Effectiveness of Conservation Areas for Protecting Biodiversity and Ecosystem Services: A Multi-Criteria Approach - Notes

Introduction

Current environmental conservation strategies are shifting towards ecosystem services (ESs) approaches, moving beyond purely biodiversity and landscape-scale management.
This shift is evidenced by increased research and publications focusing on ESs, which develop tools, methods, and models and promote ESs as a basis for area prioritization and land planning.
International agreements like The Aichi Targets, assessment initiatives such as the Millennium Ecosystem Assessments (MA) and The Economics of Biodiversity (TEEB), and governmental platforms like the Intergovernmental Platform on Biodiversity and Ecosystem Services (IPBES) reflect this tendency.
Colombia, as a member of the Convention on Biological Diversity (CBD), is committed to the Aichi Targets, integrating ESs into new environmental policies, land planning, and decision-making tools.
The National Development Plan (DNP 2015) and the National Policy for Biodiversity and Ecosystem Services Management (MADS 2013) in Colombia explicitly integrate the ESs approach.
This incorporation aims to define the ‘environmental conservation and protection category,’ grouping protected areas, ecosystems of special interest, areas of high ESs provision, and key biodiversity indicators such as endemic, threatened, and migratory species.
Areas of high species richness are not always the most effective for conserving species with particular characteristics like endemic or migratory species, necessitating consideration of different key biodiversity indicators.
Examples of management strategies and decision support for ESs in land use planning are scarce compared to those addressing biodiversity or conventional landscape conservation.
Knowledge of ESs quantity, distribution, and valuation is limited in Latin America, including Colombia, due to weak governmental and local environmental authorities and a lack of expertise in applying available technical and methodological frameworks.
The ability of current conservation areas to protect both key biodiversity indicators and ecosystem services in Colombia has not been thoroughly evaluated.
Research questions whether traditional conservation areas simultaneously protect high-value biodiversity and ESs areas and identifies additional areas for consideration.
Studies show that protecting ESs within existing protected area networks is important but limited, indicating the need to develop prioritization frameworks specifically for this purpose.
Spatial correlations between different ESs and between ESs and biodiversity are not necessarily high, often involving trade-offs where protecting one service compromises others.
Spatial multi-criteria analysis (sMCA) is a commonly used methodology in land planning, defining suitable areas by overlaying information layers, quantifying criteria, and assigning weights to reflect relative importance.
Assigning criteria weights is a major source of uncertainty in multi-criteria analysis, influenced by stakeholder biases, hierarchies, and political agendas.
This study investigates the effectiveness of Colombia's protected areas in protecting biodiversity and ESs, identifying potential priority areas, and applying a multi-criteria approach considering uncertainties in criteria weighting to inform local land planning.

Methods

Study area

The study area is the Corporación Autónoma Regional de Cundinamarca (CAR) jurisdiction in central Colombia, located in the eastern mountain range (cordillera oriental) of the Andes.
Elevation ranges from 4000 m.a.s.l. in the páramos ecosystems to 150 m.a.s.l. in the Magdalena River basin.
The area includes a large flat terrain where Bogotá is located (2500 m.a.s.l.).
The CAR manages natural resources in rural areas and ecosystems surrounding Bogotá, part of the Magdalena river basin.
The total population in Bogotá and surrounding cities is about 12 million.
The study area covers 19,411 km^2, with 18% (3679 km^2) designated as conservation and protection areas (cCPAs), including national, regional, and local protected areas and key ecosystems like páramos and wetlands.
cCPAs are densely located in the eastern part of the study area around Bogotá and along the high plateau.
The western part of the study area, sloping down to the Magdalena River, has few areas with protection status.

