Forest Policy and Economics 125 (2021) 102403 Contents lists available at ScienceDirect Forest Policy and Economics journal homepage: www.elsevier.com/locate/forpol Stochastic simulation of restoration outcomes for a dry afromontane forest landscape in northern Ethiopia Yvonne Tamba a,*, Joshua Wafula a, Cory Whitney b,c, Eike Luedeling b,c, Negusse Yigzaw d, Aklilu Negussie d, Caroline Muchiri a, Yemane Gebru d, Keith Shepherd a, Ermias Aynekulu a a World Agroforestry (ICRAF), UN Avenue, P.O. BOX 30677, Nairobi, Kenya b Center for Development Research (ZEF), University of Bonn, Genscherallee 3, D-53113 Bonn, Germany c University of Bonn, INRES – Horticultural Sciences, Auf dem Hügel 6, D-53121 Bonn, Germany d WeForest, P.O.BOX 25450/1000, Addis Ababa, Ethiopia A R T I C L E I N F O A B S T R A C T Keywords: Forest and Landscape Restoration (FLR) is carried out with the objective of regaining ecological functions and Forest and landscape restoration enhancing human well-being through intervention in degrading ecosystems. However, uncertainties and risks Decision analysis related to FLR make it difficult to predict long-term outcomes and inform investment plans. We applied a Sto- Monte Carlo risk analysis chastic Impact Evaluation framework (SIE) to simulate returns on investment in the case of FLR interventions in a Value of information Uncertainty degraded dry Afromontane forest while accounting for uncertainties. We ran 10,000 iterations of a Monte Carlo simulation that projected FLR outcomes over a period of 25 years. Our simulations show that investments in assisted natural regeneration, enrichment planting, exclosure establishment and soil-water conservation struc- tures all have a greater than 77% chance of positive returns. Sensitivity analysis of these outcomes indicated that the greatest threat to positive cashflows is the time required to achieve the targeted ecological outcomes. Value of Information (VOI) analysis indicated that the biggest priority for further measurement in this case is the maturity age of exclosures at which maximum biomass accumulation is achieved. The SIE framework was effective in providing forecasts of the distribution of outcomes and highlighting critical uncertainties where further measurements can help support decision-making. This approach can be useful for informing the man- agement and planning of similar FLR interventions. 1. Introduction landscape restoration (FLR), may offer effective and integrated strate- gies for sustainable and integrated landscape management. According to the United Nations Environmental Programme, degra- FLR is a planned process where forest landscapes are restored with dation of terrestrial and marine ecosystems undermines the well-being the goal of ecological integrity and improved human well-being. In of 3.2 billion people and costs about 10% of the annual global gross practice, FLR projects follow guiding principles that dictate a focus on product in loss of species and ecosystem services (UNEP, 2019). In landscapes and natural ecosystems, participatory governance, context- Ethiopia, land and forest resource degradation across the different specific approaches, adaptive management and restoration of multiple production systems of the country is considered a major impediment for functions for multiple benefits (Gitz et al., 2020). The definition of FLR is sustainable development, causing considerable negative impacts on the broad, allowing for flexibility in how the process is implemented in local national economy (Gashaw et al., 2014). A rapidly growing population, landscapes, while the underlying set of guiding principles were devel- combined with increasingly frequent droughts, prevalent poverty and oped to ensure restoration quality. Despite being adopted as a vehicle for lack of alternative employment opportunities, is leading to over- transformation in multiple initiatives that target degraded landscapes exploitation of the country’s natural resources (Tesfaye et al., 2014). (such as the Convention on Biological Diversity (CBD, 2010), the United The traditional customary resource management systems that commu- Nations Framework Convention on Climate Change’s REDD+ goals nities have relied on for generations are therefore being challenged (COP 16, 2011), the United Nations Conventions to Combat Desertifi- (Scull et al., 2017). Novel approaches to restoration, such as forest and cation (Chotte et al., 2019), and the United Nation’s Decade of * Corresponding author. E-mail address: Y.Tamba@cgiar.org (Y. Tamba). https://doi.org/10.1016/j.forpol.2021.102403 Received 15 September 2020; Received in revised form 15 January 2021; Accepted 18 January 2021 Available online 6 February 2021 1389-9341/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 Ecosystem Restoration (FAO, 2019), there is still a need for empirical even-aged secondary forest, hosting about 90 tree and shrub species, and evidence to support scaling-up efforts. Case studies that meet all the dominated by Wild African wild Olive (Olea europaea subsp. cuspidata) criteria of FLR are few due to the recency of the concept, and the lack of a and African Juniper (Juniperus procera) (Aynekulu et al., 2009). The standard method for assessing FLR outcomes (Stanturf et al., 2019). A forest is of high ecological and socio-economic importance as it has the variety of methods have been developed to address the need for scien- potential to conserve biodiversity and soils, supply biomass for fuelwood tific tools to support decision-making on specific components of resto- and construction, regulate water and carbon cycles and offer a host of ration outcomes, such as soil health (land degradation surveillance other ecosystem services (Teklay et al., 2013). Despite its protected framework; (Vågen et al., 2013), soil nutrient deficiencies (Munialo status as a state forest, about 70% of dense forest, with a canopy cover of et al., 2019), soil organic matter content (Zomer et al., 2017), biomass more than 40%, has been deforested and degraded since the 1970s accumulation (Romijn et al., 2019), rangeland/grazing management (WeForest, 2018). This is mainly due to forest land conversion to agri- and governance (Sircely et al., 2019), as well as the economics of land culture land and settlements, over-extraction of woody biomass for fuel degradation (Nkonya et al., 2015). However, assessment metrics that and timber, fire, and free grazing (Aynekulu et al., 2011). integrate both socioeconomic and biophysical outcomes are still lacking (Chomba et al., 2020). In Ethiopia, there have been several interventions that meet the FLR 2.2. Methods criteria of sustainable land management, including the Integrated Food Security Project and, more recently, the landscapes for people, food, and 2.2.1. FLR interventions nature initiatives (Nigussie et al., 2017; Weldesemaet, 2015). Substan- To restore the degraded Desa’a forest, WeForest, a non-profit orga- tial investment is required but often cannot be secured due to evidence nisation with support from the Ethiopian government, launched a long- gaps that threaten the success of management strategies. Another reason term FLR programme that proposed investments in a portfolio of scal- for limited investments is the long-term planning horizon of FLR, which able restoration and livelihood interventions. The interventions are ex- dampens enthusiasm for funding (Kusters et al., 2018; McGonigle et al., pected to achieve socioeconomic benefits by promoting economic 2020; Pistorius et al., 2017). To evaluate and justify investments in resilience of vulnerable communities and incentivizing improved natu- sustainable land management, development practitioners commonly ral resource governance. The targeted beneficiaries of the interventions employ deterministic cost-benefit analysis approaches that are hinged were subsistence farmers. The following FLR interventions aimed to upon precise models of system functions, such