Key Activities

    • Ziulu, Virginia, and James Garforth. "Advancing Deepfake Detection in RGB Satellite Imagery Through Domain-Specific Ensembles." In IGARSS 2025-2025 IEEE International Geoscience and Remote Sensing Symposium, pp. 5821-5827. IEEE, 2025.

    • Naeher, Dominik, and Virginia Ziulu. "Does adoption of zero tillage reduce crop residue burning? Evidence from satellite remote sensing and household survey data in India." Applied Economics (2025): 1-15.

    • Ziulu, Virginia. “Leveraging Imagery Data in Evaluations. Applications of Remote-Sensing and Streetscape Imagery Analysis.” IEG Methods and Evaluation Capacity Development Working Paper Series. Independent Evaluation Group. Washington, DC: World Bank (2024).

    • Naeher, Dominik, Raghavan Narayanan, and Virginia Ziulu. "Cash for Coolers or Sustainable Lighting? Assessing Different Components of a Large-Scale Energy Efficiency Program in Mexico." The Journal of Development Studies (2024): 1-15.

    • Ziulu, Virginia, Jessica Meckler, Gonzalo Hernández Licona, and Jozef Vaessen. "Poverty Mapping: Innovative Approaches to Creating Poverty Maps with New Data Sources." IEG Methods and Evaluation Capacity Development Working Paper Series.Independent Evaluation Group. Washington, DC: World Bank (2022).

    • Naeher, Dominik, Raghavan Narayanan, and Virginia Ziulu. "Impacts of energy efficiency projects in developing countries: Evidence from a spatial difference-in-differences analysis in Malawi." Energy for Sustainable Development 73 (2023): 365-375 (Working paper version: World Bank Policy Research Working Paper 9842).

    • Naeher, Dominik, Raghavan Narayanan, and Virginia Ziulu. "The relevance of development policies to confront crisis situations: World Bank’s early response to Covid-19.” Journal of Policy Modeling (2024). (Working paper version: World Bank Policy Research Working Paper 9935).

    • Chase, Claire, Virginia Ziulu, Priya Lall, Phyrum Kov, Susanna Smets, Virak Chan, and Yeng Lun. "Addressing the behavioralconstraints to latrine uptake: effectiveness of a behavior-change campaign in rural Cambodia." Waterlines (2015): 365-378.

    • “Spatial Evidence for Climate Action: Generating Insights through Geospatial Analysis”, 2025 United Nations Climate Change Conference, Evidence for Climate Action (E4CA), Virtual,  17 November 2025.

    • “Remote Sensing for Evaluations”, InterAmerican Development Bank, Office of Evaluation and Oversight, Washington DC (United States), 17 October 2025.

    • “Advancing Deepfake Detection in RGB Satellite Imagery Through Domain-Specific Ensembles”, International Geoscience and Remote Sensing Symposium (IGARSS) 2025, Brisbane (Australia), 7 August 2025 (as part of the session “Data Security in Remote Sensing”, for which I acted as chair).

    • “Geospatial Artificial Intelligence in Evaluation Practice”, GeoField conference, Agence Française de Développement (ADF), Paris (France), 28 May 2025.

    • “Leveraging AI for Enhanced Imagery Analysis in Evaluations​”, Workshop on Artificial Intelligence, Global Environment Facility (GEF), Washington DC (United States), 16 April 2025.

    • “From Pixels to Insights: Image Analysis for Evaluation”, Eval-GIS Community of Practice, United Nations Development Programme (UNDP), Virtual, 9 May 2025.

    • “Harnessing Image Data for Efficient Evaluation”, National Evaluation Capacities (NEC) Conference 2024, Beijing (China), 14 October 2024.

    • “Leveraging Text Data and Generative AI in Complex Thematic Evaluations”, 10th Annual Measuring Development Conference: AI, the Next Generation, Center for Effective Global Action (CEGA) and University of Chicago, Washington DC (United States), 2 May 2024.

