As countries worldwide rally to protect 30% of land by 2030, the question isn’t just how much land is protected, but how well it’s protected. Are these “protected” areas truly shielding ecosystems from human pressures, or are they just “paper parks” — protected in name only, without real enforcement or ecological impact?

While marine conservation efforts have benefited from assessing quality through tools like Marine Conservation Institute’s MPAtlas and ProtectedSeas’ Navigator, both of which are integrated into SkyTruth’s 30×30 Progress Tracker, measurement of terrestrial protected area effectiveness remains uneven. There are currently no comprehensive, global, cross-ecosystem indicators of the effectiveness of terrestrial protected areas. SkyTruth tackled this critical question head-on in a pilot project that developed a scalable framework to measure the effectiveness of terrestrial protected areas (PAs) over time.

What We Did

Using 40 years of satellite imagery and land cover data, we analyzed 10 protected areas in Brazil to test how well these landscapes resisted human-driven land use change. The results were revealing: some areas clearly limited habitat loss, while others continued to lose habitat rapidly even after designation as PAs. Inspired by peer-reviewed methods, we sought to design a globally scalable method for a transparent, data-driven foundation for governments, NGOs, and funders to use to identify what’s working, what’s not, and how to direct future efforts.

The early results from this pilot confirm a key point: it’s not enough to declare land as protected — we need to understand if protections help truly resist habitat loss.

Our team utilized remote sensing data products from NASA, USGS, the European Space Agency (ESA), and peer-reviewed literature:

Annual Landsat Composites (1984-2024)

The foundation of our analysis rested on four decades of Annual Landsat Composites spanning from 1984 to 2024. These composites incorporated comprehensive spectral bands and vegetation indices, including NDVI (Normalized Difference Vegetation Index, a measure of vegetation greenness) and NDMI (Normalized Difference Moisture Index, a measure of vegetation moisture). Each composite was carefully cloud-masked and processed to median per pixel per year, ensuring data quality and consistency across the temporal range. The 30-meter resolution of these composites provided the detailed landcover analysis necessary for assessing changes within protected areas, many of which are relatively small in scale.

Annual Land Cover Data

To complement the spectral analysis, we developed annual land cover maps using a type of AI called a Random Forest machine learning model, which was trained on the Landsat composites. This approach was enhanced by combining our model with established global datasets like WorldCover2020 and CroplandAgreement2020, creating a more robust classification system. Our methodology specifically addressed and overcame limitations of existing datasets like MODIS, which at 500-meter resolution could misclassify smaller protected areas — a critical concern when assessing the effectiveness of conservation efforts that often operate at fine spatial scales.

Land Cover Change Detection

Beyond static land cover classification, our analysis focused on identifying sudden shifts in land cover that indicated conversion from natural to anthropogenic cover types. This change detection approach expanded beyond traditional forest-only datasets to capture changes across diverse ecosystems, including grasslands, wetlands, and other critical habitats. By detecting these transitions, we could quantify not just what land cover existed at any given time, but precisely when and where human activities were altering natural landscapes, providing a dynamic view of conservation effectiveness over time.

Analytical Approach

To assess Protected Area effectiveness, we developed a dual-comparison framework, informed by peer-reviewed methods recently published in Nature Communications (Li et al., 2024). Our framework consisted of three primary approaches:

Temporal Analysis: Before vs. After Protection

We compared vegetation conditions in the regions ten years before and ten years after PA designation. We used autoregressive modeling, an AR(1) model, to evaluate expected steady states in vegetation greenness and moisture, and detected subtle shifts and identified stable conditions over time. This methodology adapts techniques we originally developed for detecting mining activity, demonstrating the flexibility of remote sensing for conservation applications.

Spatial Analysis: Inside vs. Outside Protection

We contrasted conditions inside PAs with surrounding buffer zones extending 10 kilometers outward. This “donut” approach helped us identify whether protection displaced threats to nearby regions, protected habitats better than nearby unprotected areas, or extended benefits to surrounding areas.