as the Restoration Op- restore degraded forest functions: portunities Assessment Methodology (ROAM) manual and the Restora- tion Diagnostic (IUCN and WRI, 2014; World Resources institute, 2015). • Assisted natural regeneration (ANR) of degraded forest. The ANR However, deterministic models often fall short of adequately supporting intervention involved restricting access to the forest products decisions when data are scarce or of low quality (Wendt, 1975), or through social fencing facilitated by local by-laws and governance where complex system functions introduce risk and uncertainty (Lued- structures in a process termed “exclosure”. Social fencing was eling and Shepherd, 2016). Effective planning and prioritization of in- enforced by community members trained as forest guards, and terventions may be compromised by uncertainty in the definition of community participation was encouraged through livelihood devel- restoration objectives, failure to identify the most efficient practices and opment interventions. failure to identify the socio-economic and cultural drivers of deforesta- • Enrichment planting and assisted natural regenertation of up to 1000 tion (Cortina et al., 2011; McGonigle et al., 2020; Yet et al., 2020). At- native trees per hectare in the open forest areas with canopy cover of tempts by managers to value restoration outcomes also face difficulties less than 40% but more than 10%, where assisted natural regener- when assigning monetary values to ecosystem services with low market ation has low potential to restore vegetation. values, such as carbon sequestration, regulation of hydrological cycles • Grazing land exclosure, where communally owned grasslands were and improved micro-climates (de Groot et al., 2010). protected from free grazing to encourage natural regeneration of Decision support approaches that holistically evaluate decision op- woody vegetation. The community was allowed access to harvest tions based on plausible ranges of costs and benefits while accounting grass for livestock feed (cut and carry method). for uncertainties and risks could overcome these knowledge barriers. • Soil and water conservation, where gully restoration and in-situ Furthermore, they could strengthen the capacity of managers to use water harvesting structures were established to reduce soil erosion continuous learning and monitoring systems to track their progress to- and improve water infiltration. wards their goals (Rumpff et al., 2011). It is also possible to take stock of the successes and failures of restoration policies and efforts undertaken The project also implemented a set of livelihood improvement and to learn lessons for improved natural resource management and interventions: protection (Cronkleton et al., 2018). Through these approaches, we can prioritize critical uncertainties where targeted research could enhance • Beekeeping, where two to three modern beehives were distributed clarity on expected outcomes. In this study, we demonstrated the among 3280 beekeepers to establish apiaries around their home- application of a stochastic impact evaluation (SIE) framework to (i) steads with the aim of promoting non-timber forest products. predict bio-physical and socio-economic outcomes of FLR practices, (ii) • Sheep rearing, where three to five sheep were distributed among identify knowledge gaps that constrain effective decision making and, 7650 female-headed households to provide alternative sources of (iii) provide insights that aid in adaptive management of FLR efforts. income to forest products. Small ruminants were chosen due to their resilience to harsh climatic conditions, ease of liquidity to meet 2. Materials and methods household financial needs and the proximity of animal feed in the form of fodder from exclosures established on communal grazing 2.1. Study area lands. • High-value fruit trees, where eight to thirteen apple tree seedlings Desa’a forest is one of the oldest remaining dry Afromontane forests were distributed among 5465 targeted farmers (whose farms were along the western escarpment of the Great Rift Valley in northern located within the FLR restoration project area) with the aim of Ethiopia (Lat. 13◦ 53′ – 13◦ 56′ N and Long. 39◦ 48′ - 39◦ 51′ E) (Fig. 1). diversifying incomes and reducing demand for forest commodities. It lies between 900 and 3100 m above sea level. Based on rainfall data • Poultry farming, where ten poultry birds were distributed among from the Ethiopian Meteorological Agency for 2006 to 2015, the mean 7650 impoverished female-headed households to provide livelihood annual rainfall was about 602 mm (Mokria et al., 2015). Desa’a is an benefits through sale and consumption of poultry products. 2 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 Fig. 1. Map of Desa’a state forest in Ethiopia (adapted from WeForest, 2018). Forest zones are demarcated based on vegetation density and human influence. The core zone is an area of dense forest with canopy cover ≥40%. Buffer zone 1 includes areas categorized as open forest where vegetation cover is greater than 10% but less than 40%. Buffer zone 2 denotes areas that are communally owned and made up of fragmented open forests and grazing lands where vegetation cover is ≤10%. Development zone denotes areas covered by community settlements. 3 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 • Energy efficient stoves that reduce demand for fuelwood from the practitioners and one academic staff of Mekelle University. We applied forest were distributed among 10,390 households to promote alter- purposive sampling to identify interviewees who had expertise in native energy sources. environmental management, forest resource management and agricul- tural value chains. We also held a focus group discussion with 12 male 2.2.2. Stochastic simulation of FLR outcomes and eight female members of the local community, two development We used the SIE framework (Fig. 2), based on Luedeling and Shep- agents and two community leaders in the project area. The objective of herd (2016) to simulate FLR outcomes. SIE is a mixed methods approach the discussion was to elicit perspectives from members of the commu- that has been widely used to simulate outcomes under uncertainty and nity on the historical trends in land use and land cover in the forest risk for investments in honey value chains (Wafula et al., 2018), water landscape, how the community expected the FLR interventions to supply (Luedeling et al., 2015), irrigation development (Yigzaw et al., change the trends in use of forest resources, and the potential barriers to 2019), management of reservoir sedimentation (Lanzanova et al., 2019), implementation that they could foresee. and to determine the value of ecosystem services in rangelands (Favretto Step 2: Conceptual modelling et al., 2017). In this study, we applied SIE as an iterative five-step pro- We followed a participatory model development process with the cess that supports decisions by integrating evidence and expert opinion aim of conceptualizing the decision’s impact pathways and identifying in quantitative simulations of decision impact pathways (Fig. 2). the cost, benefit and risk variables that would be parameterized in a Step 1: Decision framing simulation model. We held a workshop with 17 stakeholders from six Decision framing is a crucial step where decision-makers need to development agencies, five representatives of state agencies and two explicitly define their problem and target outcomes. This first step ad- researchers from Mekelle University, Ethiopia and elicited relevant dresses questions regarding the short-term and long-term outcomes, the variables. We then consolidated the resulting impact pathways and targeted beneficiaries and the type of decision under consideration causal relationships between costs, benefits, and risks to generate the (prioritizing vs planning) (Luedeling and Shepherd, 2016). To clearly overall conceptual