1. Introduction
Environmental preservation and economic development are widely understood to be mutually reinforcing goals that must advance in tandem. This principle is enshrined in key global agreements such as the Rio Declaration on Environment and Development (United Nations, 1992), the 2030 Agenda for Sustainable Development, and the Paris Agreement, all of which call for integrative approaches that balance growth with ecological sustainability and climate resilience (United Nations, 1992; United Nations, 2015). At the core of this vision lies the political and institutional architecture of environmental regulation, wherein governments at all levels must establish and enforce appropriate laws, regulations, and codes to effectively advance sustainable development (Sarabdeen, 2024). While regulatory frameworks are not always the most socially equitable or technically effective mechanisms for achieving environmental justice (Sachs, 2010; Kotzé and Adelman, 2023), they provide essential legal instruments and procedural safeguards for managing potentially harmful activities. These frameworks typically arise from deliberative policy processes and reflect consensus over how to normatively structure environmental risk and compliance (Lemos and Agrawal, 2006).
Environmental licensing systems are guided by foundational principles such as the precautionary principle, public participation, and the need for technical assessments—principles widely consolidated in countries with robust environmental governance systems (Handoyo, 2024). In Brazil, however, these principles are currently being redefined as part of ongoing legislative reforms to environmental licensing. A central component of this reform is the General Environmental Licensing Law, approved by the Federal Senate and, subsequently, by the Chamber of Deputies, before being signed into law by the President in 2025 (No. 15.190/2025), which will enter into force in 2026 (Brazil, 2025). The law introduces a major restructuring of the existing licensing framework, proposing more flexible mechanisms — including self-declaratory licenses — for specific categories of projects (see Section I, Art. 5, Subsection V, Law 15.190/2025). These instruments aim to simplify and expedite licensing procedures by reducing the need for case-by-case assessments. While the law is framed as a modernization measure intended to streamline bureaucratic processes, it lacks a clear implementation strategy that details how the necessary data systems, institutional capacity, and technical infrastructure will be developed or mobilized to support this transition. Previous analyses, including Athayde et al. (2022), had already noted the absence of such planning when the bill was under consideration in the Chamber of Deputies. Among the changes, the law permits the use of self-declaratory licenses for projects classified as low to moderate environmental impact, according to Federal Ordinance No. 8.437/2015 and Federal Law No. 140/2011. This includes activities such as small hydroelectric plants, road construction, sewage treatment facilities, agrosilvopastoral systems, and certain types of mining operations. These licenses would be issued directly by developers and could bypass prior environmental impact assessments, eliminate the analysis of locational alternatives, and allow for automatic renewals. The law also does not include provisions for climate adaptation or governance instruments that support climate change coping strategies (Schramm and Fishman, 2010; Fernandes et al., 2025), which are important for long-term resilience. Between the parliamentary approval of bill and its enactment as law, the Presidential Office issued partial vetoes aimed at restoring key environmental safeguards and ensuring legal consistency. Broadly, these vetoes curtailed provisions that excessively simplified licensing procedures, weakened federal oversight, and reduced social participation mechanisms. Their intent was to reestablish minimum standards of environmental protection and reaffirm the Union’s authority to define general licensing norms. Nevertheless, despite these corrections, the final text of the law retained clauses granting municipalities extensive autonomy to conduct environmental licensing. The law advances the decentralization of environmental licensing from federal and state agencies to municipal governments without clearly ensuring that local authorities possess the legal, technical, and fiscal capacities required for effective environmental oversight, thereby reinforcing the systemic risks of decentralization without adequate institutional support (Athayde et al., 2022).
Here, our argument is not against the decentralization of environmental licensing per se. From an applied standpoint, local enforcement of environmental regulations may offer certain advantages (Bakhsh et al., 2018). Local governments often possess context-specific knowledge, as environmental concerns are closely intertwined with urban planning (Sjoberg, 2015), landscapes management, and local social-ecological dynamics (Dawson et al., 2017). Moreover, decentralization can enable cross-sector collaboration and foster more responsible, place-based solutions to complex environmental challenges (Yuerong et al., 2024), particularly in large, territorially diverse nations where centralized policies often fail to reach local realities with sufficient efficacy and timeliness. However, decentralization without adequate safeguards can also generate negative externalities. Lenient enforcement may help municipalities appear business-friendly and attract investments (Sjoberg, 2015), while the environmental and social costs are externalized and borne by society and natural assets. Fredriksson et al. (2010), through a theoretical model, demonstrated that this dynamic leads to a “home bias”, where local officials favor short-term gains and adopt suboptimal regulatory standards. In this context, two critical issues emerge: legal certainty and administrative efficiency. While national legislation is designed to provide a harmonized framework for environmental governance (Esty and Porter, 2005), it presumes that local governments have the institutional capacity to implement and enforce such regulations effectively (Borrás et al., 2024). This is precisely where the danger lies: the new General Environmental Licensing Law in Brazil delegates excessive autonomy to municipalities — in our view, without ensuring adequate technical, legal, or operations preparedness — thereby posing significant risks to environmental protection and governance integrity.
A
From a theoretical standpoint, this tension reflects a broader pattern long identified in decentralization studies. As Ribot (
2004) observes, many reforms labeled as “decentralization” often amount to
deconcentration, that is the mere transfer of responsibilities without corresponding authority, fiscal resources, or decision-making power. Such incomplete reforms tend to reproduce central control while transferring management burdens downward, leading local governments vulnerable and under-resourced. In these conditions, environmental functions frequently fail, as responsibilities are devolved without the fiscal or institutional capacity to exercise discretion effectively. Similarly, Agrawal and Gibson (
1999) argue that successful decentralization requires local institutions capable of making, implementing, and enforcing rules over resource use. When the implementation is devolved but rule-making and adjudication remain in little assurance of local institutional capacity, governance becomes largely procedural rather than transformative. In this sense, decentralization without genuine empowerment of fiscal support risks producing only the appearance of participation while entrenching structural inequality and administrative fragility. The Brazilian case thus exemplifies this paradox: while the legal reform of decentralization promises proximity and efficiency, its practice, absent institutional and fiscal scaffolding, may weaken rather than strengthen environmental governance.
