Joshua Dimasaka

Hello, I am Joshua Dimasaka, a researcher with a professional background in civil & structural engineering, public policy, and deep learning using satellite imagery. I explore AI methods (probabilistic, deep, graph, state-space) and time-series Earth Observation data (embeddings, Sentinel, Landsat, any from Google Earth Engine) to audit regional spatiotemporal changes in exposure, physical vulnerability dynamics, and regional climate & disaster risk measurement at large scales.

Research Interests: Climate & Disaster Risk Assessment, Earthquake Risk & Loss Modelling, Deep Probabilistic Modelling, Geospatial Data Science, Spatiotemporal Analysis

News

May 29, 2026 - Our research, "GraphCSVAE: Graph Categorical Structured Variational Autoencoder for Spatiotemporal Auditing of Physical Vulnerability Towards Sustainable Post-Disaster Risk Reduction", has been published in Progress in Disaster Science.
May 17, 2026 - Hello! It's Mental Health Awareness Week (MHAW), an annual event in the UK dedicated to raising awareness of mental health and promoting mental wellbeing for all. My "Take Action" for this year's theme is that I have just completed 50,000 pushups since starting a year and a half ago, doing 100-300 pushups every day! To 100,000!
May 5, 2026 - I presented "METEOR 2.5D: An Open Geospatial Dataset of the Spatiotemporal Evolution of Physical Vulnerability in UN-recognized Least Developed Countries (as of 2020) at Five-year Intervals, 1975-2030" at the EGU General Assembly 2026 in Vienna, Austria. Preprint available here and a short animation of the PICO presentation available here.
Apr 25, 2026 - I am expressing my support to preserve EM-DAT by signing this open letter. "For more than 30 years, EM-DAT has served as a foundational dataset supporting thousands of peer-reviewed studies, national risk assessments, and international policy processes. Its loss would represent a significant gap in the global community’s ability to assess disaster impacts and trends. At a time of increasing climate-related risks, access to reliable data is essential for informed decision-making." Please, feel free to share this!
Apr 14, 2026 - Kon'nichiwa! Our group, Cambridge University Centre for Risk in the Built Environment (CURBE) with Prof. Emily So and Aline Van Driessche, presented our research to Prof. Saki Yotsui's research group and her students of the University of Tokyo. Arigatou gozaimasu!
Apr 9, 2026 - How has the physical vulnerability in Least Developed Countries (LDCs) evolved and redistributed over the past five decades from 1975 to 2030, and what story does it tell about the trajectory of our global disaster risk reduction efforts? Our long-version preprint "Quantifying the Regional Dynamics and Redistribution of Physical Vulnerability in Least Developed Countries" is now available on EarthArXiv. This will be a highlight of our upcoming EGU and GEM presentations. Happy reading!
Apr 6, 2026 - What if we could estimate the number of affected people that needs immediate evacuation depending on the intensity of an upcoming typhoon so that we can better prepare for such events? Derived from a decadal record of post-disaster reports, our work on the "Analysis of Causal Drivers and Predictive Modelling of the Typhoon-Affected Population Rate in the Philippines" has been accepted for oral presentation at the 16th International Conference of IDRiM Society to be held in Bali, Indonesia, 5-9 September 2026. Preprint coming soon!
Apr 4, 2026 - Guten Tag, I have begun my Digital GreenTalents Research stay at German Aerospace Center (DLR) and in the exciting Bavarian alps!
Mar 16, 2026 - Our research on quantifying the regional dynamics and redistribution of physical vulnerability in Least Developed Countries (METEOR 2.5D: An Open Geospatial Dataset of the Spatiotemporal Evolution of Physical Vulnerability in UN-recognized Least Developed Countries (as of 2020) at Five-year Intervals, 1975-2030) has been accepted as a PICO presentation of the "Session NH6.7 Exposure modeling for natural hazards: creation, analysis and application" at EGU General Assembly 2026 to be held in Vienna, Austria, May 5, 2026. PICO presentation is "short, precise, scientific" and stands for "Presenting Interactive COntent®". Preprint coming soon!