structure of the decision model (Do et al., 2020) define the intervention’s social, economic and biophysical impacts, we (Fig. 3). carried out semi-structured interviews with representatives from the The variable estimates (Tamba et al., 2020) were used to feed the implementing agency. The interviews provided insights on the objec- mathematical model, which was then run as a Monte Carlo (MC) tives of the FLR project, implementation strategies and the targeted simulation with 10,000 iterations using the decisionSupport package outcomes. A 25 year horizon was chosen to support long-term planning (Luedeling and Whitney, 2018; Luedeling et al., 2020) in the R pro- that incorporates key uncertainties. Through these interactions, the gramming language (R Core Team, 2017). For each run, the model decision problem that emerged was whether the selected FLR in- produced a projection of the NPV, computed by adding up discounted terventions will be able to restore the degraded forest to provide sus- net benefits over a 25-year simulation period. tainable socioeconomic and biophysical benefits. Step 3: Developing a mathematical model To further clarify the decision context, we conducted semi-structured In the third step, the conceptual model was translated into a math- interviews with five government officers, seven development ematical model to quantify the impact of nine FLR interventions as the Fig. 2. Sequence of activities in the Stochastic Impact Evaluation approach (adapted from Yigzaw et al., 2019). Step 1 defines the decision context by identifying stakeholders and engaging them in a participatory research process. Step 2 creates a conceptual model of the decision’s impact pathways and describes the re- lationships between the cost, benefit and risk variables identified by stakeholders. Step 3 translates the conceptual model into a mathematical model with causal relationships between variables rewritten as equations. In step 4, the values of model parameters are estimated by calibrated subject matter experts, and a Monte Carlo (MC) simulation of the cost-benefit analysis is run to project the distribution of returns. To analyse the sensitivity of the model, the Variable Importance in the Projection (VIP) is computed based on the results of a Partial Least Squares regression analysis. Expected Value of Perfect Information (EVPI) analysis serves to identify variables with high information value for the specific decision. Step 5 is where the model is refined when necessary. The process is iterative and allows for multiple cycles until the decision maker has sufficient information to make a decision. 4 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 Fig. 3. Conceptual model developed by stakeholders showing the expected costs, benefits and risks of programme activities. We collected estimates of model variables (yellow bubbles) from calibrated subject matter experts and passed the inputs through the model to arrive at the value of outcome variables (blue bubbles) under risk and uncertainty (red rectangles). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) change in outcomes with intervention relative to existing land use Benefits: For livelihood interventions, we quantified the expected systems. increase in income per household as the main benefit. The beekeeping Risks: Risk factors were considered as the probability of occurrence intervention would provide revenues from the sale of honey. For energy- of risk (used to generate a binomial distribution describing whether the efficient cookstoves, we accounted for the benefits indirectly as house- respective events occur or not in the simulations) and the magnitude of hold health cost savings and reduced dry wood harvesting costs. Sheep impacts on expected benefits, if these events occur. We identified two rearing would mostly provide revenues from the sale of sheep. Poultry classes of risks. The first was a class of risks that had a random chance of farming would provide revenues from the sale of poultry products. occurring in any given year (random risks). These were simulated by Apple trees would generate revenues from the sale of fruits. computing the annual likelihood of occurrence and the impact on out- For restoration interventions, the benefits were expected to accrue to comes when the risks do occur using the chance_event function of the the entire community and therefore quantified as communal benefits. decisionSupport package. For example, for anthropogenic risk, we This assessment targeted provisioning and regulating ecosystem ser- computed the chance that community members would encroach into the vices, since these have direct use values. We then applied a mix of forest (for various reasons, such as illegal logging, charcoal burning, market and non-market pricing strategies. The main benefit expected fuelwood harvesting). We then estimated the magnitude of loss of pre- from ANR and enrichment planting would be the increase in carbon viously computed benefit streams. The second class of risk factors was stocks as vegetation regenerates. In addition to regeneration, there conditional risks. For these, the risk events were associated with other would be agricultural benefits from a favourable microclimate, resulting events whose occurrence was uncertain. For instance, the risk of the in improved yields for surrounding farmers. To simulate the carbon community encroaching on the forest to graze their livestock was sequestration benefit of enrichment planting, ANR and exclosure determined by the probability that exclosures were ineffective given the establishment, we used the gain-loss method that sums up changes in community’s demand for fodder and the probability of poor imple- biomass stock for the specific land-use category (IPCC, 2006). We mentation of social fencing. determined biomass stocks using two approaches: Costs: We categorized costs into ‘individual costs’ incurred by ben- -An exponential function adapted from guidelines from the Inter- eficiaries and ‘program costs’ incurred by the implementing agency. For governmental Panel on Climate Change was used to simulate the annual the livelihood development activities, the cost of acquiring assets was increment in biomass per replanted tree. borne by the implementing agency and calculated only for the first year, a while operating costs were borne by the individuals and considered biomass per tree = (1) 1 + exp − (BCEF*(b − c) ) annually over the 25-year period. Opportunity costs were added to the cost per individual. The cost of restoration interventions included the where a = maximum marketable volume, BCEF = biomass conversion cost of acquiring materials (which was a one-off investment) and and expansion factor, b = simulation period, c = stem maturity age. recurrent expenditure on technical labour and maintenance. As some -We computed biomass growth as a function of the mean annual labour was provided as in-kind payment by the community, we increment per hectare of regenerated forest area and exclosure: considered that only a part of the labour costs was paid in cash. 5 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 biomass per ha = measurably improve an expert’s ability to provide accurate estimates mean annual increment*mean biomass per ha (2) (Hubbard, 2014). We then quantified the change in carbon stocks that would result Step 5: Model Refinment from restoration activities (ANR, enrichment planting, and exclosure) We used Value of Information (VOI) analysis to identify important using the gain-loss method. Emphasis was placed on gains to estimate knowledge gaps where further measurement efforts could provide the impact of successful restoration. With this approach, we determined clarity on the best decision (Wilson, 2015). We did this by computing the the change by the product of the area of land and the incremental expected value of perfect information (EVPI). EVPI represents the op- biomass stock per unit of land area. The impact of social fencing was portunity loss that could be incurred by a decision-maker due to lack of used as a proxy for forest areas gained from avoided deforestation and information on a specific variable (Felli and Hazen, 1998; Hubbard, degradation (Eqs. 3& 4). 