Anthropogenic Change Metrics

We calculated annual rates of human-driven land cover change and identified high-confidence shifts from natural to human land uses (forest to cropland, grassland to urban), while filtering out natural fluctuations to focus on human impacts.

Case Studies: Real-World Insights from Brazil

We selected 10 pilot sites across Brazil’s most valuable ecosystems, focusing on areas designated between 2008-2015 to ensure adequate observation periods (10 years pre and post designation). These sites represented various IUCN protection categories and national designations, including Indigenous areas, and faced different anthropogenic development patterns. Below are three examples which reflect the range of PA effectiveness we found across our pilot test cases:

Success Story: Baú Indigenous Area (designated 2008)

This indigenous area in the Amazon demonstrated highly effective protection. We found:

• Minimal internal impact: Less than 1 km² of anthropogenic land cover change both before and after designation
• Clear boundary effect: Steady vegetation health inside the PA, while the surrounding areas experienced significant agricultural conversion along the western boundary
• Stable ecosystem indicators: NDVI and NDMI steady states remained consistent, indicating maintained ecosystem health

Mixed Results: Pequizal do Naruvôtu (designated 2009)

This case highlighted the complexity of effectiveness assessment. There was significant pre-existing disturbance in the region – approximately 44 km² was already disturbed before designation; but the designation seemed to effectively halt the degradation, revealing near-zero anthropogenic change after protection establishment.

We noted both positive regional impact effects (extending to the 10-kilometer buffer zone) and restoration potential (areas showed re-greening of previously cleared fields). We also encountered instances of complexity in interpretation in this region; for example, vegetation greenness decreased while moisture increased, possibly indicating succession from cropland to natural vegetation.

Concerning Example: Reserva Extrativista Jaci-Paraná (designated 2011)

This extractive reserve revealed the challenges of certain protection categories. Our analysis revealed significant ongoing deforestation, with over 500 km² of habitat lost within the PA since designation. We also saw increasing anthropogenic pressure with annual rates of land cover change accelerating post-designation.

This reserve is classified as IUCN Category VI, which per IUCN guidelines, allows “low-level non-industrial use of natural resources compatible with nature conservation” such as traditional forestry and small-scale extraction, but this area showed industrial-scale deforestation incompatible with the category’s sustainable use principles.

Management questions: These discrepancies raise important questions about what should count toward 30×30 goals when extraction becomes unsustainable. Under Brazil’s SNUC system, extractive reserves are classified as “sustainable use” protected areas and are included in the country’s official protected area statistics reported to the World Database on Protected Areas (WDPA), contributing to Brazil’s current 30.6% terrestrial protection coverage. However, when these areas experience industrial-scale deforestation rather than the intended low-impact traditional extraction, it calls into question whether they should count toward global conservation targets.

Why This Matters

This pilot begins to fill a critical gap in global conservation by providing a transparent, replicable framework to evaluate protection effectiveness anywhere on Earth. The early results from this pilot confirm what many conservationists have long suspected: it’s not enough to declare land as protected — we need to understand if it’s working.

Some protected areas, like Brazil’s Baú Indigenous Territory, showed a strong buffer against deforestation, with stable ecological indicators and even positive effects spilling over into nearby lands. Others, such as the Jaci-Paraná Extractive Reserve, struggled to limit habitat loss despite contributing towards Brazil’s official 30×30 statistics. This variation reinforces the importance of quality over quantity. A global goal like 30×30 only delivers conservation benefits if the areas protected are effectively managed, adequately enforced, and situated in ecologically meaningful places. It’s not just about hitting a target on paper — it’s about making those protections real.

Key Insights

Quality Over Quantity: If many PAs function like ineffective “paper parks,” then reaching 30% coverage won’t achieve conservation goals. In our Brazil pilot, effectiveness varied sharply — even among PAs in the same biome and category, underscoring that focusing merely on the extent of PAs can mask critical performance gaps.

