Journal list menu
COSST: A tool to facilitate seed provenancing for climate-smart ecosystem restoration
Abstract
en
- Selecting the best seed sources is a key step in ecological restoration planning especially under climate change. Seed-provenancing strategies include composite, aiming to reproduce natural gene flow; predictive, focusing on future climate adaptation; and climate-adjusted, a combination of composite and predictive. Yet, implementing different seed-provenancing principles remains a challenge.
- To fill this methodological gap, we developed the Climate-Oriented Seed-Sourcing Tool (COSST), a tool built in R capable of suggesting priority areas for seed sourcing according to composite, predictive, or climate-adjusted strategies, as well as the restoration site and focal species.
- The tool derives its inputs from species distribution models, which require occurrence and climate data only. COSST accommodates multiple climatic variables, weights the variables according to species-specific sensitivities, and accounts for uncertainties between climate forecasts.
- We demonstrated the flexibility of COSST using Caryocar brasiliense (pequi), a tree native to the Brazilian Cerrado, as a case study. The tool identified optimal areas for collecting C. brasiliense seeds and estimated the proportion of seeds to be sourced from various suppliers. We made available an R code for running COSST along with a Shiny application for data visualization.
- Synthesis and applications. Our tool can guide where to source seeds for species lacking range-wide information on genetic structure, which is the case for a substantial proportion of the tropical flora, where ecosystem restoration is of paramount importance.
Resumo
pt
- Selecionar as melhores matrizes de sementes é uma etapa fundamental no planejamento da restauração ecológica, especialmente frente às mudanças climáticas. As estratégias de proveniência de sementes incluem: composta, visando reproduzir o fluxo gênico natural; preditiva, focada na adaptação ao clima futuro; e ajustada ao clima, que combina as abordagens composta e preditiva. No entanto, implementar diferentes princípios de proveniência de sementes continua sendo um desafio.
- Para preencher essa lacuna metodológica, desenvolvemos a Ferramenta de Fornecimento de Sementes Orientada pelo Clima (COSST, na sigla em inglês), uma ferramenta construída no R capaz de sugerir áreas prioritárias para coleta de sementes de acordo com estratégias composta, preditiva ou ajustada ao clima, além de considerar o local de restauração e as espécies-alvo.
- A ferramenta utiliza como dado de entrada Modelos de Distribuição de Espécies (SDMs, na sigla em inglês), que requerem apenas dados de ocorrência e climáticos. O COSST acomoda múltiplas variáveis climáticas, atribui pesos às variáveis conforme as sensibilidades específicas de cada espécie e considera as incertezas entre diferentes projeções climáticas.
- Demonstramos a flexibilidade do COSST usando a espécie Caryocar brasiliense (pequi), uma árvore nativa do Cerrado brasileiro, como estudo de caso. A ferramenta identificou áreas ideais para a coleta de sementes de C. brasiliense e estimou a proporção de sementes a ser obtida de diferentes fornecedores. Disponibilizamos o código em R para rodar o COSST, juntamente com uma aplicação Shiny para visualização dos dados.
- Síntese e aplicações. Nossa ferramenta pode orientar onde obter sementes para espécies que carecem de informações abrangentes sobre sua estrutura genética, uma realidade para uma proporção substancial da flora tropical, onde a restauração de ecossistemas é de importância crucial.
Resumen
es
- La selección de las mejores fuentes semilleras es un paso clave en la planificación de la restauración ecológica, especialmente bajo el cambio climático. Las estrategias de procedencia de semillas incluyen la procedencia compuesta, que busca reproducir el flujo genético natural; la predictiva, enfocada en la adaptación al clima futuro; y la ajustada al clima, que combina las estrategias compuesta y predictiva. Sin embargo, implementar diferentes principios de procedencia de semillas sigue siendo un desafío.