2014). Applied in this way, the EVPI computation can help to determine where further measurements may help reduce uncertainty on decision ∆Carbon gain from avoided deforestation and degradation = (3) outcomes. We also applied Partial Least Squares (PLS) regression anal- biomassstock per ha*∆areasocialfencing+enrichment*carbonfraction ysis to the MC simulation results and used Variable-Importance-in-the- Projection (VIP) sensitivity analysis to assess the input parameters Where carbon fraction of dry matter = 0.47 (IPCC, 2006) (Luedeling and Gassner, 2012). The VIP statistic represents the direction ∆areasocial fencing+enrichment = avoided loss in forest area due to agricultural and settlement expansion+ (4) avoided loss in forest area due illegal commercial logging We also quantified carbon losses from random events of fire and and strength of each input variable’s relationship with the output vari- disease outbreaks and calculated carbon accumulation per hectare of able (Wold et al., 2001). restored and conserved forest based on the mean annual increase in carbon stocks (Eqs. 1, 2). We then used the benefit transfer method to 3. Results determine the market price for carbon. The impact of exclosure establishment was assessed by valuing the 3.1. Returns from livelihood interventions change in the quantity of grass produced when land use shifted from communal grazing to exclosures with cut-and-carry harvesting. The use Model results for livelihood interventions showed that most in- value of establishing exclosures was determined by the amount of fodder terventions would have positive NPVs for the 25-year simulation period. produced in exclosures relative to the amount produced by grazing Returns on investments in fruit trees and beekeeping had a 0.4% chance lands. The non-use value of carbon sequestration was found by calcu- of loss but beekeeping had a wider range of returns than fruit trees lating the mean annual increment of above- ground biomass (Eq. 2). For (Table 1). Poultry farming and efficient cooking stoves both had no investments in soil and water conservation, the avoided-cost method chance of loss but were less profitable than fruit trees and beekeeping. was used to quantify the primary benefit of reducing costs to the com- The net present value of returns to sheep rearing had the highest pos- munity related to removal of sediments from a community dam (Pan- sibility of loss (60%) and the lowest profits of all livelihood agos et al., 2015; Cheboiwo et al., 2018). interventions. For each intervention, we quantified the expected net benefits by subtracting the aggregate costs from risk-adjusted benefits (Eq. 5) and 3.2. Return from restoration interventions then discounted the net benefit to find the net present value (Eq. 6). risk scaler probability of risk occuring impact of risk (5) The simulated NPV of ANR cashflows had a possibility of negative = × returns (Table 1). The distribution showed minor variation over time, ∑n ∑t with the median return for each year progressively increasing but never Net Benefiti = [Total Benefiti ×(1 − risk scaler) ] − Total costsi exceeding 4000 USD ha− 1. VIP analysis of outcomes revealed 9 variables 1 1 that the projected returns were sensitive to. The impact of ANR on yields (6) in the surrounding agricultural area, market price of carbon, annual rate where n number of targeted beneficiary households, of deforestation, and viability of carbon marketing were the 4 most = t number of simulation years. highly ranked variables correlated with ANR outcomes (Fig. 4d). VOI = analysis revealed that there were no critical knowledge gaps to be filled NetBenefit NPV i (7) (Fig. 4b). i = (1 + r t.) The model simulated positive returns on the enrichment planting intervention in 89.8% of model runs. Annual outcomes varied signifi- where NPV = Net present value, r = discount rate, and t = year cantly with a high likelihood of losses in the first 5 years after planting. If Step 4: Model parameterization and simulation further clarity on this outcome is needed, priority should be given to We used expert knowledge elicitation and literature review to assign reducing uncertainty related to carbon markets (Fig. 5b). VIP analysis probability distributions for all model variables and operationalize the highlighted 13 variables with a coefficient value above the threshold of model. However, expert opinion can be subjective and susceptible to 0.8 (Fig. 5d). The most sensitive variables in this case were strongly biases such as overconfidence or under-confidence (Hubbard, 2014; Yet related to carbon markets (cost of carbon and risk of lack of carbon et al., 2016). To reduce these biases, we conducted a calibration training markets) and the tree population (annual rate of deforestation, number of subject matter experts during a model validation workshop. The of replanted trees per ha and risk of wildfires) (Fig. 5d). Grazing land training aimed to improve the capacity of subject matter experts to make exclosure was the riskiest restoration intervention with a 77.2% likeli- estimates for which they are 90% confident that the actual values lie hood of positive returns. Annual cashflows (Fig. 6c) revealed possibil- within the provided ranges. We used Klein’s Pre-mortem (Klein, 2007) ities of net losses in the initial years and and expectation of breaking and the equivalent-bet technique, which have been proven to even in the 10th year. 6 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 Table 1 Summary of returns on Forest Landscape Restoration (FLR) interventions. The range represents the 90% confidence interval of the total Net Present Value (NPV), considering a 25-year simulation period. Also shown are the chance of loss for each intervention and the value of information (VOI) for further measurement for each intervention, expressed as the Value of Perfect Information (EVPI). NPV in USD (n = 10,000, 90% C.I.) VOI Intervention Lower bound Median Upper bound Chance of loss EVPI (USD) Critical knowledge gap Beekeeping 1594 4517 10,961 0.1% 0.4 Honey yield per hive Cookstoves 1165 2008 3140 0% 0 – Sheep rearing − 1258 − 165 1013 60.0% 209 Cost of Labour Poultry farming 624 1053 1569 0.0% 0 – High Value trees 1482 4292 8023 0.2% 0.4 Max. fruit yield potential Grazing land exclosure − 13,119 9800 50,785 22.9% 2000 Biomass maturity age Assisted natural Regeneration 13,231 20,215 30,286 0% 0 – Soil water conservation 1104 4141 7401 1.2% 7.9 Rate of soil loss Enrichment planting − 492 3212 13,852 10.2% 56 Market price per ton of carbon Fig. 4. Projected outcome of the decision to implement ANR in Desa’a (a), high decision-value variables (b), the respective cashflows (c) and important variables (determined by VIP analysis of PLS regression models) (d). The results were produced through MC simulation (10,000 model runs) of ANR performance over 25 years. In the PLS plot, green bars indicate positive correlations of uncertain variables with the outcome variable, while red bars indicate negative correlations. Blue bars indicate variables that did not meet the threshold of the model sensitivity analysis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 7 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 Fig. 5. Projected outcome of the decision to undertake enrichment planting in Desa’a (a), high decision-value variables (b), the respective cashflows (c) and important variables (determined by VIP analysis of PLS regression models) (d). The results were produced through MC simulation (10,000 model runs) of enrichment planting outcomes over 25 years. In the PLS plot, green bars indicate positive correlations of uncertain variables with the outcome variable, while red