Cost‐effectiveness of management strategies in a nucleation experiment in a tropical dry forest

Cost‐effective, large‐scale strategies are needed to restore degraded ecosystems worldwide. Applied nucleation is one technique that can accelerate succession in tropical forests. However, the effectiveness of irrigation and fertilization in the context of large‐scale applied nucleation in tropical dry forests (TDFs) has not yet been tested. To this end, we established a large‐scale experiment in southwestern Colombia on abandoned pastures. We planted 11,382 seedlings of 11 native species coupled with six management treatments that varied in the amounts of fertilizer and water. We monitored survival, height, and resprouting ability seedlings over 2 years. We also estimated the cost of seedling production, planting procedures, management, and monitoring, and assessed their cost‐effectiveness using seedling survival as an indication of effectiveness. After 2 years, 73% (8,266) of seedlings planted survived, and species survival ranged from 5 to 99%. Individuals that received the lowest amount of fertilizer (25 g of Nitrogen, Phosphorus, and Potassium [NPK]) with additional irrigation had the highest survival rates (>76%). Final height ranged from 52 to 330 cm across species. Seedlings that received the highest amount of fertilizer (50 g of NPK) without additional irrigation had the highest mean final height (174 ± 9.0 cm). The control was cheapest ($7313/ha) and the most cost‐effective method in terms of seedling survival. Our findings suggest that the best way to establish cost‐effective, large‐scale restoration projects in TDFs is to plant native species mixtures of locally adapted species without fertilization and irrigation and to invest in long‐term monitoring.