Feb 18, 2026 - I have been awarded a travel grant to present our PhD research at Global Earthquake Model (GEM) Conference, From Faults to Future Scenarios, to be held at the University of Zagreb in Croatia, June 23-25, 2026. We are excited to share our work on "Spatial Disaggregation and Temporal Projection of Building Exposure and Physical Vulnerability using Deep Constrained Clustering and Probabilistic Graph Deep Learning".
Jan 23, 2026 - I presented "GraphVSSM: Graph Variational State-Space Model for Probabilistic Spatiotemporal Inference of Dynamic Exposure and Vulnerability for Regional Disaster Resilience Assessment" at Special Track on AI for Social Impact (AISI), 40th AAAI Conference on Artificial Intelligence 2026 in Singapore.
Dec 16, 2025 - I presented "A Data-Driven Probabilistic Approach to Regional Dynamics of Building Exposure and Physical Vulnerability Towards Global Disaster Risk Quantification Audit" at American Geophysical Union Meeting 2025.
Nov 25, 2025 - I am deeply honored to be one of the 20 selected young scientists all over the world (43 countries) to receive the “Digital GreenTalents - High Potentials in Sustainable Development Award 2025” by the German Federal Ministry of Research, Technology, and Space (BMFTR), Projektträger Karlsruhe (PTKA) am Karlsruher Institut für Technologie (KIT), & German Informatics Society. Check my biography page here.
Nov 21, 2025 - Our AAAI AI for Social Impact paper on "GraphVSSM: Graph Variational State-Space Model for Probabilistic Spatiotemporal Inference of Dynamic Exposure and Vulnerability for Regional Disaster Resilience Assessment" has been selected as one of the 52 oral presentations and one of the four oral presentations under the "Disasters" subtopic. Schedule to follow! See you all in Singapore!
Nov 5, 2025 - Our 'spatial auditing' paper, "DeepC4: Deep Conditional Census-Constrained Clustering for Large-scale Multitask Disaggregation of Urban Morphology", is currently under consideration, review, and revision for the special issue (Interpretability and Explainability in GeoAI for Geospatial Science and Earth Observations) of the ISPRS Journal of Photogrammetry and Remote Sensing (P&RS). To facilitate early sharing of this work, our non-peer-reviewed preprint is available in arXiv. Supported by the Helmholtz Visiting Researcher Grant, our proposed novel DeepC4, a deep constrained clustering model, advances the existing classical spatial disaggregation techniques using deep learning and EO data to understand the distribution of various urban morphology characteristics at large scales, using cluster-level constraints, multitask optimization, and multilabel conditionality, under a weak supervision setup.
Nov 3, 2025 - Our full paper, "GraphVSSM: Graph Variational State-Space Model for Probabilistic Spatiotemporal Inference of Dynamic Exposure and Vulnerability for Regional Disaster Resilience Assessment" -- a spatiotemporal model combining graph deep learning, state-space modeling, and variational inference to improve the dynamic modeling of building exposure and physical vulnerability across three case studies from city to multi-country scale -- has been accepted to the Special Track on AI for Social Impact at the 40th AAAI Conference on Artificial Intelligence (AAAI), to be held in Singapore from January 20–27, 2026. This track received a record number of submissions, with 693 papers sent for review and 167 accepted, resulting in an acceptance rate of 24.1%.
Oct 31, 2025 - Our oral presentation, "GraphCSVAE: Graph Categorical Structured Variational Autoencoder for Spatiotemporal Auditing of Physical Vulnerability Towards Sustainable Post-Disaster Risk Reduction", received the Best Overall Paper of the Conference at the 8th International Disaster and Risk Conference (IDRC 2025) - AI, Emerging Tech & Immersive Solutions for Disaster & Emergency Response, in Nicosia, Cyprus.