- Para llenar este vacío metodológico, desarrollamos la Herramienta para la Selección de Semillas Orientada al Clima (COSST, por sus siglas en inglés), una herramienta desarrollada en R capaz de sugerir áreas prioritarias para la selección de semillas según las estrategias compuesta, predictiva o ajustada al clima, así como el sitio de restauración y las especies objetivo.
- La herramienta utiliza como insumo Modelos de Distribución de Especies (SDMs, por sus siglas en inglés), que solo requieren datos de ocurrencia y clima. COSST acomoda múltiples variables climáticas, pondera estas variables según las sensibilidades específicas de las especies y considera las incertidumbres entre pronósticos climáticos.
- Demostramos la flexibilidad de COSST utilizando a Caryocar brasiliense (pequi), un árbol nativo del Cerrado brasileño, como caso de estudio. La herramienta identificó áreas óptimas para la recolección de semillas de C. brasiliense y estimó la proporción de semillas a obtener de diversos proveedores. También proporcionamos un código en R para ejecutar COSST junto con una aplicación Shiny para la visualización de datos.
- Síntesis y aplicaciones. Nuestra herramienta puede orientar sobre dónde obtener semillas para especies que carecen de información genética a nivel de distribución, lo cual es el caso para una proporción sustancial de la flora tropical, donde la restauración de ecosistemas es de vital importancia.
1 INTRODUCTION
Ecosystem restoration is central to reducing and reversing biodiversity loss and the erosion of ecosystem services (IPBES, 2018; Leclère et al., 2020). However, positive biodiversity outcomes require restoration projects to be successful over long periods in the face of continued climate change (Prober et al., 2019; Zabin et al., 2022). Extreme weather events can push a restoration site back to a degraded state (Suding et al., 2004), making climate change a key challenge to restoring ecosystems worldwide (Frietsch et al., 2023). Seeds are the main basis for active restoration on land and they carry part of the genetic pool of the population they were sourced from. Practitioners can take advantage of natural genetic variability to select seed genotypes more resilient to future climates and climate extremes (Broadhurst et al., 2008; Hancock & Hughes, 2014; Havens et al., 2015). The origin of seeds is known to affect seed germination rates (Lortie & Hierro, 2022), as well as survival (Gross et al., 2017), growth (Gellie et al., 2016), and the phenology of adult plants (Bucharova et al., 2022; Pizza et al., 2023; Rushing et al., 2021). Therefore, seed-provenancing decision-making has the potential to climate-proof restoration projects (Vitt et al., 2022).
Seed-provenancing guidelines have been debated in the ecological restoration community (Dupré la Tour et al., 2020). Prioritizing seeds from the single geographically closest population (i.e., local provenancing) is a longstanding principle based on the assumption that local genotypes are adapted to the restoration site's local conditions. The local provenancing concept is often subjective, as nativity is a gradient rather than a discrete unit (Dupré la Tour et al., 2020). Therefore, arbitrary buffers around the restoration site representing the ‘local population’ may constrain seed supply capacity to conform with local provenancing principles (Gibson-Roy et al., 2021) or even lead to overharvesting (Broadhurst et al., 2008). Furthermore, the strict use of local seeds may come at the cost of inbreeding depression due to deleterious allele proliferation and loss of genetic variation (McKay et al., 2005). Finally, local seeds might instead show maladaptation as the climate is changing and may differ from the conditions the genotypes evolved in (Wilczek et al., 2014), jeopardizing long-term restoration success.
Other strategies have been proposed as an alternative to local provenancing (Breed et al., 2018). Composite provenancing addresses the genetic diversity issue by allowing the contribution of several populations to the seed mix (Aitken & Whitlock, 2013; Breed et al., 2018). In this strategy, the contribution of donor sites decreases with the geographical distance to the restoration site, mimicking natural genetic flow (Havens et al., 2015). Predictive provenancing addresses the maladaptation issue by favouring seed collection in populations theoretically adapted to the future climate at the restoration site (Broadhurst et al., 2008; Havens et al., 2015). Yet, the predictive approach has been criticized due to the risk of outbreeding depression (Bucharova et al., 2019). To reduce this risk, climate-adjusted provenancing was developed aiming to mix local seeds with non-local seeds from populations that match the predicted climate. Climate-adjusted provenancing is a combination of composite and predictive strategies, maximizing climate adaptiveness and genetic variation while minimizing genetic risks (Prober et al., 2015).