Biodiversity and ecosystem services indicators

Indicators were selected and parameterized during a workshop with technical actors, based on the quality and availability of spatial information and CAR’s short-term environmental management priorities.
Five indicators were chosen to evaluate the degree of representativeness when ESs are considered in prioritization approaches:
- Sensitive species
- Ecological systems
- Habitat quality
- Scenic beauty
- Water provision

Sensitive species

'Sensitive species' encompass endemic, migratory, and endangered species.
A database was constructed from museum collections, publications, and online databases, with taxonomical nomenclature checked and corrected.
The indicator used was species richness (number of sensitive species in a 1 km^2 cell).
In total, 75 sensitive species were considered: 41 birds, 14 mammals, 14 amphibians, and 6 reptiles.
Species richness is a well-established criterion in Colombia’s environmental frameworks.
Distribution ranges were estimated using species distribution models (SDMs), combining species locations with environmental variables.
Species with fewer than five records were removed due to accuracy requirements.
Environmental variables included climatic (Hijmans et al. 2005) and terrain variables from a digital elevation model (DEM) (USGS 2008).
Variables with a Pearson correlation coefficient less than 0.7 were selected: elevation, annual temperature range, mean temperature of the driest quarter, annual precipitation, and precipitation of the driest month.
Support vector machines (SVMs) (Cortes & Vapnik 1995) were used to calculate species distribution ranges.
SVMs employ optimization algorithms to locate boundaries between classes by finding a hyperplane in the feature space.
Absence data were represented by randomly selecting locations outside the multivariate environmental similarity surface (MESS) of the presence data.
A 10-fold cross-validation was applied to measure model performance and calculate the ʋ parameter.
The kernel used was radial, implemented in R using the ‘e1071’ library (Meyer et al. 2014).
Species richness was calculated by summing the species probability values obtained for each cell, giving a probability-weighted sum of species present.

Ecological systems

Ecological systems (landscape units combining ecosystems and biogeographical units) were used to rank land based on four criteria:
- Representation (amount within protected areas)
- Rarity (national and local levels)
- Remanence (percentage remaining relative to historic distribution)
- Rate of loss (calculated over the previous six years)
This ranking, or offset ratio calculation (Saenz et al. 2013), serves as the prioritization value of each ecological system.
Values range from 10 (maximum) for ecological systems with very low representation, high rarity, low remanence, and a high rate of loss, to 5 for ‘natural’ ecological systems and 2.5 for ecological systems with secondary vegetation.
Ecological systems without natural or secondary vegetation are assigned a value of zero.

Habitat quality

Habitat quality represents a land cover’s ‘naturalness,’ with different land covers denoting habitats.
Naturalness indicates the availability of ecosystem services.
The indicator is calculated by relating an ‘integrity value’ to each land cover with factors threatening its persistence.
Integrity values range from 0 to 1, assigned by project members based on the land cover’s potential to serve as habitat.
Threats include roads, railways, mining, urban areas, settlements, and combinations of agricultural and pasture lands.
Each threat is related to each land cover by a value (0 to 1) describing its negative effect (Supplementary Material III).
The model was run 10 times with different maximum impact distances (1 km to 10 km).
Weights were assigned by discussion and consensus with project members.
The model used is available in the Integrated Valuation of Environmental Services and Trade-offs (InVEST) (Sharp et al. 2014).
CORINE level 2 categories of the land cover map (IDEAM 2010) were used to describe land cover types.
Official cartography provided by CAR was used to generate the threat layers.
Geoprocessing was done in R and Quantum GIS (QGIS).

Scenic beauty

Scenic beauty quantifies the quality of a place as a source of natural views, potentially inhibited by human disturbances.
A DEM for the study area was extracted from the Global Land Cover Facility (USGS 2008).
Human disturbance elements included roads, railways, urban and rural settlements, hydroelectric plants, electrical power lines, mining, and quarry areas.
A default distance of 8 km was used.
Data were integrated in the ‘Unobstructed Views: Scenic Quality Provision’ module of InVEST (Sharp et al. 2014).

Water provision

The water balance, calculated as rainfall plus fog inputs minus actual evapotranspiration, was used as the indicator.
The WaterWorld model (Mulligan 2013) was used, with details about parameterization and algorithms found in Mulligan and Burke (2005).
This model was chosen due to scarce field data and its reliance on globally available data.