Effective environmental licensing must balance developer, sustainable, and social interests, which is inherently difficult to achieve (Enríquez-de-Salamanca, 2021). Projects and their associated development processes, whether led by private actors or government authorities can heighten pressures on evaluators, as noted by Kolhoff et al. (2013). In addition, the independence and functioning of environmental licensing system varies between countries or regions, and in all cases the final decision rests with government agencies (Morrison-Saunders, 2018), which must respond to all interested parties. Nevertheless, political control is a constant (Enríquez-de-Salamanca, 2021). From the standpoint of environmental safeguards, licensing needs broad requirements to be effective, fair, and development-enabling: (i) prior and proportionate assessment that explicitly considers reasonable locational and technological alternatives, cumulative effects, and climate-risk appraisals (Bragagnolo and Geneletti, 2012; Schramm and Fishman, 2010); (ii) verified institutional capacity and decisional independence at the competent authority (Loomis and Dziedzic, 2018); (iii) meaningful, timely, and inclusive participation consistent with procedural-justice standards (Ulibarri et al., 2022); and (iv) binding conditions, compliance monitoring, and adaptative management (Barr et al., 2021). Accordingly, reforms to the licensing process should, at a minimum, maintain rather than weaken these socio-environmental safeguards.
Rooted in sound environmental governance, effective institutional arrangements can lead to improved environmental management outcomes (Young, 2003). Public institutions must be capable of enforcing regulations, facilitating public participation, and ensuring transparency and accountability in decision-making processes (Paavola, 2007). As a contribution to understanding the structural foundations of environmental governance, this study evaluates the institutional capacity of municipalities in Brazil to undertake the complex and high-stakes responsibilities of environmental licensing — particularly in light of new environmental reform, which will offload significant regulatory authority to the local level without assuring the necessary institutional readiness. To this end, we evaluated the organizational and institutional environmental frameworks of 1,270 municipalities across all twenty-six Brazilian states. This assessment was based on key indicators, including the existence of a municipal environmental secretariat, the number of active staff, the presence and nature (deliberative or consultative) of municipal environmental councils, the participation of civil society in these councils, and the existence of local environmental legislation. These elements provided a comprehensive overview of municipal-level capacity to address the governance challenges posed by decentralization, particularly whether local structures are equipped to uphold and strengthen environmental protection in the face of potentially widespread regulatory weakening under the law reform. Finaly, this study addresses a critical research gap concerning the complex relationship between public environmental governance and the need for large-scale, quantitative assessments of institutional structures responsible for implementing environmental regulation.
2. Material and methods
2.1. Rationale and Analytical Framework
Public governance and operational effectiveness of institutions encompass multiple dimensions, including voice and accountability, political structure, government effectiveness, regulatory quality, and control mechanisms (Oyewo et al., 2024). These elements not only provide the institutional scaffolding for environmental policy decisions but also influence a government’s capacity to address complex environmental challenges (Handoyo, 2024). Nevertheless, this relationship is neither linear nor homogeneous across government tiers (Nordin et al., 2024). Variability in institutional maturity and structural development can result in divergent governance outcomes; in other words, a well-defined institutional architecture does not always translate into consistent levels of environmental performance (Magsi et al., 2018). To assess the readiness of Brazilian municipalities to handle high-complexity environmental licensing processes — particularly under the decentralization effects of new environmental law — we compiled a dataset integrating demographic, territorial, economic, and institutional variables, with particular emphasis on attributes related to local environmental governance.
2.2. Sampling Design and Exclusion Criteria
Out of the total 5,570 municipalities distributed across Brazil’s twenty-six states, we adopted a stratified random sampling with proportional allocation, selecting from each state in proportion to its number of municipalities, resulting in 1,270 evaluated cases. This sampling design provides a balanced representation of the national municipal structure, minimizing geographic and administrative bias. Because each stratum (state) contributed observations in proportion to its total number of municipalities, no additional weighting or post-stratification adjustments were necessary. This approach ensured that the sample accurately captured Brazil’s regional and biogeographical diversity while maintaining analytical coherence across all models. Furthermore, all state capitals (26) were evaluated within the study framework; however, they were intentionally retained from the macro-regional and biome-level comparative analyses and were maintained in the main national-level analyses. This decision was grounded on preliminary assessments indicating that capitals consistently demonstrate higher levels of environmental governance and institutional capacity relative to non-capital municipalities (de Moura, 2016). Including these outliers could artificially inflate the average capacity indicators and obscure the underlying heterogeneity across the broader set of municipalities, which are ultimately responsible for implementing decentralized environmental licensing at the local scale. Nevertheless, capitals serve as important benchmarks for interpreting institutional disparities in the national context.
2.3. Territorial, Socioeconomic, and Environmental Characterization of Municipalities
Territorial area (km2), population size, predominant biome, gross municipal product (GMP), Human Development Index (HDI), urbanized area (km2), and area under agro-pastoral use (hectares) were collected to situate municipalities within Brazil’s vast regional heterogeneity. Additionally, we considered population exposure to environmental risk, defined as residents located in areas prone to floods, flash floods, and landslides according to CEMADEN’s classification of municipalities critical to natural or climate-related disasters. This dataset establishes a foundation baseline for analyzing how structural characteristics shape the institutional capacity for environmental governance (environmental indicators are detailed in the next section) at the local level. Existing literature emphasizes that territorial scale, population density, and economic complexity influence not only the demand for environmental governance but also the ability of subnational governments to respond effectively to such demands (Jordan et al., 2005; Evans et al., 2017). Moreover, the spatial distribution of environmental risks, coupled with disparities in development indicators such as HDI and gross domestic product per capita, has been shown to correlate with uneven institutional performance and environmental enforcement capabilities (Hickel, 2019), as well as are indicators of economic well-being (Handoyo, 2024). By accounting for these structural variables, we aim to capture the underlying context that mediates the effectiveness of environmental regulation and the feasibility of decentralizing licensing authority to municipalities.
