Public-interest research and decision-making increasingly draw on digital traces, earth observation, community-generated information, and other sources created outside conventional statistical and administrative systems. The cases examined in this update show how these non-traditional data sources can reveal conditions that are difficult to observe through established channels alone—from methane emissions and forest loss to patterns of mobility, financial inclusion, infrastructure vulnerability, and access to public services. They also show the importance of combining non-traditional data with surveys, official records, field observations, and geospatial information to interpret these signals and connect them to real-world conditions.

This post is the latest in our series tracking uses of non-traditional data in the public interest (see the previous update here, and see a compilation of 100 use cases here). The 25 examples explored below, primarily published or released between May and August 2026, span different stages of application, including demonstrated applications and emerging analytical capacity alike. Some document operational systems, while others inform planning and resource allocation, test substantive pilots, or develop methods for future use. 

The update is organized into four sections: 

• Mobility, Cities, and Infrastructure

• Crisis Response, Health, and Public Services

• Social and Economic Conditions, Markets, and Financial Inclusion

• Environment, Climate, Natural Resources, and Development Operations

Each entry describes the problem being addressed, explains how non-traditional data contributes to the work, and considers its public-interest significance.

Mobility, Cities, and Infrastructure 

Moulton, Cyrus. “‘Giving Back’ Cellphone Data so Communities Can Plan.” Northeastern Global News. May 22, 2026. https://news.northeastern.edu/2026/05/22/cellphone-data-mobility-for-communities 

Focus: Introduces Mobility Data for Communities (MD4C), a platform for examining who visits places in Massachusetts, where visitors come from, and how they use those areas.

Role of Non-Traditional Data: MD4C analyzes aggregated and anonymized cellphone location data to estimate movement within and between communities. The platform describes visitor origins, travel patterns, activities, visit duration, demographic characteristics, and likely transportation modes. These measurements create a localized picture of the population present in an area that complements the residential information available through the U.S. Census.

Why This Matters: Detailed mobility data is often expensive, technically demanding, and inaccessible to municipalities and community organizations. MD4C translates it into a community-oriented planning resource that could help local users assess infrastructure needs, plan events, support businesses, allocate resources, and understand economic activity. At publication, the platform was in its initial stage and contained only 2022 data for Massachusetts. 

Oughton, Edward J., Tom Russell, Jeongjin Oh, Sara Ballan & Jim W. Hall. “Global Vulnerability Assessment of Mobile Telecommunications Infrastructure to Climate Hazards Using Crowdsourced Open Data.” Nature Communications. August 3, 2026. https://doi.org/10.1038/s41467-026-76197-w 

Focus: Assesses the exposure of mobile telecommunications infrastructure to coastal flooding, riverine flooding, and tropical cyclones under current and future climate conditions.

Role of Non-Traditional Data: The researchers used crowdsourced OpenCelliD records covering approximately 17 million 2G, 3G, and 4G base stations worldwide. They combined these locations with global flood and simulated cyclone hazards, estimates of infrastructure vulnerability, and reconstruction costs to calculate exposed assets and potential direct damage under different probabilities and emissions scenarios. The inputs, results, and analytical code are publicly available.

Why This Matters: Under a high-emissions scenario, an extreme tropical cyclone in 2050 could affect 3.7 million base stations and cause US$4.22 billion in direct damage. An equivalent coastal flood in 2080 could affect 268,000 stations and cause US$6.44 billion. The World Bank has used the research as background evidence for green digital transformation and country climate reports. 

Kennebeck, Kathryn A., Mason Smetana, Igor Sukharev & Lev Khazanovich. “AI-Powered Community Insights for Strategic Physical Transportation Infrastructure Management.” Safety and Mobility Advancements Regional Transportation and Economics Research Center, Report SM09. May 5, 2026. https://rosap.ntl.bts.gov/view/dot/92774 

Focus: Examines how transportation agencies can use public comments to identify infrastructure concerns that conventional asset-condition measurements may overlook.

Role of Non-Traditional Data: The study analyzed 925 comments collected by the Southwestern Pennsylvania Commission concerning roads, bridges, transit, sidewalks, and traffic safety. AI-assisted analysis organized the comments into 14 themes, including road repairs, missing sidewalks, dangerous crossings, and bicycle safety. Researchers linked 864 comments to census tracts, combined them with neighborhood-income and roadway-roughness measurements, and manually reviewed the automated groupings.

