Conference Agenda

This paper introduces an explainable machine-learning framework for environmental valuation using stated-preference data. We analyse recreation demand for the Cansiglio Forest in northeastern Italy, where congestion issues have prompted policy interest in parking fees as a management tool. Using data from a contingent-behaviour experiment, we estimate visit demand with a Random Forest model that flexibly captures nonlinearities, thresholds, and heterogeneous responses without imposing functional-form or distributional assumptions. To recover economic interpretability, we combine the Random Forest with Shapley values, which decompose each prediction into contributions of individual variables. The results show that seasonality and visitor characteristics dominate demand, while parking fees have a nonlinear and heterogeneous effect: moderate prices have little impact, whereas high fees reduce visits for a subset of more price-sensitive users. Our framework reveals threshold effects and heterogeneity that standard parametric models would be prone to mask. Overall, the approach offers a powerful tool for environmental valuation and support of policy design.

The global transition to low-carbon energy depends on energy transition minerals (ETMs). Yet, the extraction of these minerals often occurs in biodiverse and carbon-rich forests, potentially undermining their climate benefits. Here, we provide global, causally identified estimates of deforestation and related greenhouse gas (GHG) emissions attributable to ETM mining, combining nearly 3,000 projects with satellite-based forest change data. Using a staggered difference-in-differences design, we find that ETM mining causes sustained forest loss—averaging ~20% within 10 km buffers over 15 years—comparable in magnitude to traditional minerals such as coal and gold. These losses are disproportionately concentrated in tropical forests with high climate mitigation potential. Incorporating deforestation-related emissions increases the mining-stage carbon footprint of ETMs by 63% on average, and up to 98% for certain minerals. Our findings reveal mining-induced land-use change as a major but overlooked source of emissions in global energy transition.

We estimate of the economic value of a globally significant biome: the tropical moist forest of the Congo Basin. Using open data and reproducible workflows, we find a gross value for the Congo Basin forests of US$7.8 billion in 2019 (6.3% of regional GDP). This value flows from forest ecosystem services with existing markets (artisanal and industrial timber, fuelwood, bushmeat, and tourism), establishing a verifiable lower bound. Since the success of policies targeting the intertwined challenges of development and conservation in the Congo Basin (including novel financial instruments such as the Tropical Forests Forever Facility) depends on how interventions impact local economic trajectories, trade-offs, and incentives, we estimate how the value of each ecosystem service is distributed across rightsholders. We apply our distributional approach to value non-marketed climate regulation services, identifying a trade-off between uncertainty and equity in ongoing efforts to monetize this service.

We undertake the first systematic analysis comparing and contrasting U.S. improved forest management (IFM) projects participating in the California regulated and global voluntary carbon offset markets. We link a novel geospatial dataset of IFM offset projects located in the U.S. Forest Northern Region to measurements of above-ground live carbon collected through U.S. Forest Service Forest and Inventory Analysis (FIA). Statistical and calibrated simulation analyses both provide evidence for market entrance selections and excessive offset credit issuance. As revealed by event study analysis and two-stage logistic regression with carbon trends, IFM projects on regulated and voluntary markets differ significantly in pre-market forest management relative to their statistical counterfactual forestlands with similar carbon storage capacities. Payoff estimations based on forest simulation modeling replicate this entrance outcome, illustrating that regulated market projects realize greater revenues under the regulated market rules than they would under the voluntary market rules, and likewise voluntary market projects realize greater revenues under the voluntary market rules. Using the simulation model, we also find that real-world market entrance outcomes match with the generous lower baselines implied in the IFM projects' registry documentation, rather than the more appropriate business-as-usual baselines. Because of that, regulated market projects are non-additional and voluntary market projects also issue about 1.8 times the offset credits of their justified emission reduction. While current offset markets promote forest-based projects as nature-based climate solutions, they also raise serious integrity concerns that may weaken overall incentives for carbon reduction by corporate buyers. Based on the results from a nested logistic model with simulation approximated project payoffs, switching to stricter business-as-usual baselines would solve the overcrediting problem, but also lower the market participation of forest-based projects, and reduce the potential of carbon offset markets to mitigate emissions through motivating forest conservations.