Protection Categories Matter: IUCN categories are not all equal. Some extractive reserves in our sample showed habitat loss rates on par with surrounding unprotected lands, suggesting that industrial activities can undermine the intent of “sustainable use” designations.

Landscape-Level Dynamics: Effective PAs don’t just protect their own boundaries — they can positively or negatively influence surrounding areas.

Temporal Context is Critical: Many areas were already disturbed before designation, highlighting the importance of considering historical context when evaluating current effectiveness.

Combining Multiple Evidence Streams Improves Accuracy: Combining independent data sources, including land cover maps and spectral change detection, produced cleaner signals in our pilot analyses than any single measure alone. This ensemble approach reduced noise, improved reliability, and is likely to become even more powerful as additional independent measures of landscape change are incorporated.

Looking Forward: Scaling and Integration

Based on stakeholder and expert feedback on our pilot results, we’re pursuing several exciting directions:

Technical Enhancements

• Testing additional data sources to create additional, independent indicators of PA effectiveness, including aboveground biomass, height, and cover data from GEDI, biodiversity intactness, and other measures of ecosystem structure, function, and quality. These will complement existing land cover and spectral change analyses to provide a more complete picture of conservation outcomes.
• Refine baseline (counterfactual) selection. We used a simple 10 km buffer in this pilot. We intend to explore more sophisticated ways to select counterfactuals for each PA.
• Distinguish disturbance types (e.g., agriculture, fire, etc). Landscape change can be driven by anthropogenic land clearance and degradation (e.g., agriculture) or natural disturbance, such as wind events and fires. This distinction matters because natural events do not necessarily indicate management failure, while human-driven clearances often do.

Integration with Global Goals

The work aligns directly with global conservation targets established under the Kunming-Montreal Global Biodiversity Framework adopted at COP15 in December 2022. Target 3 calls for protecting 30% of terrestrial and marine areas by 2030, and this framework provides the analytical backbone for SkyTruth’s 30×30 Progress Tracker terrestrial assessments. While the marine component already incorporates effectiveness metrics, the terrestrial side has lacked equivalent global frameworks for distinguishing between areas that are merely designated as protected versus those that demonstrably conserve ecosystems.

The methodology also creates synergies with Target 2’s goal of restoring 30% of degraded ecosystems by 2030. Many restoration efforts naturally occur within the 10-kilometer buffer zones the team was already analyzing, offering a potential “two-for-one” approach where the same framework could track both protection effectiveness and restoration progress simultaneously. Beyond individual targets, the framework enables landscape-scale connectivity analysis, revealing whether protected areas function as connected networks supporting ecosystem integrity or as isolated conservation islands — a distinction critical to achieving the broader vision of reversing biodiversity loss.

The Bigger Picture

As we race toward 2030, this work offers more than new numbers — it begins to provide a foundation for observing and measuring meaningful protection.

The stakes couldn’t be higher. If we’re serious about safeguarding 30% of the planet’s land by 2030, we need tools that distinguish between protection in name and protection in practice. This framework offers:

• Transparent accountability: Repeatable measures of protection effectiveness, not just coverage
• Evidence-based policy: Data-driven foundations for conservation decisions
• Global scalability: Consistent methodology applicable anywhere on Earth
• Real-time monitoring: Capacity for ongoing assessment as threats and protections evolve

We Want to Hear From You

This pilot is just the beginning. SkyTruth is actively seeking partnerships, feedback, and collaboration from anyone working on terrestrial conservation, protected area management, remote sensing, or 30×30 implementation. Early conversations have revealed strong interest in applications ranging from conservation planning based on threat assessments to adaptive management strategies, but we know there’s much more to explore.

The path forward requires continued collaboration between technologists, conservationists, policymakers, and communities. Help us ensure that “protected” truly means protected — not just a line on a map, but a driver of resilient ecosystems and thriving biodiversity. Reach out to share your perspective, point out what we might have missed, or explore how this framework could support your work.