Why This Matters: Engineering measurements can identify pavement deterioration, while community feedback reveals how infrastructure affects safety, accessibility, and daily travel. Comments highlighted missing or narrow sidewalks, dangerous crossings, drainage and erosion risks, and bicycle safety concerns. AI can help agencies process this input and identify issues requiring inspection or engagement. 

Verma, Priyanka, Dan Qiang & Grant McKenzie. “Expanding the 15-Minute City through E-Micromobility Services.” Journal of Geographical Systems. August 10, 2026. https://doi.org/10.1007/s10109-026-00513-7

Focus: Examines whether shared e-bikes and e-scooters expand access to supermarkets, schools, hospitals, parks, and entertainment within a 15-minute journey.

Role of Non-Traditional Data: The researchers reconstructed approximately 24.8 million trips in Berlin, London, Paris, Washington, D.C., and Wellington using public vehicle-availability data from Tier, Lime, and Flamingo. They inferred origins, destinations, and travel times by tracking vehicle identifiers through operator interfaces at 60-second intervals. These trips were combined with OpenStreetMap amenities and street networks and compared with estimated walking times.

Why This Matters: E-micromobility produced the largest accessibility gains in less compact cities and peripheral neighborhoods. In Wellington, the share of the city within 15 minutes of entertainment increased from 6% on foot to 44% by e-micromobility, while hospital access rose from 11% to 33%. Gains were smaller in compact cities such as Paris and Berlin. 

Fork, David, Elizabeth J. Wesley, Salil Banerjee et al. “Estimating High-Resolution Albedo for Urban Applications.” Nature Communications, Volume 17, Article 4815. June 22, 2026. https://www.nature.com/articles/s41467-026-73436-y 

Focus: Develops a method for measuring individual rooftops’ reflectivity and identifying where reflective “cool roofs” could contribute to urban cooling.

Role of Non-Traditional Data: The researchers combined free Sentinel-2 satellite imagery, higher-resolution commercial imagery, and digital building footprints to estimate rooftop reflectivity at 30-centimeter resolution. They compared the results with airborne measurements over Boulder, Colorado, and applied the method across 12 cities. The resulting maps supported comparisons between converting all roofs, the largest roofs, or the darkest roofs to more reflective materials.

Why This Matters: Targeting roofs above the 90th percentile in building area often produced more than half of the modeled reflectivity increase from converting every roof, suggesting that fewer projects could capture much of the potential benefit. Cooling of up to 0.5°C applies to the modeled all-roof scenario and uses a relationship from earlier studies; it was not observed after implementation. 

Yang, Jinming, Shaoyu Huang, Zongyuan Huang, Yaohui Jin, Xiaokang Yang, Marta C. González & Yanyan Xu. “Transferable Human Mobility Network Reconstruction with neuroGravity.” Nature Computational Science, Volume 6, pp. 630–641. June 12, 2026. https://www.nature.com/articles/s43588-026-01003-y 

Focus: Develops neuroGravity, a deep learning model that reconstructs movement flows from limited observations and transfers learned patterns to cities without detailed mobility data.

Role of Non-Traditional Data: The researchers constructed mobility networks for six cities using anonymized mobile phone call records and location observations from mobile applications. They combined these traces with population estimates and 52 OpenStreetMap indicators describing buildings, land use, roads, and points of interest. The model learned how these local characteristics relate to observed movement and then estimated flows using only a small sample of local observations.

Why This Matters: Many cities lack the detailed mobility information needed for transport, infrastructure, and public-health planning. The study generated estimated mobility networks for more than 1,200 cities and compared global estimates with travel surveys in two sub-Saharan African regions. Transferability weakened where cities differed in spatial income segregation, and incomplete OpenStreetMap coverage can affect the inputs. Local observations and validation therefore remain important before the estimates inform decisions.

Crisis Response, Health, and Public Services

Marín Villagrana, Mar, Camila Garzon-Ruiz, Jessica Pechmann et al. “The Night Venezuela Shook: How Open Mapping Data Made the Emergency Response Faster.” Humanitarian OpenStreetMap Team. July 9, 2026. https://www.hotosm.org/en/news/the-night-venezuela-shook-how-open-mapping-data-made-the-emergency-response-faster/ 

Fernández, Daniella & Andrea Paola Hernández. “AI Powers Citizen-Led Disaster Relief from Afar for Venezuela.” Rest of World. July 15, 2026. https://restofworld.org/2026/venezuela-ai-citizen-disaster-response/ 

Focus: Documents how mapping groups, volunteer developers, and affected communities created emergency information after the June 2026 earthquakes in Venezuela.