2.4. Institutional Environmental Governance Indicators
To evaluate municipal readiness for environmental licensing responsibilities under the decentralized regulatory model, we incorporated a comprehensive set of indicators representing the institutional environmental governance structures at the local level (Table 1). These metrics reflect both formal administrative capacities and participatory governance mechanisms, which together define a municipality’s ability to plan, deliberate, and enforce environmental policies effectively (Giest, 2023). Indicators were selected based on their relevance to key dimensions of environmental governance: administrative structure, regulatory autonomy, participatory deliberation, and policy planning (Lemos and Agrawal, 2006). Together, they provide an evidence-based foundation for assessing whether municipalities are institutionally equipped to take on licensing processes involving complex socio-environmental risks.
Table 1
Institutional environmental governance indicators used to assess municipal readiness for decentralized environmental licensing. Indicators capture dimensions of administrative structure, technical capacity, participatory governance, and regulatory autonomy.
Environmental Governance Indicator | Indicator Relevance to Policy Planning, Deliberation, and Enforcement |
|---|
Existence of a Municipal Environmental Secretariat (yes/no) | Indicates whether the municipality has an official administrative body dedicated to environmental management. |
Number of Active Environmental Staff | Reflects the human resource capacity available to implement and monitor environmental policies. |
Outsourcing of Environmental Services (yes/no) | Shows reliance on outsourced services, which may affect continuity and technical autonomy. |
Existence of a Municipal Environmental Council (yes/no) | Demonstrates the existence of participatory governance mechanisms for environmental decision-making. |
Council Meetings Held in the Past 12 Months (yes/no) | It indicates the activity and functionality of the environmental council. |
Council’s Legal Status (Deliberative vs. Consultative) | Defines whether the council has the authority to make binding decisions or serves only an advisory role. |
Percentage of Civil Society Representation on the Council (< 50%, 50%, or > 50%) | Measures the level of civil society’s engagement in environmental governance. |
Administrative Agreements Delegating Environmental Responsibilities to State Agencies, specifically for environmental licensing (yes/no) | Indicates whether the municipality has formalized environmental responsibilities through agreements with state-level bodies. |
Presence of Local Environmental Legislation (yes/no) | Shows whether the municipality has its own legal instruments for regulating environmental issues. |
Inclusion in of a Regional Ecological-Economic Zoning Plan (ZEE) within municipality (yes/no) | Indicates integration into regional environmental planning frameworks that guide sustainable land use and conservation. |
All data used in this study, across both the territorial-socioeconomic descriptors and the institutional environmental governance indicators are publicly available and sourced from the MUNIC (2025) official database of the Brazilian Institute of Geography and Statistics (IBGE). This national agency is responsible for producing and disseminating official statistical, geographic, and environmental information to support public policy and research. The datasets can be accessed at: https://cidades.ibge.gov.br/brasil. By integrating these indicators, this study not only maps the existing institutional heterogeneity across Brazilian municipalities but also helps to identify structural gaps that may compromise the integrity of decentralized environmental licensing in the context of new law.
3. Data Analysis
3.1. Construction of the Environmental Institutional Capacity Index (ICI)
To synthesize the multidimensional information embedded in the set of governance indicators for all municipalities including capital states, we applied a Factor Analysis of Mixed Data (FAMD). This method is designed to handle datasets contained both categorical (qualitative) and continuous (quantitative) variables (Pagès and Camiz, 2008). The FAMD was applied using the factominer package (Husson et al., 2025) following the methodological framework developed by Pagès (2014). FAMD combines the principles of Principal Component Analysis (PCA) for numerical variables and Multiple Correspondence Analysis (MCA) for categorical variables, projecting both types into a shared factorial space where their contributions are balanced (Kenkel, 2006; Abdin and Valentin, 2007). The variables selected for the analysis represent key institutional components of municipal environmental governance (see Table 1). The decomposition can be formally represented as:
Where: X* is the standardized data matrix, combining centered and scaled quantitative variables with disjunctive-coded qualitative variables; U is the matrix of observation scores (municipalities); D is the diagonal matrix of singular values (i.e., square roots of eigenvalues), and V contains the loadings across the principal dimensions.
Categorical indicators were encoded using disjunctive (dummy) coding prior to the FAMD, in which each category was transformed into binary variables representing the presence or absence of that attribute. This standard procedure ensures that categorical variables contribute equally to the factorial space relative to continuous variables. All coding procedures were double-checked against the original dataset to ensure interpretive validity and semantic consistency, particularly for ordinal variables with policy relevance, which were verified through cross-referenced with municipal legal records and the MUNIC database.
The optimal number of retained dimensions was determined using the scree plot of eigenvalues and the cumulative explained variance criterion. For the construction of the Environmental Institutional Capacity Index (ICI), we retained the first three dimensions, which together captured for over 84% of the total variance. These dimensions preserved the contribution of all original variables on grounds of parsimony and interpretability, since subsequent axes add only minor incremental variance and would likely introduce noise into the composite ICI without improving its explanatory power (Table S1). To assess the robustness of the index structure, we performed sensitivity checks by recalculating the ICI using four and five retained dimensions. These additional axes accounted for approximately 10% of the cumulative variance combined, and the resulting patterns and rank correlations with the original ICI remained highly consistent (see Table S3 in which indicates stable patterns and high rank correlations with the original index). The ICI for each municipality, including the capital states, was computed by aggregating its scores across the retained dimensions through a weighted linear combination, where the weights correspond to the proportion of variance explained by each dimension, a common approach in composite index construction (Zhou et al., 2010):
Where: ICI
j is the index score for municipality
j; F
ij is the score of municipality
j on dimension
i;
wi =
is the normalized eigenvalue (weight) for dimension I; and
k = 3 is the number of retained dimensions.
Finally, to enhance interpretability and comparability, ICI values were standardized across all municipalities using z-score transformation:
Where µ and σ represent the national mean and standard deviation of the raw ICI scores, respectively, and * denotes standardized values. Negative ICI values are not only statistically valid but also conceptually meaningful, denoting municipalities with below-average institutional capacity relative to the national profile. Conversely, positive values indicate municipalities with institutional readiness above the national mean.