Introduction
The United Nations' "Decade on Ecosystem Restoration" has focused a much-needed spotlight on the challenges and possibilities of restoring one billion hectares of degraded lands worldwide.This global initiative aims to boost previous international (the Bonn Challenge, the New York Declaration on Forests) and regional efforts (initiative 20 Â 20) to restore forest landscapes.Given that nearly 2 billion ha of terrestrial ecosystems have been converted to alternative uses, and every year 10 million ha more are lost (FAO 2020), these large-scale restoration initiatives seek to slow down biodiversity loss and land use change (i.e.large-scale agriculture, cattle ranching, etc.), enhance climate change mitigation and adaptation, and secure the welfare of future generations (Alexander et al. 2011;Griscom et al. 2017).Most of these current restoration efforts are focused on restoring highly threatened ecosystems such as tropical forests, which have a high carbon storage Authors contributions: LT, FTR, JSP designed the experiment; LT, SMS implemented the experiment and collected data; LT performed data management and statistical analysis; JSP, SG, AAM, SS provided insight into the cost-effectiveness analysis and interpretation; LT, JSP wrote the manuscript and all other authors contributed to revisions.
Multiple strategies have been implemented to restore areas in tropical regions.These approaches include natural regeneration, that is, the removal of the cause(s) of degradation and leaving secondary succession to proceed on its own (Chazdon & Guariguata 2016), assisted natural regeneration, that is, the removal of the cause(s) of degradation and implementation of active interventions to correct abiotic and biotic damage (Guariguata & Dupuy 1997), and active restoration, that is, the reintroduction or planting of multiple desired tree species (Holl & Aide 2011).The results obtained and costs vary among methods (Brancalion et al. 2016(Brancalion et al. , 2019)).Therefore, it is critical to identify the target system's desired state, establish restoration goals, and have a clear budget to be able to determine the best management practices for the area that will be intervened (Dimson & Gillespie 2020;Bodin et al. 2022).Passive restoration strategies like natural regeneration tend to be inexpensive and comparatively easy to implement and scale up compared to more active methods (Zahawi et al. 2014;Uriarte & Chazdon 2016).However, passive restoration practices cannot always be implemented.Their effectiveness in most tropical areas is still uncertain because most passive restoration projects tend to be established in areas where secondary succession is ongoing, and usually rapidly (Reid et al. 2018).Therefore, active, and more costly interventions, like high density plantations or planting in islands (applied nucleation, here after "nucleation") (Corbin & Holl 2012), tend to be favored where succession has been arrested (Holl & Aide 2011;Chazdon et al. 2017) or in areas far from forest remnants (Chazdon 2013).Nucleation has proven to be a cost-effective way to stimulate natural recovery in multiple tropical ecosystems (Holl et al. 2011;Bechara et al. 2021).However, in ecosystems that have low biomass recovery rates and have been extensively deforested, like tropical dry forest (TDF) (Dirzo et al. 2011;Buchadas et al. 2022), the nucleation method has not been extensively tested.
TDF is a challenging ecosystem to restore (Werden et al. 2020;Negoita et al. 2022).This ecosystem experiences strong seasonality (Murphy & Lugo 1986) that limits seed and seedling establishment because water and nutrients are only available during certain times of the year and for short periods of time (Vieira & Scariot 2006).The dominance of non-native grasses in TDFs soil seed banks also reduces the survival of native seedlings (Cordell et al. 2008;Treuer et al. 2017).Therefore, projects should invest more resources in establishment and maintenance, which increase the costs of the restoration (project establishment ranges from $105 to 25,830/ha, and annual maintenance ranges from $167 to 2421/ha) (Bodin et al. 2022).Some of the highest costs of restoring TDFs include purchasing seeds, seedling production, site preparation, tree planting, management (i.e.irrigation, fertilization, and weed control), and monitoring (Brancalion & Holl 2020).Irrigation and fertilization are the most common and expensive management practices implemented in TDFs (Dimson & Gillespie 2020).Seedlings in TDF restoration projects are often irrigated after being planted to increase their survival (Dimson & Gillespie 2020).Even though TDF species might be adapted to certain nutrient conditions (Lambers et al. 2008), fertilizers are frequently applied because most soils are either highly degraded, or simply because projects have adopted methods from intensive forestry and agriculture to quickly establish trees (Brancalion et al. 2019).
Even though the addition of water and fertilizers are intended to increase plant survival and growth, these management practices do not always achieve the expected outcomes, and their costs may exceed the benefits obtained (Brancalion et al. 2016(Brancalion et al. , 2019).Fertilizers, for example, can be toxic to plants or can increase seedling mortality if they are applied in excess (Fajardo et al. 2013).Although agronomic research has extensively addressed the impact of fertilization practices on crop production (Steiner et al. 2007), restoration studies still lack information about how fertilizers impact native tree species and non-native grasses performance (Fajardo et al. 2013;Cole et al. 2021).Finally, the complexity of estimating restoration costs (Kimball et al. 2015;Brancalion et al. 2019), difficulty of replicating the methodologies of experimental studies in projects driven by local communities (Garen et al. 2009), lack of cost reports for large-scale projects, conflicting results among the few existing cost-effectiveness analyses for similar management practices (Holl et al. 2011;Ramírez-Soto et al. 2018), and the handful of projects that have implemented mid-and longterm (at least 2 years) monitoring plans (Dimson & Gillespie 2020) make it hard to predict when a strategy is worth applying and its potential cost.This uncertainty impacts the ability of restoration practitioners to successfully restore a habitat at the lowest cost (Robbins & Daniels 2012), and prevents researchers from extrapolating their findings to long-term and large-scale restoration projects (Wortley et al. 2013).Therefore, more cost-effectiveness analyses of different management practices are needed to distribute limited project funds, improve restoration policies (Brancalion & Chazdon 2017;Bodin et al. 2022), restore tropical forests in the next decade (Holl 2017), and support practitioners in their decision-making process in diverse settings (Kimball et al. 2015).
Therefore, the goal of our study was to evaluate the effects of different fertilization and irrigation practices on tree seedling performance during the initial phase of a large-scale nucleation experiment, and to compare the cost-effectiveness of these management practices in terms of the cost of seedling survival per hectare.We monitored the growth, survival, and resprouting ability of 11 native tree species including four N-fixing legumes that received one out of six fertilization and irrigation treatments (50 g of NPK plus irrigation, 25 g of NPK plus irrigation, 43 g of phosphoric rock plus irrigation, 50 g of NPK without additional irrigation, irrigation, and control) for 2 years.More specifically, we addressed the following questions: (1) How do management practices impact the performance of planted TDF tree species?(2) What are the trade-offs across plant performance traits?And (3) How does the cost-effectiveness of management practices vary among species?We tracked the costs of management practices-including monitoring-so that we could compare performance on a cost-effectiveness basis.Finally, we collected pretreatment soil samples to better understand nutrient and texture variability across plots.

