversity Framework.Strategic forest restoration planning will play an important role in achieving Colombia's biodiversity conservation and climate mitigation goals.We provide quantitative evidence to inform planning for environmentally and economically sensible restoration policy and practice in the country.Our framework and results can help guide Colombia towards meeting its ambitious forest restoration targets cost-effectively.K E Y W O R D S biodiversity, carbon, conservation, ecosystem services, natural regeneration, optimisation, spatial planning, sustainable development, trade-offs, tropics Colombia has proactively committed to and produced a portfolio of restoration-focussed policies and initiatives that aim to reverse biodiversity declines and mitigate climate change.These include global commitments such as the Nationally Determined Contribution action plan as part of the Paris Agreement which defines restoration and climate reduction targets to be undertaken by different sectors (such as the cattle federation -FEDEGAN) (Colombian Government, 2020b), and the Convention on Biological Diversity's Global Biodiversity Framework which seeks to ensure that 20 percent of degraded ecosystems are under restoration by 2030 (CBD, 2021).At the national level, policy commitments include the Comprehensive Sectoral Climate Change Management Plan (Colombian Government, 2020a), the National Policy for the Integrated Management of Biodiversity and its Ecosystem Services (Colombian Government, 2021b), and law 2173 which promotes ecological restoration and provides economic benefits to young people who get involved with restoration activities (Colombian Government, 2021a).

1
Simplified land use map (a) based on (IDEAM, 2018) and area available for forest restoration (b).

(
Strassburg et al., 2019Strassburg et al., , 2020;;Williams et al., 2020), z = 0.25).The e = 1 − a∕A 0 z F I G U R E 2 Methodological flow diagram of the input data, targets, and weighting requirements used to develop exact spatially explicit land use solutions solved using Gurobi Optimizer version 8.1.0(Gurobi Optimisation 2019).benefit to biodiversity conservation (B) of forest restoration among species is then estimated as: where e c is the extinction risk based on current area, e r is the projected extinction based on the habitat area following restoration, and N is the number of species included in the model.We included 2779 species range models that were developed by the Humboldt Institute, a governmental Colombian research institute.We used five different suites of species distribution models all of which follow the 'Biomodelos' methodology which relies on an expert elicitation process (Velásquez-Tibatá et al., 2019).Models curated and validated by experts were for amphibians, primates, reptiles, plants (Zamia and Magnolia) and birds.We used a sixth batch of models (n = 1168) that were developed by Londoño et al., 2015 using a Maximum Entropy algorithm (Maxent).When duplicates among species existed, the models curated by experts were prioritised, followed by group-specific datasets, followed by the Londoño et al., 2015 suite of models.This resulted in species distributions under the classes Amphibia (n = 94), Aves (n = 1390), Cycadopsida (n = 14), Gnetopsida (n = 1), Liliopsida (n = 47), Magnoliopsida (n = 832), Mammalia (n = 298), Pinopsida (n = 7), Polypodiopsida (n = 1), and