Following standardization, ICI scores were expressed on a z-score scale, enabling direct comparison across municipalities. These standardized values served as the basis for defining the interpretative classes shown in Table 2, where one-unit intervals around the national mean were used to delineate ordinal categories of institutional capacity.
3.2. Machine Learning Modeling and Explainability Approach
To explore how structural, demographic, and socioeconomic characteristics of municipalities influence their ICIs regard to predictive accuracy, we implemented a supervised machine learning approach based on the Random Forest (RF) algorithm. The ICI, previously calculated via a factorial analysis (FAMD), was used as the response variable in a regression modeling framework. The RF algorithm was selected due to its robustness for non-linear modeling, its ability to handle high-dimensional data without strict assumptions regarding variable distributions (Scornet, 2017), and its capacity to automatically model complex interactions among predictors (Breiman, 2001). Model fitting was performed using the ranger package (Wright and Ziegler, 2017), with the number of trees and variable importance computed via permutation-based measures (i.e., MSE increase). The keep.inbag = TRUE argument was specified to enable the computation of Shapley values, which require access to the in-bag samples from the fitted trees.
Critically, to avoid circularity and endogeneity issues, the set of predictors excluded all variables directly involved in the construction of the ICI (i.e., those related to environmental governance structures and institutional capacity). The final set of features thus consisted exclusively of structural, demographic, and socioeconomic attributes of each municipality, including total area (km²), urban area (km²), population size, population exposed to environmental risks, agro-pastoral land use (hectares), Gross Municipal Product (GMP) per capita, and Human Development Index (HDI) (more details
A
in Supplementary material). Notably, while the number of active environmental staff was originally incorporated as an operational indicator within the ICI formulation, it was retained in the RF model as a human resource parameter reflecting the broader structural capacity of municipalities, rather than a direct proxy of institutional governance. To verify that its inclusion did not introduce endogeneity or circularity, we conducted a robustness test by re-estimating the model without this variable using the same optimized hyperparameters. The comparative results showed negligible differences in explanatory performance (ΔR² < 0.03) and nearly identical SHAP-based variable importance rankings (ρ = 0.95), confirming that the model’s explanatory structure remains stable and theoretically coherent even when the staff variable is excluded (see Table S4). Missing data handling was only required for the variable
population exposed to environmental risks, since not all municipalities have populations residing in officially designated risk-prone areas. All other predictors contained complete data, as they represent structural, demographic, or socioeconomic attributes universally available across the national territory. The missing values for this variable were imputed using the
imputeTS package (Moritz and Bartz-Beielstein, 2017), which applies time-series-based interpolation methods adapted here to spatial municipal datasets.
3.3. Model Explainability via SHAP Values
To ensure interpretable results beyond standard feature importance metrics, we employed a SHAP (SHapley Additive exPlanations) framework to assess both global and local contributions of each predictor to the model output (Lundberg and Lee, 2017). The importance of features in conventional machine learning only indicates the degree of influence of input variables on model output but does not reveal how input variables influence the final model (Wang et al., 2022). SHAP values provide a theoretically grounded approach based on cooperative game theory, offering consistent and locally accurate attributions of feature contributions for each individual prediction (Strumbelj and Kononenko, 2014). Therefore, the SHAP method was embedded and applied to interpret the trained model (Lundberg and Lee, 2017). The primary advantage of the SHAP values lies in their ability to reflect the impact of features on each sample, illustrating both positive and negative effects (Molnar et al., 2020).
SHAP computations were performed using the treeshap package (Maksymiuk et al., 2020), which is specifically designed for tree-based models such as RF and offers significant computational efficiency over general-purpose model-agnostic explainers. The ranger.unify function was first used to convert the ranger model into a unified tree structure compatible with treeshap, followed by the computation of exact SHAP values across the entire dataset using the treeshap function. Two complementary explainability analyses were carried out: Global interpretability, by aggregating the absolute mean SHAP values of each predictor to determine their overall importance in driving ICI variability across municipalities; and Local interpretability, through beeswarm plot that visualize the distribution and directionality (positive or negative) of SHAP values for individual predictions, highlighting heterogeneity in predictor effects among municipalities. SHAP values were interpreted as associative indicators of feature importance within the model rather than causal influences. They describe how each predictor contributes to shifts in the model output, conditional on the presence of other variables. Thus, SHAP analysis reveals the strength and direction of associations learned by the Random Forest but does not imply mechanistic or causal relationships.
4. Results
4.1. General Panorama of Municipal Environmental Governance
Among the 1,270 municipalities assessed across all Brazilian regions and states, 47% have a Municipal Environmental Secretariat, while 53% lack this institutional structure. Furthermore, 58% of the municipalities do not have an Environmental Council (EC), highlighting the widespread absence of participatory governance mechanisms for environmental decision-making at the local level. Among the municipalities with an established EC, 50.4% operate solely in a consultative capacity, indicating limited deliberative power in most councils. Regarding the presence of local environmental legislation, the distribution was nearly balanced, with 49.8% of municipalities reporting the existence of specific environmental legislation, while 50.2% reported its absence. There is a high asymmetry in the distribution of technical staff within the municipal environmental sector. A total of 28.3% of municipalities reported having no active environmental staff, while 46.8% had between one (19.3%) and four (6.6%) professionals (see Figure S1).
Pronounced spatial asymmetries were also observed when the institutional capacity attributes were analyzed at the state level (Fig. 1). The presence of Municipal Environmental Secretariats exhibited marked variability across states, with some achieving proportions close to 90%, while others remained below 30% (Fig. 1A). Similarly, the percentage of municipalities with local environmental legislation was also heterogeneous; however, most states demonstrated relatively high coverage, suggesting a wider institutionalization of normative frameworks (Fig. 1B). A clear regional gradient emerged for the presence of Environmental Councils (Fig. 1C), with higher prevalence concentrated in municipalities from the Central-West, Southeast, and South regions, in contrast to consistently lower levels in the North and Northeast (with exception for Bahia state). The most critical scenario was observed for the existence of Municipal Ecological-Economic Zoning Plans, with 12 states showing no municipalities with such instruments (score zero), and the highest coverage reaching only 18% in the best-performing state (Ceará state) (Fig. 1D).