Role of Non-Traditional Data: The Humanitarian OpenStreetMap Team combined satellite imagery, AI-assisted building detection, and volunteer review to map affected areas. About 590 contributors traced nearly 97,000 building footprints. WhatsApp submissions added photographs, videos, and locations of damaged buildings, roads, and shelters, which trained volunteers checked against imagery. Other citizen-led platforms organized social media posts and public submissions concerning missing people, hospital capacity, shelters, supplies, and requests for assistance.

Why This Matters: The resulting maps and datasets filled gaps in official information during the response. HOT reported more than 360 downloads of its open data, including use by the International Organization for Migration and MapAction to support food-delivery logistics. The case shows how satellite imagery, community reports, AI-assisted analysis, and volunteer verification can rapidly produce actionable emergency information. It also illustrates the need for security controls and human oversight when platforms process sensitive location, health, or biometric information.

Abdul Latif Jameel Poverty Action Lab. “Using Machine Learning and Mobile Phone Data to Improve the Speed and Cost-Effectiveness of Social Protection.” June 2026. https://www.povertyactionlab.org/case-study/using-machine-learning-and-mobile-phone-data-improve-speed-and-cost-effectiveness-social

Focus: Examines how mobile phone data, satellite imagery, and machine learning have helped identify vulnerable households and deliver emergency cash assistance.

Role of Non-Traditional Data: In Togo, researchers combined surveys with high-resolution satellite imagery to estimate wealth across small areas and identify poor cantons. They then linked living-conditions surveys with anonymized mobile phone records, including calling patterns and airtime purchases, to predict household consumption. The government used these estimates to target the rural expansion of its Novissi cash transfer program. Related approaches were later used in Malawi, Bangladesh, and the Democratic Republic of the Congo.

Why This Matters: The system helped identify and enroll more than 138,500 recipients in Togo and reduced the erroneous exclusion of poor households by an estimated 4–21% compared with geographic targeting. A Malawi pilot reached 12,800 households, shortened delivery from nine months to 3.5 months, and reduced operational costs by 40%. These applications show how digital traces can accelerate assistance when registries are outdated, while complementary enrollment methods remain necessary for households without mobile phones.

HeiGIT. “Sketch Map Tool Use-Case in the EVCA Context.” July 8, 2026. https://heigit.org/sketch-map-tool-use-case-in-the-evca-context/

Focus: Examines how the Colombian Red Cross incorporated participatory mapping into community assessments and local disaster risk management plans.

Role of Non-Traditional Data: Residents of Pacelli, Monserrate, and Cubís marked flood-prone areas, historical water levels, evacuation routes, safe zones, supply infrastructure, and community capacities on printed OpenStreetMap maps. The Sketch Map Tool converted these markings into georeferenced digital information. Red Cross staff combined the results with historical risk profiles, interviews, seasonal calendars, and field observations and checked the maps through additional site visits.

Why This Matters: The approach enabled communities without specialized equipment or mapping expertise to contribute local knowledge in an analyzable and reusable form. Following an initial pilot, the Colombian Red Cross applied the method in three communities over two years. HeiGIT reports that the information helped identify critical areas and response resources and entered formal Community Risk Management Plans. Automated conversion still required manual correction, field validation, and coordination between facilitators and analysts to preserve the meaning of residents’ contributions.

Clark, Benjamin Y. & Shahinur Bashar. “How IoT Enables the Protective Action Decision Model (PADM) During a Wildfire Event.” Risk Analysis, Volume 46, Issue 9, Article e70319. August 20, 2026. https://doi.org/10.1111/risa.70319 

Focus: Examines whether access to hyperlocal air-quality information was associated with protective behavior during Oregon’s severe 2020 wildfire season.

Role of Non-Traditional Data: Low-cost PurpleAir sensors operated by residents, organizations, and public agencies generated real-time measurements of fine particulate matter. These readings appeared on PurpleAir and were incorporated into the U.S. government’s AirNow maps alongside regulatory measurements. Researchers surveyed 1,200 Oregon residents, including targeted rural and Spanish-speaking samples, about their use of this information and actions involving filtration, masking, and temporary relocation.