Study Site
Our experiment was established in December of 2019 in a TDF area at El Quimbo Environmental Offsetting and Hydroelectric Project-(here after "El Quimbo") in southwestern-central Colombia (2 17 0 47.60 00 N, 75 40 0 52.12 00 W).Mean annual temperature at El Quimbo is approximately 24 C and mean annual precipitation is approximately 1036 mm with two periods with little to no rain (June-August and December-January) (Avella-M et al. 2019).Soils in the area include sandy inceptisols with moderate soil development and clayey mollisols with high organic matter content (Torres-Rodríguez et al. 2019).In 2014, the Enel-Emgesa company built a large dam at El Quimbo.The construction of this dam impacted approximately 1200 ha of TDFs, and as part of Colombia's biodiversity offset plan (MADS 2012) the company agreed to restore approximately 12,000 ha of degraded TDFs in the next 20 years (Torres-Rodríguez et al. 2019).This area has been exposed to extensive agriculture (rice and corn) and cattle grazing since members of the Spanish army and religious groups established in Colombia (Ducuara Manrique 2011), so there has been little evidence of spontaneous natural regeneration on these formerly cultivated lands.The dominant species growing in the area are native (Aristida gibbosa

Seedling Production
We collected seeds from nine native tree species near the site and purchased seeds of Handroanthus ochraceus (Cham.)Mattos and Cordia alliodora (Ruiz & Pav.) Oken from NeoTerra, Bogot a, Colombia.The selected species commonly occur in TDFs in Colombia (Pizano & García 2014).We included four N-fixing species from two subfamilies of Fabaceae (Caesalpinioideae and Papilionoideae) and seven non-fixing species (Table S1).In April 2019 prior to planting, all seeds were soaked in water for 24 hours to soften the seed coats.Pseudosamanea guachapele (Kunth) Harms and Vachellia farnesiana (L.) Wight & Arn.seeds were scarified to accelerate germination by placing the seeds in 70 C water for 4 minutes before soaking them in ambient water temperature (Villota-Ojeda et al. 2019).All seeds were then sown into seedbeds that had 60% ant peat, 30% soil, and 10% sand.The seedbeds were located in a nursery that was covered with 80% polyethylene shade cloth that excluded approximately 80% of photosynthetically active radiation.Seeds spent on average 15-20 days on the seedbeds and were watered every day until germination.Once seedlings were 6-8 cm tall, they were placed into conical containers (290 cm 3 ) that contained 60 g of ant peat, 30 g of soil, and 10 g of rice husk.The potted seedlings also received potassium nitrate and indolebutyric acid to stimulate root growth (1 L of water for 0.5 g indolebutyric acid, 2 g of potassium nitrate).These seedlings spent 5-6 months (early May to late December) in an area that was covered with 60% polyethylene shade cloth and were watered every 1.5 days.Seedlings that showed herbivory damage were sprayed with a diluted insecticide (1 g/L Lorsban-Chlorpyrifos, 1 L per 40 seedlings).Prior to transportation to the field, seedlings spent 3 weeks in full sun to harden.During this time, we recorded plant height and added a plastic numbered tag to each seedling (20-29 November, 2019).Plant height was measured from the base of the stem to the apical meristem.
Experimental Design, Seedling Planting, and Maintenance In November 2019, we cleared with machetes and scythes the existing vegetation of a 7 ha patch of grassland, which is 1.5 km away from the nearest forest remnant.Then, we tilled the area twice with a tractor to reduce soil compaction (30-40 cm).After tilling, we set up 42 hexagonal plots (35 m Â 35 m) that each contained nine concentric rings (Fig. 1).Each plot was separated by approximately 15 m from other plots.In each plot, we dug 271 holes (30 Â 30 Â 30 cm) that were 1.5 m apart from each other and from the plot edge.Before planting, seedlings greater than 40 cm tall were submerged in indolebutyric acid for a minute to activate root growth.A total of 11,382 seedlings were planted over a 30-day period (December 2019-January 2020) that encompasses the dry season (December-February).The number of individuals planted per species and their position in the plot varied according to their light requirements and nursery availability (Table S1; Fig. S1).The mid-and late-successional species were interplanted in the first six inner rings, while the early successional species were planted in the last three rings.
Upon planting, seedlings in each hexagonal plot received 1 kg of ant peat, 10 g of hydrogel (to extend the period of favorable soil moisture) and one of four fertilization treatments (50 g of NPK per seedling [equivalent to 81.3 kg ha À1 yr À1 -14% N, 25% P, 8% K, 3% Ca, 2% Mg, 1% S, and micronutrients], 25 g of NPK per seedling [40.5 kg ha À1 yr À1 ], 43 g of phosphoric rock per seedling [69.9 kg ha À1 yr À1 -29% P and 40% Ca]) or no fertilizer, depending on the plot (Table 1).Plots that were not irrigated nor fertilized were considered control plots.A random generator number was used to decide the treatment each plot would get.Every treatment was replicated seven times (6 treatments Â 7 replicates = 42 plots).The NPK fertilizer used in this experiment was locally produced and specifically prepared for El Quimbo.We chose the fertilization rates based on previous essays in the area conducted by Fundaci on Natura (2018).The fertilization treatments were applied around the seedlings, and then the holes were refilled with soil.Then, a layer of organic mulch was applied around every seedling to prevent desiccation.Fertilized seedlings received a second dose of the initial granular fertilizers in December 2020.The granular fertilizers (NPK and phosphoric rock) were broadcasted, and each seedling received the same original dose.All seedlings received 7 L of water once they were planted to activate the hydrogel (December 2019-January 2020).Seedlings in irrigation treatments received an additional 7 L of water during two more periods of time.The control and the 50 g of NPK without additional irrigation treatments did not receive any additional irrigation to independently test the effects of fertilizers and water on seedling performance.In February 2020, individuals of Ceiba pentandra (L.) Gaertn and Gliricidia sepium (Jacq.)Kunth ex Walp showed herbivory damage, so we sprayed all individuals of these two species with a diluted insecticide (1 g/L Lorban, 1 L per 40 seedlings).To reduce competition from weeds, we cleared all grasses and woody vines growing close to the planted seedlings with machetes and hoes in June 2020, March 2021, and January 2022.