4.2. Environmental Institutional Capacity Index
The Environmental Institutional Capacity Index (ICI) reveals a structurally critical scenario of municipal environmental governance in Brazil, with the majority of municipalities exhibiting low institutional readiness. Derived from FAMD, the ICI integrates three retained dimensions that capture complementary aspects of local environmental governance (see Table S2). The resulting pattern is especially pronounced along a geographic gradient, with more acute institutional fragility concentrated in the North and Northeast regions, while comparatively stronger capacities are observed in southern municipalities (see Fig. 2). The ICI distribution shows a strong concentration of municipalities within the lower tiers of the index, indicating fragile, insufficient, or even entirely absent institutional structures (Fig. 3). Approximately half of the assessed municipalities fall below an ICI score of 1.5, reflecting the widespread absence of fundamental governance elements such as environmental secretariats, active environmental councils, municipal legislation, and a minimum number of technical staff (Figure S2). Table 2 defines the ICI range categories, providing a technical description and corresponding institutional situation for each.
Alarmingly, a substantial proportion of municipalities are located within the negative range of the ICI which is a clear indication that these territories are not only below the national average but entirely devoid of minimally functional institutional frameworks. On the other hand, only a small subset of municipalities achieves scores above 2.5, representing cases of consolidated governance and institutional capacity or, in rare instances, institutional excellence (Fig. 3). This disparity becomes even more evident when compared to the consistently higher ICI scores observed for state capitals, which serve as benchmarks of stronger institutional infrastructure. These are municipalities equipped with comprehensive institutional frameworks, robust technical staff, functional participatory councils, and strong environmental legislation—conditions that enable them to effectively conduct environmental licensing, enforcement, and management processes independently. When disaggregated by macroregion (i.e., geopolitical clusters) (Fig. 4), the data reveal pronounced territorial inequalities. Municipalities in the North and Northeast regions exhibit markedly lower ICI distributions, with a significant accumulation in the lowest capacity tiers. In contrast, the South, Southeast, and Central-West regions show a more balanced distribution, with a relatively higher proportion of municipalities attaining intermediate and high levels of institutional capacity. When analyzed by biome (Fig. 5), the pattern remains equally concerning, with the overall institutional environmental capacity across all biomes still critically low. The situation is particularly alarming in the Amazon and Caatinga biomes, where the concentration of municipalities in the lowest governance tiers is even more pronounced, underscoring profound territorial asymmetries in institutional capacity linked to Brazil's ecological regions.
Table S1
Results of the factor extraction procedure used to determine the most suitable dimension for constructing the Environmental Institutional Capacity Index (ICI). The table presents the eigenvalues, the percentage of variance explained by each dimension, and the cumulative variance.
| | Eigenvalue | Percentage of variance (%) | Cumulative variance (%) |
|---|
Dimension 1 | 4.1293861 | 59.659206 | 59.659206 |
Dimension 2 | 1.1530219 | 12.811355 | 72.470561 |
Dimension 3 | 1.0999605 | 12.221783 | 84.692344 |
Dimension 4 | 0.9598121 | 5.664579 | 90.356923 |
Dimension 5 | 0.8647390 | 4.608211 | 94.965134 |
Table 2
Interpretative framework of the Environmental Institutional Capacity Index (ICI) classification. The table defines the ICI range categories, providing a description and corresponding institutional situation for each. This classification supports the interpretation of governance and institutional capacity thresholds to guide understanding of Figs. 3, 4, and 5.
ICI Range | Technical Description | Institutional Situation |
|---|
< 0 | Virtually nonexistent | Complete or near-complete absence of institutional capacity. No environmental secretariat, no council, no legislation, no technical staff, no intergovernmental agreements, no ecological-economic zoning plan (EEZ). Environmental governance is virtually nonexistent. |
0 to 0.5 | Very low | Extremely fragile governance. Some isolated minimal structures (e.g., an environmental council, but no secretariat and almost no staff). |
0.5 to 1.5 | Low | Partial presence of governance structures. Some combination of environmental secretariat, council, or technical staff, but incomplete, inconsistent, or non-functional. |
1.5 to 2.5 | Medium / Under development | Governance under development. Active secretariat, functional council, reduced technical staff, and legislation that may be absent or only recently implemented. |
2.5 to 3.5 | High / Consolidated | Consolidated governance. Active secretariat, deliberative and participatory council, local environmental legislation in force, reasonably proportional technical staff, and active intergovernmental agreements. |
> 3.5 | Very high / Institutional excellence | Excellence standard. All institutional structures fully formalized, high technical staff, active participatory and deliberative council, robust environmental legislation, and strong intergovernmental integration (including EEZ and formal agreements). |
Table S2
Factor loadings (FC) of the environmental governance indicators on the first three retained dimensions obtained through the Factorial Analysis of Mixed Data (FAMD) used to construct the Environmental Institutional Capacity Index (ICI).
Environmental Governance Indicator | FC (Dim 1) | FC (Dim 2) | FC (Dim 3) |
|---|
Existence of a Municipal Environmental Secretariat | 0.81 | 0.14 | 0.07 |
Number of Active Environmental Staff | 0.78 | 0.19 | 0.11 |
Outsourcing of Environmental Services | −0.62 | 0.27 | 0.21 |
Existence of a Municipal Environmental Council | 0.84 | 0.11 | 0.09 |
Council Meetings Held in the Past 12 Months | 0.79 | 0.18 | 0.12 |
Council’s Legal Status | 0.73 | 0.36 | 0.25 |
Percentage of Civil Society Representation on the Council | 0.68 | 0.41 | 0.23 |
Administrative Agreements Delegating Environmental Responsibilities to State Agencies | 0.71 | 0.25 | 0.28 |
Presence of Local Environmental Legislation | 0.77 | 0.29 | 0.22 |
Inclusion in a Regional Ecological-Economic Zoning Plan (ZEE) | 0.65 | 0.48 | 0.31 |
Table S3
Sensitivity analysis of the Environmental Institutional Capacity Index (ICI) to the number of retained FAMD dimensions. The table presents the cumulative variance explained, rank correlation (Spearman’s ρ) with the original three-dimensional ICI, and qualitative assessment of spatial pattern consistency.
