Biodiversity for a Livable Planet: An Evaluation of World Bank Group Support for Biodiversity, Fiscal Years 2010–2024

Appendix A. Evaluation Methods

This appendix explains the methodological approach used in this evaluation. It includes a description of the evaluation design and questions, scope, portfolio identification and description, methods applied, and limitations.

Evaluation Design and Questions

This evaluation asked the overarching question: How well is the World Bank Group supporting clients to address biodiversity loss? This question was examined through two main evaluation questions (EQs):

EQ1: How well is the World Bank addressing biodiversity challenges through conservation-focused activities?
EQ2: How well are the World Bank and the International Finance Corporation (IFC) supporting activities with potential biodiversity benefits in key production sectors, and are those activities likely to achieve such benefits?

A third EQ—How well is the Bank Group supporting clients to manage risks affecting biodiversity at the project level?—was subsumed within the analyses of EQ1 and EQ2 and is also addressed through a dedicated analysis of biodiversity offsets that covers the World Bank, IFC, and the Multilateral Investment Guarantee Agency (MIGA).

To answer these questions, we triangulated findings generated through a set of mixed methods, both qualitative and quantitative. The questions were also further subdivided into a set of subquestions, as per the Approach Paper.

The subquestions for EQ1—on conservation-focused activities—are as follows:

EQ1a. How well is the World Bank applying good practice approaches in its biodiversity conservation activities?
EQ1b. To what extent are biodiversity conservation activities designed to leverage the World Bank’s advantages?
EQ1c. How well are biodiversity projects achieving their biodiversity goals?
EQ1d. How well are biodiversity projects articulating and achieving their multiple benefits (economic, development, climate)?

The subquestions for EQ2—on integrating biodiversity into key production sectors—are as follows:

EQ2a. What has worked to enable the integration of activities with biodiversity benefits in engagements in key production sectors—in the Bank Group, with clients, and with resource users?
EQ2b. Do projects with potential biodiversity benefits include evidence on proxies for biodiversity benefits, and are they achieving those proxies?
EQ2c. How have activities with potential biodiversity benefits contributed to climate change benefits?

The subquestions for EQ3—on biodiversity risk management—are as follows:

EQ3a. How well have biodiversity risk management policies been used to inform the design and support the effective implementation of projects that could have an adverse effect on biodiversity?
EQ3b. To the extent that evidence is available, has the application of biodiversity-related risk management policies mitigated biodiversity loss?

Figure A.1 presents the evaluation design. For EQ1, which assesses the extent to which the World Bank is achieving biodiversity outcomes through conservation-focused activities, methods include focused literature reviews, strategy reviews, portfolio review and analysis, geospatial analysis, and deep dives on (i) Indigenous Peoples and local communities (IPLCs), (ii) land and natural resource rights (LNR), and (iii) biodiversity offsets. For EQ2, which examines how well the World Bank and IFC are supporting biodiversity outcomes in key production sectors, the methods include focused literature reviews, portfolio reviews and analyses, and exploratory case studies. EQ3, on how well the Bank Group is supporting clients to manage risks affecting biodiversity at the project level, was addressed through the dedicated analysis of biodiversity offsets (covering the World Bank, IFC, and MIGA), the integration of environmental and social risk-related analysis into the portfolio review and deep dives under EQ1, and the sectoral portfolio and case study reviews under EQ2. This appendix also describes the methodology used to analyze core country-level diagnostics—Country Partnership Frameworks (CPFs) and Country Climate and Development Reports (CCDRs).

Source: Independent Evaluation Group.

Note: EQ3—on how well the World Bank Group is supporting clients to manage risks affecting biodiversity—was integrated into the analyses of EQ1 and EQ2. Specifically, EQ3 was examined through a dedicated analysis of biodiversity offsets (covering the World Bank, IFC, and MIGA), and integrated into the portfolio review and deep dives under EQ1 and the portfolio and case study reviews under EQ2. CCDR = Country Climate and Development Report; CPF = Country Partnership Framework; EQ = evaluation question; IFC = International Finance Corporation; IPLC= Indigenous Peoples and local communities; LNR = land and natural resource rights; MIGA = Multilateral Investment Guarantee Agency; PRA = portfolio review and analysis.

Evaluation Scope

This evaluation was scoped along three dimensions: time frame, institutional coverage, and level of engagement.

Time frame. The evaluation includes World Bank, IFC, and MIGA projects approved during FY15–24, except for the conservation-focused portfolio, which includes projects approved from FY10–24. This extended time frame was used to ensure the inclusion of closed projects with evaluative evidence on results, as EQ1 examines both the evolution of the World Bank’s engagement in conservation and the extent to which environmental and development results were achieved.

Institutional coverage. As outlined in the Approach Paper, EQ1 on conservation-focused activities covers the World Bank only, as IFC and MIGA do not implement biodiversity conservation operations. EQ2 on biodiversity integration into key production sectors covers both the World Bank and IFC. EQ2 excludes MIGA because the evaluation scope, as per the Approach Paper, did not include these activities. MIGA may undertake some relevant activities, but consultations with MIGA during the Approach Paper stage suggested that these are relatively few, and the evaluation might not add much value by covering them. The biodiversity offsets analysis includes all three Bank Group institutions: the World Bank, IFC, and MIGA.

Engagement level. To manage scope, the evaluation focuses on national-level issues rather than the global convening power of the World Bank (for example, its activities at the Conferences of the Parties), as specified in the Approach Paper. Evaluating the Bank Group’s global convening efforts would have required a distinct methodological approach. Lessons on convening were also generated by a separate Independent Evaluation Group (IEG) evaluation of the Bank Group’s global convening efforts in 2020. In addition, the evaluation does not assess IFC’s Biodiversity Finance Reference Guide, which was launched alongside the evaluation and is too recent to evaluate.

Evaluation Portfolio Identification and Classifications

This section has two parts: the first describes the process used to identify the portfolios included in the evaluation for each EQ, and the second provides a description of these portfolios.

EQ1. Biodiversity Conservation

World Bank

First, we identified the universe of potentially relevant projects. To identify the World Bank’s conservation-focused portfolio (FY10–24), we first began by identifying projects tagged with the biodiversity theme code (theme code 834). Using this output, we manually screened projects (that is, their project development objectives, component titles, and indicator titles) to develop a search taxonomy of biodiversity-related terms, as outlined in box A.1. We then used this taxonomy for text mining to supplement the theme code search to ensure comprehensiveness. To do this, a string search was conducted in key text descriptions of projects (that is, project titles, project development objectives, key lending project document abstracts, project descriptions, activity summaries, component titles, component descriptions, and indicator titles).

Box A.1. Search Taxonomy Used for Text Mining to Identify Conservation-Focused Portfolio

biodiversity, biological corridor, biological divers, conservation area, conservation corridor, critical habitat, ecological corridor, ecosystem value, fauna, flora, marine reserve, natural capital, natural habitat, payment for ecosystem service, payment for environmental service, payments for ecosystem service, payments for environmental service, poaching, protected area, specie, WAVES, wildlife.

Source: Independent Evaluation Group.

Next, to determine the relevant in-scope portfolio from the universe of potentially relevant projects, we used AI-assisted manual screening supported by an off-line, open-source, Mistral 7b model running on a World Bank power desktop machine with an NVIDIA GPU card.

We developed specific prompts to categorize projects as in or out of scope based on a set of instructions and examples. We developed three prompts tailored to the three lending instruments, and for each lending instrument, we used different text fields (that is, project development objectives, components, and indicators for investment project financing; project development objectives and disbursement-linked indicators for Program-for-Results; and project development objectives and prior actions for development policy operations). The prompts and text data were fed to the model systematically for the projects for which all necessary text data were available. The model was instructed to provide an “in/out” categorization for each project, along with a brief explanation—grounded in the data—for its decisions. The model’s generation parameters and the prompts were optimized for accuracy (as opposed to creativity) through iterative testing with examples.

This preliminary AI-assisted categorization served to efficiently narrow the pool of projects by identifying those that potentially met or did not meet our evaluation scoping criteria. Subsequently, we conducted a manual verification to ensure the accuracy of the AI categorization and to adjust for any nuances or specific details the model might have missed. The manual verification involved reviewing the relevant text, including project development objectives, components, and indicators for each of the projects identified by the AI categorization process, to ensure that false positives were omitted from the portfolio. To reduce the chances of false negatives occurring, the search taxonomy used (see box A.1) was deliberately broad, with spot checks being conducted on projects to reduce the likelihood of relevant projects being omitted from the portfolio. This blended approach helped achieve both efficiency and thoroughness in our identification process.

During the portfolio review and analysis process, the portfolio was further refined to exclude false positives and negatives, with projects omitted or included based on the results of a detailed review and analysis of project documents including the Project Appraisal Documents (PADs) and Implementation Completion and Results Reports (ICRs).

International Finance Corporation and Multilateral Investment Guarantee Agency

As per the evaluation Approach Paper, IFC, and MIGA were not included in the EQ1 portfolio because they do not undertake conservation-focused activities.

EQ2. Biodiversity Integration in Key Production Sectors

This section outlines the methods used to identify the portfolio of World Bank and IFC projects with relevant production sector activities. As per the Approach Paper, the analysis focused on three sectors: (i) agriculture and agribusiness, (ii) forests, and (iii) fisheries and aquaculture.

World Bank Lending

Agriculture. To conduct the analysis of World Bank agriculture projects, we identified the universe of projects approved by the Agriculture and Food Global Practice during the evaluation time frame (FY15–24).

Forests. Because forest-related projects are financed across Global Practices, we used relevant sector codes (AT—Forestry; AK—Public Administration—Agriculture, Fishing, and Forestry; AZ—Other Agriculture, Fishing, and Forestry) and theme codes (831—Forests Policies and institutions) to identify the forests portfolio. We also pulled all projects that included relevant keywords (that is, forest, wood, timber) in their project name, project development objective, component title, and indicators. Using this combined output, we manually screened projects (that is, their project names, project development objectives, component titles, and indicator titles) to identify those with in-scope activities such as forest carbon and forest production.

Fisheries and aquaculture. To identify the fisheries and aquaculture portfolio, we used relevant sector codes (AF—Fisheries; AZ—Other Agriculture, Fishing and Forestry) and theme codes (832—Fisheries Policies and institutions). Using this output, we manually screened projects (that is, their project development objectives, component titles, and indicator titles) to identify those with in-scope activities such as fisheries management and aquaculture.

International Finance Corporation Investments

Step 1: Data extraction. We extracted project text data from the IFC Disclosure Portal for projects approved during the evaluation period covering the following: (i) project description, (ii) overview fund, (iii) risk impact, (iv) ESAP, (v) review scope, (vi) environmental social information, (vii) impact, (viii) result, (ix) risk assessment, (x) role, (xi) contribution, (xii) environmental social issues, (xiii) sponsor, (xiv) cost nature, (xv) investment, (xvi) location, (xvii) environmental social categorization rationale, (xviii) risk fund, (xix) risk impact, (xx) stakeholders, and (xxi) mitigation measures.

Step 2: Data consolidation. We consolidated the text data, originally stored in multiple columns, into a single comprehensive text field for each project.

Step 3: Keyword search. Using R, we conducted a search for predefined biodiversity-related keywords and phrases across all the text columns mentioned in step 1 within the aggregated text. The list of keywords can be found in box A.2.

Box A.2. Additional Keyword Taxonomy Used for Text Mining to Identify International Finance Corporation Portfolio

climate smart ag; climate-smart ag; deforestation; Eucalyptus; fish; forest; good agricultural practice; good agriculture practice; Pulp mill; smallholder; sustainable ag; traceability.

Source: Independent Evaluation Group.

Step 4: Initial project inclusion. Projects that contained at least one occurrence of the keywords and phrases were included in our initial longlist.

Step 5: Sector-based inclusion. In addition to keyword-triggered projects, we included projects from relevant sectors in our longlist. The list of relevant sectors is available in table A.1 below.

Step 6: Manual preliminary screening (i): intercoder reliability exercise. We manually reviewed the text column of projects in the longlist, focusing on those that proactively adopted sustainable measures benefiting biodiversity in land- or seascape production sectors. To ensure consistency, the team conducted a simultaneous review of 10 randomly selected projects, comparing results to ensure intercoder reliability.

Step 7: Manual preliminary screening (ii): coding and validation. After ensuring consistency among coders through the intercoder reliability exercise, we divided the longlist among team members. Each member coded the projects assigned to them.

Step 8: Final selection. After the manual validation from steps 6 and 7, we identified 87 investment projects out of the initial 948 projects in the longlist.

International Finance Corporation Advisory

Step 1: Data extraction. We extracted text data for advisory projects approved during the evaluation period, covering the following: (i) objectives statement, (ii) statement of market failure, (iii) statement of market failure original, (iv) strategic relevance, (v) expected development impact for public disclosure, (vi) project description for public disclosure, (vii) project description, (viii) IFC role and additionality, (ix) context, (x) upstream comments, and (xi) MFD comments.

Step 2: Data consolidation. We consolidated the text data, originally stored in multiple columns, into a single comprehensive text field for each project.

Step 3: Keyword search. Using R, we conducted a search for predefined biodiversity-related keywords and phrases within the aggregated text. The list of keywords can be found in box A.1.

Step 4: Initial project inclusion. Projects that contained at least one occurrence of the keywords or phrases were included in our initial longlist.

Step 5: Primary business line-based inclusion. In addition to keyword-triggered projects, we included projects from relevant primary business lines in our longlist. The list of relevant primary business lines is provided in table A.1.

Step 8: Final selection. After the manual validation from steps 6 and 7, we identified 63 advisory projects out of the initial 218 projects in the longlist.

Table A.1. International Finance Corporation Business Lines and Sector Codes Used to Identify Evaluation Question 2 Portfolio

Business Line for IFC Advisory Projects

Tertiary Sector Name for IFC Investment Services

ESG

ESG—E&S Risk Management

ESG—Integrated ESG and Sustainability

MAS

MAS—Agri Expertise

MAS—Agribusiness: Industry Standards

MAS—Agribusiness: SME Productivity

MAS—Agribusiness: Strategic Community Investment

MAS—Agribusiness Other

MAS—Crop Production

MAS—Sustainable Protein

MAS—Forestry and Land Uses

Animal Aquaculture

Cattle Farming

Coffee, Cocoa, Tea

Dairy Products

Diversified Edible Agricultural Crops Production

Fishing

Fruits and Vegetables

Horticultural Products (Flowers)

Grains and Beans

Natural Fibers (Cotton, Sisal, Jute, and so on)

Other Animal Production

Other Vegetable Oil Crops (Coconut, Rapeseed, Peanut, Sunflower, and so on)

Palm Oil Plantations

Palm Vegetable Oil

Paperboard (Including Boxboard, Fiberboard)

Plantation Forests

Poultry Farming

Sugarcane and Beets

Source: Independent Evaluation Group.

Note: E&S = environmental and social; ESG = environmental, social, and governance; IFC = International Finance Corporation; MAS = Manufacturing, Agribusiness, and Services; SME = small and medium enterprise.