Oct 8, 2025 - I wrote a personal reflection essay 'Connecting the Dots of Vulnerability: The Story of a Filipino Wood Carver and Tricycle Driver’s Son — from a Laguna Lake Squatter Settlement to UP, Stanford, and Cambridge' about our family's struggle in light of September's Brain Aneurysm Awareness Month. Special: I'm also sharing here, for the first time publicly, my Stanford Knight-Hennessy Scholar essay. Read it here. Hope you find it inspiring!
Sep 21, 2025 - Understanding the risks and impacts of disasters is not just an academic topic -- it goes beyond theory and affects the lives of many hardworking people na lumalaban nang patas at may dangal. As someone who used to wade through waist-deep floods just to access education in our rural area, today is a very important moment to join our fellow Filipinos in the #TrillionPesoMarch. Follow my undergraduate organization (CEO)'s series of posts for some informative explainer!
Sep 19, 2025 - Before this week ends, we are happy to share with you all that I have been selected as one of the 20 awardees to receive the prestigious "Digital GreenTalents – High Potentials in Sustainable Development" Award 2025 by the German Federal Ministry of Research, Technology and Space (BMFTR) "for highly motivated and talented young scientists from all over the world to find solutions to urgent issues and contribute to achieving the United Nations Sustainable Development Goals (SDGs)". See you in 2026, Germany!
Sep 14, 2025 - In the years following a disaster, do affected communities reduce their future risk by avoiding the construction of vulnerable structures, or do they rather return to settle in pre-identified danger zones, even after experiencing their devastating impacts firsthand? Examining the spatiotemporal changes in physical vulnerability by applying advanced machine learning to satellite-derived datasets with prior expert belief systems, the arXiv copy of our work "GraphCSVAE: Graph Categorical Structured Variational Autoencoder for Spatiotemporal Auditing of Physical Vulnerability Towards Sustainable Post-Disaster Risk Reduction", accepted at IDRC 2025, is now available here.
Sep 13, 2025 - Last 10th of September was our 'World Suicide Prevention Day', and it's very important for our leaders today in academia and industry to recognize their role in "changing the narrative on suicide (and, in general, mental wellbeing)" in their workplace. To celebrate this day, I'm happy to share that I've reached my 30,000 pushups that greatly helped my mental wellbeing! Also, feel free to check out my new website tab here 'WELLBEING' that contains my recently read self-help books! It's both beneficial for ourselves as we become leaders in our own career, and how we could help others by creating 'listening spaces' for many 'tender conversations'.
Sep 5, 2025 - We're happy to start our Friday with a news on acceptance and €400-registration scholarship for our work "GraphCSVAE: Graph Categorical Structured Variational Autoencoder for Spatiotemporal Auditing of Physical Vulnerability Towards Sustainable Post-Disaster Risk Reduction" at 8th International Disaster and Risk Conference (IDRC 2025) in Nicosia, Cyprus on 23–24 October 2025, with this year's theme on "AI, Emerging Tech & Immersive Solutions for Disaster & Emergency Response." Many thanks to the support and guidance of our mentors and co-authors: Dr. Emily So (CURBE), Dr. Christian Geiß (DLR) and Dr. Robert Muir-Wood (Moody’s RMS)! Our preprint will be available soon. See you in Cyprus!
Sep 4, 2025 - Our collaborative research "Modernizing Quezon City’s Building Risk Data with AI and Earth Observation", through a blind review process, has been accepted for oral presentation in the 20th Association of Pacific Rim Universities Multi-hazards Conference and Symposium (APRU-MHS), which will take place from 26th to the 29th of November, 2025 at the University of the Philippines Bonifacio Global City Campus, Philippines. Many thanks to the advice from our industry and practitioner partners: Dr. Fouad Bendimerad and Dr. Renan Ma. Tanhueco (EMI) and with the kind support of Enp. Ma. Bianca D. Perez, MPA of the Quezon City Disaster and Risk Reduction Management Office (QCDRRMO)! If you are also attending and would like to chat about our work for possible future collaborations and projects, please feel free to send me an email to schedule a coffee chat around the area. See you again, Pilipinas!