Soil Characterization
In November 2019, we collected five 10 cm deep volumetric soil samples (456 cm 3 ) in each plot using a steel ring at the four corners and center of the plot.Soil samples were air-dried and pooled per plot.A subset of the composite samples was analyzed for pH, texture, and nutrients at the soil laboratory of the International Center for Tropical Agriculture (CIAT), Palmira, Colombia.Initial soil pH ranged from 4.77 to 5.48 among plots.Soils were classified either as sandy clay loams or clay loams United States Department of Agriculture (USDA).Sand concentrations ranged from 39.76 to 51.46%, whereas silt concentrations ranged from 24.73 to 38.89%.Soil clay concentrations varied from 21.73 to 38.89%.Soil organic carbon ranged from 15.95 to 33.94 g/kg.Total N ranged from 1.26 to 1.72 g/kg, while soil available P (Bray II) varied from 3.20 to 11.06 mg/kg among the 42 plots (Table S2).

Seedling Survival, Growth, and Resprouting
We monitored plant survival, height, and resprouting ability of every planted seedling for 2 years.We recorded survival six times: February, July, August, December 2020, March 2021, and January 2022.Individuals that were reported dead in one or more surveys but were alive during the last census (January 2022) were recorded as alive in previous censuses.Plant height and resprouting ability were recorded three times (July 2020 [dry season], March 2021 [rainy season], and January 2022 [rainy season]).Due to the COVID-19 pandemic, we were not able to collect data at the beginning of the dry season (June 2020, June, and December 2021).Plant height was measured from the base of the stem to the apical meristem for each seedling.Finally, resprouting ability was defined as seedling's ability to produce a new stem when in a previous census the main stem was recorded dead.This variable was summarized for each species (Pausas & Keeley 2014).

Herbivory
We devised a method to indirectly account for herbivory due to the presence of cows in some of the plots.We calculated the percentage of seedlings that presented signs of herbivory per plot.The percentages ranged from 0 to 40%.We then used the Jenks natural breaks (Jenks & Coulson 1963) to classify herbivory damage into four ordinal classes to be able to synthesize the variation across plots.

Cost Estimation
To compare the costs among the management strategies, we divided them into four categories: seedling production, planting procedures (cleaning, plot establishment, tilling, and digging), management practices (fertilizers, hydrogel, ant peat, irrigation, pesticides, fertilizer application, and post-planting maintenance) and monitoring.We documented the labor time, labor cost, and material costs required to establish and maintain each treatment for 2 years.Fundaci on Natura provided all the cost and labor information in Colombian pesos, and we converted them to U.S.$ using the average exchange rate from 2020 to 2022 (COL$3,311).An overview of the cost estimates is provided in Table 2 and the items that explain each category can be found in Table S3.Finally, the cost-effectiveness was calculated by first estimating the total cost of planting a hectare of grassland with a single tree species (a hectare is equivalent to six nuclei = 1,626 individuals) under a specific treatment, and then dividing that cost by the % survival of that species under that specific treatment after 2 years, this variable was reported in U.S.$/ha Â % survival.The most cost-effective method was the one that yielded the lowest cost and had the highest % survival.