Biodiversity Offsets

The biodiversity offsets portfolio across the World Bank, IFC, and MIGA was identified through a combination of counterpart collaboration and systematic portfolio reviews.

World Bank. For the World Bank, 16 projects under ESF and 5 projects under the safeguard policies were identified as having offsets. The ESF projects were identified using the ESF 5.0 Master Report by filtering for projects that marked biodiversity offsets as relevant. Safeguard policies projects were identified through a structured document review process, leveraging AI-assisted keyword searches to detect offset-related language in PADs and Integrated Safeguards Data Sheets, followed by a manual verification by the evaluation team. The manual verification process involved reviewing the text excerpts identified by the keyword search tool to ensure projects that referred to offsets or other relevant terms in project documents included offset activities within the project’s scope. False positives, such as instances where offsets or related terms were mentioned in a cursory fashion or in an irrelevant context, were screened out. The final list was also validated in semistructured interviews with environmental and social specialists during data collection activities.

IFC. The IFC environmental, social, and governance (ESG) team provided a list of 22 active and 7 closed projects approved within the evaluation timeline (FY15–24) that include biodiversity offsets as part of Performance Standard on Environmental and Social Sustainability (PS6) requirements. The evaluation team requested no-objection from investment officers and managers to access project documents. We received a few questions on how the project information would be used and clarified that the evaluation would present only portfolio-level trends and findings, with no reference to specific projects or partners. After this no-objection process, the IT/ESG360 team shared a bulk download of the available environmental and social files from the ESG dashboard. Several environmental and social documents related to biodiversity offsets were not available in the ESG dashboard, and the team requested these directly from project teams in coordination with the IFC ESG team. The evaluation team followed IFC’s privacy protocols throughout the evaluation process to ensure the confidentiality of sensitive information, including by following established guidelines on data protection, and ensured that all data were handled securely.

MIGA. MIGA counterparts provided a list of 44 active and 12 terminated projects that applied PS6. The evaluation team identified 12 projects that involved biodiversity offsets. The final list was confirmed by MIGA’s ESG team.

Portfolio Description

EQ1. Biodiversity Conservation

There were 139 World Bank lending projects (and 23 additional financing projects) approved between FY10 and FY24 within the scope of biodiversity conservation activities, of which 47 are active and 92 are closed. Most of these projects are financed by the Environment, Natural Resources, and Blue Economy Global Practice (n = 114; 83 percent). Of the 92 closed projects, 73 projects have ICRs, of which 62 have Implementation Completion and Results Report Reviews (ICRRs), according to data from the IEG DataHub (as of March 2025). Eighty-four percent of the closed and validated projects were rated moderately satisfactory or higher. A breakdown of the portfolio is provided in figures A.2–A.4.

Source: Independent Evaluation Group.

Note: AGR = Agriculture and Food; ENB = Environment, Natural Resources, and Blue Economy; MTI = Macroeconomics, Trade, and Investment; TRA = Transport; URL = Urban, Disaster Risk Management, Resilience, and Land; WAT = Water.

Source: Independent Evaluation Group.

Note: EAP = East Asia and Pacific; ECA = Europe and Central Asia; ESA = East and Southern Africa; LAC = Latin America and the Caribbean; MENA = Middle East and North Africa; SA = South Asia; WCA = West and Central Africa.

Source: Independent Evaluation Group.

EQ2. Biodiversity Integration in Key Production Sectors

World Bank Lending

Agriculture. There were 247 World Bank lending projects financed by the Agriculture Global Practice approved between FY15 and FY24. Of these, 145 are active and 102 are closed. Of the 102 closed projects, 72 projects have an ICR, of which 53 have an ICRR (data from IEG DataHub as of March 2025). Eighty-seven percent of the closed and validated projects were rated moderately satisfactory or higher. A breakdown of the portfolio is provided in figures A.5–A.7.

Source: Independent Evaluation Group.

Note: DPF = development policy financing; IPF = investment project financing; PforR = Program-for-Results.

Source: Independent Evaluation Group.

Note: MENA = Middle East and North Africa.

Source: Independent Evaluation Group.

Forestry. There were 83 World Bank lending projects approved between FY15 and FY24 with in-scope forest production activities, of which 41 are active and 42 are closed. Most of these projects (n = 66; 80 percent) are financed by the Environment, Natural Resources, and Blue Economy Global Practice. The remainder are divided among the Agriculture and Food (n = 9; 11 percent); Macroeconomics, Trade, and Investment (n = 5; 6 percent); Energy and Extractives (n = 2; 2 percent); and Poverty (n = 1; 1 percent) Global Practices. Of the 42 closed projects, 35 projects have ICRs, of which 28 have ICRRs (data from IEG DataHub as of April 2025). Eight-six percent of the projects were rated moderately satisfactory or higher. A breakdown of the portfolio is provided in figures A.8–A.10.

Source: Independent Evaluation Group.

Note: AGR = Agriculture and Food; EAE = Energy and Extractives; ENB = Environment, Natural Resources, and Blue Economy; MTI = Macroeconomics, Trade, and Investment; POV = Poverty.

Source: Independent Evaluation Group.

Note: AFE = East and Southern Africa; AFW = West and Central Africa; EAP = East Asia and Pacific; ECA = Europe and Central Asia; LCR = Latin America and the Caribbean; MNA = Middle East and North Africa; SAR = South Asia.

Source: Independent Evaluation Group.

Fisheries and aquaculture. There were 51 World Bank lending projects approved between FY15 and FY24 with in-scope fisheries activities, of which 26 are active and 25 are closed. Most of these projects (68.6 percent, n = 35) are financed by the Environment, Natural Resources, and Blue Economy Global Practice, while 20 percent (n = 10) are financed by the Agriculture and Food Global Practice. The remainder are divided among the Macroeconomics, Trade, and Investment (n = 3), Water (n = 2), and Finance Competitiveness and Innovation (n = 1) Global Practices, respectively. Of the 25 closed projects, 20 projects have ICRs, of which 18 have ICRRs (data from IEG DataHub as of March 2025). Fifty percent of the projects were rated moderately satisfactory or higher. A breakdown of the portfolio is provided in figures A.11–A.13.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

International Finance Corporation Investments and Advisory

Agriculture. During FY15–24, IFC approved 107 investments in the agricultural crop production and animal production (excluding aquaculture and fishing) sectors. Of these, 60 projects are active and 47 are closed. IEG validated the results of 21 of these investment projects. Moreover, there were 37 advisory projects in these sectors, with implementation ongoing for 17 projects and completed for 20 projects. Among the completed projects, IEG has validated results for 9 projects.

Forestry. Between FY15 and FY24, IFC approved 15 investments in the forestry sector. These included 4 investments in plantation forests, 8 in pulp and paper manufacturing, and 3 in wood panel/wood product manufacturing. Of these, IEG has evaluated 2 investments, both of which were in the wood panel and wood product and pulp and paper manufacturing sectors. Additionally, there was one IFC advisory project mapped to the plantation forests (90 percent) and paperboard (10 percent) sectors, which has not been validated by IEG. No other advisory projects were identified in the pulp, paper, or wood manufacturing sectors.

Fisheries and aquaculture. During FY15–24, IFC approved eight investments in the fisheries and aquaculture sector. This included seven investments in aquaculture and one in fishing. Of the seven aquaculture projects, five were in Ecuador and were covered in our Ecuador case study (refer to the exploratory case study section). The remaining two aquaculture projects were in China and focused on aquafeed production. The single fishing investment was in the Solomon Islands, where IFC financed the purchase of an additional fishing vessel, which triggered PS6. Additionally, there were three IFC advisory projects mapped to the aquaculture and fishing sectors. IFC self-evaluated these advisory projects at completion, but only one of them has been validated by IEG.

Biodiversity Offsets

World Bank. For the World Bank, 16 projects under ESF and 5 projects under the safeguard policies were identified as having offsets. A breakdown of the portfolio is provided in figures A.14–A.16.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

Note: AFE = East and Southern Africa; AFW = West and Central Africa; EAP = East Asia and Pacific; ECA = Europe and Central Asia; ESA = East and Southern Africa; LCR = Latin America and the Caribbean; MENA = Middle East and North Africa; SAR = South Asia; WCA = West and Central Africa.

IFC. For IFC, 29 projects were identified as having offsets as part of PS6 requirements (22 active and 7 closed) approved within the evaluation timeline (FY15–24). A breakdown of the portfolio is provided in figures A.17–A.18.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

MIGA. Twelve MIGA projects in scope involved biodiversity offsets. A breakdown of the portfolio is provided in figures A.19–A.21.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

EQ1 Methods: How Well Is the World Bank Achieving Biodiversity Aims Through Its Conservation-Focused Activities?

To assess the relevance and effectiveness of World Bank support for biodiversity conservation, we used several triangulated methods, including literature review, portfolio review and analysis, the geo-location of conservation activities, and three thematic deep dives. We expand on these methods below.

Focused Literature Review and Review of World Bank Group Strategies and Commitments

First, we identified good practice conservation approaches enshrined in the global biodiversity conventions and associated targets, which were developed based on scientific literature. This included (i) the Convention on Biological Diversity (CBD) adopted in 1992; (ii) the Aichi Biodiversity Targets 2011–2020; and (iii) the Kunming-Montreal Global Biodiversity Framework (GBF) adopted in 2022, and associated targets. In total, 196 countries are members of the CBD, signifying near-universal participation among nations, which is also the case for the Kunming-Montreal GBF.

Second, we identified the Bank Group’s advantages by reviewing its strategies and corporate commitments. This included (i) the World Bank’s Environment Strategy (2012–22), which lays out the World Bank’s commitments for supporting client countries to conserve and restore critical biodiversity, and (ii) the 2021 Approach Paper entitled Unlocking Nature-Smart Development: An Approach Paper on Biodiversity and Ecosystem Services, which placed a renewed emphasis on biodiversity and nature. We also reviewed the commitments enshrined in the World Bank’s Global Challenge Program Forests for Development, Climate, and Biodiversity, which seeks to scale sustainable forest landscapes and ecosystem solutions to enhance development, climate, and biodiversity outcomes.

Portfolio Review and Analysis

The literature review yielded five themes that reflect a combination of global best practice with the World Bank’s comparative advantage. The themes identified were as follows:

Geo-representativeness, or the equitable and proportional representation of different biomes, ecosystems, and taxonomic groups. Analyzing the portfolio through the lens of geo-representativeness is important for determining the relevance of World Bank conservation activities in relation to the protection and conservation of critical biomes and areas of high endemism.
Ecological connectivity, or the unimpeded movement of species and flow of natural processes that sustain life across ecosystems. Research indicates that enhancing the connectivity of protected areas and conservation areas is critical in improving genetic diversity, enabling the movement of wildlife species and enhancing ecosystem resilience through the reduction of habitat fragmentation.
Ecological monitoring and reporting of biodiversity outcomes. Quantitative ecological monitoring is crucial for understanding biodiversity outcomes. It is a priority within the GBF, providing essential data to assess and guide conservation actions.
Engagement and protection of IPLCs. The World Bank’s biodiversity Approach Paper highlights the essential role of IPLCs in achieving biodiversity goals. Local communities have deep, place-based ecological knowledge developed over millennia, enhancing biodiversity preservation. Good governance, including land tenure security, equitable benefit sharing, and preserving local knowledge, is crucial for their contribution to effective conservation.
Sustainable financing. Sustainable financing is crucial for biodiversity conservation because it ensures long-term funding to protect ecosystems, species, and natural resources. Many conservation initiatives face financial shortfalls, leading to habitat degradation, species loss, and reduced ecosystem services.

These five themes were incorporated into a portfolio coding protocol that also included basic project data, project theory, and other relevant variables. The result of this process was a detailed codebook that included all underlying portfolio variables, any relevant extracts from within project documents relevant to those themes, available results and indicator data for each project, and a series of flags for emergent issues identified in inductive coding. The enumerated parameters included in the coding protocol are outlined in box A.3.

Box A.3. Coding Protocol for Conservation-Focused Projects

Basic project data

Project ID, name, country, source of finance, amount of finance

Project development objective

Project components

Project indicators

Intended beneficiaries, also disaggregated

Location of biodiversity areas (if possible)

Project lessons

Project theory as it pertains to biodiversity aims

Stated theory or implicit theory

Explicit or implied outcomes not included in the results framework

Identification of good practice approaches, including measurement thereof

Ecological connectivity (for example, corridors, transfrontier areas)

Use of climate science, or other scientific evidence to make conservation planning decisions

Use of Indigenous knowledge to make conservation planning decisions

Sustainable financing at the national and park level (for example, recurrent finance/national budget; endowment funds; payment for environmental service)

Inclusion of Indigenous Peoples and local communities

Support for land tenure security, access, or resource-use rights

Use of relevant ecological monitoring techniques and indicators (for example, management effectiveness tracking tools, quantitative monitoring of species or species markers, remote sensing/habitat assessments using satellite/LIDAR)

Other (open cell for observations)

Source: Independent Evaluation Group.

This coding protocol was then used as the basis for an in-depth portfolio review exercise across each variable, with each following a similar approach of keyword search, extraction, and content analysis.

Keyword search and extraction. To test the keyword search and extraction process, the team undertook a pilot exercise with a selection of project documents (PADs, ICRs) from a subset of the portfolio to determine the suitability of using a proprietary keyword search tool developed by the evaluation team. The tool is a script, built with a Python backend, and a graphical user interface, built with React, that allows for systematic searches for keywords using defined parameters within a collection of PDF documents. The tool is based on both the String-Similarity JavaScript library and proprietary searching and matching processes to account for fuzzy matching (where a specified percentage of characters in a word needs to match the chosen keyword to be considered a keyword mention) and intraword matching. The other features of the tool include data visualization, Excel exporting, and PDF viewing.

The tool allowed for a parametric search based on identified keywords, including exact text, whole words, and proximity words, extracting all relevant data from the documents uploaded. The excerpts extracted by the tool were then exported to an Excel file and organized into a matrix that was combined with the portfolio codebook variables in box A.3. As a control measure, a manual review of the same subset of projects was undertaken using typical qualitative coding and extraction, with the results of the two exercises being compared. This allowed for the refinement of the parameters included in the keyword search tool.

Having undertaken this pilot, we built out a finalized series of keywords for each thematic area. These keywords were searched for in PADs or equivalents (for example, program documents) and ICRs or equivalents of all 139 projects in the conservation portfolio using a proprietary keyword search and extraction tool developed by the evaluation team. The excerpts for each thematic area were exported to Excel and consolidated into a comprehensive portfolio matrix.

For projects with completed and validated ICRs, the extraction and analysis of results indicators was undertaken. Where relevant, data from the geospatial analysis on variables such as tree cover change or crop change were also extracted.