Aug 18, 2025 - Our short volunteering work to understand Myanmar-Thailand earthquake damage has been featured on the Department of Engineering website. Even though our volunteering capacities are sadly limited, we have truly learned a lot from this educational experience "what it's like in this very unique setting, very urgent, no data availability, and any other problems that we may have". Read more here.
Aug 5, 2025 - If "DeepC4" is for "spatial auditing" from our first year, we now present our efforts towards "spatiotemporal auditing" from our second year, with a non-peer-reviewed preprint "GraphVSSM: Graph Variational State-Space Model for Probabilistic Spatiotemporal Inference of Dynamic Exposure and Vulnerability for Regional Disaster Resilience Assessment" now available on arXiv! Across our three case studies ranging from city-level to multi-country scale, our proposed novel GraphVSSM tackles the equally important yet uniquely challenging aspect of modeling physical vulnerability at large spatiotemporal scales. It uniquely integrates graph deep learning, state-space modeling, and variational inference using time-series data and prior expert belief systems in a coarse-to-fine-grained manner, contributing to broader urban studies that require compositional data analysis in weakly supervised settings. Read the preprint here.
Jul 31, 2025 - Our non-peer-reviewed preprint "DeepC4: Deep Conditional Census-Constrained Clustering for Large-scale Multitask Disaggregation of Urban Morphology" is now available on arXiv! This work is the highlight contribution of my first-year PhD, which deals with "spatial auditing". Our proposed novel DeepC4, a deep constrained clustering model, advances the existing classical spatial disaggregation techniques using deep learning and EO data to understand the distribution of various urban morphology characteristics at large scales, using cluster-level constraints, multitask optimization, and multilabel conditionality, under a weak supervision setup. Read the preprint here.
May 21, 2025 - Many thanks to our attendees in our AI4ER Annual Showcase 2025 yesterday! It's so motivating to see our community back again together, from the afternoon talks with interesting AI algorithmic tricks to our evening dinner filled with fun stories and delicious food + wine.
May 8, 2025 - Next week, it'll be Mental Health Awareness Week (May 12th-18th) and I hope you are all taking care, that your work is as equally important as what's affecting your work. This is not work-related news, but I'm very happy to share that I have successfully reached 10,000 push ups in 40 days. Did you know that there's a global and systemic mental-health crisis in academia?
May 6, 2025 - I will be presenting some of the newest and exciting updates of our PhD work (Probabilistic Regional Dynamics of Building Exposure & Physical Vulnerability) in our annual AI4ER CDT Showcase 2025 on Tuesday 20 May 2025 from 12.00 - 19.00 at the Department of Computer Science and Technology, William Gates Building, 15 JJ Thomson Ave, Cambridge, CB3 0FD. Join us and connect! RSVP here.
Apr 24, 2025 - I participated (and enjoyed) a roundtable discussion with other Cambridge CDT researchers and our EPSRC Executive Chair, Professor Charlotte Deane. I am very grateful to have shared our exciting research on using AI and Earth observation data in understanding large-scale disaster risk spatiotemporally here at the University of Cambridge.
Apr 21, 2025 - Our volunteering collaboration to understand the impact after Mw-7.7 earthquake in Myanmar and Thailand is now available on Zenodo and MapBox (a web map visualization). Many thanks to our volunteers (Shagun, Lisanne, Lorenzo, Grace, Fer, Han, Orlando, Shaomin, Andrea, and Sakthy) and many thanks for this leadership role.
Apr 18, 2025 - Due to many exciting developments in our research on spatiotemporal patterns of building exposure and physical vulnerability using probabilistic AI, geometric/graph deep learning, and geospatial datasets, our preprint for DeepC4 will most likely be available in the third quarter of 2025. Many thanks for patience! You may always check our DeepC4 work here and send me an email for any inquiries.
Dec 10, 2024 - I presented our poster presentation entitled "Deep Conditional Census-Constrained Clustering (DeepC4) for Large-scale Multitask Disaggregation of Urban Morphology" at American Geophysical Union Meeting 2024 in Washington, D.C. Many thanks for all those kind words and early requests for preprints! I also appreciate many fruitful conversations on AI and DRR. Many thanks to our AI4ER, Filipino attendees, Yue, Victor, Charlie, Robert, Sabine, Chenbo, and Susu's PhD students (who've been continuing our other AI and DRR work) for making this day a memorable one! We're still working on a preprint, which is expected to be shared publicly in the first quarter of 2025. Check my Bluesky posting here.