Statistical Analysis
First, we determined the effect of management practices over 2 years on plant performance in terms of % survival, final height, and resprouting ability.To evaluate how survival changed across time and treatments, we used a logistic regression with a binomial distribution (alive = 1, dead = 0).In this model, % survival was the response variable.Treatment, time (0, 30, 180, 330, 420, and 730 days), and their interactions were included as fixed effects.Herbivory was included as a covariable to control for differences in grazing across plots.We obtained the F-values of a two-way analysis of variance (type III) for the fixed effects and performed multiple comparisons among treatments and time using a Tukey post hoc test.To determine if species had distinct responses to management treatments after 2 years, we used a mixed effects logistic regression with a binomial distribution (alive = 1, dead = 0).In this model, % survival after 2 years was the response variable.Species identity, treatment, and their interactions were included as fixed effects.Herbivory was included as a covariable and plot identity as a random effect to account for variation among plots.The Fvalues and the multiple comparisons among species and treatments were obtained in the same way the logistic regression values were obtained (Supplement S1).
We identified two potential issues with final height measurements.Because we were uniquely interested in the nutrient and water effects on seedling growth after 2 years, and we wanted to separate them from other sources of variability we removed seedlings from our dataset that were dead (3,116 seedlings) during the last census, and/or those whose growth may have been affected by cattle grazing or human error, which we identified as seedlings that had height increases less than 5 cm compared to their initial height (n = 698 seedlings).To examine the effects of fertilizers and irrigation on tree species final height after 2 years, we used a linear mixed model.Species identity, treatment, and their interactions were included as fixed effects.Herbivory was included as a covariable and plot identity as a random effect to account for variation among plots.We calculated species resprouting percentage by counting the number of resprouts per species and dividing them by the number of resprouts per species plus the number of dead individuals per species (% resprouting per species = no. of resprouts per species/[no. of resprouts per species + no. of dead seedlings per species] Â 100).We fitted a linear model to explore how resprouting ability varied among species.The F-values of both the linear mixed model and the linear model were obtained in the same way the logistic regression values were obtained.Finally, we evaluated the Spearman pairwise correlations between survival, final height (all living individuals after 2 years were included), and resprouting ability, to explore trade-offs among performance traits.All statistical analyses were done using R software for statistical computing version 4.0.3(R Core Team 2022) with the following packages: lme4 (Bates et al. 2011) and car (Fox & Weisberg 2020).

Seedling Performance
After 2 years, 8,266 out of 11,382 seedlings planted survived.The plots that received the 25 g of NPK plus irrigation treatment had the highest survival rates (76.6%),whereas the ones that were fertilized with 50 g of NPK without additional irrigation had the lowest survival rates (68.8%)(Table S4; Fig. 2).Even though survival rates differed significantly among treatments, the plots where seedlings received 50 g of NPK without additional irrigation had survival rates significantly lower than the control plots ( p < 0.05).Survival rates differed significantly across treatments and species ( p < 0.001).Among the studied species, C. karstenii seedlings treated with 50 g of NPK without additional irrigation had 1.5 lower chances of surviving compared to the ones in the control treatment ( p < 0.001), whereas O. pyramidale seedlings that received 50 g of NPK with and without additional irrigation were 2.6 less unlikely to survive than seedlings that did not receive any additional management treatment (control) (p < 0.05) (Fig. 3).The other nine species planted did not have significant survival rates compared to the control treatment (p > 0.05).The main differences in survival rates were observed across species (p < 0.001), which ranged from 5 to 99% (Table S5; Fig. 3).The rank order of species survival was: Ceiba pentandra > Gliricidia sepium > Vachellia farnesiana > Maclura tinctoria > Pithecellobium dulce > Citharexylum karstenii > Tabebuia rosea > Speudasamanea guachepele > Cordia alliodora > Ochroma pyramidale > Handroanthus ochraceus (Table S6).Final height varied from 25.7 to 680 cm after 2 years.Plots that received 50 g of NPK without additional irrigation had the highest mean final height (169 AE 9.0 cm).Seedlings that received 43 g of P plus irrigation or were only irrigated had the lowest mean final height (119 AE 9.0 cm).We found that the effect of the treatment depends on the identity of the species ( p < 0.001) (Fig. S2).However, not all species responded in the same way to the treatments applied, as only three species had different final heights compared to the control ( p < 0.05).For example, C. pentandra individuals in the control had a higher final height compared to the individuals that were only irrigated or received 43 g of P plus irrigation ( p < 0.05).While O. pyramidale plants growing in the control plots had a higher final height compared to the ones that were only irrigated ( p < 0.001), received 43 g of P plus irrigation ( p < 0.01), or received the 50 g of NPK without irrigation treatment ( p < 0.01).Finally, T. rosea individuals that received 43 g of P plus irrigation had a lower final height compared to the ones growing in the control plots (p < 0.05).The identity of the species was the main factor determining the differences in seedling final height ( p < 0.001) (Table S5).The rank order of maximum final height per species was: O. pyramidale > C. pentandra > G. sepium > P. dulce > S. guachepele > T. rosea > V. farnesiana > M. tinctoria > C. alliodora > H. ochraceus > C. karstenii (Table S6; Fig. 4).