Content analysis. The consolidation of the above data into a comprehensive portfolio matrix allowed for detailed quantitative and qualitative analysis that ultimately comprised:

The relevance of the portfolio’s activities in each thematic area in terms of alignment to best practice, the World Bank’s areas of comparative advantage, and strategic priorities.
The effectiveness of included activities for closed projects in each thematic area, based on the achievements of relevant intermediate results indicators and project development objectives and any explanatory factors included in project documents (ICRs, ICRRs, and Project Performance Assessment Reports, where available).
A synthesis of lessons to identify evidence of best practices and any missed opportunities.

Geospatial Analysis

We conducted geospatial analyses to map the full range of conservation activities supported by the World Bank between FY10 and FY24, with the aim of evaluating relevance, effectiveness, and sustainability, and identifying patterns across the portfolio. This involved identifying and geocoding unique conservation sites and performing geospatial analysis to assess land cover dynamics over time. For each geo-located protected area, we calculated annual tree cover percentages, along with other land cover categories such as cropland and built-up areas, using consistent spatial and temporal resolution. This marks the first time that the full range of identifiable World Bank–financed conservation sites has been systematically mapped. We identified the presence of protected or physically conserved areas in 130 of the 139 projects, encompassing 884 unique sites, including 605 protected areas validated against the April 2025 release of the World Database on Protected Areas, which includes 305,198 registered sites. This section describes the geospatial analysis methodology in detail.

Identifying Protected Areas

First, we used the IEG bulk download tool to download available PADs (or equivalent) and ICRs (or equivalent) for each project in the conservation-focused portfolio.

Second, we undertook a project document review of projects in the portfolio to identify relevant protected areas supported by the World Bank. This involved using AI-assisted text extraction to identify relevant passages of the documents and mentions of specific protected area names. This helped narrow the process of identifying individual protected areas. To complement this process, we conducted a thorough manual review of the surrounding text, as well as key components of the project documents, including project development objectives and project components, paying close attention to review any tabular or graphic information for the presence of protected area lists or maps. The protected areas identified during this process were recorded in a matrix with adjacent columns containing additional variables (such as project ID, country, region, and approval year).

Third, any relevant details on the protected area in question were recorded in adjacent columns. These details included whether the protected area was national or regional, the protected area’s location, and in what way the protected area was supported by the corresponding project, among other things.

Fourth, using the April 2025 edition of the World Database on Protected Areas, the team searched for the relevant country to identify either a verbatim or proximate protected area match with the name of each protected area identified in the project document review. Once a match was identified, it was recorded in an adjacent column of the protected area matrix. The World Database on Protected Areas (WDPA) is the most comprehensive global database of marine and terrestrial protected areas. It is a joint project between the United Nations (UN) Environment Programme and the International Union for Conservation of Nature and is managed by the UN Environment Programme World Conservation Monitoring Centre in collaboration with governments, nongovernmental organizations, academia, and industry.

The following is the complete list of parameters in the finalized protected area data set:

The project ID and project name
The country and region
The name of the protected area found in the project documentation
Important details corresponding to the protected area, if pertinent
The official WDPA indexed name of the protected area, if available
The official WDPA ID number of the protected area, if available

Data Preparation

The methodology relies on two primary spatial data sources: (i) the Global Edge-matched Subnational Boundaries data set for administrative boundaries (https://fieldmaps.io/data), and (ii) the WDPA from Protected Planet (https://www.protectedplanet.net/en). These sources were selected for their global coverage, standardization, and regular updates. To accommodate the various combinations of location information found in project documents, a comprehensive spatial database was created through a systematic union (https://gisgeography.com/union-tool) of administrative boundaries and protected areas.

Data processing and structure. The data processing workflow involves several key steps:

Spatial data integration: Administrative boundaries and protected areas are processed through a union operation, creating comprehensive spatial units that capture all possible combinations of administrative areas and protected areas. This includes generating separate layers for:
1. Administrative boundaries (levels 0, 1, and 2)
2. Protected areas
3. Combined administrative-protected area units

Only polygons were used (that is, shapes that represent actual areas on a map, such as the full boundary of a district or a protected area), while points (which represent locations without area, like a single GPS coordinate) were excluded from the analysis.

Coordinate generation: Centroid coordinates (longitude and latitude) are calculated for each spatial unit to facilitate location data extraction (https://support.esri.com/en-us/knowledge-base/how-to-find-the-centroid-of-polygons-using-calculate-ge-000021849). This includes centroids for:
1. Individual administrative units at each level
2. Protected areas
3. Combined administrative-protected area units
Data entry system: A structured data entry template was developed using spreadsheet functionality to ensure consistent data capture. The template features:
1. Nested dropdown menus for standardized selection of administrative units and protected areas
2. Dynamic validation rules using array formulas to ensure data integrity
3. Automated coordinate assignment based on selected locations
4. Capability to record multiple locations per project

Quality assurance and data management. To maintain data quality and facilitate analysis, several control measures were implemented:

Standardization: All geographic names are standardized against authoritative sources, with relationships maintained between local and international nomenclature.
Flexibility: The data structure accommodates varying levels of geographic specificity, allowing for the recording of locations at any administrative level or protected area, either individually or in combination.
Automation: Python scripts were developed to:
1. Organize and manage project documentation
2. Clean and process spatial data
3. Convert finalized data to GeoJSON format for visualization and analysis
Visualization: Final outputs were visualized using QGIS software, enabling both quality control and analytical representation of the project portfolio’s spatial distribution.

Assessing Land Cover Dynamics

This section outlines the approach used to examine land cover dynamics within specific protected areas. The analysis focuses on assessing annual land cover changes from 2016 to 2024 (the period for which satellite imagery and data are available, as noted in the limitations section), employing a series of data processing techniques designed to ensure the accuracy and reliability of the results while addressing potential gaps and inconsistencies in the data.

Data Sources

Land cover data. The primary data source for the land cover analysis is the Dynamic World data set, which offers global near-time land cover classification at a high spatial resolution of 10 meters. This data set provides a comprehensive classification of land cover into nine distinct categories, enabling detailed assessments of land use and land cover dynamics over time (table A.2).

Table A.2. Land Cover Classes

Class	Description
Class 0	Water—Includes both permanent and seasonal water bodies, such as lakes, rivers, and wetlands
Class 1	Trees—Encompasses primary and secondary forests, as well as large-scale plantations
Class 2	Grass—Natural grasslands, livestock pastures, and parks
Class 3	Flooded vegetation—Mangroves and other inundated ecosystems
Class 4	Crops—Includes row crops and paddy crops
Class 5	Shrub and scrub—Sparse to dense open vegetation consisting of shrubs
Class 6	Built area—Low- and high-density buildings, roads, and urban open space
Class 7	Bare ground—Deserts and exposed rock
Class 8	Snow and ice—Permanent and seasonal snow cover

Source: Brown et al. 2022.

Protected areas data. The unit of analysis for this study is the protected area. The geocoded project data used in this analysis, identified 605 unique and matched protected areas. These protected areas encompass a range of categories, including marine, terrestrial, and hybrid protected areas. As the focus of the land cover analysis is solely on terrestrial ecosystems, marine areas were excluded from the analysis. Consequently, the analysis focuses on 526 protected areas, of which 469 are terrestrial and 57 are hybrid in nature. A map illustrating the location of these protected areas is provided in figure A.22.

The spatial boundaries of the protected areas were derived from the WDPA, a globally recognized repository that provides accurate geospatial data on the location and extent of protected areas, which allows for precise linking of spatial data to specific protected areas globally. Protected areas comprised of multiple polygons were dissolved for the analysis.

To capture the broader land cover dynamics surrounding the protected areas, a 5-kilometer buffer zone was applied to each protected area boundary after reprojection of the original protected area boundaries into an appropriate coordinate reference system. This buffer accounts for potential external land cover changes that could influence the protected areas, thereby providing a more comprehensive assessment of land cover dynamics.

Source: Independent Evaluation Group (based on protected area data from WDPA).

Note: This map has been cleared by the World Bank Group cartography unit.

Data Processing

The land cover analysis was performed using Google Earth Engine with JavaScript, leveraging its powerful processing capabilities for large-scale geospatial data.

Temporal range and data filtering. The analysis covers a nine-year period, from 2016 to 2024. For each year, land cover data are filtered to match the specific temporal and spatial boundaries of the protected areas. It is important to note that Dynamic World data are available from June 27, 2015. To ensure consistency and comparability across time, only complete years were included; therefore, 2015 was excluded from the analysis.

Land cover classification. For each protected area and year, the Dynamic World data set provides pixel-level classifications of land cover. To derive the dominant land cover type for each protected area, a mode composite approach was applied. This technique selects the most frequent land cover classification per pixel across the year, effectively capturing the primary land cover type within each protected area. The mode composite approach accounts for temporal variations, such as seasonal or yearly fluctuations, ensuring that the selected classification represents the most consistent land cover type during the given period.

Pixel counting and statistical aggregation. The analysis aggregates pixel-level land cover data within the spatial boundaries of each protected area. For each year and protected area, the total number of pixels corresponding to each land cover class is displayed as a histogram. This histogram approach quantifies the distribution of land cover types, allowing for the clear identification of the extent of each class within the protected area. The pixel counts are unweighted, meaning each pixel is treated equally, providing a straightforward assessment of land cover types based on the frequency of each classification within the spatial boundaries. This aggregation facilitates the quantification of land cover changes over time and enables the identification of trends across the nine-year study period.

Batch processing and data export. To manage the large scale of the data set, the data were processed in batches. This approach facilitates the efficient handling of the computational resources required for the analysis. Once the processing for each batch was complete, the results were exported in CSV format. The exported data included key information such as the protected area identifier, the year of analysis, and the pixel counts for each land cover class.

Example Analysis: Land Cover Change in Mamunta-Mayosso Protected Area

To illustrate the methodology applied in this study, this section presents a detailed analysis of land cover change in the Mamunta-Mayosso Protected Area (WDPA ID 555720436) over 2016–24. This example highlights key trends in deforestation and reforestation as well as shifts in land cover composition.

Land cover trends. Table A.3 presents the percentage distribution of land cover classes within the protected area for each year.

Table A.3. Land Cover Distribution in Mamunta-Mayosso, Sierra Leone (2016–24) (percent)

Year	Trees	Grass	Flooded Vegetation	Crops	Shrubs and Scrub	Built Up	Bare	Water	Snow and Ice
2016	46.46	6.42	0.11	1.70	43.64	0.31	0.19	1.18	0.00
2017	65.82	12.24	0.10	0.38	19.96	0.37	0.01	1.12	0.00
2018	63.05	5.45	0.09	0.32	29.70	0.29	0.02	1.09	0.00
2019	52.05	2.12	0.06	1.46	42.94	0.21	0.07	1.09	0.00
2020	53.22	3.70	0.08	1.31	40.16	0.32	0.09	1.12	0.00
2021	49.01	5.34	0.05	0.79	43.40	0.30	0.06	1.06	0.00
2022	54.94	3.39	0.06	0.84	39.11	0.26	0.31	1.09	0.00
2023	53.37	2.66	0.06	0.76	41.78	0.25	0.04	1.09	0.00
2024	56.37	2.40	0.07	1.11	38.61	0.20	0.14	1.10	0.00

Source: Independent Evaluation Group (based on Dynamic World data, accessed and processed using Google Earth Engine).

The data indicate an overall increase in tree cover from 46.46 percent in 2016 to 56.37 percent in 2024, though with notable fluctuations. The peak occurred in 2017 at 65.82 percent, followed by a decline until 2021, when tree cover reached its lowest point at 49.01 percent. Afterward, it gradually recovered. This suggests periods of afforestation or natural regeneration, potentially influenced by conservation efforts or environmental factors.

Shrub and scrub cover has generally declined over time. Initially high at 43.64 percent in 2016, it dropped sharply to 19.96 percent in 2017 before partially recovering in subsequent years. By 2024, it stood at 38.61 percent, indicating a gradual transition of shrubland into either tree cover or other land types. Grass cover has followed a similar pattern, peaking at 12.24 percent in 2017 but then decreasing steadily to 2.40 percent in 2024, further supporting the notion of vegetation succession or land-use change.

Agricultural land, represented by crop cover, fluctuates without a clear trend. It started at 1.70 percent in 2016, dropped to a low of 0.32 percent in 2018, and then gradually increased to 1.11 percent by 2024. Built-up areas remain relatively stable, ranging between 0.20 percent and 0.37 percent, showing no significant expansion of urbanization. Meanwhile, bare land, though minimal, has slightly increased over time, from 0.01 percent in 2017 to 0.14 percent in 2024, suggesting minor land degradation.

Water coverage has remained largely stable, fluctuating between 1.06 percent and 1.18 percent, while snow and ice are consistently absent throughout the data set. Overall, the trends suggest a general shift toward increasing tree cover at the expense of shrubs and grasslands, with relatively minor changes in agricultural, urban, and bare land.

Visualization of land cover change. To further illustrate these trends, the stacked bar chart in figure A.23 depicts the relative proportion of each land cover class annually.

Source: Independent Evaluation Group (based on Dynamic World data, accessed and processed using Google Earth Engine).

In addition to the statistical trends, spatial patterns of land cover change are visualized through a series of annual land cover maps (figure A.24). These maps provide insights into the geographic distribution of different land cover types and help identify specific areas where deforestation, regeneration, or conversion to other land uses has occurred.

A collage of images of a green and yellow object

AI-generated content may be incorrect.

Source: Independent Evaluation Group (based on Dynamic World data, accessed and processed using Google Earth Engine).

Deep Dives

To supplement and deepen the portfolio analyses, we commissioned industry and legal specialists to conduct thematic deep dives on (i) IPLCs; (ii) LNR; and (iii) biodiversity offsets. The selection of these three issues reflects their significance to the achievement of biodiversity outcomes, and the relative specialization associated with evaluating the aims and results of associated activities. We used emerging findings from these deep dives to guide interviews with senior experts from the World Bank (and IFC and MIGA in the specific case of offsets) in these associated fields to add explanatory power to these sections of the report.

Indigenous Peoples and Local Communities

To capture a holistic view of the engagement of IPLCs and other project-affected groups in the conservation-focused portfolio (see box A.4 for definitions and criteria), we employed a two-pronged approach that evaluates (i) the application of the Operational Policies (Safeguards) and the Environmental and Social Framework (ESF) to assess IPLC engagement, empowerment, and protection from a risk management perspective; and (ii) geospatial analysis to provide insights into the spatial distribution of World Bank–supported conservation activities in relation to IPLC lands.

Box A.4. Definitions and Criteria

Indigenous Peoples: The World Bank generally understands Indigenous Peoples to be distinct social and cultural groups that share collective ancestral ties to the lands and natural resources they inhabit or occupy, or from which they have been displaced. Under the Environmental and Social Framework, a group must possess four characteristics for Environmental and Social Standard (ESS)7 to be applicable. These criteria can be summarized as follows: self-identification as a distinct Indigenous group that is recognized by others, collective attachment to geographically distinct areas, distinct customary institutions, and a distinct language or dialect (ESS7, para. 8).