    • “From Pixels to Geospatial Insights: IEG’s Experience Leveraging Image Data in Evaluations”, Geospatial Symposium: Unlocking the Potential of Geospatial Analysis for Evaluations, World Bank, Washington DC (United States), 11 April 2024.

    • “Exploring the Potential of Data Science for Evaluations”, African Development Bank, Virtual, 30 January 2024.

    • “From Pixels to Geospatial Insights”, European Evaluation Society (EES), Virtual, 16 November 2023.

    • “Deep Learning Applications for Geospatial Analysis”, World Bank, Washington DC (United States), 21 September 2023.

    • “Exploring the Potential of Generative Models for Evaluative Tasks at the World Bank”, Applied Physics Laboratory (APL), Johns Hopkins University, Laurel MD (United States), 5 October 2023.

    • “Data Science in Evaluations”, Global Evaluation Initiative (GEI), Virtual, 8 March 2023.

    • “Geospatial 2.0? Advances in the Use of Imagery Data for Evaluation”, Asian Evaluation Week, Virtual, 15 September 2022.

    • “Opportunities for Innovation in M&E: Data Science” (delivered in Spanish), gLOCAL 2022, Virtual, 30 May 2022.

    • “Use of Image Data to Assess Urban Growth and Informality”, European Evaluation Society (EES), Copenhagen (Denmark), 10 June 2022.

    • “Leveraging Image Data Analysis for Evaluations”, World Bank, Washington DC (United States), 26 January 2022.

    • “Using Nighttime Lights Data to Evaluate the Impact of Energy Efficient Projects in Developing Countries”, World Bank, Washington DC (United States), 6 October 2021.

    • Ziulu, Virginia, Harsh Anuj, Ariya Hagh, Estelle Raimondo, Jos Vaessen. "Extracting Meaning from Textual Data for Evaluation – Lessons from Recent Practice at the Independent Evaluation Group of the World Bank" (Chapter 5 of book "Artificial Intelligence and Evaluation. Emerging Technologies and their Implications for Evaluation").

    • Anuj, Harsh, Virginia Ziulu, Ariya Hagh, Estelle Raimondo, and Jos Vaessen. "World Bank IEG evaluations and the role of data science: reflections from recent experiences" (Chapter 11 of book "Artificial Intelligence and Big Data").

    • Ziulu, Virginia, and James Garforth. "Toward Efficient, Explainable Satellite Deepfake Detection: Colour and Frequency Band Contributions". To be submitted to IGARSS 2026.

    • Singh, Kunwar, Amy Frazier, Virginia Ziulu, and Jordan Landrum. “Harnessing drones to incept and evaluate climate-smart agriculture interventions for sustainable development”. To be submitted to Nature Sustainability.

    • Ziulu, Virginia. “Geospatial Artificial Intelligence in Evaluation Practice” (Chapter of textbook “Geospatial Impact Evaluation in Practice”). To be published in 2026.

    • Ziulu, Virginia, Estelle Raimondo, and Claire Zanuso. “Broadening the use of geospatial analysis for evaluation practice: the way forward” (Chapter of textbook “Geospatial Impact Evaluation in Practice”). To be published in 2026.

    • Ziulu, Virginia, Maria de las Mercedes Vellez, Elena Bardasi, and Xiaoxiao Peng. “Leveraging Classic and Generative AI for Text Classification in Evaluation: Insights from Recent IEG Applications”. IEG Methods Paper Series. To be published in 2026.

    • Ziulu, Virginia , and Taru Meshram. “Super-resolution approaches for satellite imagery” (working title). IEG Methods Paper Series. To be published in 2026.

    • Detecting Synthetic Satellite Imagery. I developed an ensemble deep learning framework to detect deepfake RGB satellite images by combining spatial-domain and frequency-domain (Fourier transform–based) CNN classifiers. The approach leverages domain-specific fine-tuning and model ensembling to improve robustness against increasingly sophisticated synthetic imagery.