Nov 23, 2024 - The data and code repositories in Zenodo and GitHub of our "Deep Conditional Census-Constrained Clustering (DeepC4) for Large-scale Multi-task Disaggregation of Urban Morphology" are now available for public access. See you at AGU24 in Washington, D.C.!
Oct 18, 2024 - I attended 2024 UNDRR Asia-Pacific Ministerial Conference on Disaster Risk Reduction (APMCDRR) from 14 to 18 October 2024 at the Philippine International Convention Center in Manila. Check my post-event reflection here. Thankful for new connections and our EMI community gathering together!
Oct 2, 2024 - Our work "Spatial Disaggregation of Rwandan Building Exposure and Vulnerability via Weakly Supervised Conditional Census-Constrained Clustering (C4) using Earth Observation Data" has been accepted as a poster presentation at American Geophysical Union Meeting 2024 in Washington, D.C., on 10th of December 2024.
Sep 30, 2024 - New paper alert! My work with Andrea and Sakthy on "Enhancing assessment of direct and indirect exposure of settlement-transportation systems to mass movements by intergraph representation learning" in Environmental Research Letters has just been accepted.
Jun 30, 2024 - I started my research stay in DLR Earth Observation Center, Germany. Thankful for the Helmholtz Association. Let's explore Munich and Germany!
Jun 25, 2024 - I gave an invited presentation to the Innovations Group WG5 of LandAware, an international network on early warning systems. Many thanks for the good questions from our community. See their LinkedIn post here and Twitter/X post here.
May 22, 2024 - I delivered an oral presentation version of our ICLR ML4RS workshop poster at the Cambridge Hazard and Risk Research Day. Access the slides here.
May 11, 2024 - I presented our developing work "Global Mapping of Exposure and Physical Vulnerability Dynamics in Least Developed Countries using Remote Sensing and Machine Learning" at 2nd Machine Learning for Remote Sensing Workshop, 12th International Conference on Learning Representations (ICLR) in Vienna, Austria. Access the poster here, the 4-page paper here, the animated poster here, and the OpenSendaiBench Version 1 here.
Apr 29, 2024 - Joseph (University of Bath), Lois (Philippine Space Agency) and I (University of Cambridge) are building an online community PAIRL (Philippine AI Research and Learning Community) to democratize AI for the Philippine public and support PH AI researchers & learners to ignite meaningful collaborations and knowledge sharing. Follow us on a Twitter/X or join our growing community on Slack (free to join)!
Mar 4, 2024 - I won the Natural Hazards - Outstanding Student Presentation Award (OSPA) for my AGU 2023 presentation of "Near-real-time Country-wide Estimation of Susceptibility and Settlement Exposure from Norwegian Mass Movements via Inter-graph Representation Learning", which is now available here on Earth and Space Science Open Archive. Official press release and publication to follow.
Mar 3, 2024 - Our developing work "Global Mapping of Exposure and Physical Vulnerability Dynamics in Least Developed Countries using Remote Sensing and Machine Learning" has been accepted as a poster presentation at 2nd Machine Learning for Remote Sensing Workshop, 12th International Conference on Learning Representations (ICLR) in Vienna, Austria, on 11th of May 2024. More details to follow.
Feb 8, 2024 - I have been awarded a three-month fully-funded Helmholtz Visiting Researcher Grant to do an exciting summer research with the "Team Geo-risks" at the German Aerospace Center in July, August, and September 2024. If you'd be around and would like to chat about our work on the intersection of machine learning, disaster risk, and remote sensing, please feel free to send me an email.
Jan 12, 2024 - I have been accepted to attend the Bayesian Machine Learning as a tool for Climate Scientists Workshop at the Institute of Computing for Climate Science, University of Cambridge on March 25-27, 2024.