Why This Matters: About 28% of respondents used AirNow or PurpleAir. After accounting for demographic, geographic, and economic factors, users had approximately 2.6 times the odds of filtering indoor air and masking indoors, and 1.5 times the odds of masking outdoors. Use was not associated with evacuation, and older and rural residents were less likely to access the information. The findings support combining localized readings with accessible communication and targeted assistance. Because the survey was retrospective and cross-sectional, it demonstrates association, not causation.

Ma, Evelyn, Rama Kumar Pasumarthi, Kishwar Shafin et al. “Planetary Prediction Engine: Autonomous Geospatial Prediction via Intelligent Data Selection and Foundation Model Embeddings.” arXiv preprint. August 26, 2026. https://arxiv.org/abs/2608.26088

Focus: Develops an AI system that assembles geospatial data and builds predictive models for disease outbreaks, food security, public health, and environmental risks.

Role of Non-Traditional Data: The Planetary Prediction Engine automatically combined official records with satellite observations, mobility estimates, digital maps, news indicators, and general-purpose representations derived from geospatial imagery. For the 2026 Ebola outbreak in the Democratic Republic of the Congo, it integrated case reports with Flowminder relocation estimates, OpenStreetMap roads, health facility locations, population and vulnerability indicators, and satellite-derived climate and built-environment information. The work involved the country’s national biomedical research institute.

Why This Matters: Across five weekly forecasts, the system placed 15 of 18 health zones that subsequently reported their first Ebola cases among its ten highest-risk predictions. It also produced more detailed food security estimates in Nigeria using surveys, prices, rainfall, vegetation, nighttime lights, and news sentiment. Automated data integration could accelerate localized evidence where conventional information is sparse. 

Juma, Rachel. “Heard, Not Just Counted: Moving Beyond the Dashboards…” Open Institute. August 18, 2026. https://openinstitute.africa/2026/08/18/moving-beyond-the-dashboards/

Focus: Examines how citizen-generated assessments of water, health, and education projects in Nandi County, Kenya, entered government oversight and budgeting processes.

Role of Non-Traditional Data: A citizen scorecard and dashboard covered 418 public projects and facilities in water, health, and education, with 3,876 ratings across sector-specific measures of access, quality, and service availability. Community validation sessions added context that the standardized ratings missed, including informal arrangements for managing water schemes and shortcomings in disability-inclusive design.

Why This Matters: The findings were converted into a policy brief and presented to the County Assembly’s Deputy Clerk for referral to relevant committees during the supplementary budget process. Recommendations included allocating resources across subcounties using population and service accessibility. The case shows citizen-generated evidence entering a defined government process and reaching officials responsible for budget scrutiny. 

Bahadorizadeh, Hossein & Mohammad Reza Malek. “Leveraging Social Media and Vulnerability Maps for Post-Flood Event Localization.” Journal of Computational Social Science, Volume 9, Article 47. June 29, 2026. https://link.springer.com/article/10.1007/s42001-026-00478-z 

Focus: Tests whether analysts can estimate the locations of flood damage and related events described in social media posts that lack geographic tags.

Role of Non-Traditional Data: The researchers retrospectively examined 150,000 posts published during the April 2019 floods in southwestern Iran. They identified flood-related reports and estimated their locations by combining place references in the text with posting time, nearby events, flood-vulnerability indicators, and a structured database of geographic relationships. The method clustered related observations before estimating locations for reports without geographic tags.

Why This Matters: The method identified flood-related events with 70% precision and 77% recall and located selected reports with an average error of 2.15 kilometres. Social media could help analysts identify reports requiring verification when conventional information is delayed or incomplete. 

Social and Economic Conditions, Markets, and Financial Inclusion 

International Finance Corporation. “Cracking the Credit Code: Alternative Data and AI for Financial Inclusion.” May 7, 2026. https://www.ifc.org/en/insights-reports/2026/cracking-the-credit-code-alternative-data-and-ai-for-financial-inclusion 

Focus: Examines how lenders use alternative data and AI to assess people and small businesses without conventional credit histories in emerging markets.

