Implementing land acquisitions for watershed services in the tropical Colombian Andes produces marginal progress for advancing area‐based conservation objectives

The biodiversity crisis necessitates a global implementation of effective, equitable, and feasible conservation strategies. Public land acquisitions (PLAs) for watershed protection can produce co‐benefits for basic human needs and biodiversity, but there are concerns that acquiring land is not scalable or may not protect threatened biodiversity. PLA programs are rare, allowing for limited opportunities to assess these concerns. We investigate a unique policy that has legally mandated PLAs in the Colombian Andes, a tropical biodiversity hotspot, since 1993. Using a novel dataset of >7000 PLAs, along with surveys and in‐person interviews, we study the scale, spatial allocation, and constraints on PLA implementation, and examine its potential conservation benefits. More than 300,000 ha were acquired through PLAs, but economic and institutional factors such as land cost, government income, and land tenure informality were main drivers in the degree of PLA implementation. Acquisitions were thus directed mostly toward cheaper, less threatened ecosystems.


INTRODUCTION
A variety of strategies and funding sources will be needed to realize ambitious area-based conservation targets, such as the vaunted goal to protect 30% of the Earth's surface by 2030 (Dinerstein et al., 2019). Meeting these targets implies considerable implementation costs, while conservation resources are often insufficient (McCarthy et al., This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2022 The Authors. Conservation Letters published by Wiley Periodicals LLC. 2012). Accordingly, many scientists and the 2018 Convention on Biological Diversity are calling on governments and conservation organizations to capitalize on other effective area-based conservation measures (OECMs), whereby management of areas that are not legally defined as protected areas (PAs), regardless of the objective, results in de facto sustained biodiversity protection (Convention on Biological Diversity, 2018; Geldmann et al., 2021; IUCN-WCPA Task Force on OECMs, 2019; Kroetz et al., 2014). Ecosystem protection can receive political and financial support when ecosystem function and benefits to people are closely linked, even in areas with high demands on natural resources. Protecting watersheds is an example of demand for water-related ecosystem services (ES) effectively driving investment in nature-based solutions because the beneficiaries (the downstream population) and the geographical areas requiring protection (upstream watersheds) are well-defined and localized. Governments, the private sector, and nongovernmental organizations are increasingly making investments in watershed services (IWS) with the intention of generating biodiversity co-benefits, including increased species or ecosystem representation within protected lands and increased area of protection coverage for threatened biodiversity (Forest Trends, 2014).
Tropical montane biodiversity can especially stand to benefit from IWS because it is both threatened and able to provide water-related ES through known mechanisms. Ecosystems such as páramos (neotropical alpine grassland ecosystems), cloud forests, and riparian forests contribute to water-related ES by increasing yields to watersheds through the capture of mist and fog, maintaining the hydrological properties of upland climates, holding and storing water, reducing surface runoff and soil erosion, and keeping areas free of polluting land uses, including mining, farming, or grazing (Buytaert et al., 2007;FAO, 2008;Ochoa-Tocachi et al., 2016).
But what are the potential increases in species and ecosystem representation and protection coverage from sustained local public spending on IWS? The literature on incentive-based watershed protection instruments in Latin America has predominantly focused on programs that establish temporary contracts with landowners, often under the label "payments for ecosystem services" (PES) (Bremer et al., 2016). An alternative instrument, public land acquisitions (PLAs), permanently transfers all property rights to the government in exchange for a one-time payment. The long-term nature of PLAs can provide the management continuity needed for degraded ecosystems to recover and for planners to account for IWS protection within PA networks. However, PLAs can require lengthy negotiations, high up-front costs, and formal documentation of land ownership. Mismatches between locally available funding and threat might further attenuate the potential for PLAs to contribute to biodiversity protection. Understanding the conditions in which diverse conservation instruments can be feasible, effective, and equitable is a key concern for environmental planners (Maxwell et al., 2020); yet due to the scarcity of PLA as IWS in the tropics, PLAs are not well-studied in this region.