(
Londoño et al., 2015) high altitude, Grassland and savanna dry, Grassland and savanna seasonally wet/flooded, Grassland high altitude (Paramo), Wetlands (Inland), Continental rocky areas and deserts, and Marine/Coastal Supratidal (see SM1.4 and SM1.5 for reclassifications).Through this reclassification and the IUCN database (IUCN, 2022), which links species to their habitat types, we were able to determine the benefit of restoring an area for the suite of species.2.5.3 | Climate change mitigation benefitTo quantify and map potential carbon uptake, we utilised the Global Aboveground Biomass Potential (GAP; v45) maps, as outlined inBroadbent and Zambrano (2021).Detailed methods can be found at the download link provided in the citation.The GAP dataset offers maps at a 100 × 100 m resolution, comprising: (a) a map of predicted old-growth forest aboveground biomass (OGF-AGB), with accuracy to within ±20 Mg ha -1 , as verified using a global validation dataset of 870 field-based forest AGB plots, and (b) an estimate of the year 2017 AGB, including its minimum and maximum bounded uncertainty, serving as a baseline for existing biomass.The GAP products are generated through an iterative Random Forest model, integrating a suite of potential predictor variables, such as topographic, soil (edaphic), and climatic (bioclimatic) parameters.The final GAP map is produced by calculating the difference between the OGF-AGB and the 2017 Quantifying costs2.6.1 | Opportunity costOpportunity costs were based on estimates of annual land rent (a measure of net income generated by land) for cropland and pastureland (2017 USD$ values) at a resolution of approximately 10 km developed for the UN Food and Agriculture Organisation's (FAO) se.plan model(SEPAL, 2023).As an initial step, gross annual revenue was determined separately for cropland and pastureland using existing gridded data sources cropland: MapSPAM (MapSPAM, 2023); pastureland: Gridded Livestock of the World(Robinson et al., 2014), augmented by national data from FAOSTAT (FAO, 2023).In the final step, detailed farm budget data from large-scale household surveys conducted by the World Bank and FAO in several dozen developing countries were used to estimate gross annual revenue attributed to land (i.e. annual land rent).The cropland opportunity cost layer is based on more robust data and thus took precedence over the pastureland opportunity cost in areas where the two overlap.To match the establishment cost data described in the next section, we inflated the se.plan values from 2017 to 2020 by using an international price index from the World Bank's cross-country World Development Indicators database (The World Bank, 2023).Further details can be found at the se.plan website (SEPAL, 2023) and(Busch   et al., in review).We explored the possibility of developing a nationallevel model to calculate opportunity costs (see SM1.6) but found it to be infeasible due to issues related to data accuracy and availability.2.6.2 | Establishment costEstablishment costs were derived from an expert elicitation process with restoration practitioners, and from information on projects from different institutions across Colombia.The experts were representatives from projects implemented by Instituto Humboldt through the 180 million tree planting program in the páramos and high andean and humid forest, the Regional Environment Authority (Regional Autonomous Corporation of the Negro and Nare River Basins -CORNARE), the UNPD and Environmental Ministry of Colombia program in wetlands (Mojana region), and the Marine and Coastal Research Institute of Colombia (INVEMAR).Costs are at the ecosystem level, and represent active planting and a single follow up round of maintenance, focus and enhance investments and management of the agricultural and rural development sector, promote efficient use of land, streamline social ordering of the rural property, and contribute to stabilising and reducing the loss of ecosystems of environmental importance by dictating where agriculture should and should not expand (MADR, 2016) while the Exclusion Zone are areas set aside for restoration and conservation (UPRA, 2018).These zones were established as part of the Colombian government peace agreement signed between the government and the Revolutionary Armed Forces of Colombia (FARC) militia group (Gobierno Nacional de Colombia, 2016).The zones are an agreement regarding where agricultural expansion stops to leave areas for conservation or restoration (UPRA, 2018).