    • Urban Morphology & Land Use – Dar es Salaam, Tanzania. Using Microsoft/Bing building footprint data, I applied unsupervised clustering to classify land use patterns based on parcel morphology. By extracting geometric and spatial features from building layouts, I identified functional urban structures without relying on labeled training data.

    • Electricity Reliability & Anomaly Detection – Democratic Republic of Congo. Using daily nighttime lights data from NASA’s Black Marble, I identified spatial and temporal anomalies in electrification reliability. The analysis incorporated viewing angle normalization and classified electrification states to detect service disruptions in data-scarce environments.

    • Urban Environmental Indicators – Bathore, Albania. I quantified greenness and sky view factor at the neighborhood scale using semantic segmentation applied to street-level imagery. The project translated pixel-level classifications into environmental indicators relevant for urban livability and planning assessments.

    • Agricultural Burn Scar Estimation – India. Using multispectral imagery from Sentinel-2, I estimated agricultural burn scars through spectral index thresholding, enabling seasonal monitoring of land management practices.

    • Mangrove Cover & Vegetation Health – Belize. I quantified changes in mangrove extent and vegetation health across three coastal regions using vegetation indices and existing mangrove datasets, supporting ecosystem monitoring in climate-vulnerable coastal zones.

    • Flood Risk & Annual Flood Mapping – Dar es Salaam, Tanzania. I developed high-risk flood masks by integrating digital elevation models and surface imperviousness layers to model urban flood susceptibility.

    • Annual Flood Modelling – Surakarta, Indonesia. Using Sentinel-2 imagery, I generated annual high-resolution flood maps based on the Modified Normalized Difference Water Index (MNDWI), producing a multi-year flood dynamics time series.

    • Protected Area Land Cover Monitoring – Global (10-year portfolio). I produced annual land cover time series for supported protected areas, with a focus on tree cover change. The analysis relied primarily on the AI-generated dataset Dynamic World to assess ecosystem dynamics over time.

    • Water Budget & Groundwater Stress – Oum Er Rbia Basin, Morocco. I developed a remote sensing–based water balance model estimating evapotranspiration, runoff, and precipitation at the pixel level to assess changes in arable land and groundwater stress at the basin scale.

    • Supervised Land Cover Change – Bathore, Albania. Using Landsat imagery, I implemented supervised machine learning classification to estimate annual land cover transitions and urban expansion dynamics.

    • Energy Efficiency Impact – Malawi. I conducted a spatial econometric difference-in-differences analysis combining nighttime lights data from NOAA with survey data to assess the impact of energy efficiency interventions.

    • Electricity Reliability Monitoring – Sub-Saharan Africa (10-year portfolio). I estimated long-term trends in electricity reliability across grid electrification projects using Black Marble time series, quantifying blackouts and brownouts in infrastructure-scarce regions.

    • From Geo-Pixels to Evaluation Insights, What We Have Learned Podcast, June 13, 2024. Link.

    • Meet Virginia Ziulu, IEG's Data Science Trailblazer, March 18, 2024. Link.

    • Innovations in Geospatial Analysis for Evaluations, 2024. Link.

    • Unlocking the Potential of Geospatial Analysis for Evaluation, June 14, 2024. Link.

    • Setting up Experiments to Test GPT for Evaluation, August 13, 2023. Link.

    • Fulfilled Promises: Using GPT for Analytical Tasks, August 23, 2023. Link.

    • Unfulfilled Promises: Using GPT for Synthetic Tasks, August 30, 2023. Link.

    • Geospatial Analysis in Evaluation, March 25, 2022. Link.

    • Dataset of synthetically generated satellite images (deepfakes). Link 1. Link 2.

    • Python script for generating deepfake satellite images using the pix2pix architecture and evaluating the quality of generated images. Link.

    • Python script for finetuning InceptionV3 model on a remote sensing dataset (RSSCN7 dataset). Link.

    • Python script to build, run, and evaluate a binary classification model (using the spatial domain or the frequency domain as input) to separate real and fake satellite images. Link.