Model configuration | Retained dimensions | Cumulative variance (%) | Spearman’s ρ with original ICI | Regional pattern consistency |
|---|
Original ICI | 3 | 84.69 | 1.000 | – |
Alternative ICI | 4 | 90.36 | 0.983 | Stable |
Alternative ICI | 5 | 94.97 | 0.975 | Stable |
Table S4
Random Forest model robustness to exclusion of the variable “Number of Active Environmental Staff”. The model was re-estimated using the same optimized hyperparameters reported in Table S4 to verify potential conceptual overlap with the Environmental Institutional Capacity Index (ICI).
Model configuration | Predictor set | R² (out-of-bag) | RMSE | Spearman’s ρ (SHAP ranking) | Observations |
|---|
Full model | Includes “Active Environmental Staff” | 0.78 | 0.23 | — | 1,270 |
Alternative model | Excluding “Active Environmental Staff” | 0.76 | 0.24 | 0.95 | 1,270 |
Table S5
Hyperparameters and structure of developed machine learning model.
Model | Parameters, functions, and typical range | Optimal parameter |
|---|
Random Forest | Estimator number: 100–3,000 | 1,500 |
N_jobs: 10–80 | 25 |
Minimum_sample_split: 5–40 | 17 |
Minimum_sample_leafs: 20–60 | 28 |
4.3. Drivers of Environmental Institutional Capacity at the National Scale
Random forest was tuned to predict ICI using territorial, socioeconomic, and environmental characterization of municipalities. The optimized machine learning model was retrained for the best-select hyperparameters and evaluated through the simulation of ICI during the validation process (Hyperparameters: estimator number, n jobs, minimum sample split, and minimum sample leafs; see Table S5) and RF yielded satisfactory accuracy at ICI distribution from assessed municipalities (MSE increase = 0.81; r range = 0.86) with the model accurate and sufficiently generalizable (observed ICI vs. predicted ICI) to reproduce the ICI over entire validation process (see Figure S3).
The SHAP-based global interpretability analysis (Fig. 6A) reveals that the ICI is predominantly explained by structural and socioeconomic factors. The Human Development Index (HDI) stands out as the most influential predictor, followed by urban area and total population size, indicating that municipalities with higher levels of socioeconomic development and urbanization tend to exhibit stronger institutional capacity (Figure S4A and C). The number of active environmental staff also plays a relevant, albeit secondary, role as an operational driver. The SHAP dependence plot (Fig. 6B) provides a detailed view of the marginal contributions of each predictor to the ICI across municipalities. Consistent with the global interpretation, socioeconomic variables — including HDI, population size, urban area, and the number of active environmental staff, and additionally population exposed to the climate risk — exert consistently positive impacts on the ICI, with relatively little overlap between positive and negative contributions. Notably, HDI displays a clear directional effect, where higher HDI values are systematically associated with positive SHAP values, while lower HDI drives negative impacts on the ICI. Similarly, municipalities with larger populations and greater urban areas show strong positive contributions. Interestingly, the total municipality area shows an inverse pattern: while its overall effect is moderate, larger municipalities tend to face negative contributions to ICI, suggesting that greater territorial extension may pose logistical, operational, or governance challenges that undermine environmental institutional capacity. Lastly, the area dedicated to agro-pastoral use plays a relatively minor role, with a slightly negative but diffuse effect, indicating that this variable does not strongly differentiate institutional capacity across municipalities. This is likely due to the limited variability in agro-pastoral area among most municipalities. Nevertheless, municipalities from the Central-West and North regions exhibit consistently higher agro-pastoral use areas (Figure S4B).
5. Discussion
From a municipal perspective, environmental governance and the capacity to implement it in Brazil, are fragile and must be carefully considered before delegating greater autonomy to municipalities. Accordingly, our findings indicate that, in general, municipalities are not adequately prepared to assume greater responsibilities for environmentally sound and precautionary licensing processes. Proponents argue that decentralization can reduce bureaucracy, speed up decision-making, and enhance place-based responsiveness; however, without substantial, targeted investment municipal staffing, training, data systems, and enforcement, these benefits are unlikely to materialize. The novelty of this study lies in its comprehensive assessment of the multidimensional aspects of public environmental governance and institutional capacity, and how these intersect with the territorial, socioeconomic, and environmental characteristics of municipalities. As environmental governance has become increasingly fragmented and complex, the growing interactions among institutions under polycentric governance warrant further investigation (Elsässer et al., 2022). Our findings contribute to this debate by showing how subnational institutional weaknesses may undermine the effectiveness of decentralized governance arrangements, particularly in contexts marked by rapid regulatory reform. In this context, Brazil moves against a growing body of evidence suggesting that environmental regulation tends to improve when intergovernmental interaction and cooperation are strengthened at state and substate levels (i.e., municipalities) (Weiland, 1998; Bastos Lima et al., 2017; Elsässer et al., 2022; Handoyo, 2024). This quantitative diagnosis, therefore, provides an empirically grounded overview that should inform both academic debate and the design of public policies aimed at strengthening local institutional capacities — a sine qua non condition for achieving equitable and functional environmental governance.
Over the past several decades, countries around the world have enacted and progressively improved legislation aimed at protecting their natural biological heritage (Armitage et al., 2020). Furthermore, national environmental performance, which reflects how well countries manage their environmental resources and mitigate environmental harms, has been scrutinized in light of global sustainability goals (Handoyo, 2024). Unfortunately, Brazil appears to be on the verge of moving in a dangerously regressive direction with the deregulatory environmental law. Revisions and adaptations of the legal and operational framework for environmental governance are indeed necessary. As development pressures evolve and the synergistic nature of environmental impacts become more complex (Reis-Filho et al., 2024), governance systems must adjust accordingly (Fried et al., 2022). However, as Schramm and Fishman (2010) argue, such improvements must incorporate, for example, adaptative management — a strategy that promotes resilient and decision-making frameworks capable of responding dynamically to new information and changing ecological conditions. In contrast, top-down decisions by monocratic political debates, often justified by the need to streamline procedures, reduce bureaucracy, and expand local autonomy (e.g., municipality governments), can lead to significant setbacks.
