Local conditions and policy design determine whether ecological compensation can achieve No Net Loss goals

Industrial development places substantial pressure on ecosystems, prompting widespread adoption of ecological compensation policies aiming for No Net Loss (NNL) of biodiversity and related goals. Compensation follows the mitigation hierarchy: avoid, minimize, remediate, then offset residual impacts. Despite hundreds of policies globally, their contribution to conservation goals remains uncertain and failures are common. Policy design varies widely (e.g., Improvement via restoration vs Averted Loss via protection), and local conditions (e.g., restoration potential, background trends) strongly influence outcomes. Developing realistic counterfactuals is challenging because biodiversity trajectories differ among places. This study investigates how policy design and local conditions interact to influence compensation performance—how close policies come to NNL—by systematically testing 18 policy options across four case studies (Australia, Brazil, Indonesia, Mozambique), measuring impacts on biodiversity (native vegetation extent) and two ecosystem services (carbon storage, sediment retention) relative to counterfactual scenarios.

The paper highlights that compensation policies differ greatly in design, implementation, and scope, complicating comparisons. Two main approaches dominate: Improvement (restoring cleared land) and Averted Loss (protecting existing biodiversity to prevent future loss). Prior studies show mixed outcomes, including cases where offsets enable ongoing biodiversity loss and overlook ecosystem services. Multipliers used in practice are often arbitrary and not grounded in counterfactual losses or restoration uncertainty. Restoration success rates vary and can be low, especially when creating new habitat comparable to reference conditions. Policies often target biodiversity only, assuming co-benefits for ecosystem services, despite evidence of trade-offs. Baseline and counterfactual specification critically affect evaluations of NNL.

The authors developed a spatial simulation framework to assess 18 ecological compensation policy designs across four case studies: Brigalow Belt (Australia), Iron Quadrangle (Brazil), East Kalimantan (Indonesia), and Cabo Delgado (Mozambique). Biodiversity was proxied by native vegetation extent; ecosystem services included above-ground carbon storage and sediment retention. Key elements:

• Policy design options (18 combinations):
  • Gain approaches: Averted Loss (protect unprotected existing vegetation) vs Improvement (revegetate and protect cleared land).
  • Trade types: Out-of-Kind, In-Kind, Trading-up: Additional Gains (prioritize areas most at risk of loss or least likely to recover), Trading-up: Rarity (prioritize rare vegetation types).
  • Spatial prioritization: Outside PAs, Near PAs (within 10 km), Within PAs (only for Out-of-Kind due to limited opportunities inside PAs).
• Multipliers and assumptions: Averted Loss multiplier = 4 (implies 20% avertable loss); Improvement multiplier = 2 (assumes 50% restoration success). Compensation was allocated at the start of the simulation period. Improvement did not target built-up/industrial land. No leakage was assumed, and protection was assumed to avert all future losses in protected sites.
• Step 1 (Development impacts): Regulated development footprints differed by case. Mining and infrastructure (Brigalow Belt, Cabo Delgado); mining (Iron Quadrangle); mining and oil palm (East Kalimantan). Development was assumed to fully remove vegetation where overlapping.
• Step 2 (Compensation allocation): Using Dinamica EGO, compensation areas were allocated according to each design’s constraints (approach, trade type, prioritization). For In-Kind, compensation matched vegetation type lost; for Out-of-Kind, any type was eligible. Allocation ceased if opportunities were exhausted (insufficient unprotected vegetation for Averted Loss or cleared land for Improvement).
• Step 3 (Counterfactuals): Land-use change models (Dinamica EGO) simulated unregulated biodiversity losses and gains over case-specific timeframes (Brigalow: 2011–2020; Iron Quadrangle: 2010–2020; Cabo Delgado: 2015–2040; East Kalimantan: 2015–2040). Models were calibrated with historical transitions, spatial determinants, and validated against observed changes; all outperformed null models.
• Step 4 (Impact quantification): For Averted Loss, gains equaled counterfactual losses within compensation sites (assuming full avoidance, no leakage). For Improvement, gains equaled half of revegetated area minus any counterfactual gains within compensation sites (reflecting 50% success). Ecosystem services were quantified by comparing current, pre-clearing, and future landscapes: carbon from compiled datasets; sediment retention via InVEST SDR (parameters: flow accumulation 2000 [500 in Cabo Delgado], Borselli’s kb 1.8, ICO 0.5, SDR max 0.8). Outcomes combined regulated development losses, compensation gains, and counterfactual trends to assess proximity to NNL.