Role of Non-Traditional Data: The report identifies 448 firms using sources such as mobile-money records, digital-wallet transactions, utility payments, point-of-sale activity, telecommunications indicators, platform histories, geolocation, and psychometric assessments. It also analyzes more than one million microloan applications assessed by Eshandi in Zambia and surveys over 7,000 Vexi customers in Mexico. These cases show how digital behavior and transactions inform approvals, loan amounts, and credit limits.

Why This Matters: People working in informal economies may have stable income and reliable financial behavior without appearing in conventional credit records. Alternative data can make these patterns visible and expand access to formal credit. Women received more repeat loans in Zambia and experienced faster credit-limit growth in Mexico, although these findings are descriptive. 

von Carnap, Tillmann, Reza M. Asiyabi, Paul Dingus & Anna Tompsett. “Using Satellite Imagery to Map Rural Marketplaces and Monitor Their Activity at High Frequency.” Nature Communications, Volume 17, Article 6180. May 9, 2026. https://doi.org/10.1038/s41467-026-72865-z 

Focus: Develops a method for locating periodic rural marketplaces and tracking their activity in remote regions where reliable market information is limited.

Role of Non-Traditional Data: The researchers used repeated satellite images to detect visual differences between market and non-market days, including changes in crowds, stalls, and vehicles. After testing the approach with known markets in Kenya, Malawi, and Mozambique, they combined Sentinel-2 and higher-resolution PlanetScope imagery to identify 1,776 markets across Ethiopia and monitor activity as frequently as once a week between 2017 and 2024.

Why This Matters: Rural markets support livelihoods and food access but are often absent from official maps and statistics. The satellite-derived measurements followed agricultural seasons and rainfall patterns and registered declines during conflict and COVID-19 restrictions. These signals could complement food security and humanitarian monitoring where surveys are delayed, unavailable, or unsafe. 

Australian Bureau of Statistics. “Household Consumption of Illicit Tobacco and Nicotine Products.” June 3, 2026. https://www.abs.gov.au/articles/household-consumption-illicit-tobacco-and-nicotine-products. See also Noonan, Andie, Josh Byrd, and Georgina Piper. “Charting the Rise of Illegal Tobacco in Australia.” ABS News, August 9, 2026. https://www.abc.net.au/news/2026-08-10/charting-the-rise-of-illegal-tobacco-in-australia/106941130. 

Focus: Develops experimental estimates of illicit tobacco and nicotine consumption in Australia to improve understanding of an economic activity largely absent from official statistics.

Role of Non-Traditional Data: The Australian Bureau of Statistics used quarterly measurements of nicotine metabolites from 60 wastewater treatment plants covering approximately 60% of Australia’s population. Wastewater provided an estimate of total nicotine consumption but could not identify the products consumed. The ABS therefore used supermarket scanner data, tax records, surveys, and statistical modelling to estimate legal consumption, allocate the remaining nicotine across illicit cigarettes, e-cigarettes, and other products, and convert these quantities into household-expenditure estimates.

Why This Matters: Illicit markets are difficult to measure because transactions are deliberately concealed and remain excluded from Australia’s official national accounts. The experimental estimates indicate that the share of nicotine consumption attributed to illicit sources increased from 12% in 2017 to 80% in 2025. The work will inform the Illicit Tobacco and E-cigarette Commissioner’s next report and could support economic, public health, taxation, and enforcement analysis. 

Iacus, Stefano M., Devika Jain, Andrea Nasuto, Giuseppe Porro, Marcello Carammia & Andrea Vezzulli. “The Human Flourishing Geographic Index: A County-Level Dataset for the United States, 2013–2023.” Scientific Data, Volume 13, Article 1234. August 26, 2026. https://www.nature.com/articles/s41597-026-07803-1 

Focus: Introduces the Human Flourishing Geographic Index, which tracks 48 dimensions of well-being across U.S. counties and states between 2013 and 2023.

Role of Non-Traditional Data: The researchers analyzed approximately 2.6 billion geolocated U.S. posts from X. A language model classified the posts across indicators covering happiness, health, purpose, relationships, financial stability, migration attitudes, and perceived corruption. A human-reviewed sample was used to train and validate the classifications, and the results were aggregated into monthly and annual county- and state-level measures without releasing individual information.

Why This Matters: Surveys remain important for measuring well-being but are costly and often provide limited local or frequently updated results. The index offers a complementary way to examine how expressions of well-being vary across places and change over time, supporting research on social conditions, inequality, health, and major events. Because people who post geolocated content are not representative of the wider population, the indicators measure online expression and should be interpreted alongside surveys and other established evidence.