In this paper, we examine a unique case of sustained government spending on PLAs as IWS. In 1993, the Colombian legislature mandated that all departments (analogous to states or provinces) and municipalities spend 1% of their annual income on acquiring land for water conservation through article 111 of law 99 (Fondo Acción, Fundepúblico & WCS, 2017). 1 The article stipulated that the biodiversity and ecosystems on these lands were to be conserved or restored (see ibid. for policy details). The instrument should have contributed funding equivalent to 33% of the budget for the national system of PAs (Rudas, 2010). Most of Colombia's population and economic productivity is concentrated in the Andes, producing the bulk of the public demand and financial resources for nature-based water solutions. Despite variable compliance (ibid.), the 1% mandate provides a rich empirical context to investigate the general potential of PLAs as OECMs. This case is also worthy of study on its own; Colombia is a megadiverse country that is frequently listed as a conservation priority, with its Andean ecosystems harboring especially high levels of endemic and threatened species (Brooks et al., 2006;Forero-Medina & Joppa, 2010). In the 27 years that Article 111 has been in place, this is the first spatially explicit analysis on the potential conservation benefits of these PLAs because spatial data had not previously been compiled at the national level.
We compile a new dataset to ask the following questions: (Q1) What is the scale at which PLAs have been implemented? (Q2) What additional contributions to ecosystem and species coverage and representation do PLAs provide within the existing PA network? (Q3) Which factors influence the number of PLAs, the total area acquired, and the amount spent on PLAs at the municipal level? We lay the groundwork for future assessments of PLA outcomes (e.g., causal effects on hydrological services, land cover, or landscape connectivity, which are outside the scope of this paper).

METHODS
We combine spatial analysis (Q1 and Q2), statistics (Q3), and survey analysis (Q3) to answer our research questions. Data collection began in 2017 with a survey of 31 departmental governments and 33 regional environmental agencies (CAR for their Spanish acronym) to assess PLA activity and data availability. We subsequently requested spatially explicit information on the extent and location, date of purchase, and cost of PLAs through letters, email, phone calls, or in person. Because data from departments and CAR did not always include all acquisitions made by municipal governments, we also requested data from a stratified sample of municipalities (of 1122) in 2018 (see Supporting Information). We received data from 24 CARs, 14 continental departments, and 107/219 contacted municipalities. We then consolidated and cleaned the tabular and spatial data to uncover the spatial extent (Q1) and distribution (Q2 and Q3) of PLAs.
To assess contributions to ecosystem and species representation (Q2), we ran categorical zonal statistics using our consolidated spatial data. We measured additional ecosystem coverage (area and percentage of its distribution) and representation relative to endangered status using the red list of Colombian Ecosystems (Etter et al., 2018). For 2252 species, we leveraged species distributions maps in Colombia refined by altitudinal range, including bird distributions (Quiñones, 2019;Vélez et al., 2021), all maps available in Biomodelos (Velásquez-Tibatá et al., 2018), and a complementary set of mammal distributions in the 2017 IUCN Red List of Threatened Species. Because PLAs often overlapped with existing national or regional PAs or indigenous reserves, we separately compute increases in protection for (1) all PLAs and (2) only PLAs outside of PAs and indigenous lands. See the Supporting Information for details on data processing.
To assess the factors affecting PLA implementation at the municipality level (Q3) quantitatively, we constructed statistical models that predicted three outcomes: the number of parcels acquired, number of hectares acquired, and total amount spent. As predictors, we considered nine variables that we expected to affect levels of acquisition activity (details in Table 1). The variables included in the statistical models were selected through stepwise selection (in both directions), optimizing for the lowest Aikake information criterion. We employed generalized linear models with a negative binomial error distribution for the number of parcels because these were count data with overdispersion and a high likelihood of clustering. We applied linear models for the number of hectares purchased and the amount spent, using the inverse hyperbolic sine transformation for these outcome variables because their relationships to the explanatory variables were nonlinear, and because they contained zeroes. We limited the statistical analysis to municipalities overlapping Andean ecoregions (Olson et al., 2001) and to PLAs occurring between 1993 and 2016 (ending at the start of data collection).