Cost-Effectiveness
C. pentandra ($5.48-$8.17)was the most expensive species planted across treatments followed by G. sepium ($5.37-$8.01).The other nine species had similar costs ($4.19-$7.0).After 2 years, the total cost of restoring a hectare of TDF, varied from $7,313/ha (control) to $11,689/ha (50 g of NPK plus irrigation) (Table 3).In the case of treatments with irrigation, the operational and management procedures were the costliest activities, accounting for 73 to 76% of the total cost.Irrigation was the most expensive management procedure, adding $3,512/ha during the first year of the project.On the other hand, the most expensive activities for the nonirrigated treatments (50 g of NPK without additional irrigation and the control) were the operational and monitoring procedures.They represented 57 to 63% of the total cost.Monitoring a hectare of TDF seven times in this area cost $1,614/ha ($231 per measurement), which includes measuring the height, diameter, survival, and resprouting ability of 1,626 individuals, manually removing weeds close to the planted seedlings, and data entry (Table 3).This task required three people working for 1.6 days.Finally, the values of cost-effectiveness analysis significantly varied across treatments and species.The control treatment was on average the most cost-effective strategy ($13,505/ha), while the 50 g of NPK plus irrigation treatment was the least cost-effective strategy ($41,550/ha).V. farnesiana, G. sepium, and M. tinctoria were on average the most costeffective species to plant per hectare ($10,141, $11,977, and $11,825/ha) across treatments, whereas O. pyramidale and H. ochraceus were the least cost-effective species per hectare ($42,167 and $115,632/ha) (Table S3).