Local communities: The term “local communities” refers to groups of people residing in the same local area, possessing a long association with lands and natural resources, and embodying traditional lifestyles. The Environmental and Social Framework refers to local communities as project stakeholders, project-affected communities, and Sub-Saharan African Historically Underserved Traditional Local Communities that satisfy the criteria for ESS7.

Other rural groups: This term refers to groups actually or potentially affected by a World Bank–financed conservation-focused project that are not Indigenous Peoples or local communities. They may include rural communities that have recently immigrated to a project area, displaced persons, and other rural residents in the area of a World Bank–supported conservation project.

Source: ESF ESS7 (Indigenous Peoples/Sub-Saharan African Historically Underserved Traditional Local Communities) Guidance Note; Independent Evaluation Group.

Safeguard policies and ESF analysis. To assess the engagement, empowerment, and protection of IPLCs in conservation-focused projects, we conducted an analysis from a risk management perspective. This involved (i) examining the application of the World Bank’s Operational Policy 4.10 on Indigenous Peoples and the subsequent ESF Environmental and Social Standard (ESS) 7 on Indigenous Peoples/Sub-Saharan African Historically Underserved Traditional Local Communities and (ii) identifying trends and factors influencing their application.

First, we examined the incidence of conservation-focused projects that apply these policies by analyzing data exported from the World Bank’s Standard Reports. While the ESF data on standards application was found to be reliable, safeguard policies data were found to be incomplete and manually gap-filled from the operations portal and relevant project documentation.

We then performed a trend analysis, comparing the results from the two cohorts (that is, projects under the safeguard policies versus those under the ESF), including regional and country-specific trends. This analysis identified the regions and countries where Indigenous Peoples are more consistently recognized and included in conservation projects.

Additionally, we conducted project document reviews (for example, PADs, Environmental and Social Review Summaries, Environmental and Social Commitment Plans, Implementation Status and Results Reports, ICRs, and a select number of additional environmental and social analyses and risk management documents) to collect qualitative data and deepen our understanding of how projects engaged, empowered, and protected Indigenous Peoples.

Lastly, to validate our findings and explore explanatory factors, we conducted interviews with relevant staff and management.

Geospatial analysis. To further understand the relationship between the conservation-focused portfolio and IPLCs, we conducted a geospatial analysis to assess the extent to which World Bank–supported conservation activities are situated within IPLC territories. We used geospatial data from the LandMark platform and conducted spatial analyses to estimate the proportion of protected areas located within IPLC lands. We applied a spatial overlay technique to categorize project-supported protected areas based on their proximity to IPLC lands, distinguishing among sites located within IPLC territories, within 10 kilometers, 10–30 kilometers, or more than 30 kilometers away from IPLC lands.

Recognizing the limitations of the LandMark data set, particularly its underidentification of Indigenous lands in Africa, we conducted additional analyses using a data set constructed by Garnett et al. (2018). This data set provided a broader perspective on Indigenous land stewardship by incorporating de facto land management. We mapped the project sites using this data set to identify those within Indigenous Peoples’ lands, ensuring a comprehensive understanding of the spatial relationship between conservation-focused projects and IPLC territories. It is important to note that this data set does not distinguish whether the Indigenous Peoples’ areas are formally recognized by the government (box A.5).

Box A.5. Technical Details of Data Sources

LandMark: The LandMark data delineating Indigenous areas is sourced from a combination of official government records, community maps, and other verified sources to ensure accuracy and reliability. It includes spatial boundaries of Indigenous and community lands worldwide, distinguishing between lands that are legally recognized by governments and those where Indigenous Peoples have a claim but lack formal recognition. The data set is regularly updated and refined through collaboration with local organizations, governments, and Indigenous communities to reflect evolving land tenure status. By providing clear distinctions between recognized and unrecognized lands, LandMark highlights gaps in legal recognition and supports efforts to secure Indigenous land rights. Furthermore, the data set helps visualize the extent of Indigenous land stewardship, which is critical for conservation, climate resilience, and sustainable resource management.

Garnett et al. (2018): The Indigenous land-use data set created by Garnett et al. (2018) compiles data from 127 sources, including cadastral records, participatory mapping efforts, census-based models, and scholarly publications. It identifies Indigenous lands in 87 out of 235 countries or administratively independent entities, excluding uninhabited areas. The data set helps quantify the extent of Indigenous land management and its intersection with conservation priorities. By mapping these lands at a global scale, it provides a crucial foundation for recognizing Indigenous contributions to biodiversity protection and policy development.

The LandMark data set offers a conservative estimate of Indigenous Peoples and local community lands, as it includes only areas that meet strict criteria for documented tenure rights. It prioritizes legally recognized and publicly available data, drawing from government records, nongovernmental organization reports, and contributions from Indigenous organizations. In contrast, the Garnett et al. (2018) data set takes a more expansive approach, integrating multiple sources to provide a broader estimate of Indigenous lands. It relies on peer-reviewed literature, academic books, and reputable data providers while also incorporating spatial data from platforms like LandMark Global. By aggregating information from a wider range of sources, the Garnett et al. (2018) data set identifies more Indigenous Peoples and local community lands than LandMark, but it may also include territories with varying levels of formal recognition. These methodological differences highlight a key distinction: the LandMark data set offers a more conservative yet authoritative view of Indigenous Peoples and local community lands, whereas the Garnett et al. (2018) data set captures a wider spectrum of Indigenous land tenure, including areas with less formal documentation.

Sources: Garnett et al. 2018; LandMark (https://landmarkmap.org/data-methods/methodology).

Land and Natural Resource Rights

To further explore the extent to which LNR were considered and supported in project design, we conducted an in-depth analysis of a subset of World Bank conservation-focused projects that explicitly pursued a landscape approach (58 projects out of 139). This cohort was selected because landscape-level projects inherently involve complex spatial interactions among multiple stakeholders (see box A.5), integrating protected areas with restricted access alongside production zones, resource-use areas, and reforestation, afforestation, and restoration activities. As per the literature, these characteristics amplify potential LNR risks and opportunities, necessitating explicit consideration of land tenure, access rights, and resource governance within project design.

We conducted a review of project documents—including PADs, ISRs, and ICRs—focusing on identifying explicit considerations of, and activities supporting, land and resource tenure and rights within project design. Our review centered on several key criteria and LNR activities. First, we assessed the extent to which projects identified potential risks related to restrictions on rural communities’ access to land and resources by examining project documents for explicit acknowledgments of such risks, as well as the application of relevant environmental and social risk management policies, specifically Operational Policy 4.12 (Involuntary Resettlement) and ESS5 (Land Acquisition, Restrictions on Land Use, and Involuntary Resettlement).

We also coded project-financed activities according to the following LNR activities: (i) strengthening community participation and community-led land and resource governance; (ii) supporting tenure regularization or formalization; and (iii) reinforcing legislative or regulatory reforms related to land and resource rights. Additionally, we examined the presence and types of indicators used by projects to track achievements in line with these same activity bundles.

To identify trends over time, we analyzed the temporal distribution of projects that financed activities to strengthen community land and resource rights by categorizing projects based on their approval fiscal years. This analysis helped us understand the evolving commitment to these issues over different periods.

To illustrate the practical implementation and outcomes of LNR considerations, we drew on our Mozambique case study (see also “Exploratory Case Studies” section) and a desk-based review of a project in Tanzania. These illustrated achievements in strengthening community land and resource rights, participatory governance, and biodiversity protection, as well as the potential risks and consequences of insufficient consideration of community land and resource rights in project implementation.

Biodiversity Offsets

We conducted a review of biodiversity offsets to explore the Bank Group’s efforts in addressing significant residual adverse biodiversity impacts arising from development projects. Biodiversity offsets were analyzed as a deep dive under EQ1, given their relevance to conservation-focused activities and their role in mitigating residual impacts on critical habitats and natural ecosystems. Offsets are most often applied in contexts where development activities overlap with areas of high biodiversity value—such as critical natural habitats or restoration landscapes—which are central to the conservation objectives examined under EQ1.

To derive lessons from these initiatives, we employed a sequenced approach. First, we conducted a focused literature review to identify biodiversity offset good practice principles. Next, we conducted a portfolio review and analysis that (i) assessed the alignment of biodiversity offset development with the Bank Group’s environmental and social requirements, as well as the supervision of client implementation (including survey work, stakeholder engagement, and documentation); and (ii) evaluated the outcomes based on available monitoring and reporting data, focusing on principles such as no net loss or net gain. Finally, we derived explanatory factors through interviews with senior environmental and social experts.

Focused literature review. The literature review primarily referenced peer-reviewed articles published in reputable academic journals. The geographic scope of biodiversity offset literature remained somewhat limited since most studies have been conducted in Australia, North America, and Western Europe, where offset legislation had been established for many years. The review followed a partially chronological and thematic structure, beginning with the definition of biodiversity offsets and their theoretical foundations. It then examined regulatory frameworks before addressing offset challenges, highlighting the methodological and implementation difficulties that kept much of the literature theoretical. Finally, the review covered emerging research on real-world implementation, particularly in regions with long-standing offset mandates.

Portfolio review and analysis. The portfolio analysis involved systematically reviewing project documentation to assess the extent to which biodiversity offset good practice principles were met. Key questions included whether projects consistently identified and quantified residual impacts by applying the mitigation hierarchy; how offset activities were designed to achieve no net loss or net gain; the extent to which alternatives to offsetting were considered; whether stakeholder consultations reflected offset discussions; whether projects engaged independent experts; and the extent to which projects disclosed offsets-related documents. Additionally, it assessed financial provisions, long-term site security, and progress reported on offset implementation. The findings helped determine alignment with best practices, including compliance with ESS6 and PS6, and whether offset outcomes were achieved.

Stakeholder interviews. The evaluation team conducted interviews with 23 environmental and social specialists and coordinators. Interviews were conducted to gather internal perspectives on the implementation and effectiveness of biodiversity offsets and views on practical challenges and outcomes. Interview questions focused on the adequacy of guidance in safeguard policies, upstream work, implementation and monitoring after project approval, offset due diligence, and institutional learning. The interview data were then analyzed thematically to identify recurring patterns and key insights.

EQ2 Methods. How Well Are the World Bank and IFC Supporting Activities with Potential Biodiversity Benefits in Key Production Sectors, and Are Those Activities Likely to Achieve Such Benefits?

We assessed the World Bank and IFC’s support for integrating biodiversity into three key production sectors: (i) agriculture, including agroforestry and agribusiness; (ii) forestry; and (iii) fisheries and aquaculture. For each of these sectors, we gathered evidence using focused literature reviews, portfolio reviews and analyses, and to provide more explanatory power, we conducted exploratory case studies.

Focused Literature Reviews

We conducted AI-assisted focused literature reviews to identify good practices in the key production sectors covered—agriculture, including agroforestry and agribusiness; forestry; and fisheries and aquaculture—that have been empirically or theoretically linked to positive biodiversity outcomes. These reviews included academic research, technical guidance from reputable organizations, and relevant gray literature to establish a taxonomy of practices associated with biodiversity benefits within each sector. The resulting taxonomy served as an essential reference to define and inform coding protocols used in the subsequent portfolio reviews and analyses for each sector.

Portfolio Reviews and Analyses

The portfolio reviews systematically assessed the World Bank and IFC-supported projects within each sector covered to determine the extent to which biodiversity considerations were integrated into project design and implementation. Drawing directly from the taxonomies developed through the literature reviews, detailed coding protocols were established, focusing on activities indicative of good practice for biodiversity integration. For each project—focusing particularly on project-financed activities and results framework indicators measuring associated results—we analyzed the presence and nature of biodiversity-related activities and assessed alignment with identified good practices. For active projects, the review focused on project design, while for closed projects, both design and evidence of results were assessed.

In addition, closed projects were reviewed for the application of the safeguard policies and ESF requirements to respond to EQ3b: To the extent that evidence is available, has the application of biodiversity-related risk management policies mitigated biodiversity loss?

Exploratory Case Studies

We conducted exploratory case studies to provide explanatory power about the ways that the Bank Group integrates biodiversity considerations into key production sectors. The unit of analysis was the technical or policy mechanisms identified in projects in each country. We used an exploratory case design because of the nascency of this work in the Bank Group. Cases conducted in Brazil, Côte d’Ivoire, Ecuador, Peru, and Viet Nam included local engagement with resources users; case studies conducted in Mozambique were conducted on desk with interviews.

We selected case studies with interventions that were explicit in their intent to achieve biodiversity outcomes, or proxies thereof, to maximize the learning potential from the cases. Other case selection criteria included (i) timing: activities needed to be mature enough for evaluation; (ii) geography: regional and country typology coverage; (iii) production sectors, covering agriculture, forestry, fisheries, and aquaculture; and (iv) project mechanisms. The presence of IFC activities was also included as a criterion in a subset of cases.

The case studies were guided by a detailed case protocol (see box A.6) to ensure data collection and analytic consistency across cases. Case authors were experienced researchers with a combination of biodiversity, production sector, country/regional and evaluation expertise. For each case study, interviews were conducted with relevant World Bank staff; government ministries and agencies; local government; project management and implementation units; regional organizations; local subject matter experts; donor agencies; nongovernmental organizations; civil society; and associations.

The purpose of the exploratory case studies was to contextualize and derive explanatory factors on integrating biodiversity into production sectors, given the nascency of these activities. The case studies were designed to capture evidence on (i) the level of adoption of biodiversity-sensitive approaches by client governments, firms, and resource users across relevant activities; (ii) the factors that supported or challenged this adoption (including development outcomes); (iii) the extent to which adopted approaches led to biodiversity proxies and, where feasible, climate change outcomes; and (iv) explanatory factors that influenced these outcomes. Case studies adopted a systems approach using triangulated evidence: they identified the drivers of biodiversity loss in each landscape and how World Bank and IFC interventions addressed those drivers, assessed resource governance and incentives, identified trade-offs faced by clients and resource users in adopting biodiversity-relevant practices, and examined how the World Bank and IFC worked with other partners to achieve shared goals (where relevant). Case studies were designed to identify specific “mechanisms” used by the Bank Group to produce biodiversity benefits in each country and sectoral context. The case studies followed a structured approach outlined in box A.6.

Box A.6. Exploratory Case Study Approach

Step 1: Identification of the commitments made by the country to the goals pertaining to sustainable production outlined in the Global Biodiversity Framework.

A review of relevant country literature, including National Biodiversity Strategies and Action Plans, was undertaken to assess the extent of alignment with the commitments set out in the Global Biodiversity Framework. The review included an assessment of the recency of relevant commitments, their scale, and any financing committed for their fulfilment. This review was supplemented in primary data collection through semistructured interviews with in-country officials during the country visit.

Step 2: Analysis of the role of the World Bank Group in the country in financing or contributing to sustainable production activities that enhance biodiversity outcomes.

This step involved multiple components, including the following:

The selection of projects identified for data collection as part of the case studies was based on a selection criteria exercise described in the section above.

An analysis of World Bank efforts in country to engage with government clients on activities at the policy or regulatory level that might contribute to enhancing biodiversity outcomes such as adjustments to policies that drive biodiversity loss, among other things.