Uneven data reporting was a challenge in the statistical analysis: for some municipalities, we received data from all three sources (CAR, department, and municipality), whereas for others, data were available from only one or two. We implemented a range of alternative modeling strategies to examine the robustness of our results to different assumptions and data quality. For PLAs where cost information was missing, we imputed the purchase price using a spatialized cost map (Nolte et al., in review). See the Supporting Information for details on modeling and data handling methods.
We supplement our statistical analysis with a survey of government officials in charge of managing land acquisitions (managers) to further explore the importance of unquantified drivers, and to identify new drivers we had not considered (Q3). Complementing earlier surveys aimed at departments and CARs, in 2020 we targeted survey requests to municipalities. The surveys asked to what extent predefined factors presented an obstacle or a limitation to PLA implementation from municipal-level managers' perspectives.
It also included open-ended questions about any other factors managers thought were important. Predefined factors were chosen through previous knowledge of the study area and themes that arose in interviews. These informal interviews arose during in-person data requests from departments, CARs, and municipalities that sometimes resulted in conversations. The lead author discussed PLA implementation experiences, opportunities, and constraints with dozens of managers in four departments (many in Antioquia). Twenty-three CARs, six departments, and 55 of the 164 contacted municipalities completed surveys. See the Supporting Information for survey questions, details, interpretation methods, and sample requests.

Q1:
We received data for 7348 individual PLAs and were able to map 60.6% of them. Mapped PLA area totaled 219,252 ha ( Figure 1) and unmapped PLAs contributed an additional 107,505 ha (Q1). A total of 114 PLAs lacked information on spatial extent. Most of the mapped PLA area (215,352 ha) fell within Andean ecoregions (Olson et al., 2001). One third of the PLAs with spatial data (1487) and 53% of all the mapped acquired area overlapped with official PAs or indigenous lands. Most of this overlapping area (72.7% of mapped and overlapping PLAs) fell within PAs with lesser use restrictions, including Regional Integrated Management Districts, Regional Natural Parks, and Regional Forest Reserves (Table 2). Around half of PLAs within PAs occurred after PA creation (Table S1).
PLAs covered larger areas of less threatened ecosystems (in Vulnerable and Least Concern categories) than of Endangered or Critically Endangered ones (about triple or double the area, respectively), with some outliers (Q2) (Figure 2a). For example, PLAs covered 8.1% of an Endangered Andean páramo ecosystem when including PLAs within PAs, and 2.4% of the ecosystem outside of PAs TA B L E 1 Variables considered in statistical models aiming to explain public land acquisition (PLA) outcomes (number of hectares acquired, number of parcels, and amount spent) at the municipal level in the Colombian Andes  ( Figure 2b). PLAs tended to cover slightly higher percentages of Critically Endangered and Endangered species distributions (although these also tended to have smaller range sizes) (Figure 2d). PLAs outside of PAs covered over 1% of the species' distribution area in Colombia for six of the species we examined that did not already have 30% of their range protected ( Figure S4). Municipalities varied considerably in the number of PLAs, the number of acquired hectares, and the amount of money spent (Figure 3). Differential outcomes were mostly driven by economic factors (Q3) ( Table 3). Municipalities with higher incomes spent more on PLAs, purchased a higher number of PLAs, and acquired more PLA area. These findings were robust across a range of assumptions on how department-level funds had been allocated across municipalities. Municipalities with higher land costs had smaller areas of PLAs. Notably, the average water use index in a dry year and the average threat to ecosystems were not strongly associated with municipality-level metrics of PLA activity. All metrics were always positively associated with the number of sources that we received data from, but results were similar when excluding this variable, and results from the statistical models are largely consistent with survey responses.