• No compensation policy achieved NNL of biodiversity in any case study. Even best-performing designs failed to offset regulated development losses fully.
• Determinants of performance:
  • Land availability for compensation constrained outcomes. Example: East Kalimantan’s regulated development cleared 6,311 km² of forest; with an Improvement multiplier of 2, 12,622 km² of restoration was required, exceeding available cleared land for restoration (6,408 km²).
  • Counterfactual trends within compensation sites limited additionality. Averted Loss performed better where counterfactual losses were high (e.g., Brigalow Belt), but still insufficient; Improvement performance declined where natural recovery was likely (e.g., Brigalow Belt).
• Policy design effects:
  • Under the chosen multipliers, Improvement often out-performed Averted Loss for biodiversity, largely due to optimistic 50% restoration success and low counterfactual gains; however, Averted Loss could outperform where restoration success is low and counterfactual gains are high.
  • Trading-up: Additional Gains generally performed best among trade types by targeting high-risk areas (for Averted Loss) or areas least likely to recover (for Improvement). In East Kalimantan and Cabo Delgado (Improvement), Out-of-Kind and Trading-up performed similarly because compensation demands restored all available land.
  • Prioritizing Outside PAs typically outperformed Near or Within PAs due to higher counterfactual losses outside; exceptions occurred where threats were near PAs (e.g., Cabo Delgado).
  • In-Kind trades were limited where impacted vegetation types had few compensation opportunities (e.g., Cabo Delgado’s deciduous miombo savannah woodland WSW28 needed >2.6× available restoration area).
• Ecosystem services:
  • Some scenarios achieved NNL for carbon storage (Brigalow Belt and Iron Quadrangle), as compensation targeted more carbon-dense areas than those developed. Improvement tended to perform worse than Averted Loss for carbon in some contexts due to counterfactual clearing of carbon-dense lands.
  • Sediment retention outcomes were mixed: some policies outperformed biodiversity outcomes (e.g., Improvement in Brigalow Belt, Averted Loss in East Kalimantan), while others underperformed (e.g., Iron Quadrangle; Improvement in East Kalimantan).
  • Performance for ecosystem services depended on both biodiversity outcomes and spatial links between development and compensation sites; policies should explicitly target each service to ensure NNL.
• Multipliers required to reach NNL were often impractically large (e.g., >60 for Averted Loss Out-of-Kind in Cabo Delgado). In at least two case studies (Cabo Delgado, East Kalimantan), achieving NNL was impossible given land constraints even with large multipliers.
• Regional outcomes: Policies reduced region-wide vegetation loss by <10% in three case studies (up to 37% in Cabo Delgado) relative to counterfactuals, with overall regional losses of 3–13% of native vegetation over the analysis periods, reflecting narrow policy scope and large unregulated losses.

The study shows that compensation policy performance is jointly governed by policy design and local conditions. Limited land for restoration or protection and counterfactual trends within compensation sites prevent compensation from achieving NNL of biodiversity at regional scales under realistic multipliers. Although certain designs (e.g., Trading-up: Additional Gains) and spatial strategies can improve outcomes, they do not overcome structural constraints, including narrow policy scope that regulates only a subset of sectors. Ecosystem service outcomes diverge from biodiversity outcomes, underscoring the need for explicit multi-goal design. Achieving NNL via compensation alone is unlikely where unregulated losses are high and land availability is limited; avoidance and minimization of impacts therefore remain essential. Broadening policy scope could improve alignment with conservation goals but would further increase land requirements for compensation, often beyond availability. Target-based compensation frameworks may provide clearer contributions to conservation outcomes than counterfactual-based NNL framing in declining systems.

Across four diverse regions, no tested compensation policy achieved NNL of biodiversity, primarily due to limited land for compensation and counterfactual dynamics that reduce additionality. While certain policy designs and prioritizations improve performance and can deliver benefits for some ecosystem services, compensation typically yields only modest reductions in regional biodiversity loss where unregulated pressures persist. Policy implications include: prioritizing impact avoidance once compensation opportunities are exhausted; setting multipliers informed by counterfactual losses and restoration uncertainty; explicitly targeting multiple goals (biodiversity and ecosystem services); and considering a shift toward target-based compensation to align with broader conservation objectives. Future research should improve empirical estimates of restoration success across ecosystems, refine counterfactual modeling (including leakage), incorporate vegetation condition and other biodiversity proxies, and evaluate governance and implementation factors that affect real-world outcomes.

Key limitations and assumptions include: using native vegetation extent as a proxy for biodiversity (excluding condition and species-level changes); assuming development fully removes vegetation within footprints; assuming no leakage/displacement of losses; assuming 50% restoration success (multiplier = 2) which may be optimistic; allocating all compensation at the start without dynamic market or temporal responses; holding governance, compliance, and management effectiveness constant; potential errors in land-use and ecosystem service datasets and propagation of uncertainty in simulations; differences in regulated development definitions among case studies; and not assessing temporal dynamics of benefit accrual in detail. Results are intended to guide high-level policy insights rather than prescribe site-specific policies.