bars indicate negative correlations. Blue bars indicate variables that did not meet the threshold of the model sensitivity analysis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Further measurements to pinpoint the maturity age of woody the cost of extracting fuelwood from the forest, improve health and biomass in exclosures (EVPI = 1330 USD) and, to a lesser extent, the reduce carbon emissions from forest degradation (Grieshop et al., 2011). maximum carbon stock that can be accumulated in the exclosures and Indirect income from reduced fuelwood needs and savings on health the rate of deforestation would reduce ambiguity for the decision- costs might not be enough financial incentive to encourage community makers (Fig. 6b). The NPV for soil water conservation efforts had a members to adopt energy-saving stoves (Okuthe and Akotsi, 2014). For 98.84% chance of positive outcomes (Fig. 7a). If further clarity is households targeted for sheep rearing, the enterprise is risky with a 60% necessary, the analysis identified the rate of soil loss in the forest as a chance of incurring losses. This outcome indicates uncertainty, as it does source of uncertainty (Fig. 7b). Despite this uncertainty, outcomes are not offer sufficient evidence to support the decision to roll out the most likely to be positive. Sensitivity analysis revealed five variables intervention. Measurements to gain a better understanding of labour with a significant correlation with the projected outcomes (Fig. 7d). requirements of sheep rearing can help eliminate uncertainty about this outcome. Identifying the ideal number of sheep to distribute to com- 4. Discussion munity members for the intervention to make economic sense and the effect of a drought event on sheep rearing ventures can also help to gain 4.1. Livelihood interventions clarity on outcomes. Poultry farming is profitable and could effectively provide an alternative source of income for the most resource poor Beekeeping is the most profitable among the livelihood interventions households (Pica-Ciamarra and Otte, 2010). Investment in planting of that were investigated. Energy-efficient cookstoves are also promising, fruit trees would also generate positive returns, but these returns will not although this intervention would not provide direct income, but save on be realised in the first few years, since apple trees take several years to 8 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 Fig. 6. Projected outcome of improved exclosure management on the economic value of ecosystem goods and services in exclosure (a), high decision-value variables (b), the respective cashflows (c) and important variables (determined by VIP analysis of PLS regression models) (d). The results were produced through MC simulation (10,000 model runs) of 25 years of exclosure performance. In the PLS plot, green bars indicate positive correlations of uncertain variables with the outcome variable, while red bars indicate negative correlations. Blue bars indicate variables that did not meet the threshold of the model sensitivity analysis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) reach their maximum production potential. projected returns on enrichment planting relative to ANR is also explained by the differences in costs, as lower implementation costs are incurred with ANR as compared to enrichment planting (Chazdon et al., 4.2. Restoration interventions 2016; Shono et al., 2007). Nevertheless, projected NPVs per ha for both interventions were higher than those of most alternative land uses. This Model results showed that the restoration strategy with ANR accel- makes both interventions effective and profitable to achieve biophysical erates recovery of natural forest by reducing disturbance. Enrichment outcomes (Pistorius et al., 2017). planting achieves the same outcome but through active replanting in the VOI analysis revealed that there were no high value variables for the fragmented forest zones. While our simulations showed that both in- ANR intervention. However, there were variables of importance that terventions were likely to have promising outcomes, the likelihood of would determine the magnitude of positive cashflows. When the effect positive monetary returns was higher with ANR than enrichment of restoration on yields in adjacent agricultural lands was considered, planting. This agrees with the results of a meta-analysis of forest resto- we found that the projected returns increased.A study on the effect of ration interventions in tropical forests. Crouzeilles et al. (2017) report increased tree cover on agriculture in southern Ethiopia simulated a 5% that restoration outcomes measured by vegetation structure and biodi- increase in wheat production on lands adjacent to reforested forests and versity were higher for natural regeneration than for tree planting. Our hedgerows (Yang et al., 2020). This result is attributed to improved soil findings on the differences in quantities of sequestered carbon for moisture, temperature regulation and increased soil nutrient availability replanting compared with regeneration are explained by slower rates of for agricultural lands bordering forest. accumulation in replanted trees and high quantities of sequestered Exclosure establishment would also generate substantial benefits carbon in old trees (Köhl et al., 2017). Furthermore, the difference in 9 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 Fig. 7. Projected outcome of introducing soil and water conservation measures in Desa’a. (a), high decision-value variables (b), the respective cashflows (c) and important variables (determined by VIP analysis of PLS regression models) (d). The results were produced through MC simulation (10,000 model runs) of the performance of conservation structures over 25 years. In the PLS plot, green bars indicate positive correlations of uncertain variables with the outcome variable, while red bars indicate negative correlations. Blue bars indicate variables that did not meet the threshold of the model sensitivity analysis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) when compared to the alternative, a communal free grazing land use greatest difference in older exclosures. system. However, while regulating services are expected to improve with the establishment of exclosure, the impact on feed resources will be negative. Exclosures could create competition for livestock feed re- 4.3. Implications for FLR actors and policy-makers sources among community members by restricting access to grazing land (Birhane et al., 2018). Results of VOI analysis showed that While FLR is widely expected to have positive socioeconomic out- achieving greater precision in estimating the time required to achieve comes, measurements of ecosystem benefits are sorely lagging behind maximum biomass accumulation should be prioritized. This knowledge the recognition that they exist (Matzek, 2018). This is the result of a could potentially improve management of exclosures by making the shortage of technical experts who can address the methodological con- valuation of carbon stock more precise. Nonetheless, the range of out- cerns and philosophical objections that come up when attempting to comes projected by the model brackets the deterministic result projected monetize nature’s services. Uncertainty in measurement results from by Mekuria (2013) (about 3000 USD ha− 1) when assessing the changes practical challenges in monitoring ecosystem services (de Groot et al., in regulating ecosystem services following establishment of exclosure on 2010), and the acknowledgement that restoration efforts cannot fully communal grazing lands in Ethiopia. The trend in returns showed that recover the natural ‘pre-disturbance’ ecosystem functions (Crouzeilles over time, the cash flows increase with exclosure age and level off after et al., 2016). the exclosure reaches its production potential. This was also found by For long-term