A review of relevant World Bank country-level documents (Systematic Country Diagnostics, Country Climate and Development Reports, and Country Partnership Frameworks) to assess the extent to which they integrate biodiversity considerations.

Identification of all World Bank analytics and investments in the country that incorporate biodiversity considerations, and selection of a targeted group of projects within this portfolio to be explored in detail.

Analysis of the key sectoral events that coincided with and influenced the World Bank’s engagement in the country during the evaluation period (FY 2010 and FY24). This analysis was underpinned by a review of literature, project documentation, and consultation with key stakeholders in country.

Step 3: A focused literature review of the biodiversity considerations relevant to each subsector(s) selected in each country.

The literature review included several components: a synthesis of the predominant challenges to achieving biodiversity outcomes in the chosen subsector, a description of the social, economic, and governance constraints to achievement of biodiversity outcomes, the integration of climate change as a cross-cutting consideration, an analysis of the chief long-term constraints to producing sustainable biodiversity outcomes in the subsector (for example, lack of financing, institutional capacity), and good practice examples of achievement of biodiversity outcomes in the chosen subsector(s).

Step 4: An analysis of the specific mechanisms or approaches used in each relevant project by the World Bank that are likely to have biodiversity benefits.

Efforts were made for each of these mechanisms to identify causal pathways through which they are expected to lead to behavior change and to document any relevant outcomes resulting from the use of this mechanism or approach. For example, this could entail the adoption of a climate and biodiversity-friendly agricultural practice likely to reduce the use of water and other natural resources, enhance soil health, and prevent land degradation.

Step 5: The case analysis culminated in a case synthesis and overarching analysis, drawing on the multiple lines of evidence identified.

This synthesis responded to the evaluation subquestions for EQ2, which included the documenting of good practice examples, innovations, lessons learned, any evidence of use of proxies for biodiversity benefits, and the contribution of biodiversity benefits to climate change benefits.

Source: Independent Evaluation Group.

Review of Core Country Diagnostics

To assess the extent to which the World Bank is using the country engagement process to identify integrated actions while building national capacity for biodiversity planning, we undertook a review of CPFs and CCDRs. We reviewed 113 CPFs—representing the most recent CPF available for every country—and all 57 CCDRs (covering 69 countries) disclosed at the time of this analysis. See appendix C for the list of CPFs and CCDRs included in the analysis. CPFs are the primary strategic document that outlines the Bank Group’s engagement with a client country. CCDRs are designed to help countries prioritize actions to reduce greenhouse gas emissions, enhance adaptation, and achieve broader development goals.

Country Partnership Framework

Alignment with global climate and biodiversity commitments. We conducted parametric keyword searches and content analysis within the 113 CPFs to assess their alignment with global climate and biodiversity agreements. Each CPF was analyzed to determine whether it articulates climate priorities, including nationally determined contributions, the Paris Agreement, or Paris Alignment. Additionally, we examined whether CPFs explicitly mention global biodiversity agreements such as the GBF, the CBD, or the Kunming-Montreal and Aichi biodiversity targets. We also identified whether CPFs articulate support linked to their respective National Biodiversity Strategies and Action Plans and Natural Capital Accounting.

CPF high-level outcomes and objectives. We manually extracted, and thematically coded, CPF high-level outcomes and subobjectives in line with relevant themes. Relevant CPF high-level outcomes and subobjectives were categorized into themes based on the dominant focus (including in their rationale), while recognizing that there is overlap across these themes (table A.4). This thematic coding allowed us to understand the distribution of thematic focus areas and assess the extent to which these focus areas provide opportunities—both taken and not taken—to engage with clients on biodiversity and capture biodiversity benefits.

Table A.4. Coding Categories for Country Partnership Framework Subobjectives

Thematic Code	Description
Climate Adaptation and Disaster Risk Management	Enhancing resilience against climate change impacts, natural hazards (floods, droughts, hurricanes, earthquakes, and so on), and related risks. Strengthening preparedness, early warning systems, recovery, and adaptive capacity of infrastructure, livelihoods, and ecosystems. Examples: “Enhance resilience to natural shocks”; “Build resilience to climate-related events”; “Strengthen multihazard disaster resilience.”
Climate Change Mitigation and Low-Carbon Transition	Reducing GHG emissions, promoting decarbonization, supporting low-carbon industries, and contributing to global climate goals (for example, the Paris Agreement, NDC implementation). Examples: “Scale up climate mitigation measures”; “Support the energy transition/reduce energy intensity.”
Natural Resource Management	Sustainable management of forests, fisheries, land, ecosystems, and so on. Includes preserving habitats, reforestation, sustainable forestry, land degradation prevention, or conservation of wildlife. Examples: “Improved management of natural resources”; “Preserve and restore natural capital”; “Sustainable landscape management.”
Water, Sanitation, and Waste Management	Ensuring water security, water resource management, sanitation, solid waste management, wastewater treatment, and related infrastructure. Examples: “Improve access to water, sanitation, and solid waste management”; “Enhance water security and sustainability”; “Efficient water resource management for resilience.”
Energy Sustainability and Efficiency	Increasing renewable energy capacity (solar, wind, hydro, geothermal), promoting energy efficiency, and ensuring a sustainable energy supply. Often linked to low-carbon development but can be distinguished if the focus is specifically on energy systems. Examples: “Expand clean energy matrix”; “Enhanced energy sustainability and renewable energy resources”; “Energy efficiency improvements in utilities and industries.”
Pollution and Air Quality Management	Reducing ambient air pollution, industrial pollution, marine plastic pollution, and soil contamination, or improving broader environmental quality (apart from GHG-focused mitigation). Examples: “Reduce air pollution in urban centers”; “Address plastic waste in coastal regions”; “Improve air, soil, water pollution control frameworks.”
Resilient Infrastructure and Urban Resilience	Resilience and sustainability in roads, transport, housing, and urban planning. Ensuring cities adapt to climate change impacts (sea-level rise, flooding), reduce congestion/pollution, and strengthen livability. Examples: “Improved transport connectivity and safety” (with an explicit climate/disaster lens); “Promote green and resilient cities”; “Sustain and strengthen urban infrastructure to withstand climate shocks.”
Financial and Macroeconomic Resilience	Strengthening a country’s financial capacity to cope with climate-related or disaster-related shocks. Building fiscal buffers, insurance mechanisms, contingent financing, or macrofiscal strategies for climate resilience. Examples: “Enhanced financial resilience”; “Strengthen capacity for macro-financial sustainability / reduce vulnerability to external shocks”; “Risk financing for disasters / catastrophe insurance frameworks.”
Governance and Institutional Capacity for Environment/ Climate	Public sector reforms, policies, regulations, or institutional frameworks specifically geared toward better environmental or climate outcomes, for example, setting up climate governance structures, environmental agencies, or cross-sector coordination. Examples: “Improve government’s effectiveness, efficiency, and transparency in climate/natural resource management”; “Strengthen institutional and financial framework for risk management”; “Enhance capacity for climate finance and green budgeting.”
Productivity, Climate-Smart Agriculture and Resilient Livelihoods	Focus on making agriculture, fisheries, or rural livelihoods more resilient and sustainable under climate variability (crop diversification, water-efficient irrigation, sustainable land practices, and so on). Examples: “Promote climate-resilient agriculture”; “Strengthen rural livelihoods through sustainable land and water use.”
Biodiversity	Explicit mention of biodiversity in objective. Examples: “Improved management of mining, natural resources and biodiversity”; “Improved and climate-adaptive management of forests, biodiversity and protected areas.”
Social Protection and Safety Nets	Strengthening household and community resilience to shocks by enhancing social protection systems, improving livelihoods, and building crisis preparedness and response capacities, particularly for vulnerable and conflict-affected populations. Examples: “Strengthen crisis resilience for vulnerable, displaced, and conflict-affected populations”; “Improve efficiency and effectiveness of the social protection system”; “Strengthen mechanisms to protect people against shocks.”
Land Administration	Example: “Improve Land Titling and Administration.”

Source: Independent Evaluation Group.

Note: GHG = greenhouse gas; NDC = nationally determined contribution.

CPF metrics. We also manually extracted and analyzed the metrics measured in CPFs—at both the high-level outcome and sublevel objective—to identify those metrics that could serve as proxies for biodiversity-related outcomes.

Country Climate and Development Reports

We reviewed each CCDR to extract and analyze relevant information on climate mitigation, adaptation strategies, and the climate–biodiversity risk nexus. First, we focused our content analysis on identifying how each CCDR addressed the use and conservation of natural ecosystems, such as forests and coastal and marine ecosystems, to achieve national climate mitigation goals. Additionally, we examined the recommended practices for mitigation, for example, habitat restoration versus reforestation or afforestation activities. Second, we assessed the adaptation strategies presented to determine the extent to which CCDRs recommended using and conserving biodiversity-rich natural habitats to build resilience. Third, we explored the extent to which CCDRs diagnose the risks posed by climate change to biodiversity and the recommended measures to safeguard it.

Limitations

Portfolio Review and Analysis

Missing documentation. Some identified lending operations that were found to be relevant lacked sufficient evidence and documentation in the operations portal and could not therefore be included in the portfolio review and analysis (especially smaller trust funded activities). We excluded these projects because the information required to determine and code variables relevant to biodiversity efforts was not available.

Geospatial Analysis

Missing protected area names. In 25 of the 130 conservation-focused projects that include support to protected areas, it was not possible to identify the names of the specific protected areas as they were not included in any project documentation. Furthermore, in 279 cases (out of 884), we could not match the names of the protected areas as found in the respective project documentation with the WDPA (despite extensive efforts to do so, including searches online). Lastly, 19 protected areas did not have shape files and were therefore excluded from the land cover analysis.

Land cover change analysis. The geospatial analysis of land cover change relies on several key assumptions and is subject to certain limitations inherent in the data and methodology. These considerations are important for interpreting the results accurately and understanding the potential sources of uncertainty.

Probabilistic nature of the Dynamic World model. The land cover classifications used in this study are derived from the Dynamic World data set, which is a probabilistic model rather than a deterministic classification. This means that each pixel is assigned a probability of belonging to different land cover classes, and the final classification represents the most likely class rather than a definitive ground-truth label. As a result, some classification uncertainty is expected, particularly in transitional or heterogeneous landscapes where multiple land cover types coexist within a single pixel. Several recent studies have analyzed recently developed global land-use/land cover models (Kerner et al. 2024; Venter et al. 2022; Wang et al. 2023), and we refer to these resources for a more detailed assessment of the limitations.
Impact of missing data and cloud cover. Optical satellite data, such as those used in Dynamic World, are susceptible to cloud cover, haze, and other atmospheric interferences, which can result in missing data for certain time periods. Although compositing techniques help mitigate these gaps, some areas may have fewer valid observations, potentially introducing noise or inconsistencies in the temporal analysis. This issue is particularly relevant in tropical regions, where persistent cloud cover can reduce the frequency of high-quality observations.
Although Dynamic World, which relies on Sentinel-2 L1C images, excludes images with cloud cover greater than 35 percent from its land cover calculations, we observed that in some years the coverage was incomplete, likely due to persistent cloud cover. To address potential data quality issues, we implemented the following decision rule: (i) if a protected area-year had fewer than 80 percent valid pixels, it was excluded from the analysis; and (ii) if a protected area-year had 80 percent or more valid pixels, it was included, with weights assigned according to the percentage of valid pixels. This approach allowed us to calculate the weighted average percentage of land cover for two periods, 2016–19 and 2020–24, and to estimate the change in land cover as the difference between these two periods.
Influence of mixed pixels and land cover transitions. Given the 10-meter spatial resolution of the data set, some pixels are likely to contain a mixture of multiple land cover types. This can lead to misclassifications, especially in areas where tree cover, shrubs, and grasslands intermingle. Additionally, land cover transitions (for example, forest degradation, regrowth, or seasonal changes in vegetation) may not be fully captured if the dominant class in a pixel does not change significantly over time.
Spatial and temporal resolution considerations. Although the 10-meter spatial resolution of Dynamic World allows for relatively fine-scale analysis, it may not detect small-scale changes such as selective logging, understory degradation, or small agricultural clearings. Additionally, the temporal resolution of the analysis is limited by the availability of cloud-free observations, meaning that some short-term disturbances or rapid land cover changes may not be fully captured.

Biodiversity Offsets Analysis

Due to resource constraints, the evaluation team could not conduct site visits or engage directly with client counterparts or beneficiaries on the topic of offsets. This limited the collection of firsthand data and hindered the ability to verify information or assess the contextual factors influencing project outcomes.

The evaluation team relied on different approaches to identify offset projects across institutions (IFC, MIGA, and the World Bank) because of limitations in access to documents and resource constraints. Although the list of MIGA and IFC projects with offsets was retrieved from ESG counterparts, the team relied on the Standard Reports and a combination of manual and large language model (LLM) assessment. The World Bank does not have a centralized tracking tool for offsets, but the evaluation team met with several biodiversity experts in the World Bank and received their input on the list of projects.

References

Brown, C. F., S. P. Brumby, B. Guzder-Williams, et al. 2022. “Dynamic World, Near Real-time Global 10 m Land Use Land Cover Mapping.” Scientific Data 9: 251.

Garnett, S. T., N. D. Burgess, J. E. Fa, et al. “A Spatial Overview of the Global Importance of Indigenous Lands for Conservation.” Nature Sustainability 1: 369–74.

Kerner, H., C. Nakalembe, A. Yang, et al. 2024. “How Accurate are Existing Land Cover Maps for Agriculture in Sub-Saharan Africa?” Scientific Data 11 (1): 486.

Venter, Z. S., D. N. Barton, T. Chakraborty, T. Simensen, and G. Singh. 2022. “Global 10 m Land Use Land Cover Data Sets: A Comparison of Dynamic World, World Cover and Esri Land Cover.” Remote Sensing 14 (16): 4101.

Wang, Y., Y. Sun, X. Cao, Y. Wang, W. Zhang, and X. Cheng. 2023. “A Review of Regional and Global Scale Land Use/Land Cover Mapping Products Generated from Satellite Remote Sensing.” ISPRS Journal of Photogrammetry and Remote Sensing 206: 311–34.

World Bank. 2018. “ESF Guidance Note 7: Indigenous Peoples/Sub-Saharan African Historically Underserved Traditional Local Communities (ESS7).” Washington, DC: World Bank.

Evaluation Design and Questions

EQ1: How well is the World Bank addressing biodiversity challenges through conservation-focused activities?
EQ2: How well are the World Bank and the International Finance Corporation (IFC) supporting activities with potential biodiversity benefits in key production sectors, and are those activities likely to achieve such benefits?

The subquestions for EQ1—on conservation-focused activities—are as follows:

EQ1a. How well is the World Bank applying good practice approaches in its biodiversity conservation activities?
EQ1b. To what extent are biodiversity conservation activities designed to leverage the World Bank’s advantages?
EQ1c. How well are biodiversity projects achieving their biodiversity goals?
EQ1d. How well are biodiversity projects articulating and achieving their multiple benefits (economic, development, climate)?