When environmental licensing is accelerated without adequate safeguards, it often leads to poor risk assessments (Dias et al., 2022), particularly in municipalities lacking institutional capacity and clear legal instruments — which, surprisingly, is the case for the majority of the assessed municipalities in Brazil. This concern is particularly urgent in regions where local governments are neither structurally equipped nor supported by well-defined environmental governance mechanisms — with amplified effects in the North and Northeast regions, which encompass the Amazon Rainforest and the Caatinga biome, respectively. The Amazon biome, 60% of which lies within Brazilian territory, harbors nearly one-third of the world’s remaining tropical rainforests and holds some of the highest biodiversity levels on the planet (Guayasamin et al., 2024). The Caatinga, an exclusively Brazilian semiarid biome, sustains supports rich socio-biodiversity and is central to local and regional livelihoods, yet paradoxically remains the least protected biome in the country (Barbosa and Gomes Filho, 2022; Teixeira et al., 2021). Our results reinforce this territorial inequality: municipalities in regions with higher HDI scores consistently exhibit better environmental and institutional capacity, while those with lower HDI, often overlapping with these critical biomes, show pronounced institutional deficits. Therefore, any reform aimed at decentralizing environmental responsibilities must be accompanied by deliberate efforts to strengthen local institutional capabilities, ensure legal and procedural clarity, and uphold environmental protection as a fundamental and non-negotiable commitment — especially in territories where ecological significance and institutional vulnerability intersect most acutely.
Municipalities in Brazil have held distinct traits of autonomy since the federalist model was inaugurated in the late 1980s, following the end of the military dictatorship (Melo, 1996). Under the current federal arrangement, municipal autonomy is defined as the power to govern local affairs within the boundaries established by higher authority in political, administrative, and fiscal spheres, supported by both self‑generated and transferred resources (Federal Constitution, art. 29). Within this framework, the implementation and enforcement of environmental laws at municipal level are constitutionally legitime function of Brazilian democracy. However, effective environmental governance at the local level requires more than formal autonomy. It demands institutional mechanisms, locally enacted legislation, and a comprehensive set of enforcement and regulatory procedures that enable municipalities to effectively exercise their responsibilities — particularly with regard to environmental licensing. This need becomes even more critical as municipalities are expected to assume greater responsibilities due to projected legal reforms introduced at the federal level. As evidenced in this study, the expectation that such a system can function effectively across Brazil presumes a level of uniformity that does not reflect reality. The country’s extreme heterogeneity in terms of administrative capacity, legal infrastructure, and institutional maturity in the environmental governance poses significant challenges to implementing this vision. What may appear to be technically sound and democratic process may, in practice, encounter serious limitations that jeopardize its effectiveness and legitimacy.
The geographical asymmetry regarding environmental governance and deep socio-economic heterogeneity, combined with the overall poor environmental institutional capacity of Brazilian municipalities, underscores a critical challenge for the ongoing decentralization of environmental governance. As highlighted by Sands et al. (2012) from the perspective of non-implementation of national environmental laws, the mere existence of laws — and, by extension, broader environmental governance frameworks — often prove insufficient when countries lack adequate institutional and human resources, suffer from legal ambiguity that opens space for misuse or misinterpretation, and face structural difficulties in meeting the demands of numerous actors involved. This resonates directly with the fragile environmental institutional capacity observed in Brazilian municipalities. More seriously, as emphasized by Kassie (2024), non-compliance driven by policy ambiguity, conflicting interests, and the absence of binding legal obligations becomes an imminent risk.
Notably, however, our approach reveals a positive relationship between the proportion of population exposed to socio-environmental risks — such as droughts, floods, or deforestation — and higher values of ICI. This suggests that in municipalities where populations are more vulnerable to environmental hazards, there may be better institutional mobilization or investment in governance structures. Yet, this adaptive response appears to be uneven and insufficient at the national scale. The new law, which also introduces mechanisms such as self-declaratory environmental licenses (i.e., License by Adhesion and Commitment) and automatic license renewals without prior review by competent authorities, notably omits climate adaptation instruments, elements that are essential for supporting climate change response at the subnational level. Without integrating such tools, substantial investments, and targeted public policies aimed at strengthening local institutional capacities, particularly in the most vulnerable and ecologically critical regions, such as the Amazonian and Caatinga related states, the decentralization process risks further exacerbating existing territorial inequalities and severely undermining the overall effectiveness of environmental governance nationwide.
Before embracing a sweeping decentralization of environmental licensing, Brazil could benefit from learning from countries such as India and Indonesia, where decentralization has advanced considerably. These decentralization reforms, each characterized by distinct institutional architectures and implementation trajectories, offer valuable lessons for Brazil’s environmental licensing reform. In India, despite constitutional mandates, decentralization has faced structural challenges, including mismatches between responsibilities and local capacities, with outcomes varying according to regional governance regimes (Véron et al., 2024). In Indonesia, decentralization has enabled subnational innovation in climate governance, particularly in land-use sectors, yet has also introduced policy fragmentation and perverse incentives that undermine mitigation goals (Di Gregorio and Moeliono, 2023). In sum, both India and Indonesia show that decentralization yields benefits only when subnational mandates are matched with resources and autonomy, via ring-fenced funds and protected space for local innovation. For Brazil, the most relevant lesson is the need to front-load targeted funding and accountability to build municipal capacity prior to extending authority to licensing complex, high-impact projects. These international experiences underscore the importance of balancing local autonomy with strong institutional safeguards. Without such an equilibrium, Brazil risks replicating similar pitfalls — fragmented authority, uneven implementation, and ultimately weakened environmental governance in the face of mounting socio-environmental challenges.