Standardization

All indicators were rescaled to values between 0 and 10 by dividing original values by each indicator's maximum value and multiplying the result by 10.

Multi-criteria analysis

A sMCA (Malczewski 2006) was performed by overlaying the five indicators.
Weights (0 to 1) were assigned to each criterion (indicator).
Weights were calculated randomly from a uniform distribution and rescaled to sum to 1.
1000 simulations were run, including cases with equal weights and single-indicator preferences.
For each simulation, the weighted mean of all indicators was calculated.

Representativeness of cCPAs

Simulated Conservation and Protection Areas (sCPAs) were extracted from each simulation run by selecting grid cells with the highest values until reaching 3679 km^2.
Spatial congruence was calculated as the proportion of cCPAs area that spatially intersects with sCPAs.

Indicator correlation analysis

Spearman’s rank correlation coefficient was used to evaluate pair-wise correlations between indicators.

Complementary areas

A consensus and an uncertainty map were calculated based on simulations.
The consensus map was determined by calculating the median of the 1000 simulations.
The central 90% confidence interval was used as the uncertainty measure.
The median and confidence interval were calculated on a cell-by-cell basis.
Areas of high consensus and low uncertainty were deemed the best multifunctional areas.
A final ranking was established by plotting consensus and uncertainty values on a scatter plot.
Four zones were identified:
- Multifunctional areas: high consensus and low uncertainty
- Candidate areas: high consensus and high uncertainty
- Uncertain bad areas: low consensus and high uncertainty
- Worst case areas: low consensus and low uncertainty
Thresholds were defined by project members as the third quartile for consensus values and the second quartile for uncertainty values.
The four areas were mapped and overlain with current conservation areas to evaluate overlap and identify multifunctional areas without current protection status.

Results

Representativeness: overlap between current and simulated conservation and protected areas

The representativeness of cCPAs in protecting biodiversity and ESs ranges from low to intermediate (3% to 56% overlap between sCPAs and cCPAs depending on the weights assigned.
Priority areas for conservation vary significantly depending on stakeholder relevance weights.
Even under the most positive scenario (56% overlap) , it's only moderate.
When considering each criterion independently:
- Ecological systems: 52% overlap
- Water provision: 3% overlap
- Habitat quality: 48% overlap
- Scenic beauty and sensitive species: low overlap (30% and 20%, respectively)

Indicator correlation analysis

Distribution patterns of indicators show no evident geographical overlap.
Correlation coefficients are medium to low.
High sensitive species richness around Bogotá contrasts with low scores for other indicators.
High water provision areas have low scores for other indicators.
The highest positive correlation (r = 0.56, p < 0.001) exists between ecological systems and habitat quality, coinciding at high altitudes with páramos ecosystems.
79% of the study area has zero provision for ecological systems, corresponding to low habitat quality values.
The correlation between sensitive species distribution and other indicators is always negative, especially strong with water provision (r = −0.51, p < 0.001).

Complementary areas

Despite weak global correlations, some areas of high value for all indicators are evident in the consensus map, largely distributed over the páramos ecosystems.
The lowest consensus scores are concentrated on the high plateau with high urban settlement density.
26.3% of high consensus score areas are highly uncertain (candidate areas).
73.7% of high consensus score areas are effective multifunctional areas (3573 km^2).
Overlaying multifunctional areas with cCPAs shows that 47.8% of the latter are multifunctional areas.
The remaining cCPAs consist of:
- 8.0% candidate areas
- 28.3% uncertain bad areas
- 15.9% worst case areas
Of the multifunctional areas, 49.2% corresponds to cCPAs, and 50.8% (1815 km^2) are considered complementary areas.
Some complementary areas are isolated patches, while others are contiguous to cCPAs, such as in the Sumapaz area or next to the DMI Cuchilla de San Antonio.