Discussion
During the initial phase (2 years) of this large-scale restoration project, we evaluated how practitioners could invest their limited funds to restore TDF areas.Plant survival, our indicator of restoration effectiveness, was on average 73% across treatments.However, survival rates had high variation among species and little to modest variation among treatments even though the fertilizer rates applied were much higher than the typically used (Alvarez-Clare et al. 2017;Waring et al. 2019).The cost of total planting a hectare (1,626 individuals/ha) of a TDF in Galapagos ranged from $7,317 to 251,266/ha depending on the species planted and treatment applied (Negoita et al. 2022), whereas a nucleation at El Quimbo ranged from $7,313 to $11,689/ha, suggesting that the cost of a nucleation and a total planting can be similar.We found that the cost-effectiveness of irrigation and fertilization strategies can be 1.6-5 times higher than the control treatment, making the cost-effectiveness per hectare highly variable across species and among treatments.Overall, the control treatment was the most cost-effective management practice across species, in part due to the low costs of no additional management beyond planting and monitoring, and the modest benefits to survival of costly irrigation and fertilization practices.Our results highlight the relevance of testing the outcomes of different management strategies and the importance of monitoring them when large-scale restoration projects are established.Below, we explore how different management practices impacted tree performance, and the cost-effectiveness of implementing these practices.
The average survival rates of the tree species planted in this study were similar to those reported by Wishnie et al. (2007) reported in TDF in Panama, but much higher than those reported by Werden et al. (2018) and Negoita et al. (2022) in TDFs in Costa Rica and Ecuador, respectively.As expected, seedlings survival rates decreased after the first year of being planted because their root systems were establishing and competition for resources with grasses was high (Cole et al. 2011;Werden et al. 2018).It is not clear why seedlings in our experiment enjoyed comparatively high survival rates among treatments.It is possible that regular weeding (every 6 months) reduced competition with non-native grasses particularly for slow-growing species (Zimmerman et al. 2000;Bhadouria et al. 2020), and/or the selected species were well adapted to former agricultural field and cattle pasture conditions.The use of hydrogel might have also increased species survival rates (Fajardo et al. 2013;Werden et al. 2018).However, the follow-up irrigation that seedlings received twice during the first year in four out of the six treatments did not have a significant effect on seedling survival.This result is similar to those reported by Ferreira and Vieira (2017), as the control and the treatments that received additional water had similar survival rates (p > 0.05).
Total survival is a useful metric to report when comparing restoration outcomes among studies; however, multispecies projects could increase their successes if they examine speciesspecific survival rates (Cabin et al. 2002).In our large-scale nucleation experiment, survival rates ranged from 5 to 99% across species.The large differences observed in species survival could be explained by species selection, seed provenance, soil characteristics, light availability, competition with non-native grasses, and presence of herbivores (Medeiros et al. 2014).For example, the high mortality of Handroanthus.ochraceus and Cordia.alliodora may be explained by the fact that these species are not adapted to full-sun conditions and low soil fertility (Wishnie et al. 2007;Potvin & Gotelli 2008), or that their seeds were collected in wetter forests.These species can be found across broad precipitation and altitudinal gradients (Pizano & García 2014).It is possible that the genotypes planted were not adapted to the microclimatic conditions of the area that could have reduced the species chances of surviving (Bischoff et al. 2008).Moreover, non-native grasses dominate these highly managed areas (Craven et al. 2009; Cole  Cost-effectiveness of tropical dry forest restoration et al. 2011).These introduced species can take up seedling resources and modify soil characteristics resulting in non-native species outcompeting native species (Holl 1998;Cole et al. 2011).Finally, grazing by cows and other herbivores can have a negative effect on seedlings survival and growth (Griscom et al. 2005;Negoita et al. 2022).Even though, the presence of herbivores did not directly affect overall species survival rates, we observed how the presence of cows impacted the growth rates of Gliricidia.sepium.
There are multiple metrics to assess restoration outcomes, and the selection of this metric will influence the cost-effectiveness of a specific management practice.In addition to seedling survival, we included final height and resprouting as a success variable.Fast-growing species with broad permanent crowns (recovery species) increase shade and therefore modify microclimatic conditions (Rodrigues et al. 2009).They can reduce the number of non-native grasses (Cole et al. 2011), allowing native species to colonize.However, this trait can come with a cost, as fast-growing species might have lower wood density and can be more sensitive to pathogen attacks, which can increase their mortality risk (Zhu et al. 2018).Implying that the cost-effectiveness in terms of survival of fast-growing species can be higher compared to slow-growing species.The treatments added did not impact species resprouting ability.This characteristic is a property of the species that can increase its cost-effectiveness (Pausas & Keeley 2014).These results emphasize the importance of planting species with different performances like in secondary succession (Pickett et al. 2009), and the relevance of evaluating trade-offs among survival and final height to estimate the most cost-effective way to restore an area.Other measurements of success such as carbon sequestration, plant reproduction, natural regeneration of native species, and additional ecosystem services can be folded into costeffectiveness decision tools to support the sustainability of long-term, large-scale restoration projects (De Oliveira et al. 2021).