It remains a challenge to implement climate-oriented seed-provenancing strategies, such as climate-adjusted and predictive provenancing. Conventionally, seed transfer zones (STZ) have been used to support provenancing decision-making (Durka et al., 2017; Jørgensen et al., 2016). STZ can be defined as areas designed to limit genetic contamination from seed exchange and progress has been made in accounting for climate change when designing such zones (Fremout et al., 2021; Marinoni et al., 2021). Another approach is mapping contemporary climates that are analogous to the predicted future climate at the restoration site (Shryock et al., 2018). However, both STZ and climate match approaches often weigh climatic variables evenly (e.g., annual rainfall, temperature; but see Shryock et al. (2018)), which are known to affect each species differently (Harrison, 2021; Harrison et al., 2017; St. Clair et al., 2022). Furthermore, climatic forecasts vary considerably between Global Circulation Models (GCMs), which generates uncertainty in designing climate-smart seed mixes. Effective restoration planning requires, therefore, a novel seed-provenancing approach which encompasses species-specific sensitivities to different climatic variables, controls for climatic forecast uncertainties across the space, and is practical to implement.
Here we introduce the Climate-Oriented Seed Sourcing Tool (COSST), a tool built in R designed to operationalize seed-provenancing strategies for ecosystem restoration (https://github.com/silva-mc/COSST). The tool is based on Species Distribution Models (SDMs) and provides seed-provenancing guidance in the absence of genetic and experimental data. COSST identifies priority areas (ranging from 0 to 1) for sourcing seeds across the species range to restore a site specified by the user. When collection sites of commercial species are known, COSST can estimate the percentage of seeds to be purchased from different vendors. The tool allows the user to generate predictions based on three seed-provenancing strategies alternative to local provenancing: composite (not climate-oriented), predictive (fully climate-oriented), or climate-adjusted (balance between the previous ones). COSST weights climatic variables by their relative importance derived from SDMs and controls for the uncertainty in climate projections in the case of climate-adjusted provenancing. First, we describe the mathematical basis of the tool. Then, we demonstrate its applicability in the Brazilian Cerrado, a tropical global biodiversity hotspot. Specifically, we applied the tool to two actual restoration sites (~650 km apart), for single and multiple restoration-priority species (N = 3), and under the three focal provenancing strategies.
2 MATERIALS AND METHODS
2.1 The Climate-Oriented Seed-Sourcing Tool (COSST)
The COSST framework (Figure 1) generates a raster layer where the cell values (i.e., COSST priority index) correspond to the priority of the pixel as a seed source given the species of interest and target restoration site. The restoration site is defined as the approximate centroid of the location to be restored (coordinates). Using gridded climatic data (raster), COSST produces three raster layers, one for each seed-provenancing strategy (composite, climate-adjusted, and predictive, see introduction for definitions). When seed-sourcing site coordinates are known, the user can obtain the fraction of seeds to be purchased from each site by dividing the COSST index of each site by the total sum.

2.1.1 Input data
The first input to COSST is SDMs. Presence-only SDM algorithms, such as MaxEnt, require only species occurrence and bioclimatic data. Testing for multicollinearity and retaining only independent bioclimatic variables is essential to avoid overfitting SDMs. COSST uses two SDM outputs: the species range map (R) and the relative importance of the bioclimatic variables (v) to predict R. R is the binary projection of the SDM, representing the range of the species inferred from its climatic requirements. R sets the COSST spatial extent by restricting it to the species' range, but it is not an essential input (hence, its absence in Figure 1). In the case of MaxEnt, v corresponds to the permutation importance of the bioclimatic variables. COSST also requires the same baseline bioclimatic data (B) used to run the SDMs (e.g., 1981–2010) and the same data for a future timeframe (F, e.g., 2011–2040). The last input to COSST is the restoration site coordinates (s).