Shao, Yijia, Dora Zhao, Vishakh Padmakumar, Jennifer Wang & Diyi Yang. “Human–AI Collaboration at Scale: Task Criticality, Agency, and Friction Across 250,000 Conversations.” Preprint. August 21, 2026. https://www.alphaxiv.org/abs/2608.human-ai-collaboration-at-scalev1. See also Handa, Kunal, Miranda Zhang, Gabriel Nicholas et al. “Enabling Independent Research on How People Use Claude.” Anthropic. August 26, 2026. https://www.anthropic.com/research/enabling-independent-research. 

Focus: Examines how people collaborate with a commercial generative-AI system and tests a controlled-access mechanism for external research using proprietary platform data.

Role of Non-Traditional Data: Stanford researchers analyzed 249,834 real Claude.ai conversations from April and May 2026 to examine task criticality, human agency, learning, and interaction friction. Anthropic Insights applied researcher-defined classifications to the conversations and returned aggregate results without giving researchers access to transcripts or identifiers. Researchers at Oxford and METR used the same mechanism for separate studies, and Anthropic released the projects’ aggregate outputs under a CC BY 4.0 licence.

Why This Matters: Evidence about real-world AI use is largely held by technology companies. This pilot shows how external researchers can investigate sensitive platform data without receiving the underlying conversations. The Stanford study found that 56% of conversations involving actionable tasks concerned consequential work, human-led collaboration predominated, and friction occurred in approximately half of conversations.

Fletcher, Daniel, Gavin Long, Joanne Parkes, Evgeniya Lukinova, John Harvey, James Goulding, and Alexa Spence. “Consumers’ Environmental Impact Perceptions Are Sensitive to Broad Food Categories Purchased but Not Within-Category Purchasing Patterns.” Research Square, August 3, 2026. https://doi.org/10.21203/rs.3.rs-9901985/v1 

Focus: Examines how accurately consumers understand the environmental footprint of their grocery purchases and where more specific information might improve those judgments.

Role of Non-Traditional Data: The researchers linked voluntarily donated Tesco Clubcard records from 947 UK shoppers with environmental estimates for thousands of products. Twelve months of purchases were converted into an index covering greenhouse gas emissions, land use, water pollution, and water scarcity. The transaction-based results were then compared with survey responses about participants’ diets and perceived environmental impact.

Why This Matters: Self-reported and transaction-based impacts were only weakly related. Shoppers recognized broad differences between meat-heavy and plant-heavy baskets but often missed important distinctions within categories, especially between beef and chicken and between cheese and milk. The findings can inform more precise labels and personalized feedback, including the project’s food footprint calculator. 

Environment, Climate, Natural Resources, and Development Operations 

United Nations Environment Programme. “Spotlighting Opportunity: How Artificial Intelligence Is Accelerating Methane Action.” Nairobi: UNEP. July 2026.  https://wedocs.unep.org/items/dccb17ae-c9d2-4f50-bd0e-82bba9bfa8af 

Focus: Examines how UNEP’s Methane Alert and Response System (MARS) identifies major methane releases, alerts governments and companies, and verifies subsequent repairs.

Role of Non-Traditional Data: MARS integrates observations from more than 30 satellite instruments monitoring methane emissions worldwide. AI models screen this large stream of imagery for likely methane plumes, allowing specialists to process 12–15 times more data than manual analysis alone. UNEP analysts verify detections before sending alerts, while later satellite observations and information from operators are used to assess whether mitigation occurred.

Why This Matters: MARS connects remote detection with organizations capable of stopping emissions. Since becoming operational in 2024, the system has facilitated or verified mitigation in more than 40 cases. In Kazakhstan, repeated satellite detections identified a leak from the Kamenistoe gas pipeline. After notification, the operator transferred gas flow to another pipeline, and subsequent monitoring found no further significant emissions. The case documents a chain from satellite observation and expert review to notification, operator action, and verification.

Serajuddin, Umar, Chris Aubrecht, Fabio Cian, David Taverner, Dominic Palazzolo & Giulia Costella. “From Satellite Data to Development Finance: How Earth Observation Is Scaling Impact across World Bank Operations.” World Bank Data Blog. May 8, 2026. https://blogs.worldbank.org/en/opendata/from-satellite-data-to-development-finance--how-earth-observatio 

Focus: Reviews how the World Bank and European Space Agency have integrated earth observation into development projects, financing decisions, and national information systems.