Irrigation is the most expensive and challenging management strategy to implement in TDFs (Cabin et al. 2002).Most restoration projects transport water from places far away from the intervention areas, increasing the labor and transportation costs of the projects.Even though this strategy can increase seedling survival (Dimson & Gillespie 2020), the costs of irrigation might exceed the benefits provided.Therefore, if restoration projects want to lower costs associated with irrigation, they should try to plant at the beginning of the rainy season to reduce the need for supplemental irrigation during seedling establishment (Cabin et al. 2000).However, when logistical and environmental factors do not allow for planting during the rainy season, we recommend the use of hydrogel or other management strategies that help increase the exposure of root systems' to water (Fajardo et al. 2013;Werden et al. 2018).Testing how hydrogel influences species establishment and survival, and the cost of this management strategy is key to establish more cost-effective restoration projects in TDFs.
From a cost-effectiveness perspective, fertilization and irrigation provide little benefits in terms of plant survival after 2 years.These two management strategies can also cost 1.6-5 more per hectare than not implementing any extra management practice.This information provides guidance for scaling-up and reducing the costs of restoration projects in TDFs (Vieira & Scariot 2006;Bonilla-moheno & Holl 2010;Fajardo et al. 2013).Even though we tested different management practices in a nucleation experiment, we see the potential for incorporating other strategies, e.g.variation in weeding intensity, multiple hydrogel doses, etc., that have been successful in other TDF restoration projects (Fajardo et al. 2013).For example, using direct seeding to introduce late-successional or large-seeded species once there is an established canopy could be a cost-effective way to accelerate the restoration process (Bonilla-moheno & Holl 2010;Mangueira et al. 2019) and to reduce restoration costs (Sampaio et al. 2007).We encourage practitioners and researchers to keep collecting the cost of every management practice implemented and to use this information to calculate the effectiveness of these practices in terms of multiple measures of success (Bodin et al. 2022).
Monitoring and recording the costs of nucleation projects can be critical to determine if this strategy is a cost-effective way to restore TDFs (Corbin & Holl 2012).Monitoring costs in particular are poorly documented and tend to be estimated only in the initial stages of a project (Drayton & Primack 2012;Brancalion et al. 2019).In most cases, the available costs do not reflect the true cost of a project, which according to our results could be particularly high and time-consuming to collect for large-scale projects.Thus, long-term monitoring and standardize protocols could help compare results across projects and design less costly restoration programs and monitoring schemes that could balance the need to frequently sample large numbers of individuals, with the need to sample for extended periods of time (Zahawi et al. 2014).The development of new technologies and social collaborative monitoring will open new avenues for scaling up monitoring activities and making them more affordable (Evans et al. 2018;De Almeida et al. 2020).Ensuring effective monitoring will secure the success of next decade's restoration targets (Brancalion & Holl 2020).
The main goal of restoration projects is to catalyze forest regeneration.Even though practitioners deal with a lot of uncertainty, there are multiple ways to improve restoration outcomes.We suggest paying close attention to species establishment requirements, e.g.soil fertility, light availability, actively weeding during the first year of the experiment and fencing the restoration area to ensure herbivore enclosure.Additionally, evaluating trade-offs among seedling performance, e.g.survival, height, and resprouting ability (Bond & Midgley 2001;Fajardo et al. 2013;Werden et al. 2018), and other species characteristics like seed dispersal and pollination syndromes when selecting species mixes may further improve restoration outcomes.For example, O. pyramidale, a fast-growing species that had the second lowest survival rates among the species planted, appears to decrease the presence of non-native grasses by shading them.This species also attracts several birds, bats, and non-volant mammal dispersers, which contribute to the recruitment of seedlings (Mora et al. 1999).Finally, we note the importance of considering physiological traits, e.g.drought tolerance and resistance (Werden et al. 2018;Álvarez-Cansino et al. 2022), when selecting species to restore areas that experience severe droughts.This information should be the product of partnerships between restoration practitioners and researchers.Based on our results, restoration projects should spend less resources on irrigation and fertilization and more on species selection, weed removal, and community building relations.
[Nees] Kunth, Andropogon bicornis L., Diectomis fastigiate [Sw.]P.Beauv, and Rhynchospora nervosa [Vahl]) and non-native (Hyparrhenia rufa [Nees], and Urochloa eminii [Mez] Davidse) grasses and sedges which makes these lands particularly hard to restore (Torres-Rodríguez et al. 2019).Thus, for the last 8 years Fundaci on Natura, a Colombian nonprofit organization has implemented and overseen the restoration of El Quimbo.Fundaci on Natura has established several restoration initiatives that include applied nucleation, enrichment planting of late successional species, and natural regeneration in more than 500 ha of abandoned fields around El Quimbo (Villota-Ojeda et al. 2019).

Figure 1 .
Figure 1.Aerial image of plot established in 2019 in a nucleation experiment in El Quimbo, Colombia.Each plot is a nucleus with 271 seedlings from 11 species that received one out of six treatments (50 g of NPK plus irrigation, 50 g of NPK without additional irrigation, 25 g of NPK plus irrigation, 43 g of P plus irrigation, irrigation, and no irrigation).Each small dot represents a seedling, the distance between seedlings is 1.5 m.The orange lines represent the limits of the plot (35 m Â 35 m).

Figure 2 .
Figure 2. Seedling % survival per treatment through time in a nucleation project in El Quimbo, Colombia.The different line colors represent the treatments applied to the seedlings, which were fitted with mixed effects logistic regression models.

Table 1 .
Treatments applied to 11,382 seedlings planted in 7 ha of grassland between 2019 and 2020 in a nucleation experiment in El Quimbo, Colombia.Grams are the amount of fertilizer a seedling received every time that it was fertilized.

Table 2 .
Costs associated with the establishment, management, and monitoring of a hectare (six nuclei) of a nucleation experiment in El Quimbo, Colombia after 2 years.Costs are in U.S. dollars (average rate 2020-2022 = COL$3,311).Data provided by Fundacion Natura (2018).