It is essential to adopt a historical and contextual perspective to adequately assess the capacity of local governments to assume this new frontier of environmental licensing. More than mere political-administrative units, Brazil’s 5,570 municipalities represent complex territorial, social, and historical realities. Their economic, social, demographic, extension, and particularly environmental characteristics form an extraordinarily diverse mosaic across the country (Batista et al., 2020). It is therefore critical to evaluate whether regions already experiencing intense agri-business expansion (Arias et al., 2017), or a proliferation of small-hydroelectric power plants (see Reis-Filho and Leduc, 2024), possess institutional capacity — based on the attributes examined in this study — to ensure environmentally sound governance in licensing processes. Specifically, these two types of developments are expected to have their environmental regulation relaxed under the new environmental reform in Brazil. Although the presidential sanction introduced partial vetoes to remove unconstitutional provisions (Fernandes et al., 2025), there is an active movement in Congress to overturn them, largely driven by the Agriculture and Ranching Parliamentary Front (FPA, 2025), whose legislative agenda continues to promote regulatory flexibility in favor of agribusiness interests. Additionally, our findings indicate that larger municipalities often exhibit lower levels of environmental governance and institutional capacity. This suggests that greater territorial extension, particularly when combined with intensive land use pressures such as agribusiness or hydropower expansion, may introduce logistical, operational, and coordination challenges that further compromise environmentally sound licensing practices. Moreover, as highlighted by Batista et al. (2020), transparency and accountability mechanisms remain fragile across Brazilian municipalities, with low adherence to freedom of information laws and limited association between transparency and effective government performance. These findings underscore that institutional weaknesses go beyond environmental governance alone. Thus, any move to decentralize environmental licensing must contend not only with uneven technical and administrative capacity, but also with broader structural challenges related to governance quality. Strengthening local institutions must therefore encompass not just technical and legal dimensions, but also measures to promote transparency, curb mismanagement, and enhance accountability—conditions that are indispensable for safeguarding environmental integrity under a decentralized regulatory regime (Orsini et al., 2013).
Interestingly, based on our analysis focused on the Environmental Institutional Capacity Index (ICI), agro-pastoral use areas were not significantly associated with the ICI in the model (either negatively or positively). It is important to note that SHAP values in this context indicate statistical associations embedded within the model structure rather than causal relationships; they reveal how structural and socioeconomic conditions co-vary with institutional capacity, not how they directly determine it. Beyond the interpretation of the relationships between the drivers of environmental institutional capacity and their relative importance in explaining model outputs (see Fig. 6), it is crucial to highlight the pervasive influence of agribusiness on Brazilian politics and how it has shaped the trajectory of domestic environmental policy (Amorim et al., 2023). The purely economic and expansionist interests of this sector have, for decades, guided, influencing, and often overridden the balance between development and environmental protection in the country (Rochedo et al., 2018). This dynamic urgently requires deeper investigation within the broader context of environmental governance and institutional capacity at the municipal level. It involves examining how local governments balance multiple regulatory frameworks in their daily operations, the extent of institutional discretion, and the bureaucratic complexities involved in pursuing sustainability and environmental protection goals. However, this normative vision appears far removed from the current reality. A considerable proportion of Brazilian municipalities, particularly in certain regions, still lack even the most basic policy instruments and operational governance structures necessary to support effective environmental licensing. In this context, the political and economic power of agribusiness further exacerbates the historical tension between the paradigm of “standing forests” and the relentless expansion of pastures and croplands. As projected, the new environmental law, which promote regulatory and institutional flexibility in Brazil, must be evaluated beyond the initial stages of environmental impact assessment (i.e., licensing procedures). Even in developed countries with historically well-enforced environmental governance, such as Sweden, regulatory flexibilization has emerged as a key challenge for managing complexity and systemic risk (Dawson et al., 2017). How, then, can one expect a country like Brazil to reconcile such anachronistic and deregulatory proposals with its urgent need for effective, coherent, and science-based environmental protection? This is especially concerning given Brazil’s socio-ecological relevance — from the unparalleled biodiversity of Amazon rain forest to the socio-ecological importance of traditional communities, including the largest population of traditional small-scale fishers in the Americas (Cavole et al., 2025).
This contradiction becomes even more striking given Brazil’s internal history of tensions involving top-down environmental regulations, conservation actors, scientific research, and various segments of society (Costa Neves, 2016). These tensions have often culminated in regulatory rollbacks, such as the controversial revision of the Brazilian Forestry Code (Law No. 12,651/2012), which weakened restriction on land use and forest preservation in rural areas. As our multidimensional analysis shows, expecting this General Environmental Licensing Law to simultaneously streamline licensing procedures and guarantee environmental safeguards in not only unrealistic ones, but it is also fundamentally contradictory. This concern is even more pressing given that Brazil is set to host the COP30 in 2025, a global climate summit organized under the United Nations Framework Convention on Climate Change, where countries are expected to demonstrate bold and credible commitments to climate action. Moreover, Brazil has recently approved a Bill for the Law of Sea (Bill No. 6,969/2013), which establish a national policy for the sustainable use and protection of marine resources under its jurisdiction. The law aims also to align Brazilian marine governance with international standards and biodiversity commitments. Taken together, these developments underscore another deeper question: what kind of nation does Brazil aspire to be, one that stands as a global reference for conservation, fair and science-based environmental regulation, or one where developmental interest operate in isolation from ecological stewardship?
5.1. Recommendations
Our reflections align with recent investigations on factors that drive municipalities to adopt environmental legislation and establish regulatory agencies in Brazil (Marenco and Kern, 2025). The presence of local environmental councils and environmental secretariats increases the likelihood of adopting environmental legislation. In view of the risks posed by the new environmental legal reform, our recommendations emphasize intergovernmental diffusion of environmental preparedness and municipal capacity-building, with priority to municipalities exhibiting the lowest ICI scores and belonging to sensitive biomes. Concretely, this entails: (i) enhanced staffing, training, and retention; (ii) a national training and certification program for licensing staff, paired with technical assistance hubs at state/federal levels; (iii) standardized procedures (screening, terms of reference, alternatives analysis, and climate-risk integration); and (iv) minimum national standards with monitoring and conditional transfers, linking resources to performance and periodic audits. These steps should be front-loaded before delegating complex licensing to municipalities under this legislative reform.