2.1.2 Composite provenancing strategy
2.1.3 Predictive provenancing strategy
2.1.4 Climate-adjusted provenancing strategy
2.2 Case study
2.2.1 The Brazilian Cerrado
We applied COSST to the Brazilian Cerrado, a region that covers one-quarter of Brazil's territory. Tropical savannas and grasslands are the dominant biome in the Cerrado region, representing 78% of the vegetation cover before large-scale human occupation (Rodrigues et al., 2022). About 12,000 flowering plant species are native to the Cerrado and 40% of this flora is endemic (B.F.G. Brazilian Flora Group, 2015). However, half of the Cerrado native vegetation has been lost to cattle ranching and intensive agriculture (MapBiomas, 2023). The combination of high endemism levels and rapid land-use change has made the Brazilian Cerrado a global ‘hotspot’ for biodiversity conservation (Myers et al., 2000) and ecological restoration (Strassburg et al., 2020). Brazil's ambition is to restore 2.1 Mha of Cerrado vegetation by 2030 (MMA, 2017). Native seed suppliers, cooperatives led by Indigenous peoples and local communities that harvest, process, and sell seeds of native species (Schmidt et al., 2019), play a major role in Brazil's ecosystem restoration strategy (Urzedo et al., 2020). Therefore, providing practical guidelines on seed mix design, especially seed-provenancing, will be key to achieving national restoration pledges.
2.2.2 Applying COSST to realistic scenarios
We showed two applications of the tool: seed source prioritization—aiming to map seed-sourcing priority areas to restore a particular site, and seed mix design—aiming to estimate seed demand from multiple suppliers. In both applications, we explored the outcomes of different seed-provenancing strategies (composite, climate-adjusted, and predictive) for two restoration sites using one (single species) and three species (multi-species). The two restoration sites are 653 km apart. The first is a mining site in Niquelândia (State of Goiás; 14°21′0.3168′′ S 48°24′0.0468′′ W, 1084 m a.s.l.). Mining activities in the region started approximately in 1994 and the soil remains exposed (MapBiomas, 2023). The second is an abandoned Eucalyptus plantation in Montezuma (State of Minas Gerais; 15°20′10.8852′′ S 42°24′34.6104′′ W, 1105 m a.s.l.). Eucalyptus sp. trees were planted approximately in 1997 and the plantation was abandoned in 2012 (MapBiomas, 2023).
We focused on the Caryocar brasiliense Cambess. (pequi) for single-species applications due to their ecological and socioeconomic value. C. brasiliense is a tree widespread in the Cerrado savannas and its fruit pulp and nuts are consumed across Brazil, providing income to local communities. In addition to C. brasiliense, we included Hymenaea stigonocarpa Mart. ex Hayne (jatobá-do-cerrado) and Qualea grandiflora Mart (pau-terra-grande) in the multi-species applications. These two species are widespread across Cerrado savannas (Bridgewater et al., 2004) and commonly traded by major Cerrado seed suppliers: Restauradores da RDS Nascentes Geraizeiras (RDS), Rede de Sementes do Cerrado (RSC), Rede de Sementes do Xingu (RSX), and VerdeNovo (VN; see Silva et al., 2022). Both H. stigonocarpa and Q. grandiflora have uses, including timber, medicinal, and ornamental value (Ribeiro et al., 2023). The precise polygon delimiting the seed collection areas of each species was not available, so we considered the seed-sourcing sites as the centroid of the municipalities where the seed suppliers operate and assumed that all species are collected across all sites (Silva et al., 2022). Although this was an approximation to illustrate the tool, users can provide the coordinate of the seed-sourcing area centroid to increase COSST accuracy.