planning of FLR, managers and policy-makers should Mekuria (2013), who compared biomass productivity in five, ten, fifteen pay more attention to biological factors. The time lag to production will and twenty-year old exclosures with communal grazing land, finding the affect the distribution of returns, hence low returns should be expected in the first few years and greater returns towards the end of the 10 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 simulation period. Managers therefore need to be prepared to evaluate et al., 2009). the outcomes of their interventions during the early phase of their project where implementation costs are incurred, and net losses are 5. Conclusions likely. A portfolio analysis to identify the best combinations of in- terventions could help buffer against the risk of losses in the first few Predicting FLR outcomes is a difficult endeavour, since outcomes are years. Also, varietal selection to prioritize tree species that accelerate often achieved through complex mechanisms with many uncertainties ecosystem recovery can help minimize losses. It is important for man- and risks preventing robust decision making. Development practi- agers and policy-makers to note that socioeconomic factors, i.e. the tioners, landscape restoration programme managers and researchers can drivers of deforestation and the viability of carbon trading were more overcome these challenges by applying stochastic methods and partici- likely to determine whether actual returns matched desired outcomes patory research approaches. Decision analysis tools that apply stochastic than the biophysical determinants of returns. Without strategic man- impact evaluation are suitable for decision makers who are not only agement, exclosures are expected to lead to a resource constraint for constrained by imperfect knowledge of complex systems, but also need livestock farmers, as they reduce availability of feed resources, but may to consider a range of social, economic, and biophysical factors to pre- not be able to offset this effect through positive revenues from carbon dict project impacts. The SIE framework enabled us to clearly define the credits. Even when paired with cut-and-carry harvesting, this interven- objectives of FLR activities and quantify the expected impacts on land tion may lead to a net reduction in fodder supply, which may discourage use and land cover trends. community participation. Incentivizing pastoral communities by Engagement of subject matter experts, decision-makers and com- providing a livestock insurance policy against drought could help ach- munity members enabled us to develop a decision model that incorpo- ieve community buy-in and improve revenues from carbon credits. rated priorities and beliefs of stakeholders and decision-makers. The There is also no doubt that successful implementation of FLR pro- process provided an avenue for stakeholders to express their uncertainty grammes requires key socioeconomic mechanisms be put in place to about the relevant variables, including those considered difficult to ensure there are clear rights, roles and benefit sharing arrangements measure. Thus, we conducted a robust cost-benefit analysis and pre- between different stakeholders and community members (Yami et al., sented distributions of plausible decision outcomes to decision-makers. 2013). In this way, research outcomes were translated into economic impacts Holistic and stochastic valuation of forest restoration costs and for easy integration into decision-making processes. Future studies may benefits can provide realistic estimates of the plausible ranges of returns benefit from considering the impact of social governance structures on of interventions, considering all outcome dimensions that are relevant in FLR interventions. a particular context. Since the objectives of FLR programmes can thus be better captured than in traditional evaluations that rely on precise Declaration of Competing Interest measurements, this method is suitable for accounting for costs and benefits of such programmes. To realistically value ecosystem benefits, The authors declare that they have no known competing financial FLR actors should base their predictions on expert knowledge of the interests or personal relationships that could have appeared to influence local context rather than on benchmark estimates carried over from the work reported in this paper. different contexts (Stålhammar and Pedersen, 2017). The use of distri- butions when estimating the value of variables rather than best-bet es- Acknowledgements timates avoids overly hopeful predictions that could misguide planning (Luedeling et al., 2019). This research was carried out under the CGIAR Research Program on The outcomes of this study indicate positive returns for most in- Water, Land and Ecosystems (WLE) with support from CGIAR Fund vestments. This is a clear indication that investments in FLR pro- Donors (http://www.cgiar.org/about-us/our-funders/). We also appre- grammes can succeed in reversing degradation in the long term. ciate the support accorded to us by the implementing agency (WeForest However, initial costs incurred to establish livelihood interventions, Ethiopia) and its partners. We thank community members from Kalaa- mobilize communities, strengthen social governance structures, and min village in Tigray, Ethiopia, Mekelle University and state agencies for provide capacity building and training can result in net losses in the first their input to the decision model development process. Their insights few years. Therefore, FLR actors may need significant financial support were critical in understanding complex interactions between commu- to see their interventions through to the medium and long term (Pis- nity members and Desa’a Forest. In addition, the paper benefited greatly torius et al., 2017). from the comments of reviewers. 4.4. Limitations of the study References The study did not explicitly consider the overall socio-economic Aynekulu, E., Denich, M., Tsegaye, D., 2009. Regeneration response of Juniperus procera environment and the challenges that social factors present to imple- and Olea europaea subsp cuspidata to Exclosure in a dry Afromontane Forest in menters of participatory forest management. Insecure land tenure, low northern Ethiopia. Mt. Res. Dev. 29, 143–152. https://doi.org/10.1659/mrd.1076. Aynekulu, E., Denich, M., Tsegaye, D., Aerts, R., Neuwirth, B., Boehmer, H.J., 2011. levels of technical and technological capacity and a lack of benefit- Dieback affects forest structure in a dry Afromontane forest in northern Ethiopia. sharing agreements pose an additional risk to restoration outcomes J. Arid Environ. 75, 499–503. https://doi.org/10.1016/j.jaridenv.2010.12.013. (Chazdon et al., 2016; Lemenih et al., 2014). For instance, in the case of Birhane, E., Mengistu, T., Seyoum, Y., Hagazi, N., Putzel, L., Rannestad, M.M., Kassa, H., 2018. Exclosures as forest and landscape restoration tools: lessons from Tigray land tenure, landless individuals may not have access to economic in- region, Ethiopia. Int. For. Rev. https://doi.org/10.1505/146554817822330498. centives that would discourage over-harvesting of forest resources. CBD, 2010. Aichi Biodiversity Targets (Aichi Bidiversity Targets). Incomplete consideration of land rights may mean that the returns Chazdon, R.L., Broadbent, E.N., Rozendaal, D.M.A., Bongers, F., Zambrano, A.M.A., Aide, T.M., Balvanera, P., Becknell, J.M., Boukili, V., Brancalion, P.H.S., Craven, D., presented here are only applicable to households that own land and have Almeida-Cortez, J.S., Cabral, G.A.L., De Jong, B., Denslow, J.S., Dent, D.H., direct access to the benefits. Further consideration of biotic factors may DeWalt, S.J., Dupuy, J.M., Durán, S.M., Espírito-Santo, M.M., Fandino, M.C., also be warranted. The effectiveness of restoration and choice of