The subquestions for EQ2—on integrating biodiversity into key production sectors—are as follows:

EQ2a. What has worked to enable the integration of activities with biodiversity benefits in engagements in key production sectors—in the Bank Group, with clients, and with resource users?
EQ2b. Do projects with potential biodiversity benefits include evidence on proxies for biodiversity benefits, and are they achieving those proxies?
EQ2c. How have activities with potential biodiversity benefits contributed to climate change benefits?

The subquestions for EQ3—on biodiversity risk management—are as follows:

EQ3a. How well have biodiversity risk management policies been used to inform the design and support the effective implementation of projects that could have an adverse effect on biodiversity?
EQ3b. To the extent that evidence is available, has the application of biodiversity-related risk management policies mitigated biodiversity loss?

Source: Independent Evaluation Group.

Note: EQ3—on how well the World Bank Group is supporting clients to manage risks affecting biodiversity—was integrated into the analyses of EQ1 and EQ2. Specifically, EQ3 was examined through a dedicated analysis of biodiversity offsets (covering the World Bank, IFC, and MIGA), and integrated into the portfolio review and deep dives under EQ1 and the portfolio and case study reviews under EQ2. CCDR = Country Climate and Development Report; CPF = Country Partnership Framework; EQ = evaluation question; IFC = International Finance Corporation; IPLCs= Indigenous Peoples and local communities; LNR = land and natural resource rights; MIGA = Multilateral Investment Guarantee Agency; PRA = portfolio review and analysis.

Evaluation Scope

This evaluation was scoped along three dimensions: time frame, institutional coverage, and level of engagement.

Time frame. The evaluation includes World Bank, IFC, and MIGA projects approved during FY15–24, except for the conservation-focused portfolio, which includes projects approved from FY10 to FY24. This extended time frame was used to ensure the inclusion of closed projects with evaluative evidence on results, as EQ1 examines both the evolution of the World Bank’s engagement in conservation and the extent to which environmental and development results were achieved.

Evaluation Portfolio Identification and Classifications

This section has two parts: the first describes the process used to identify the portfolios included in the evaluation for each EQ, and the second provides a description of these portfolios.

EQ1. Biodiversity Conservation

World Bank

Box A.1. Search Taxonomy Used for Text Mining to Identify Conservation-Focused Portfolio

Source: Independent Evaluation Group.

International Finance Corporation and Multilateral Investment Guarantee Agency

As per the evaluation Approach Paper, IFC and MIGA were not included in the EQ1 portfolio because they do not undertake conservation-focused activities.

EQ2. Biodiversity Integration in Key Production Sectors

World Bank Lending

Forests. Because forest-related projects are financed across Global Practices, we used relevant sector codes (AT—Forestry; AK—Public Administration—Agriculture, Fishing, and Forestry; AZ—Other Agriculture, Fishing, and Forestry) and theme codes (831—Forests Policies and Institutions) to identify the forests portfolio. We also pulled all projects that included relevant keywords (that is, forest, wood, timber) in their project name, project development objective, component title, and indicators. Using this combined output, we manually screened projects (that is, their project names, project development objectives, component titles, and indicator titles) to identify those with in-scope activities such as forest carbon and forest production.

Fisheries and aquaculture. To identify the fisheries and aquaculture portfolio, we used relevant sector codes (AF—Fisheries; AZ—Other Agriculture, Fishing, and Forestry) and theme codes (832—Fisheries Policies and Institutions). Using this output, we manually screened projects (that is, their project development objectives, component titles, and indicator titles) to identify those with in-scope activities such as fisheries management and aquaculture.

International Finance Corporation Investments

Step 1: Data extraction. We extracted project text data from the IFC Disclosure Portal for projects approved during the evaluation period covering the following: (i) project description, (ii) overview fund, (iii) risk impact, (iv) environmental and social action plan, (v) review scope, (vi) environmental and social information, (vii) impact, (viii) result, (ix) risk assessment, (x) role, (xi) contribution, (xii) environmental and social issues, (xiii) sponsor, (xiv) cost nature, (xv) investment, (xvi) location, (xvii) environmental and social categorization rationale, (xviii) risk fund, (xix) risk impact, (xx) stakeholders, and (xxi) mitigation measures.

Step 2: Data consolidation. We consolidated the text data, originally stored in multiple columns, into a single comprehensive text field for each project.

Box A.2. Additional Keyword Taxonomy Used for Text Mining to Identify International Finance Corporation Portfolio

climate smart ag; climate-smart ag; deforestation; Eucalyptus; fish; forest; good agricultural practice; good agriculture practice; Pulp mill; smallholder; sustainable ag; traceability.

Source: Independent Evaluation Group.

Step 4: Initial project inclusion. Projects that contained at least one occurrence of the keywords and phrases were included in our initial longlist.

Step 5: Sector-based inclusion. In addition to keyword-triggered projects, we included projects from relevant sectors in our longlist. The list of relevant sectors is available in table A.1.

Step 8: Final selection. After the manual validation from steps 6 and 7, we identified 87 investment projects out of the initial 948 projects in the longlist.

International Finance Corporation Advisory

Step 2: Data consolidation. We consolidated the text data, originally stored in multiple columns, into a single comprehensive text field for each project.

Step 3: Keyword search. Using R, we conducted a search for predefined biodiversity-related keywords and phrases within the aggregated text. The list of keywords can be found in box A.1.

Step 4: Initial project inclusion. Projects that contained at least one occurrence of the keywords or phrases were included in our initial longlist.

Step 5: Primary business line–based inclusion. In addition to keyword-triggered projects, we included projects from relevant primary business lines in our longlist. The list of relevant primary business lines is provided in table A.1.

Step 8: Final selection. After the manual validation from steps 6 and 7, we identified 63 advisory projects out of the initial 218 projects in the longlist.

Table A.1. International Finance Corporation Business Lines and Sector Codes Used to Identify Evaluation Question 2 Portfolio

Business Line for IFC Advisory Projects

Tertiary Sector Name for IFC Investment Services

ESG

ESG—E&S Risk Management

ESG—Integrated ESG and Sustainability

MAS

MAS—Agri Expertise

MAS—Agribusiness: Industry Standards

MAS—Agribusiness: SME Productivity

MAS—Agribusiness: Strategic Community Investment

MAS—Agribusiness Other

MAS—Crop Production

MAS—Sustainable Protein

MAS—Forestry and Land Uses

Animal Aquaculture

Cattle Farming

Coffee, Cocoa, Tea

Dairy Products

Diversified Edible Agricultural Crops Production

Fishing

Fruits and Vegetables

Horticultural Products (Flowers)

Grains and Beans

Natural Fibers (Cotton, Sisal, Jute, and so on)

Other Animal Production

Other Vegetable Oil Crops (Coconut, Rapeseed, Peanut, Sunflower, and so on)

Palm Oil Plantations

Palm Vegetable Oil

Paperboard (Including Boxboard, Fiberboard)

Plantation Forests

Poultry Farming

Sugarcane and Beets

Source: Independent Evaluation Group.

Biodiversity Offsets

The biodiversity offsets portfolio across the World Bank, IFC, and MIGA was identified through a combination of counterpart collaboration and systematic portfolio reviews.

World Bank. For the World Bank, 16 projects under the Environmental and Social Framework (ESF) and 5 projects under the safeguard policies were identified as having offsets. The ESF projects were identified using the ESF 5.0 Master Report by filtering for projects that marked biodiversity offsets as relevant. Safeguard policies projects were identified through a structured document review process, leveraging AI-assisted keyword searches to detect offset-related language in PADs and Integrated Safeguards Data Sheets, followed by a manual verification by the evaluation team. The manual verification process involved reviewing the text excerpts identified by the keyword search tool to ensure projects that referred to offsets or other relevant terms in project documents included offset activities within the project’s scope. False positives, such as instances where offsets or related terms were mentioned in a cursory fashion or in an irrelevant context, were screened out. The final list was also validated in semistructured interviews with environmental and social specialists during data collection activities.

Portfolio Description

EQ1. Biodiversity Conservation

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

Note: EAP = East Asia and Pacific; ECA = Europe and Central Asia; ESA = Eastern and Southern Africa; LAC = Latin America and the Caribbean; MENA = Middle East and North Africa; SA = South Asia; WCA = Western and Central Africa.

Source: Independent Evaluation Group.

EQ2. Biodiversity Integration in Key Production Sectors

World Bank Lending

Agriculture. There were 247 World Bank lending projects financed by the Agriculture and Food Global Practice approved between FY15 and FY24. Of these, 145 are active and 102 are closed. Of the 102 closed projects, 72 projects have an ICR, of which 53 have an ICRR (data from IEG DataHub as of March 2025). Eighty-seven percent of the closed and validated projects were rated moderately satisfactory or higher. A breakdown of the portfolio is provided in figures A.5–A.7.

Source: Independent Evaluation Group.

Note: DPF = development policy financing; IPF = investment project financing; PforR = Program-for-Results.

Source: Independent Evaluation Group.

Note: MENA = Middle East and North Africa.

Source: Independent Evaluation Group.

Forestry. There were 83 World Bank lending projects approved between FY15 and FY24 with in-scope forest production activities, of which 41 are active and 42 are closed. Most of these projects (n = 66; 80 percent) are financed by the Environment, Natural Resources, and Blue Economy Global Practice. The remainder are divided among the Agriculture and Food (n = 9; 11 percent); Macroeconomics, Trade, and Investment (n = 5; 6 percent); Energy and Extractives (n = 2; 2 percent); and Poverty and Equity (n = 1; 1 percent) Global Practices. Of the 42 closed projects, 35 projects have ICRs, of which 28 have ICRRs (data from IEG DataHub as of April 2025). Eighty-six percent of the projects were rated moderately satisfactory or higher. A breakdown of the portfolio is provided in figures A.8–A.10.

Source: Independent Evaluation Group.

Note: AGR = Agriculture, and Food; EAE = Energy and Extractives; ENB = Environment, Natural Resources, and Blue Economy; MTI = Macroeconomics, Trade, and Investment; POV = Poverty and Equity.

Source: Independent Evaluation Group.

Note: AFE = Eastern and Southern Africa; AFW = Western and Central Africa; EAP = East Asia and Pacific; ECA = Europe and Central Asia; LCR = Latin America and the Caribbean; MNA = Middle East and North Africa; SAR = South Asia.

Source: Independent Evaluation Group.

Fisheries and aquaculture. There were 51 World Bank lending projects approved between FY15 and FY24 with in-scope fisheries activities, of which 26 are active and 25 are closed. Most of these projects (68.6 percent; n = 35) are financed by the Environment, Natural Resources, and Blue Economy Global Practice, while 20 percent (n = 10) are financed by the Agriculture and Food Global Practice. The remainder are divided among the Macroeconomics, Trade, and Investment (n = 3); Water (n = 2); and Finance, Competitiveness, and Innovation (n = 1) Global Practices, respectively. Of the 25 closed projects, 20 projects have ICRs, of which 18 have ICRRs (data from IEG DataHub as of March 2025). Fifty percent of the projects were rated moderately satisfactory or higher. A breakdown of the portfolio is provided in figures A.11–A.13.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

International Finance Corporation Investments and Advisory

Forestry. Between FY15 and FY24, IFC approved 15 investments in the forestry sector. These included 4 investments in plantation forests, 8 in pulp and paper manufacturing, and 3 in wood panel and wood product manufacturing. Of these, IEG has evaluated 2 investments, both of which were in the wood panel and wood product and pulp and paper manufacturing sectors. Additionally, there was one IFC advisory project mapped to the plantation forests (90 percent) and paperboard (10 percent) sectors, which has not been validated by IEG. No other advisory projects were identified in the pulp, paper, or wood manufacturing sectors.

Fisheries and aquaculture. During FY15–24, IFC approved eight investments in the fisheries and aquaculture sector. This included seven investments in aquaculture and one in fishing. Of the seven aquaculture projects, five were in Ecuador and were covered in our Ecuador case study (refer to the Exploratory Case Studies section). The remaining two aquaculture projects were in China and focused on aquafeed production. The single fishing investment was in the Solomon Islands, where IFC financed the purchase of an additional fishing vessel, which triggered PS6. Additionally, there were three IFC advisory projects mapped to the aquaculture and fishing sectors. IFC self-evaluated these advisory projects at completion, but only one of them has been validated by IEG.

Biodiversity Offsets

World Bank. For the World Bank, 16 projects under ESF and 5 projects under the safeguard policies were identified as having offsets. A breakdown of the portfolio is provided in figures A.14–A.16.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

Note: AFE = Eastern and Southern Africa; AFW = Western and Central Africa; EAP = East Asia and Pacific; ECA = Europe and Central Asia; LCR = Latin America and the Caribbean; MENA = Middle East and North Africa; SAR = South Asia.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

MIGA. Twelve MIGA projects in scope involved biodiversity offsets. A breakdown of the portfolio is provided in figures A.19–A.21.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

Source: Independent Evaluation Group.

EQ1 Methods: How Well Is the World Bank Achieving Biodiversity Aims Through Its Conservation-Focused Activities?

To assess the relevance and effectiveness of World Bank support for biodiversity conservation, we used several triangulated methods, including literature review, portfolio review and analysis, the geolocation of conservation activities, and three thematic deep dives. We expand on these methods below.

Focused Literature Review and Review of World Bank Group Strategies and Commitments

Portfolio Review and Analysis

The literature review yielded five themes that reflect a combination of global best practice with the World Bank’s comparative advantage. The themes identified were as follows:

Geographic representativeness, or the equitable and proportional representation of different biomes, ecosystems, and taxonomic groups. Analyzing the portfolio through the lens of geographic representativeness is important for determining the relevance of World Bank conservation activities in relation to the protection and conservation of critical biomes and areas of high endemism.
Ecological connectivity, or the unimpeded movement of species and flow of natural processes that sustain life across ecosystems. Research indicates that enhancing the connectivity of protected areas and conservation areas is critical in improving genetic diversity, enabling the movement of wildlife species and enhancing ecosystem resilience through the reduction of habitat fragmentation.
Ecological monitoring and reporting of biodiversity outcomes. Quantitative ecological monitoring is crucial for understanding biodiversity outcomes. It is a priority within the GBF, providing essential data to assess and guide conservation actions.
Engagement and protection of IPLCs. The biodiversity Approach Paper highlights the essential role of IPLCs in achieving biodiversity goals (World Bank Group 2021). Local communities have deep, place-based ecological knowledge developed over millennia, enhancing biodiversity preservation. Good governance, including land tenure security, equitable benefit sharing, and preserving local knowledge, is crucial for their contribution to effective conservation.
Sustainable financing. Sustainable financing is crucial for biodiversity conservation because it ensures long-term funding to protect ecosystems, species, and natural resources. Many conservation initiatives face financial shortfalls, leading to habitat degradation, species loss, and reduced ecosystem services.