Discussion

Representativeness of protected areas

The current network of protected areas only partly represents important areas for biodiversity and ESs protection.
cCPAs for the conservation and protection of biodiversity and ESs will not exceed 56%, and it could be as low as 3%.
Focusing on areas with high consensus score and low uncertainty would lead to multifunctional areas that overlap with cCPAs at 49%.
Outcomes are sensitive to the choice of weights, highlighting uncertainty in stakeholder preferences.
Through consensus analysis and integration of uncertainty, areas are recurrently selected for high biodiversity and ESs value.
There is a limited overlap between cCPAs and multifunctional areas, particularly in the east with high cCPAs density.
Although 18% of the study area has a protection status, there is a spatial bias towards the remaining fragments of natural vegetation in and around the high plateau area.
The indicator most congruent with cCPAs is ‘ecological systems’ because areas of natural vegetation have been the main criterion in selecting protected areas.
cCPAs that do not overlap with multifunctional areas are mostly located in the east of the study area and they represent heavily transformed areas, albeit located within the páramos or wetland ecosystems.
Water provision service in the CAR jurisdiction lies outside the current network of protected areas.
These results are comparable to studies of other biological groups, ESs and biogeographical zones.

Congruence between indicators

There is no evident geographical congruence between biodiversity and ESs, with correlations ranging from moderate to low (positive and negative).
The highest positive correlation between ecological systems and habitat quality (ρ = 0.56) highlights that the natural (non-disturbed) ecosystems in the study area are under high pressure from all the threats considered in this study.
The spatial co-occurrence between different ESs and between ESs and biodiversity is highly variable, and dependent on the region and scale of analysis.
The lack of spatial congruence found in our study highlights the danger of utilizing one or few indicators for delineating conservation areas, since protecting few will not secure protection for others.
The negative correlation between sensitive species and all other indicators is noteworthy. Generally, high richness values of sensitive species are distributed over the high plateau, that is, over highly disturbed areas.
Results show that the relation can become even negative at local scales.
The negative correlation between sites of high water provision and all other services is in line with other studies.

Complementary areas

Despite the limited congruence between indicators, multifunctional areas, that is, areas of high biodiversity and provision of ESs were identified.
Almost half of them intersect spatially with the current network of protected areas but the other half has no protection status and can thus be considered complementary.
These complementary areas act as buffer zones of key ecosystems and protected areas.
We advocate for more practical actions that are feasible to implement and would account for the local communities living in these territories, like voluntary conservation agreements or payments for ESs schemes.

Methodological approach

The multi-criteria approach is one of the most common modelling techniques used in Colombia and abroad for land planning studies.
We adapted the methodological workflow proposed by Ligmann- Zielinska and Jankowski (2014) that explicitly integrates spatial uncertainty with the consensus map (the ‘average suitability score’ map in their case).
The methodological approach used in this study and the results thereof can be considered a practical application of what the Colombian legislation labels as the Main Ecological Structure (MES) (Estructura Ecológica Principal) (Andrade et al. 2008).
Challenges remain in relation to data availability and quality and the selection of robust tools to model ecological processes.
All indicators in this study were assumed to accurately characterize the aspect they represent; uncertainties related to each of the models and their outputs were not explicitly considered.

Conclusion

Current conservation and protection areas in the CAR jurisdiction have a moderate degree of biodiversity and ESs representativeness.
Management actions and land planning strategies for these areas require their consideration and acknowledgement as complex social-ecological systems and the factors that would make these areas resilient in the long term (Cumming Forthcoming 2016), as well as the harmonized work of environmental authorities and local communities, for example, through the implementation of conservation agreements or payments for ESs schemes, and not necessarily through the establishment of new protected areas.
We encourage the selection of those indicators in a joint effort with local stakeholders.
Future research should concentrate on data collection and modelling of different ESs at the local scale.
Our simulations highlight the sensitivity of protected area representativeness to stakeholder preferences of indicators.
Participatory approaches should be considered in local land planning.