2.2.3 Data processing and presentation
We used the MaxEnt algorithm to fit SDMs (Elith et al., 2011; Phillips et al., 2017; Phillips & Dudík, 2008). Refer to Silva et al. (2024) and Appendix S1 for the analytical pipeline and model specifications. We presented the seed-source prioritization application by showing the COSST index, referred to as ‘seed sourcing priority areas’. For the multi-species analysis, we binarized the COSST index per species using an arbitrary threshold of 0.75 (representing the third quartile), summing up the binary layers, and excluding pixels equal to zero. The final map shows areas that are high priority for seed sourcing across multiple species. As an additional analysis, we ran a Pearson correlation to test the association between the COSST index calculated under predictive and composite provenancing strategies. We presented the seed mix design application by extracting the COSST index at the seed-sourcing sites (i.e., approximate coordinate) and converting it into percentages by dividing it by the total. We also summed the extracted COSST index across all the sourcing sites of a given supplier to estimate the theoretical contribution of each vendor to the seed mix. Contribution is defined as the percentage of seeds originating from each supplier in the seed mix. It is worth noting that summing the COSST index may place greater emphasis on suppliers with multiple sourcing sites and using an average instead of a sum could be an alternative. All analyses were made using the R environment (v.4.2.3). The R code is available at https://github.com/silva-mc/COSST and requires Java to run.
3 RESULTS
3.1 Mapping seed-sourcing priority areas
COSST was able to generate seed-sourcing priority maps tailored to the seed provenancing and restoration site chosen by the user (Figure 2). Under predictive provenancing, the spatial distribution of COSST indexes around each restoration site differed considerably. For example, for site 1 potential seed-sourcing sites to the north of the restoration site had lower suitability (green shades) because their baseline climate does not match the restoration site's future climates. In contrast, for site 2, all nearby pixels (with savanna cover) showed high-priority values. Still considering C. brasiliense, there was a positive correlation between the COSST index calculated under composite and predictive provenancing, but the correlation coefficient was higher at site 2 (r = 0.61, p < 0.001), relative to site 1 (r = 0.41, p < 0.001; Figure S1). The COSST prioritization applied to multiple species (H. stigonocarpa, Q. grandiflora) generated similar results to the prioritization based on C. brasiliense (Figure 3; Figures S2 and S3).


3.2 Designing seed mixes with multiple suppliers
COSST was capable of mapping seed-sourcing priorities across seed suppliers, following different seed-provenancing strategies, and at different restoration sites. Considering a single species (C. brasiliense), COSST suggested RSC be the main seed supplier for site 1 (total contribution of 31%–33%) and RDS for site 2 (34.1%–36.5%) regardless of the provenancing strategy chosen (composite, climate-adjusted, or predictive; Figure 4). The greatest difference related to the provenancing optimization was in the contribution of RDS to source seeds at site 1 under composite (22.3%) versus predictive strategy (24.5%). Considering the sum of each site and between the three seed-provenancing strategies, the contribution of individual sourcing sites to the seed mix varied from ca. 4.1%–8.4% for site 1 and from ca. 3.4%–7.4% for site 2. However, it should be noted that 13 out of 18 sites sourced by RSX and one out of six sites sourced by RSC are located outside the distribution range of C. brasiliense (open points in Figure 4). Considering multiple species, RSC remained the principal vendor at site 1 and RDS at site 2 across all seed-provenancing strategies (Figure 5).


4 DISCUSSION
We present the COSST a new resource to support seed-provenancing decision-making for ecosystem restoration. When applied to the Brazilian Cerrado, the tool showed high sensitivity to the restoration location since the priority index and seed contribution per supplier were consistently higher in the restoration site surroundings. Provenancing and species choices provide a secondary refinement to the tool predictions. Below we discuss the advances, assumptions, and challenges to implement COSST in restoration projects.