resto- César, R.G., Hall, J.S., Hernández-Stefanoni, J.L., Jakovac, C.C., Junqueira, A.B., Kennard, D., Letcher, S.G., Lohbeck, M., Martínez-Ramos, M., Massoca, P., Meave, J. ration approach are linked with the regeneration potential of the species A., Mesquita, R., Mora, F., Muñoz, R., Muscarella, R., Nunes, Y.R.F., Ochoa- considered. While our analysis did not differentiate between regenera- Gaona, S., Orihuela-Belmonte, E., Peña-Claros, M., Pérez-García, E.A., Piotto, D., tion potentials of the two climax tree species, evidence from previous Powers, J.S., Rodríguez-Velazquez, J., Romero-Pérez, I.E., Ruíz, J., Saldarriaga, J.G., Sanchez-Azofeifa, A., Schwartz, N.B., Steininger, M.K., Swenson, N.G., Uriarte, M., studies indicated that J. procera has low potential for ecosystem recovery Van Breugel, M., Van Der Wal, H., Veloso, M.D.M., Vester, H., Vieira, I.C.G., and might not be effectively restored by the ANR approach (Aynekulu Bentos, T.V., Williamson, G.B., Poorter, L., 2016. Carbon sequestration potential of 11 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 second-growth forest regeneration in the Latin American tropics. Sci. Adv. https:// Luedeling, E., Shepherd, K., 2016. Decision-focused agricultural research. Solutions doi.org/10.1126/sciadv.1501639. September 46–54. Cheboiwo, J., Langat, D., Muga, M., Kiprop, J., 2018. Economic Analysis of Forest Luedeling, E., Whitney, C.W., 2018. R Package decisionSupport: Controlled Burns in Landscape Restoration Options in Kenya. Conifer Forests. Chomba, S., Sinclair, F., Savadogo, P., Bourne, M., Lohbeck, M., 2020. Opportunities and Luedeling, E., Borner, J., Amelung, W., Schiffer, K., Shepherd, K., Rosenstock, T.S., 2019. constraints for using farmer managed natural regeneration for land restoration in Forest restoration: overlooked constraints Forest restoration: expanding agriculture sub-Saharan Africa. Front. For. Glob. Chang. 3 https://doi.org/10.3389/ Forest restoration: transformative trees. Science (80-.) 366, 315. https://doi.org/ ffgc.2020.571679. 10.1126/science.aay7988. Chotte, J.L., Aynekulu, E., Cowie, A., Campbell, E., Vlek, P., Lal, R., Kapović- Luedeling, E., Göhring, L., Schiffer, K., 2020. decision support: quantitative support of Solomun, M., von Maltitz, G., Kust, G., Barger, N., Vargas, R., Gastrow, S., 2019. decision making under uncertainty. CRAN Archive. 366 (6463), 315. Realising the Carbon Benefits of Sustainable Land Management Practices: Guidelines Matzek, V., 2018. Turning delivery of ecosystem services into a deliverable of ecosystem for Estimation of Soil Organic Carbon in the Context of Land Degradation Neutrality restoration. Restor. Ecol. 26, 1013–1016. https://doi.org/10.1111/rec.12872. Planning and Monitoring. McGonigle, D.F., Rota Nodari, G., Phillips, R.L., Aynekulu, E., Estrada-Carmona, N., COP 16, 2011. Framework Convention on Climate Change / Conference of the Parties / Jones, S.K., Koziell, I., Luedeling, E., Remans, R., Shepherd, K., Wiberg, D., 2010 / 7 / Add.1. In: Report of the Conference of the Parties on Its Sixteenth Session, Whitney, C., Zhang, W., 2020. A knowledge brokering framework for integrated Held in Cancun from 29 November to 10 December 2010. Addendum. Part Two: landscape management. Front. Sustain. Food Syst. https://doi.org/10.3389/ Action Taken by the COP at Its Sixteenth Session. fsufs.2020.00013. Cortina, J., Amat, B., Castillo, V., Fuentes, D., Maestre, F.T., Padilla, F.M., Rojo, L., 2011. Mekuria, W., 2013. Changes in regulating ecosystem services following establishing The restoration of vegetation cover in the semi-arid Iberian southeast. J. Arid exclosures on communal grazing lands in Ethiopia: a synthesis. J. Ecosyst. https:// Environ. https://doi.org/10.1016/j.jaridenv.2011.08.003. doi.org/10.1155/2013/860736. Cronkleton, P., Artati, Y., Baral, H., Paudyal, K., Banjane, M.R., Liu, J.L., Tu, T.Y., Mokria, M., Gebrekirstos, A., Aynekulu, E., Bräuning, A., 2015. Tree dieback affects Putzel, L., Birhane, E., Kassa, H., 2018. How do property rights reforms provide climate change mitigation potential of a dry afromontane forest in northern incentives for forest landscape restoration? Comparing evidence from Nepal, China Ethiopia. For. Ecol. Manag. 344, 73–83. https://doi.org/10.1016/j. and Ethiopia. Int. For. Rev. 19 https://doi.org/10.1505/146554817822330506. foreco.2015.02.008. Crouzeilles, R., Curran, M., Ferreira, M.S., Lindenmayer, D.B., Grelle, C.E.V., Rey Munialo, S., Hall, O., Francisca Archila Bustos, M., Boke Olén, N., Onyango, M.C., Oluoch Benayas, J.M., 2016. A global meta-analysis on the ecological drivers of forest Kosura, W., Marstorp, H., Göran, D., 2019. Micro-spatial analysis of maize yield gap restoration success. Nat. Commun. 7, 1–8. https://doi.org/10.1038/ncomms11666. variability and production factors on smallholder farms. Agric. https://doi.org/ Crouzeilles, R., Ferreira, M.S., Chazdon, R.L., Lindenmayer, D.B., Sansevero, J.B.B., 10.3390/agriculture9100219. Monteiro, L., Iribarrem, A., Latawiec, A.E., Strassburg, B.B.N., 2017. Ecological Nigussie, Z., Tsunekawa, A., Haregeweyn, N., Adgo, E., Nohmi, M., Tsubo, M., Aklog, D., restoration success is higher for natural regeneration than for active restoration in Meshesha, D.T., Abele, S., 2017. Factors influencing small-scale farmers’ adoption of tropical forests. Sci. Adv. 3 https://doi.org/10.1126/sciadv.1701345. sustainable land management technologies in North-Western Ethiopia. Land Use de Groot, R.S., Alkemade, R., Braat, L., Hein, L., Willemen, L., 2010. Challenges in Policy 67. https://doi.org/10.1016/j.landusepol.2017.05.024. integrating the concept of ecosystem services and values in landscape planning, Nkonya, E., Mirzabaev, A., von Braun, J., 2015. Economics of land degradation and management and decision making. Ecol. Complex. https://doi.org/10.1016/j. improvement - a global assessment for sustainable development. Economics of Land ecocom.2009.10.006. Degradation and Improvement - A Global Assessment for Sustainable Development. Do, H., Luedeling, E., Whitney, C., 2020. Decision analysis of agroforestry options reveals https://doi.org/10.1007/978-3-319-19168-3. adoption risks for resource-poor farmers. Agron. Sustain. Dev. 40 https://doi.org/ Okuthe, I.K., Akotsi, E.N., 2014. Adoptions of improved biomass cook stoves by 10.1007/s13593-020-00624-5. households : an example from homabay county. Int. J. Humanit. Soc. Sci. 4 (9(1)), FAO, 2019. New UN Decade on Ecosystem Restoration Offers Unparalleled Opportunity 191–206. for Job Creation, Food Security and Addressing Climate Change (News Artic). Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K., Favretto, N., Luedeling, E., Stringer, L.C., Dougill, A.J., 2017. Valuing Ecosystem Services Montanarella, L., Alewell, C., 2015. The new assessment of soil loss by water erosion in Semi-arid Rangelands through Stochastic Simulation. L. Degrad. Dev. 28, 65–73. in Europe. Environ. Sci. Pol. https://doi.org/10.1016/j.envsci.2015.08.012. https://doi.org/10.1002/ldr.2590. Pica-Ciamarra, U., Otte, J., 2010. Poultry, food security and poverty in India: looking Felli, J.C., Hazen, G.B., 1998. Sensitivity analysis and the expected value of perfect beyond the farm-gate. Worlds. Poult. Sci. J. https://doi.org/10.1017/ information. Med. Decis. Mak. https://doi.org/10.1177/0272989X9801800117. S0043933910000358. Gashaw, T., Bantider, A., G/Silassie, H, 2014. Land degradation in Ethiopia : causes, Pistorius, T., Carodenuto, S., Wathum, G., 2017. Implementing Forest landscape impacts and rehabilitation techniques land degradation in Ethiopia : causes, impacts restoration in Ethiopia. Forests 61, 1–19. https://doi.org/10.3390/f8030061. and rehabilitation techniques. J. Environ. Earth Sci. 4, 98–104. R Core Team, 2017. R: A language and environment for statistical computing. R Found. Gitz, V., Place, F., Koziell, I., Pingault, N., van Noordwijk, M., Alexander, M., Minang, P., Stat. Comput.. ISBN 3-900051-07-0. 