Role of Non-Traditional Data: World Bank projects use satellite-derived information for land use mapping, flood monitoring, vegetation analysis, crop reporting, and infrastructure risk assessment. Applications include pasture monitoring in Paraguay, agricultural statistics in Pakistan and the Philippines, flood analysis in South Sudan, and assessments of energy infrastructure in Bangladesh and Uganda. The partnership also produces standardized geospatial datasets and reusable analytical approaches that can be transferred across countries and sectors.

Why This Matters: Since 2020, the partnership has supported more than 87 World Bank projects across approximately 70 countries and applied around 80 earth observation capabilities. The authors report that the work informed operations involving approximately $6 billion in development finance, with another $16 million mobilized for replication and capacity building. Satellite analysis can therefore contribute to project design, investment screening, and government information systems at scale. 

Associação Brasileira dos Membros do Ministério Público de Meio Ambiente. “Operação Nacional Mata Atlântica em Pé 2026 Autua Mais de 8 Mil Hectares de Desmatamento Ilegal em 17 Estados.” ABRAMPA, August 28, 2026. https://abrampa.org.br/operacao-nacional-mata-atlantica-em-pe-2026-autua-mais-de-8-mil-hectares-de-desmatamento-ilegal-em-17-estados/

Ministério Público de Minas Gerais state-level operation results.

Focus: Documents how Brazilian authorities used geospatial intelligence to identify and investigate suspected illegal deforestation across the Atlantic Forest in 17 states.

Role of Non-Traditional Data: Satellite-derived alerts from MapBiomas Alerta and other monitoring platforms identified potential deforestation and generated geographic areas for investigation. Prosecutors, regulators, and police used these alerts to prioritize remote assessments and field inspections. Minas Gerais also used the Harpia and BrasilMais monitoring systems to assess alerts and document suspected violations.

Why This Matters: During the ten-day operation, authorities investigated 1,520 alerts and identified approximately 8,370 hectares of illegal clearing. Preliminary results included about R$100 million in fines nationwide. In Minas Gerais, confirmed cases also led to embargoes and could trigger restrictions involving rural property registration and agricultural credit. The operation documents a complete chain from satellite detection and verification to confirmed violations and legal enforcement across multiple jurisdictions.

Henrich, Christoph Samba, Lono Leneuoti, Antoine De Ramon N’Yeurt, Hilda Waqa-Sakiti & Sandra Galvis Rodriguez. “From Data to Decisions? Assessing Community-Based Water Monitoring in Tuvalu.” Frontiers in Environmental Science, Volume 14, Article 1887907. August 13, 2026. https://doi.org/10.3389/fenvs.2026.1887907 

Focus: Examines whether community-based monitoring can fill information gaps and support water management across three remote islands in Tuvalu.

Role of Non-Traditional Data: Trained residents surveyed groundwater wells, rainwater tanks, and communal cisterns on Nanumea, Nui, and Vaitupu between 2022 and 2025. Using KoboToolbox, they produced 536 records covering water availability, groundwater salinity, and E. coli contamination. Researchers combined these observations with rainfall records to track freshwater conditions and examine whether the findings influenced water use.

Why This Matters: The monitoring identified widespread microbiological contamination and showed how freshwater conditions responded to rainfall. On Nanumea, dry periods coincided with higher groundwater salinity and lower cistern levels, while wetter periods brought recovery. Community reports indicate that contaminated cisterns were temporarily removed from use.

Conserva, Michelangelo, and Charlotte Stanton. “From Pixels to Planning: Earth AI for Nature Restoration.” Google Research, June 16, 2026. https://research.google/blog/from-pixels-to-planning-earth-ai-for-nature-restoration/ 

Focus: Introduces a high-resolution map of English hedgerows, shelterbelts, small woodlands, stone walls, and other landscape features often missing from conventional forest inventories.

Role of Non-Traditional Data: Google Research and the University of Oxford combined submeter imagery, one-metre LiDAR observations, and an image-classification model trained on more than 300 million satellite images. The system mapped small landscape features across more than 130,000 square kilometres of England. Google then released an open vector dataset containing measurable boundaries for millions of individual features.