The factors affecting implementation outcomes ultimately influenced the spatial distribution of the ecosystems that PLAs covered. The bulk of acquisitions happened on less expensive lands, which tended to also be less threatened in the Colombian Andes (Figure 2e). Even within critically endangered ecosystems, the less costly areas were disproportionally selected for purchase ( Figure S3). Our survey analysis revealed other key limitations for the implementation of PLAs (Q3). Respondents perceived land tenure informality to be the most limiting factor in carrying out acquisitions. Lack of short-term funding and long-term funding were the second and third most limiting factors, with institutional capacity following. Low institutional capacity can limit PLA implementation through (1) insufficient personnel to manage acquisitions, (2) inconsistent knowledge or interpretation of the law, (3) inadequate long-term data management of PLAs (this we learned through the data collection process), (4) insufficient systematic planning of areas to acquire, and (5) low incorporation of these goals into local development plans. Several municipal officials from different departments also indicated that political will was lacking in implementing PLAs despite the legal mandate. The amount of attention that was paid to PLAs as IWS varied by administrations, and some spikes in activity were attributed to individuals who placed a lot of emphasis on IWS.
When asked about actions that the municipality was taking to overcome these limitations, respondents from 51% of municipalities said that they were not taking any. Actions identified by remaining respondents include improving environmental zoning, updating management plans, and increasing cooperation with CARs or departments. Six municipalities were working to formalize land tenure within their jurisdictions.
Themes of controversies and opposing views about PLAs emerged through surveys and informal interviews with managers. Some managers worried that local people would not respect protections on government-owned lands. Because there is no permanent municipal-level legal protection status that can be given to PLAs, and because the titles of PLAs are sometimes not publicly registered as being for conservation purposes, it is possible for parcels to be repurposed once political administrations change. In addition, officials from some municipalities and CAR stated that they preferred to allocate the available public funds into temporary PES programs to avoid rural displacement or depopulation and negative social and economic outcomes for landowners and because they believed PES participants would be better land stewards than government agencies with limited enforcement capacity.

DISCUSSION
Twenty-seven years of PLA implementation have resulted in a publicly acquired area equivalent to the size of some of the country's national parks, with just the mapped area totaling to three times the size of the most recently created national park in the Andes (Figure 1). This is despite high development in the region and despite historical underfunding for PLAs; one study found that municipalities and departments only invested 0.12% of their incomes instead of the mandatory 1% (Rudas, 2010). Although PLAs likely fall short of original expectations, they are of sufficient size to merit consideration in the broader conservation portfolio. It is not immediately obvious why managers invested in acquiring lands that are already legally protected. Managers could be attempting to displace anthropogenic pressures from private landowners within PAs, which have already been identified as important areas for the protection of ecosystems and biodiversity. Additionally, landowners within PAs cannot usually sell their lands to nongovernmental buyers. PLAs can be used to provide compensation to landowners for their loss of land use rights after PA declaration and can help resolve (not always favorably for the landowners) conflicts between PA managers and landowners. Transferring ownership of parcels to the government within already-existing PAs could confer additional conservation value if anthropogenic pressures are lessened in government-owned lands compared to their inhabited counterparts.
In keeping with the finding that threatened ecosystems were more expensive to protect (Figure 2e), and that, according to our statistical models, the area of PLAs was most closely associated with higher incomes and lower land costs, PLAs covered more area in less threatened ecosystems (Figure 2a and Table 3). PLA outcomes were generally unrelated to our metrics for current ecosystem threat or water use needs at the municipality level (Table 3). For species that (without PLAs) had less than 30% of their range protected, PLAs mostly only nudged area-based protection by fractions of a percentage point ( Figure S4). We did not find that representation increased for ecosystems or for the included species.