Compared with previous tools, COSST's novelty lies in adjusting the prioritization index to the chosen provenancing strategy and focal species. The Diversity for Restoration (D4R) tool implements climate-adjusted provenancing through a dynamic STZ approach where 50% of the seed mix comes from the current seed zone and 50% from the projected future seed zones (Fremout et al., 2022). Similarly, the Climate Distance Mapper also uses a seed zone framework to implement predictive principles by estimating the match between present and future climates (Shryock et al., 2018). Our tool, on the other hand, allows the user to choose between a spectrum of strategies ranging from composite (geographically optimized) to climate-adjusted (intermediate) and predictive (climatically optimized). It is also possible to precisely adjust the weighting of geographical versus climatic optimization in COSST calculations (see Appendix S2). Furthermore, STZ are useful for handling several species at once but may fall short if there is no congruent population genetic structure among species, which is the case for Amazonian trees (Coronado et al., 2019), or if the whole distribution of a narrow-range species falls within a single seed zone. COSST avoids this issue by focusing on species-specific climatic distances constrained by the species range rather than generic polygons, being more ecologically meaningful when handling one species at a time. In fact, the ability of our tool to tailor its calculations per species using SDM-derived weights also differentiates it from Restore and Renew (Rossetto et al., 2019), another provenancing tool implementing the same provenancing strategies as COSST. Additionally, our tool also penalizes sites where the future climate is uncertain under climate-adjusted provenancing, favouring the ‘local-is-best’ logic instead. Finally, COSST complements existing species selection tools (e.g., Coutinho et al., 2023; Laughlin et al., 2018) by providing guidance on the best seed sources after the species have been chosen given the restoration targets.
The main assumption of COSST is the prevalence of intraspecific adaptation to climate, especially when the user selects climate-adjusted or predictive strategies. Evidence of climate adaptation exists for the Cerrado flora, but only for a handful of species (Appendix S3; Figure S4). The climate component of the tool may lose power if the genotypes of a species are not in equilibrium with their baseline climate (Wilczek et al., 2014) or if they are adapted to soil conditions rather than climate. However, we argue that our tool remains applicable even when local adaptation assumptions are not met. Genetic diversity tends to increase with the geographical distance between the populations (Pfeilsticker et al., 2021). In the case of C. brasiliense, COSST suggests some level of seed contribution from suppliers farther apart regardless of the seed-provenancing strategy chosen. If the practitioner follows the tool suggestion, a small fraction of seeds from distant populations should amplify genetic variation, increase adaptability, and reduce the risk of inbreeding depression (Kremer et al., 2012; McKay et al., 2005). At the same time, there was a predominance of local seeds in the C. brasiliense simulated seed mixes (RSC in site 1 and RDS in site 2), which reduces the risk of outbreeding depression due to the dilution of adaptive genes (genetic swamping) or the disruption of interacting gene networks and ploidy levels (hybrid breakdown) (Frankham et al., 2011; Hufford & Mazer, 2003). Therefore, COSST augments genetic diversity regardless of the climate match optimization, the aspect of biological diversity most relevant for evolutionary rescue under climate change (Aitken & Whitlock, 2013).
Implementing the tool will depend on overcoming two challenges starting with improving seed traceability. Several countries use wild populations for seed production (Atkinson et al., 2021; Bosshard et al., 2021; Giacomini et al., 2023), but the locations of these populations are often unavailable. Some suppliers are moving towards making these data accessible, for example, the Seeds of Success programme (Barga et al., 2020; Haidet & Olwell, 2015) and the Native Seed Vendors map (https://appliedeco.org/nativeseednetwork/find-seed/) in North America. A georeferenced map of where seeds are being collected from is the first step for applying COSST at a large scale. Aligned to this map, vendors will need to tag and separate seed batches per locality (Pedrini & Dixon, 2020), which is a logistical challenge for large seed suppliers, such as RSX, since seeds are often combined into a single mix per species and seed storage facility (Urzedo et al., 2020). Finally, strengthening seed storage technology and infrastructure is the second step to scaling up the tool. COSST encourages some level of seed transport over long distances, making it critical to develop techniques to ensure the viability of the seeds from harvesting to sowing phases (De Vitis et al., 2020; Shaw et al., 2020).