Why This Matters: Landowners, conservation organizations, and public authorities can use the dataset to measure hedgerows and small woodlands, identify breaks in ecological corridors, establish restoration baselines, and support carbon and biodiversity accounting. It may also help detect environmental losses outside project boundaries. 

Morton, Oscar, Christopher G. Bousfield, Prince Dégny Valé, Ieuan Lamb, Victor Maus, Robert G. Bryant, and David P. Edwards. “Mining Triggers Extensive Additional Deforestation in Sub-Saharan Africa.” Nature 654 (2026): 971–976. Published June 3, 2026. https://doi.org/10.1038/s41586-026-10551-2 

Focus: Measures direct and indirect forest loss associated with mining across sub-Saharan Africa and examines how impacts vary across places, minerals, and distances from mines.

Role of Non-Traditional Data: The researchers combined 20 years of satellite-derived forest-cover observations with classifications of subsequent land use, location data for 16,627 mining clusters, plantation maps, and commodity databases. They compared forest loss around operating mines with otherwise similar locations where mining had not yet begun, distinguishing clearing inside mine footprints from losses associated with surrounding agriculture, settlements, and roads.

Why This Matters: Mining accounted for approximately 187,000 hectares of direct forest loss between 2001 and 2020. For every hectare cleared within a mine footprint, the researchers estimated that another 34 hectares were lost nearby within five years, with effects detectable up to 20 kilometres away. The findings could expand impact assessments and licensing reviews beyond project boundaries and improve monitoring of critical-mineral supply chains. 

Reflections

Earlier updates in this series highlighted several recurring features of non-traditional data: its value in addressing gaps in official information, the importance of combining sources, and the institutional work needed to sustain promising applications. This collection adds several insights into how non-traditional data becomes usable in practice. 

1. Non-traditional data helps institutions allocate limited attention 

Many applications help institutions decide which possible interventions to pursue: which methane release or forest alert to investigate, which rooftop to target, or which infrastructure concern to examine. Their value lies partly in directing scarce staff, funding, and technical capacity. Transparent rankings, well-defined thresholds, and procedures for reviewing errors are therefore essential to their usefulness. 

2.Operational maturity becomes visible after an output is produced 

UNEP’s methane system connects verified detections with operator notifications, mitigation, and subsequent monitoring. Brazil’s Atlantic Forest operation links alerts with investigations, fines, and embargoes. Social protection models have informed recipient enrollment, while participatory maps have entered formal risk-management plans. Operational maturity therefore depends on who receives an output, how it is reviewed and used, and whether the resulting action is recorded. Entries that end with forecasts, maps, or recommendations show where this process remains incomplete.

3. Combining sources involves a division of roles

Different sources contribute distinct forms of evidence. The Australian tobacco estimates make this division especially clear: wastewater measured total nicotine consumption, scanner and tax records estimated legal consumption, and statistical models allocated the remaining quantity across product categories. Elsewhere, satellite imagery supplies repeated, wide-area observation; low-cost sensors capture hyperlocal conditions in near real time; social media and public comments surface emerging events and lived experiences; crowdsourced records locate infrastructure absent from official inventories; and mobile, platform, and transaction data reveals behavior at scale. Surveys and administrative records connect these signals with known populations, community reports supply local context, and field inspections confirm inferred events. Effective integration depends in part on giving each source a defined role in detection, interpretation, calibration, validation, prioritization, or follow-up.

4. Community-generated information gains influence through formal institutional entry points

In Venezuela, volunteer-generated maps supported humanitarian logistics. In Colombia, residents’ sketches entered Community Risk Management Plans. In Nandi County, citizen assessments reached committees involved in budget scrutiny. These cases indicate that participation carries greater practical weight when a planning, budgeting, or response process is prepared to receive, validate, and act on the resulting evidence. Intermittent monitoring and weak institutional links can substantially limit its influence. 

5. Reusable analytical systems support wider application 

Transferable models, automated geospatial tools, open datasets, and standardized earth observation capabilities can reduce the cost and time required to apply sophisticated analysis across locations. Their portability also introduces risks: uneven crowdsourced coverage, differences between cities, outdated imagery, and limited local validation can affect usefulness. Documentation should therefore specify where systems have been tested, which local inputs remain necessary, known sources of uncertainty, and how users can assess fitness for purpose.

Back to the Blog