Box A.3. Coding Protocol for Conservation-Focused Projects

Basic project data

Project ID, name, country, source of finance, amount of finance

Project development objective

Project components

Project indicators

Intended beneficiaries, also disaggregated

Location of biodiversity areas (if possible)

Project lessons

Project theory as it pertains to biodiversity aims

Stated theory or implicit theory

Explicit or implied outcomes not included in the results framework

Identification of good practice approaches, including measurement thereof

Ecological connectivity (for example, corridors, transfrontier areas)

Use of climate science, or other scientific evidence to make conservation planning decisions

Use of Indigenous knowledge to make conservation planning decisions

Sustainable financing at the national and park level (for example, recurrent finance/national budget; endowment funds; payment for environmental service)

Inclusion of Indigenous Peoples and local communities

Support for land tenure security, access, or resource-use rights

Other (open cell for observations)

Source: Independent Evaluation Group.

For projects with completed and validated ICRs, the extraction and analysis of results indicators were undertaken. Where relevant, data from the geospatial analysis on variables such as tree cover change or crop change were also extracted.

Content analysis. The consolidation of the above data into a comprehensive portfolio matrix allowed for detailed quantitative and qualitative analysis that ultimately comprised:

The relevance of the portfolio’s activities in each thematic area in terms of alignment to best practice, the World Bank’s areas of comparative advantage, and strategic priorities.
The effectiveness of included activities for closed projects in each thematic area, based on the achievements of relevant intermediate results indicators and project development objectives and any explanatory factors included in project documents (ICRs, ICRRs, and Project Performance Assessment Reports, where available).
A synthesis of lessons to identify evidence of best practices and any missed opportunities.

Geospatial Analysis

We conducted geospatial analyses to map the full range of conservation activities supported by the World Bank between FY10 and FY24, with the aim of evaluating relevance, effectiveness, and sustainability and identifying patterns across the portfolio. This involved identifying and geocoding unique conservation sites and performing geospatial analysis to assess land cover dynamics over time. For each geolocated protected area, we calculated annual tree cover percentages, along with other land cover categories such as cropland and built-up areas, using consistent spatial and temporal resolution. This marks the first time that the full range of identifiable World Bank–financed conservation sites has been systematically mapped. We identified the presence of protected or physically conserved areas in 130 of the 139 projects, encompassing 884 unique sites, including 605 protected areas validated against the April 2025 release of the World Database on Protected Areas (WDPA), which includes 305,198 registered sites. This section describes the geospatial analysis methodology in detail.

Identifying Protected Areas

First, we used the IEG bulk download tool to download available PADs (or equivalent) and ICRs (or equivalent) for each project in the conservation-focused portfolio.

Fourth, using the April 2025 edition of the WDPA, the team searched for the relevant country to identify either a verbatim or proximate protected area match with the name of each protected area identified in the project document review. Once a match was identified, it was recorded in an adjacent column of the protected area matrix. The WDPA is the most comprehensive global database of marine and terrestrial protected areas. It is a joint project between the United Nations Environment Programme and the International Union for Conservation of Nature and is managed by the United Nations Environment Programme World Conservation Monitoring Centre in collaboration with governments, nongovernmental organizations, academia, and industry.

The following is the complete list of parameters in the finalized protected area data set:

The project ID and project name
The country and Region
The name of the protected area found in the project documentation
Important details corresponding to the protected area, if pertinent
The official WDPA indexed name of the protected area, if available
The official WDPA ID number of the protected area, if available

Data Preparation

Data processing and structure. The data processing workflow involves several key steps:

Spatial data integration. Administrative boundaries and protected areas are processed through a union operation, creating comprehensive spatial units that capture all possible combinations of administrative areas and protected areas. This includes generating separate layers for:
1. Administrative boundaries (levels 0, 1, and 2)
2. Protected areas
3. Combined administrative-protected area units

Coordinate generation. Centroid coordinates (longitude and latitude) are calculated for each spatial unit to facilitate location data extraction (https://support.esri.com/en-us/knowledge-base/how-to-find-the-centroid-of-polygons-using-calculate-ge-000021849). This includes centroids for:
1. Individual administrative units at each level
2. Protected areas
3. Combined administrative-protected area units
Data entry system. A structured data entry template was developed using spreadsheet functionality to ensure consistent data capture. The template features:
1. Nested dropdown menus for standardized selection of administrative units and protected areas
2. Dynamic validation rules using array formulas to ensure data integrity
3. Automated coordinate assignment based on selected locations
4. Capability to record multiple locations per project

Quality assurance and data management. To maintain data quality and facilitate analysis, several control measures were implemented:

Standardization. All geographic names are standardized against authoritative sources, with relationships maintained between local and international nomenclature.
Flexibility. The data structure accommodates varying levels of geographic specificity, allowing for the recording of locations at any administrative level or protected area, either individually or in combination.
Automation. Python scripts were developed to:
1. Organize and manage project documentation
2. Clean and process spatial data
3. Convert finalized data to GeoJSON format for visualization and analysis
Visualization. Final outputs were visualized using QGIS software, enabling both quality control and analytical representation of the project portfolio’s spatial distribution.

Assessing Land Cover Dynamics

This section outlines the approach used to examine land cover dynamics within specific protected areas. The analysis focuses on assessing annual land cover changes from 2016 to 2024 (the period for which satellite imagery and data are available, as noted in the Limitations section), employing a series of data processing techniques designed to ensure the accuracy and reliability of the results while addressing potential gaps and inconsistencies in the data.

Data Sources

Table A.2. Land Cover Classes

Class	Description
Class 0	Water—Includes both permanent and seasonal water bodies, such as lakes, rivers, and wetlands
Class 1	Trees—Encompasses primary and secondary forests, as well as large-scale plantations
Class 2	Grass—Natural grasslands, livestock pastures, and parks
Class 3	Flooded vegetation—Mangroves and other inundated ecosystems
Class 4	Crops—Includes row crops and paddy crops
Class 5	Shrub and scrub—Sparse to dense open vegetation consisting of shrubs
Class 6	Built area—Low- and high-density buildings, roads, and urban open space
Class 7	Bare ground—Deserts and exposed rock
Class 8	Snow and ice—Permanent and seasonal snow cover

Source: Brown et al. 2022.

Source: Independent Evaluation Group (based on protected area data from the World Database on Protected Areas).

Note: This map has been cleared by the World Bank Group cartography unit.

Data Processing

The land cover analysis was performed using Google Earth Engine with JavaScript, leveraging its powerful processing capabilities for large-scale geospatial data.

Example Analysis: Land Cover Change in Mamunta-Mayosso Protected Area

Land cover trends. Table A.3 presents the percentage distribution of land cover classes within the protected area for each year.

Table A.3. Land Cover Distribution in Mamunta-Mayosso, Sierra Leone (2016–24) (percent)

Year	Trees	Grass	Flooded Vegetation	Crops	Shrubs and Scrub	Built Up	Bare	Water	Snow and Ice
2016	46.46	6.42	0.11	1.70	43.64	0.31	0.19	1.18	0.00
2017	65.82	12.24	0.10	0.38	19.96	0.37	0.01	1.12	0.00
2018	63.05	5.45	0.09	0.32	29.70	0.29	0.02	1.09	0.00
2019	52.05	2.12	0.06	1.46	42.94	0.21	0.07	1.09	0.00
2020	53.22	3.70	0.08	1.31	40.16	0.32	0.09	1.12	0.00
2021	49.01	5.34	0.05	0.79	43.40	0.30	0.06	1.06	0.00
2022	54.94	3.39	0.06	0.84	39.11	0.26	0.31	1.09	0.00
2023	53.37	2.66	0.06	0.76	41.78	0.25	0.04	1.09	0.00
2024	56.37	2.40	0.07	1.11	38.61	0.20	0.14	1.10	0.00

Source: Independent Evaluation Group (based on Dynamic World data, accessed and processed using Google Earth Engine).

Visualization of land cover change. To further illustrate these trends, the stacked bar chart in figure A.23 depicts the relative proportion of each land cover class annually.

Source: Independent Evaluation Group (based on Dynamic World data, accessed and processed using Google Earth Engine).

Source: Independent Evaluation Group (based on Dynamic World data, accessed and processed using Google Earth Engine).

Deep Dives

To supplement and deepen the portfolio analyses, we commissioned industry and legal specialists to conduct thematic deep dives on (i) IPLCs, (ii) LNR, and (iii) biodiversity offsets. The selection of these three issues reflects their significance to the achievement of biodiversity outcomes, and the relative specialization associated with evaluating the aims and results of associated activities. We used emerging findings from these deep dives to guide interviews with senior experts from the World Bank (and IFC and MIGA in the specific case of offsets) in these associated fields to add explanatory power to these sections of the report.

Indigenous Peoples and Local Communities

To capture a holistic view of the engagement of IPLCs and other project-affected groups in the conservation-focused portfolio (see box A.4 for definitions and criteria), we employed a two-pronged approach that evaluates (i) the application of the Operational Policies (Safeguards) and the ESF to assess IPLC engagement, empowerment, and protection from a risk management perspective and (ii) geospatial analysis to provide insights into the spatial distribution of World Bank–supported conservation activities in relation to IPLC lands.

Box A.4. Definitions and Criteria

Indigenous Peoples: The World Bank generally understands Indigenous Peoples to be distinct social and cultural groups that share collective ancestral ties to the lands and natural resources they inhabit or occupy, or from which they have been displaced. Under the Environmental and Social Framework, a group must possess four characteristics for Environmental and Social Standard (ESS) 7 to be applicable. These criteria can be summarized as follows: self-identification as a distinct Indigenous group that is recognized by others, collective attachment to geographically distinct areas, distinct customary institutions, and a distinct language or dialect (ESS7, para. 8).

Sources: ESF ESS7 (Indigenous Peoples/Sub-Saharan African Historically Underserved Traditional Local Communities) Guidance Note; Independent Evaluation Group.

Lastly, to validate our findings and explore explanatory factors, we conducted interviews with relevant staff and management.

Box A.5. Technical Details of Data Sources

The LandMark data set offers a conservative estimate of Indigenous Peoples and local community lands, as it includes only areas that meet strict criteria for documented tenure rights. It prioritizes legally recognized and publicly available data, drawing from government records, nongovernmental organization reports, and contributions from Indigenous organizations. In contrast, the Garnett et al. (2018) data set takes a more expansive approach, integrating multiple sources to provide a broader estimate of Indigenous lands. It relies on peer-reviewed literature, academic books, and reputable data providers while also incorporating spatial data from platforms such as LandMark Global. By aggregating information from a wider range of sources, the Garnett et al. (2018) data set identifies more Indigenous Peoples and local community lands than LandMark, but it may also include territories with varying levels of formal recognition. These methodological differences highlight a key distinction: the LandMark data set offers a more conservative yet authoritative view of Indigenous Peoples and local community lands, whereas the Garnett et al. (2018) data set captures a wider spectrum of Indigenous land tenure, including areas with less formal documentation.

Sources: Garnett et al. 2018; LandMark (https://landmarkmap.org/data-methods/methodology).

Land and Natural Resource Rights

We conducted a review of project documents—including PADs, Implementation Status and Results Reports, and ICRs—focusing on identifying explicit considerations of, and activities supporting, land and resource tenure and rights within project design. Our review centered on several key criteria and LNR activities. First, we assessed the extent to which projects identified potential risks related to restrictions on rural communities’ access to land and resources by examining project documents for explicit acknowledgments of such risks, as well as the application of relevant environmental and social risk management policies, specifically Operational Policy 4.12 (Involuntary Resettlement) and ESS5 (Land Acquisition, Restrictions on Land Use, and Involuntary Resettlement).

We also coded project-financed activities according to the following LNR activities: (i) strengthening community participation and community-led land and resource governance, (ii) supporting tenure regularization or formalization, and (iii) reinforcing legislative or regulatory reforms related to land and resource rights. Additionally, we examined the presence and types of indicators used by projects to track achievements in line with these same activity bundles.

To illustrate the practical implementation and outcomes of LNR considerations, we drew on our Mozambique case study (see also the Exploratory Case Studies section) and a desk-based review of a project in Tanzania. These illustrated achievements in strengthening community land and resource rights, participatory governance, and biodiversity protection, as well as the potential risks and consequences of insufficient consideration of community land and resource rights in project implementation.

Biodiversity Offsets

To derive lessons from these initiatives, we employed a sequenced approach. First, we conducted a focused literature review to identify biodiversity offset good practice principles. Next, we conducted a portfolio review and analysis that (i) assessed the alignment of biodiversity offset development with the Bank Group’s environmental and social requirements, as well as the supervision of client implementation (including survey work, stakeholder engagement, and documentation), and (ii) evaluated the outcomes based on available monitoring and reporting data, focusing on principles such as no net loss or net gain. Finally, we derived explanatory factors through interviews with senior environmental and social experts.

EQ2 Methods: How Well Are the World Bank and IFC Supporting Activities with Potential Biodiversity Benefits in Key Production Sectors, and Are Those Activities Likely to Achieve Such Benefits?

We assessed the World Bank and IFC’s support for integrating biodiversity into three key production sectors: (i) agriculture, including agroforestry and agribusiness; (ii) forestry; and (iii) fisheries and aquaculture. For each of these sectors, we gathered evidence using focused literature reviews and portfolio reviews and analyses, and to provide more explanatory power, we conducted exploratory case studies.

Focused Literature Reviews

Portfolio Reviews and Analyses

Exploratory Case Studies

The case studies were guided by a detailed case protocol (see box A.6) to ensure data collection and analytic consistency across cases. Case authors were experienced researchers with a combination of biodiversity, production sector, country and regional, and evaluation expertise. For each case study, interviews were conducted with relevant World Bank staff, government ministries and agencies, local government, project management and implementation units, regional organizations, local subject matter experts, donor agencies, nongovernmental organizations, civil society, and associations.

Box A.6. Exploratory Case Study Approach

Step 1: Identification of the commitments made by the country to the goals pertaining to sustainable production outlined in the Global Biodiversity Framework.

Step 2: Analysis of the role of the World Bank Group in the country in financing or contributing to sustainable production activities that enhance biodiversity outcomes.

This step involved multiple components, including the following:

The selection of projects identified for data collection as part of the case studies was based on a selection criteria exercise described in the section above.

Step 3: A focused literature review of the biodiversity considerations relevant to each subsector(s) selected in each country.

The literature review included several components: a synthesis of the predominant challenges to achieving biodiversity outcomes in the chosen subsector; a description of the social, economic, and governance constraints to achievement of biodiversity outcomes; the integration of climate change as a cross-cutting consideration; an analysis of the chief long-term constraints to producing sustainable biodiversity outcomes in the subsector (for example, lack of financing, institutional capacity); and good practice examples of achievement of biodiversity outcomes in the chosen subsector(s).

Step 4: An analysis of the specific mechanisms or approaches used in each relevant project by the World Bank that are likely to have biodiversity benefits.

Step 5: The case analysis culminated in a case synthesis and overarching analysis, drawing on the multiple lines of evidence identified.

Source: Independent Evaluation Group.