Accounting for seed production limitations and transport costs could increase the applicability of the tool even further. Sites will differ in the volume of seeds that can be collected there due to differences in the size of the vegetation remnants, species abundance, and number of seed collectors (Pedrini et al., 2020). Moreover, seeds can be produced ex situ (e.g., native seed farms) (Gibson-Roy, 2023) or stored over time (De Vitis et al., 2020), further increasing the seed production potential of a site. At present, COSST assumes that all sourcing sites have an equal seed production capacity. If seed production capacity is made available, it is possible to convert the COSST index into the volume/mass of seeds per sourcing site using the maximum seed production capacity as a cap. Another important consideration concerns the additional costs to the restoration project by seed transport from multiple vendors (Schmidt et al., 2019). Composite, climate-adjusted, and predictive strategies assume practitioners will purchase some degree of seeds from vendors far from the restoration site contrary to local provenancing. Sourcing seeds from multiple rather than a single site is more expensive and less practical in the short term but it can pay off in the long term (Jalonen et al., 2018). A restoration cost model revealed that augmenting genetic diversity by sourcing seeds from several populations increased seed collection costs by 33% but reduced maintenance costs by 18% (e.g. replanting) (Nef et al., 2021). Since maintenance represented more than half of total restoration costs, genetically diverse seed mixes reduced restoration costs by 11% over time. Future work can include estimated seed transportation costs in COSST, alongside costs avoided by ensuring the genetic quality of the seed mixes.
5 CONCLUSION
COSST provides a novel and generalizable tool to apply seed-provenancing principles in restoration planning. The tool can be applied to any plant species, provided there are sufficient occurrence records available to fit SDMs. The tool is likely most relevant in the tropics, where a vast number of species with different range sizes and climatic sensitivities require tailored seed-provenancing guidelines. Here, we focused on wild population seed collection, but COSST can also inform priority areas to source seeds for ex situ seed or seedling production. The tool can support not only practitioners in seed-sourcing decision-making but also suppliers in identifying priority areas for establishing new seed-sourcing sites. By connecting theory and application, we hope our tool can help practitioners maximize ecosystem restoration success under a changing climate.
AUTHOR CONTRIBUTIONS
Mateus C. Silva, Lucy Rowland, and R. Toby Pennington conceived the ideas and designed the methodology; Mateus C. Silva collated and analysed the data and led the writing of the manuscript. Peter Moonlight and Rafael S. Oliveira contributed critically to the drafts. All authors gave final approval for publication.
ACKNOWLEDGEMENTS
We are grateful to the Araticum Alliance for the Cerrado Restoration for the valuable discussions that motivated us to develop the tool. We thank Eduardo Malta Campos Filho, Anabele Gomes, and Fabian Borghetti for the feedback on the tool's concept. MCS, RTP, and LR are grateful to the WWF-UK and Exeter Alumni for supporting MS doctorate studies (710015629). RSO received support from CNPq regarding the grants 309709/2020, 303988/2018-5, and 312270/2017-8 and a productivity scholarship. RSO, LR, and RTP received support from the joint NERC-FAPESP grant 19/07773-1 and NE/S000011/1. LR acknowledges NERC for the independent research fellowship NE/ N014022/1. RTP and PM received support from NERC Newton FAPESP grant NE/N01247X/1.
CONFLICT OF INTEREST STATEMENT
The authors confirm no conflict of interest to declare.
Open Research
DATA AVAILABILITY STATEMENT
No original data were used in the article and the original data are available on Silva et al. (2022). The latest version of the code can be found at https://github.com/silva-mc/COSST, while the original version can be accessed via https://doi.org/10.5281/zenodo.14265511 (Silva, 2024).