Our statistical models explain less than half of the observed variation in outcomes across municipalities. In our municipal-level surveys, respondents indicated that land tenure informality was the most important limiting factor for implementing PLAs. While land tenure informality was not often selected or statistically significant  The model residuals were examined to ensure that the required assumptions were met for the model. We determined that spatial autocorrelation did not need to be accounted for in these models (see Supporting Information for details on methods for finding spatial autocorrelation). a The number of parcels is modeled using a negative binomial distribution with a log link function in a generalized linear model. b The number of hectares and the amount spent are inverse hyperbolic sine transformed and are modeled through a linear model. in the statistical models (although its relationship to outcomes was always negative), we suspect that our variable does not capture the aspects of land tenure informality that make purchasing parcels difficult, including a pervasive lack of consistent documentation between deeds and the cadaster and of official procedures documenting land transfers. In Colombia, where violent conflict has left many land titles contested, paperwork must be complete for the government to make a purchase. However, the time and expense required to obtain required paperwork considerably slow down the acquisition process. This coupled with limited staff to carry out PLAs is a significant rate limiter to investment in IWS. The lack of political will to implement PLAs for IWS in various locations challenges our assumption that public needs tied to ecosystem function will naturally result in more public investment in ecosystem protection. It could also indicate that PLAs are not necessarily a politically popular or expedient way to spend resources on IWS. However, increased oversight, environmental education, involvement of communities, and long-term partnerships between departments, CARs, and municipalities could help make progress on IWS a higher priority. Proposed solutions for reducing barriers to PLA implementation are detailed in a prior publication (Reboredo Segovia et al., 2021).
As this is the first nation-wide paper to spatially assess the distribution and extent of PLAs for IWS in Colombia and the factors that influence PLA implementation outcomes, many questions remain unanswered. Although the scale of PLA implementation is considerable, it is questionable whether PLAs provide true long-term benefits to species. In areas as climactically dynamic as the Andes, improving landscape connectivity and landscape conditions for biodiversity could be more important considerations (Ramirez-Villegas et al., 2014). Further study through counterfactual impact evaluation is needed to determine how effective PLAs are in improving land cover outcomes. This evaluation, along with species dispersal data, would permit quantification of the additional contributions of PLAs to PA network connectivity. Furthermore, although some municipalities and departments are working to mitigate limitations to implementing PLAs, some are switching toward investing the article 111 funds in PES. New questions emerge about the differences in implementation feasibility, effectiveness, and equitability of these two instruments.

CONCLUSIONS
Land acquisitions as IWS covered a considerable area, but PLAs mostly occurred in lower cost (and relatedly, less-threatened) ecosystems, with some exceptions. The percent of the distributions covered by PLAs for a sample of species showed mostly marginal increases, with some exceptions; yet as more accurate data are compiled for small-range species, this result could change. This research shows that PLAs as IWS could contribute to increasing protection for ecosystems, but they are constrained by important social and economic factors, including income, land costs, and land tenure informality. If governments and conservation organizations want to capitalize on OECMs such as PLAs for IWS, data on where these are implemented and why are critical for evaluating the role that they can play toward meeting conservation objectives. We leave to future research to determine whether PLAs positively impact vegetation coverage, to assess whether these result in more connected landscapes, and to compare the effectiveness of PLAs and PES in positively impacting vegetation cover.

A C K N O W L E D G M E N T S
We extend our deepest gratitude to all the municipal, departmental, and CAR officials who provided data. We thank Juliana Cortés, Camilo Correa, and Felipe Suárez at the Biological Resources Research Institute Alexander von Humboldt for sharing supporting data and comments, as well as Anne Short Gianotti, Claudia Diezmartinez, Alicia Zhang, Corrin Barnes, and Sara Cavallo for providing comments. This work was funded by Boston University and the National Aeronautics and Space Administration (grant #80NSSC20K1486).

C O N F L I C T O F I N T E R E S T
The authors declare no conflict of interest.