F
Trade-off curves for restoration targets of 1 million ha (a), 6 million ha (b), 6.47 million ha (c), 8.31 million ha (d), and all trade-off curves plotted together on the same axes (e).
provide positive returns.The recent carbon neutrality law (Gobierno Nacional de Colombia, 2021) may be a point of leverage in Colombia for engaging new stakeholders to fund restoration efforts across the country.In order to help guide implementation, all of our spatially explicit solutions are available in a free and open online platform hosted by Instituto Humboldt, a governmental Colombian research institute: http:// wepla n-colom bia.s3-websi te-us-east-1.amazo naws.com/ .
Future research might expand on our framework by considering other ecosystem services as restoration objectives.Additionally, we did not include deforestation dynamics in F I G U R E 4 Exploratory plots of the species data included in the analysis.Left -Number of species against their range as a percentage of the total area of Colombia, with a red line representing the total area available for restoration among all planning units.Middle -Extinction risk against each species range as a percentage of the total area of Colombia.Right -Number of species against the log delta extinction risk.
and the land use map(IDEAM, 2018), both of which were reclassified into a matching habitat classification which was used to quantify species extinction risk.Other logical inconsistencies exist within our analysis as the land use map used for the species distribution, carbon and potential for natural regeneration models were based on maps other than the IDEAM land use map(IDEAM, 2018).Further, we did not estimate potential revenue associated with carbon sequestration, biodiversity conservation, or other ecosystem services that local communities might gain from restoration.Some direct restoration benefits are difficult to quantify, including increased agricultural productivity through mitigating soil degradation, maintaining water supply and quality, and improving pollination services.When considering these benefits, the return on investment in restoration actions will likely offset the cost in the long term in some locations(López-Cubillos et al., 2023).Understanding and quantifying the value of these ecosystem services is a key future research need which could have far reaching implications for planning effective restoration actions(Aguilar-Garavito et al., 2017).
Colombia is increasingly committing to bolder restoration and conservation targets, such as those under the Convention on Biological Diversity.The Kunming-Montreal Global Biodiversity Framework details four long term goals with 23 targets to be achieved by 2030 (CBD, 2021).The recovery and persistence of Colombia's species, which represents more than 10% of global biodiversity, is essential to realising the Framework's vision of 'living in harmony with nature by 2050' (CBD, 2021).Global policy agendas are also being reflected through national policies.To complement the National Restoration Plan (Ospina Arango et al., 2015) the Colombian Ministry of Environment recently implemented the National Restoration Strategy 2023-2026 (Government of Colombia, 2023) as an instrument designed to guide restoration processes at a landscape scale, in a national commitment to recover the functionality of ecosystems, increase resilience against change climate, revitalise territories, generate economies for life and improve the well-being of communities.The results from our analysis inform on both the National Restoration Plan and how the objectives of the National Restoration Strategy can be achieved by 2026.With the majority of Colombia's species being forest-dependent (Arbeláez-Cortés, 2013), Colombia's biodiversity conservation commitments can only be met through strategic forest landscape restoration and reductions in deforestation.Additionally, Colombia is a signatory country of the Paris Agreement which calls for both a reduction of emissions coming from land use change and a transition to low carbon economies.Strategic planning forest restoration can play an important role in achieving Colombia's climate mitigation goals cost-effectively.With an effective national restoration program Colombia as a nation has the potential to provide insights into effective strategies for other biodiversity-rich nations with emerging economies and ambitious restoration objectives.

F
Prioritised areas for restoration within the agricultural frontier (governmental priorities for agricultural expansion) and the exclusion zone (governmental priorities for conservation) (represented as proportions of the total restoration solution).Scenarios: mc = minimum cost, mb = maximum benefit (average across all weights), and ce = cost effective (average across all weights).Error bars represent standard deviation.Colombia is well placed to take advantage of this opportunity as it is well recognised for the strides it is making towards global collaboration, policy development, and scalable approaches to address shared restoration priorities and climate change impacts.ACK N OWLED G EM ENTSWe thank Mauricio Aguilar-Garavito, David Echeverri, Alexandra Rodríguez, Ronald Ayazo, Carolina Alcázar, Patricia Velasco, Andrés Santamaria and Fundación Miramar for the data of cost of restoration implementation.This project was supported by the Ministry of Environment and Sustainable Development -Instituto de Investigación de Recursos Biológicos Alexander von Humboldt, agreement No. 21-21-resolution 0210-144-C.Figure 2 was created using https:// www.canva.com/ where "All free photos, music and video files on Canva can be used for free for commercial and noncommercial use https:// www.canva.com/ polic ies/ free-media -licen ce-agree ment-2022-01-04/ ".Open access publishing facilitated by The University of Queensland, as part of the Wiley -The University of Queensland agreement via the Council of Australian University Librarians.