Jan 4, 2024 - I have been accepted to attend the Probabilistic Numerics Spring School 2024 at the University of Southampton on April 8-10, 2024.
Dec 14, 2023 - I led our Filipino community for a courtesy call on the Philippine Consulate General in San Francisco as we (20+ Filipino researchers) came together to share the Philippine stories of Earth sciences at the American Geophysical Union Annual Meeting 2023. Read more here.
Dec 13, 2023 - I presented "Near-real-time Country-wide Estimation of Susceptibility and Settlement Exposure from Norwegian Mass Movements via Inter-graph Representation Learning" at American Geophysical Union Meeting 2023 in San Francisco, California. Access here the YouTube recording.
Oct 5, 2023 - Our work "Near-real-time Country-wide Estimation of Susceptibility and Settlement Exposure from Norwegian Mass Movements via Inter-graph Representation Learning" has been accepted as an oral presentation at American Geophysical Union Meeting 2023 in San Francisco, California, on 13th of December 2023. Learn more about the presentation here.
Oct 1, 2023 - I officially began my PhD studies with Dr. Emily So (Cambridge University Centre for Risk in the Built Environment) and Dr. Christian Geiß (German Aerospace Center) and got my third master's degree MRes.
Jul 31, 2023 - I started an exciting remote summer research project at Bonn's Center for Remote Sensing of the Land Surface (ZFL) to collaborate with Dr. Emily So (Cambridge University Centre for Risk in the Built Environment) and Dr. Christian Geiß (German Aerospace Center) for a project about quantifying our global disaster risk. If you are also based in the Philippines seeing the possiblity of AI for DRR public policymaking, feel free to connect with me.
Jun 30, 2023 - Our MRes team (myself, Andrea, and Sakthy) ended our three-month research project on "Near-real-time Country-wide Estimation of Susceptibility and Settlement Exposure from Norwegian Mass Movements via Inter-graph Representation Learning". You may access the data and code in our Zenodo and GitHub repositories.
Jun 26, 2023 - I joined the UK Alliance for Disaster Research and the Disasters Research Group (DRG) to review and discuss current disaster research practice in the UK at Wellcome Trust, London.
Jun 21, 2023 - I presented my research on artificial intelligence and disaster risk assessment at Risk Management Solutions in London.
Mar 9, 2023 - I started working with Dr. Andrea Marinoni (Engineering) and Dr. Sakthy Selvakumaran (Engineering) to characterize human-environment factors for climate change adaptation and mitigation in the Arctic by causal graph inference.
Mar 1, 2023 - I co-authored "Seismic Impacts Report for M7.8 and M7.5 Turkey-Syria Earthquake Using Near Real-time Crowdsourcing and Remote Sensing Information".
Jan 27, 2023 - Our team's work on "Classifying sea ice in high-resolution SAR imagery using deep learning" has been accepted to EGU General Assembly 2023.
Jan 21, 2023 - EMI delivered 20 reports on updated climate and disaster risk assessment of Quezon City to the local government (read more).
Dec 17, 2022 - I co-authored a Nature Communications paper "Seismic multi-hazard and impact estimation via causal inference from satellite imagery".
Sep 26, 2022 - I started MRes+PhD in Application of Artificial Intelligence to the study of Environmental Risks at the University of Cambridge with a very generous full-funding studentship from UK Research & Innovation and Department of Earth Sciences. I will be part of Sidney Sussex College postgraduate community.
Aug 31, 2022 - I completed Global Urban Resilience Fellowship at EMI (Philippines). I am very grateful for the valuable mentorship from Dr. Fouad Bendimerad and for our hardworking team - Jaja, Mam Lourdes, Doc Renan, Laarni, Hector, Jom!
Aug 4, 2022 - I participated as part of the EMI expert observers for the Quezon City Public Service Continuity Plan Tabletop Exercise preparing for the "Big One" Mw-7.2 earthquake.
Jun 12, 2022 - I graduated MS Civil & Environmental Engineering (Sustainable Design and Construction - Structures) and MA Public Policy at Stanford University.