2020. A joint stocktaking of CGIAR work on forest and landscape restoration. In: Romijn, E., Coppus, R., De Sy, V., Herold, M., Roman-Cuesta, R.M., Verchot, L., 2019. A Joint Stocktaking of CGIAR Work on Forest and Landscape Restoration. https:// Land restoration in Latin America and the Caribbean: An overview of recent, ongoing doi.org/10.17528/cifor/007669. and planned restoration initiatives and their potential for climate change mitigation. Grieshop, A.P., Marshall, J.D., Kandlikar, M., 2011. Health and climate benefits of Forests. https://doi.org/10.3390/f10060510. cookstove replacement options. Energy Policy 39, 7530–7542. https://doi.org/ Rumpff, L., Duncan, D.H., Vesk, P.A., Keith, D.A., Wintle, B.A., 2011. State-and-transition 10.1016/j.enpol.2011.03.024. modelling for adaptive management of native woodlands. Biol. Conserv. 144, Hubbard, D.W., 2014. How to Measure Anything: Finding the Value of Intangibles in 1224–1236. https://doi.org/10.1016/j.biocon.2010.10.026. Business. https://doi.org/10.1002/9781118983836. Scull, P., Cardelús, C.L., Klepeis, P., Woods, C.L., Frankl, A., Nyssen, J., 2017. The IPCC, 2006. 2006 IPCC guidelines for National Greenhouse gas Inventories Volume 4 resilience of Ethiopian church forests: interpreting aerial photographs, 1938–2015. agriculture, forestry and other land use chapter 4 forest land 2006. Forestry. L. Degrad. Dev. https://doi.org/10.1002/ldr.2633. IUCN, WRI, 2014. A guide to the Restoration Opportunities Assessment Methodology Shono, K., Cadaweng, E.A., Durst, P.B., 2007. Application of assisted natural (ROAM): Assessing forest landscape restoration opportunities at the national or sub- regeneration to restore degraded tropical forestlands. Restor. Ecol. https://doi.org/ national level. (Road-test Ed). 10.1111/j.1526-100X.2007.00274.x. Klein, G., 2007. Performing a project premortem. Harv. Bus. Rev. 85, 18. Available Sircely, J., Conant, R.T., Boone, R.B., 2019. Simulating rangeland ecosystems with G- online at: https://hbr.org/2007/09/performing-a-project-premortem. range: model description and evaluation at global and site scales. Rangel. Ecol. Köhl, M., Neupane, P.R., Lotfiomran, N., 2017. The impact of tree age on biomass growth Manag. https://doi.org/10.1016/j.rama.2019.03.002. and carbon accumulation capacity: a retrospective analysis using tree ring data of Stålhammar, S., Pedersen, E., 2017. Recreational cultural ecosystem services: how do three tropical tree species grown in natural forests of Suriname. PLoS One 12. people describe the value? Ecosyst. Serv. https://doi.org/10.1016/j. https://doi.org/10.1371/journal.pone.0181187. ecoser.2017.05.010. Kusters, K., Buck, L., de Graaf, M., Minang, P., van Oosten, C., Zagt, R., 2018. Stanturf, J.A., Kleine, M., Mansourian, S., Parrotta, J., Madsen, P., Kant, P., Burns, J. , Participatory planning, monitoring and evaluation of multi-stakeholder platforms in Bolte, A., 2019. Implementing forest landscape restoration under the Bonn integrated landscape initiatives. Environ. Manag. https://doi.org/10.1007/s00267- Challenge: a systematic approach. Ann. For. Sci. https://doi.org/10.1007/s13595- 017-0847-y. 019-0833-z. Lanzanova, D., Whitney, C., Shepherd, K., Luedeling, E., 2019. Improving development Tamba, Y., Wafula, J., Yigzaw, N., Whitney, C., Lüdeling, E., 2020. Data for the efficiency through decision analysis: Reservoir protection in Burkina Faso. evaluation of forest and landscape restoration in Northern Ethiopia. Mendeley Data Environmental Modelling & Software 115, 164–175. https://doi.org/10.1016/j. v1. envsoft.2019.01.016. Teklay, A., Abera, B., Giday, M., 2013. An ethnobotanical study of medicinal plants used Lemenih, M., Kassa, H., Kassie, G.T., Abebaw, D., Teka, W., 2014. Resettlement and in kilte awulaelo district, Tigray region of Ethiopia. J. Ethnobiol. Ethnomed. https:// woodland management problems and options: A case study from North-Western doi.org/10.1186/1746-4269-9-65. Ethiopia. L. Degrad. Dev. https://doi.org/10.1002/ldr.2136. Tesfaye, S., Guyassa, E., Joseph Raj, A., Birhane, E., Wondim, G.T., 2014. Land use and Luedeling, E., Gassner, A., 2012. Partial least squares regression for analyzing walnut land cover change, and Woody vegetation diversity in human driven landscape of phenology in California. Agric. For. Meteorol. 158–159, 43–52. https://doi.org/ Gilgel Tekeze catchment, Northern Ethiopia. Int. J. For. Res. https://doi.org/ 10.1016/j.agrformet.2011.10.020. 10.1155/2014/614249. Luedeling, E., Oord, A.L., Kiteme, B., Ogalleh, S., Malesu, M., Shepherd, K.D., de UNEP, 2019. New UN Decade on Ecosystem Restoration Offers Unparalleled Opportunity Leeuw, J., 2015. Fresh groundwater for Wajir-ex-ante assessment of uncertain for Job Creation, Food Security and Addressing Climate Change [WWW Document]. benefits for multiple stakeholders in a water supply project in Northern Kenya. Front. Environ. Sci. 3, 1–18. https://doi.org/10.3389/fenvs.2015.00016. 12 Y. Tamba et al. F o r e s t P o l i c y a n d E c o n o m i c s 125 (2021) 102403 Vågen, T-G, Winowiecki, LA, Tamene Desta, L, Tondoh, JE, 2013. The Land Degradation Yami, M., Mekuria, W., Hauser, M., 2013. The effectiveness of village bylaws in Surveillance Framework (LDSF) - Field Guide v3. World Agroforestry Centre, sustainable management of community-managed exclosures in northern Ethiopia. Nairobi, Kenya. Sustain. Sci. 8, 73–86. https://doi.org/10.1007/s11625-012-0176-2. Wafula, J., Karimjee, Y., Tamba, Y., Malava, G., Muchiri, C., Koech, G., De Leeuw, J., Yang, K.F., Gergel, S.E., Baudron, F., 2020. Forest restoration scenarios produce Nyongesa, J., Shepherd, K., Luedeling, E., 2018. Probabilistic assessment of synergies for agricultural production in southern Ethiopia. Agric. Ecosyst. Environ. investment options in honey value chains in Lamu County. Kenya. Front. Appl. Math. 295 https://doi.org/10.1016/j.agee.2020.106888. Stat. 4, 1–11. https://doi.org/10.3389/fams.2018.00006. Yet, B., Constantinou, A., Fenton, N., Neil, M., Luedeling, E., Shepherd, K., 2016. WeForest, 2018. THE GREAT RIFTVALLEY DRY AFROMONTANE Desa ’ a State Forest A Bayesian network framework for project cost, benefit and risk analysis with an Management Plan. agricultural development case study. Expert Syst. Appl. https://doi.org/10.1016/j. Weldesemaet, Y.T., 2015. Enhancing food security through forest landscape restoration: eswa.2016.05.005. lessons from Burkina Faso, Brazil, Guatemala, Viet Nam, Ghana, Ethiopia and Yet, B., Lamanna, C., Shepherd, K., Rosenstock, T., 2020. Project evaluation using Philippines, Enhancing food security through forest landscape restoration: Lessons Bayesian networks (PEBN): prioritizing agricultural development investments under from Burkina Faso, Brazil, Guatemala, Viet Nam, Ghana, Ethiopia and Philippines. climatic and socio-political risk. PLoS One 15 (6). https://doi.org/10.1371/journal. https://doi.org/10.2305/iucn.ch.2015.fr.2.en. pone.0234213. Wendt, D., 1975. Some criticism of stochastic models generally used in decision making Yigzaw, N., Mburu, J., Ackello-Ogutu, C., Whitney, C., Luedeling, E., 2019. Stochastic experiments. Theory Decis. https://doi.org/10.1007/BF00169107. impact evaluation of an irrigation development intervention in northern Ethiopia. Wilson, E.C.F., 2015. A practical guide to value of information analysis. Sci. Total Environ. 685, 1209–1220. https://doi.org/10.1016/j. Pharmacoeconomics. https://doi.org/10.1007/s40273-014-0219-x. scitotenv.2019.06.133. Wold, S., Sjöström, M., Eriksson, L., 2001. PLS-regression: A basic tool of chemometrics. Zomer, R.J., Bossio, D.A., Sommer, R., Verchot, L.V., 2017. Global sequestration In: Chemometrics and Intelligent Laboratory Systems, pp. 109–130. https://doi.org/ potential of increased organic carbon in cropland soils. Sci. Rep. https://doi.org/ 10.1016/S0169-7439(01)00155-1. 10.1038/s41598-017-15794-8. World Resources institute, 2015. The Restoration Diagnostic: A Method for Developing Forest Landscape Restoration Strategies by Rapidly Assessing the Status of Key Success Factors. 13