Review of Core Country Diagnostics

Country Partnership Framework

Alignment with global climate and biodiversity commitments. We conducted parametric keyword searches and content analysis within the 113 CPFs to assess their alignment with global climate and biodiversity agreements. Each CPF was analyzed to determine whether it articulates climate priorities, including nationally determined contributions, the Paris Agreement, or Paris Alignment. Additionally, we examined whether CPFs explicitly mention global biodiversity agreements such as the GBF, the CBD, or the Kunming-Montreal and Aichi Biodiversity Targets. We also identified whether CPFs articulate support linked to their respective National Biodiversity Strategies and Action Plans and Natural Capital Accounting.

Table A.4. Coding Categories for Country Partnership Framework Subobjectives

Thematic Code	Description
Climate Adaptation and Disaster Risk Management	Enhancing resilience against climate change impacts, natural hazards (floods, droughts, hurricanes, earthquakes, and so on), and related risks. Strengthening preparedness, early-warning systems, recovery, and adaptive capacity of infrastructure, livelihoods, and ecosystems. Examples: “Enhance resilience to natural shocks”; “Build resilience to climate-related events”; “Strengthen multihazard disaster resilience.”
Climate Change Mitigation and Low-Carbon Transition	Reducing GHG emissions, promoting decarbonization, supporting low-carbon industries, and contributing to global climate goals (for example, the Paris Agreement, NDC implementation). Examples: “Scale up climate mitigation measures”; “Support the energy transition/reduce energy intensity.”
Natural Resource Management	Sustainable management of forests, fisheries, land, ecosystems, and so on. Includes preserving habitats, reforestation, sustainable forestry, land degradation prevention, or conservation of wildlife. Examples: “Improved management of natural resources”; “Preserve and restore natural capital”; “Sustainable landscape management.”
Water, Sanitation, and Waste Management	Ensuring water security, water resource management, sanitation, solid waste management, wastewater treatment, and related infrastructure. Examples: “Improve access to water, sanitation, and solid waste management”; “Enhance water security and sustainability”; “Efficient water resource management for resilience.”
Energy Sustainability and Efficiency	Increasing renewable energy capacity (solar, wind, hydro, geothermal), promoting energy efficiency, and ensuring a sustainable energy supply. Often linked to low-carbon development but can be distinguished if the focus is specifically on energy systems. Examples: “Expand clean energy matrix”; “Enhanced energy sustainability and renewable energy resources”; “Energy efficiency improvements in utilities and industries.”
Pollution and Air Quality Management	Reducing ambient air pollution, industrial pollution, marine plastic pollution, and soil contamination, or improving broader environmental quality (apart from GHG-focused mitigation). Examples: “Reduce air pollution in urban centers”; “Address plastic waste in coastal regions”; “Improve air, soil, water pollution control frameworks.”
Resilient Infrastructure and Urban Resilience	Resilience and sustainability in roads, transport, housing, and urban planning. Ensuring cities adapt to climate change impacts (sea-level rise, flooding), reduce congestion/pollution, and strengthen livability. Examples: “Improved transport connectivity and safety” (with an explicit climate/disaster lens); “Promote green and resilient cities”; “Sustain and strengthen urban infrastructure to withstand climate shocks.”
Financial and Macroeconomic Resilience	Strengthening a country’s financial capacity to cope with climate-related or disaster-related shocks. Building fiscal buffers, insurance mechanisms, contingent financing, or macrofiscal strategies for climate resilience. Examples: “Enhanced financial resilience”; “Strengthen capacity for macrofinancial sustainability/reduce vulnerability to external shocks”; “Risk financing for disasters/catastrophe insurance frameworks.”
Governance and Institutional Capacity for Environment/ Climate	Public sector reforms, policies, regulations, or institutional frameworks specifically geared toward better environmental or climate outcomes, for example, setting up climate governance structures, environmental agencies, or cross-sector coordination. Examples: “Improve government’s effectiveness, efficiency, and transparency in climate/natural resource management”; “Strengthen institutional and financial framework for risk management”; “Enhance capacity for climate finance and green budgeting.”
Productivity, Climate-Smart Agriculture, and Resilient Livelihoods	Focus on making agriculture, fisheries, or rural livelihoods more resilient and sustainable under climate variability (crop diversification, water-efficient irrigation, sustainable land practices, and so on). Examples: “Promote climate-resilient agriculture”; “Strengthen rural livelihoods through sustainable land and water use.”
Biodiversity	Explicit mention of biodiversity in objective. Examples: “Improved management of mining, natural resources, and biodiversity”; “Improved and climate-adaptive management of forests, biodiversity, and protected areas.”
Social Protection and Safety Nets	Strengthening household and community resilience to shocks by enhancing social protection systems, improving livelihoods, and building crisis preparedness and response capacities, particularly for vulnerable and conflict-affected populations. Examples: “Strengthen crisis resilience for vulnerable, displaced, and conflict-affected populations”; “Improve efficiency and effectiveness of the social protection system”; “Strengthen mechanisms to protect people against shocks.”
Land Administration	Example: “Improve Land Titling and Administration.”

Source: Independent Evaluation Group.

Note: GHG = greenhouse gas; NDC = nationally determined contribution.

Country Climate and Development Reports

Limitations

Portfolio Review and Analysis

Geospatial Analysis

Missing protected area names. In 25 of the 130 conservation-focused projects that include support to protected areas, it was not possible to identify the names of the specific protected areas as they were not included in any project documentation. Furthermore, in 279 cases (out of 884), we could not match the names of the protected areas as found in the respective project documentation with the WDPA (despite extensive efforts to do so, including searches online). Last, 19 protected areas did not have shape files and were therefore excluded from the land cover analysis.

Probabilistic nature of the Dynamic World model. The land cover classifications used in this study are derived from the Dynamic World data set, which is a probabilistic model rather than a deterministic classification. This means that each pixel is assigned a probability of belonging to different land cover classes, and the final classification represents the most likely class rather than a definitive ground-truth label. As a result, some classification uncertainty is expected, particularly in transitional or heterogeneous landscapes where multiple land cover types coexist within a single pixel. Several recent studies have analyzed recently developed global land-use/land cover models (Kerner et al. 2024; Venter et al. 2022; Wang et al. 2023), and we refer to these resources for a more detailed assessment of the limitations.
Impact of missing data and cloud cover. Optical satellite data, such as those used in Dynamic World, are susceptible to cloud cover, haze, and other atmospheric interferences, which can result in missing data for certain time periods. Although compositing techniques help mitigate these gaps, some areas may have fewer valid observations, potentially introducing noise or inconsistencies in the temporal analysis. This issue is particularly relevant in tropical regions, where persistent cloud cover can reduce the frequency of high-quality observations.
Although Dynamic World, which relies on Sentinel-2 L1C images, excludes images with cloud cover greater than 35 percent from its land cover calculations, we observed that in some years the coverage was incomplete, likely due to persistent cloud cover. To address potential data quality issues, we implemented the following decision rule: (i) if a protected area-year had fewer than 80 percent valid pixels, it was excluded from the analysis, and (ii) if a protected area-year had 80 percent or more valid pixels, it was included, with weights assigned according to the percentage of valid pixels. This approach allowed us to calculate the weighted average percentage of land cover for two periods, 2016–19 and 2020–24, and to estimate the change in land cover as the difference between these two periods.
Influence of mixed pixels and land cover transitions. Given the 10-meter spatial resolution of the data set, some pixels are likely to contain a mixture of multiple land cover types. This can lead to misclassifications, especially in areas where tree cover, shrubs, and grasslands intermingle. Additionally, land cover transitions (for example, forest degradation, regrowth, or seasonal changes in vegetation) may not be fully captured if the dominant class in a pixel does not change significantly over time.
Spatial and temporal resolution considerations. Although the 10-meter spatial resolution of Dynamic World allows for relatively fine-scale analysis, it may not detect small-scale changes such as selective logging, understory degradation, or small agricultural clearings. Additionally, the temporal resolution of the analysis is limited by the availability of cloud-free observations, meaning that some short-term disturbances or rapid land cover changes may not be fully captured.

Biodiversity Offsets Analysis

The evaluation team relied on different approaches to identify offset projects across institutions (IFC, MIGA, and the World Bank) because of limitations in access to documents and resource constraints. Although the list of MIGA and IFC projects with offsets was retrieved from ESG counterparts, the team relied on the Standard Reports and a combination of manual and large language model assessment. The World Bank does not have a centralized tracking tool for offsets, but the evaluation team met with several biodiversity experts in the World Bank and received their input on the list of projects.

References

Brown, C. F., S. P. Brumby, B. Guzder-Williams, et al. 2022. “Dynamic World, Near Real-time Global 10 m Land Use Land Cover Mapping.” Scientific Data 9: 251.

Garnett, S. T., N. D. Burgess, J. E. Fa, et al. 2018. “A Spatial Overview of the Global Importance of Indigenous Lands for Conservation.” Nature Sustainability 1 (7): 369–74.

Kerner, H., C. Nakalembe, A. Yang, et al. 2024. “How Accurate Are Existing Land Cover Maps for Agriculture in Sub-Saharan Africa?” Scientific Data 11 (1): 486.

World Bank. 2018. “ESF Guidance Note 7: Indigenous Peoples/Sub-Saharan African Historically Underserved Traditional Local Communities (ESS7).” World Bank.

World Bank. 2024. Biodiversity for a Livable Planet: An Evaluation of World Bank Group Support for Biodiversity (FY15–24). Approach Paper. Independent Evaluation Group. World Bank.

World Bank Group. 2021. Unlocking Nature-Smart Development: An Approach Paper on Biodiversity and Ecosystem Services. World Bank Group.

Article
Blog
comment compare
Custom decscriptions
Data
Evaluation
Multimedia
Event
Expert
General Documents
Homepage spotlight feature
Home page content spotlight
ICRR Reports
IEG Timeline
MAR
News
Basic page
Podcast
Reader chapter
Reader publication
Reports
Series
Survey Banner
Topic
Upcoming Report
Upload Mar
Xml Import

Appendix A. Evaluation Methods

Evaluation Design and Questions

Figure A.1. Evaluation Design

Figure A.1. Evaluation Design

Evaluation Scope

Evaluation Portfolio Identification and Classifications

EQ1. Biodiversity Conservation

World Bank

International Finance Corporation and Multilateral Investment Guarantee Agency

EQ2. Biodiversity Integration in Key Production Sectors

World Bank Lending

International Finance Corporation Investments

International Finance Corporation Advisory

Biodiversity Offsets

Portfolio Description

EQ1. Biodiversity Conservation

Figure A.2. World Bank Conservation Portfolio by Global Practice and Project Status

Figure A.2. World Bank Conservation Portfolio by Global Practice and Project Status

Figure A.3. World Bank Conservation Portfolio by Region

Figure A.3. World Bank Conservation Portfolio by Region

Figure A.4. World Bank Conservation Portfolio by Approval Year

Figure A.4. World Bank Conservation Portfolio by Approval Year

EQ2. Biodiversity Integration in Key Production Sectors

World Bank Lending

Figure A.5. World Bank Agriculture Portfolio by Lending Instrument and Project Status

Figure A.5. World Bank Agriculture Portfolio by Lending Instrument and Project Status

Figure A.6. World Bank Agriculture Portfolio by Region

Figure A.6. World Bank Agriculture Portfolio by Region

Figure A.7. World Bank Agriculture Portfolio by Approval Year

Figure A.7. World Bank Agriculture Portfolio by Approval Year

Figure A.8. World Bank Forest Production Portfolio by Global Practice

Figure A.8. World Bank Forest Production Portfolio by Global Practice

Figure A.9. World Bank Forest Production Portfolio by Region

Figure A.9. World Bank Forest Production Portfolio by Region

Figure A.10. World Bank Forest Production Portfolio by Approval Year

Figure A.10. World Bank Forest Production Portfolio by Approval Year

Figure A.11. World Bank Fisheries Portfolio by Global Practice and Project Status

Figure A.11. World Bank Fisheries Portfolio by Global Practice and Project Status

Figure A.12. World Bank Fisheries Portfolio by Region

Figure A.12. World Bank Fisheries Portfolio by Region

Figure A.13. World Bank Fisheries Portfolio by Approval Year

Figure A.13. World Bank Fisheries Portfolio by Approval Year

International Finance Corporation Investments and Advisory

Biodiversity Offsets

Figure A.14. World Bank Biodiversity Offsets Portfolio by Global Practice and Project Status

Figure A.14. World Bank Biodiversity Offsets Portfolio by Global Practice and Project Status

Figure A.15. World Bank Biodiversity Offsets Portfolio by Approval Year

Figure A.15. World Bank Biodiversity Offsets Portfolio by Approval Year

Figure A.16. World Bank Biodiversity Offsets Portfolio by Region

Figure A.16. World Bank Biodiversity Offsets Portfolio by Region

Figure A.17. International Finance Corporation Biodiversity Offsets Portfolio by Approval Year

Figure A.17. International Finance Corporation Biodiversity Offsets Portfolio by Approval Year

Figure A.18. International Finance Corporation Biodiversity Offsets Portfolio by Region

Figure A.18. International Finance Corporation Biodiversity Offsets Portfolio by Region

Figure A.19. Multilateral Investment Guarantee Agency Biodiversity Offsets Portfolio by Global Practice and Project Status

Figure A.19. Multilateral Investment Guarantee Agency Biodiversity Offsets Portfolio by Global Practice and Project Status

Figure A.20. Multilateral Investment Guarantee Agency Biodiversity Offsets Portfolio by Approval Fiscal Year

Figure A.20. Multilateral Investment Guarantee Agency Biodiversity Offsets Portfolio by Approval Fiscal Year

Figure A.21. Multilateral Investment Guarantee Agency Biodiversity Offsets Portfolio by Region

Figure A.21. Multilateral Investment Guarantee Agency Biodiversity Offsets Portfolio by Region

EQ1 Methods: How Well Is the World Bank Achieving Biodiversity Aims Through Its Conservation-Focused Activities?

Focused Literature Review and Review of World Bank Group Strategies and Commitments

Portfolio Review and Analysis

Geospatial Analysis

Identifying Protected Areas

Data Preparation

Assessing Land Cover Dynamics

Data Sources

Figure A.22. Protected Areas (Including a 5-Kilometer Buffer) Included in the Analysis

Figure A.22. Protected Areas (Including a 5-Kilometer Buffer) Included in the Analysis

Figure A.23. Annual Land Cover Composition in Mamunta-Mayosso, Sierra Leone (2016–24)

Figure A.23. Annual Land Cover Composition in Mamunta-Mayosso, Sierra Leone (2016–24)

Figure A.24. Annual Land Cover Maps of Mamunta-Mayosso (2016–24)

Figure A.24. Annual Land Cover Maps of Mamunta-Mayosso (2016–24)

Deep Dives

Indigenous Peoples and Local Communities

Land and Natural Resource Rights

Biodiversity Offsets

EQ2 Methods. How Well Are the World Bank and IFC Supporting Activities with Potential Biodiversity Benefits in Key Production Sectors, and Are Those Activities Likely to Achieve Such Benefits?

Focused Literature Reviews