Dublin … city of writers, poets, musicians and geospatial cloud engineers. Yes, the fair city of Dublin hosted the Geo for Good Mini Summit last month, where we picked up a lot of exciting new insights from Google, met some amazingly talented people working hard to address the world’s most pressing environmental and social problems, and gained a new appreciation of Guinness.
Each year, the Geo for Good Summit brings together an awesome community of technologists, environmentalists, remote sensing scientists, AI experts, and NGOs, and showcases all things new in Google Cloud Geography and the Earth Engine platform. Until now, the event has been held in California, but this year, Google decided to hold regional summits in Sao Paulo and Dublin to enable more people from around the world to get involved.
Let’s dive into the highlights from this year’s summit:
Multiple datasets are essential to EUDR risk assessments
Unsurprisingly, given this event was in Europe and we are only 83 days away from implementation, the new EU Deforestation Regulation (EUDR) was an important topic. Despite some pushback in recent months, a panel of experts working in this area agreed that this legislation will ultimately be successful in changing the way companies manage their supply chains and will be positive in the transition to more sustainable business processes.
They agreed that the key to achieving a reliable risk assessment for EUDR compliance was to not just rely on a single dataset but instead use an “ensemble” of multiple data sources. This aligns with Earth Blox package for EUDR assessments, where we combine the Hansen, Dynamic World, JRC Forest Cover (2020), and iFORCE Tropical moist Forest datasets to assess deforestation risk.
Another theme that emerged was ensuring that reporting to EUDR is accessible to smallholder farmers. Several technical sessions discussed how deforestation analyses could be made more accessible to people outside the geospatial community by wrapping code in Google Cloud web services, which can be used in downstream applications such as the Open Foris Ground field data collection mobile app.
Earth Engine’s “Embedding Fields” dataset will speed up analyses
A technical highlight was the preview of the Earth Engine “Embedding Fields” dataset - a “foundation” model that the Earth Engine team developed in collaboration with Google Deep Mind, using the same generative AI techniques used to create Large Language Models (LLM’s) like Gemini and ChatGPT.
This dataset can be thought of as a compressed representation of multiple global-scale satellite imagery datasets such as Landsat, Sentinel 2, and Sentinel 1, where each pixel represents 64 salient features (an embedding) that best describe that place in the context of the pixels around it. For example, a set of features that characterise the pixel as being part of a gold mining site.
When this dataset is released (we think early next year), it has the potential to significantly speed up and extend many of the analyses available in Earth Blox, such as classification, change detection and regression.
New datasets coming to Earth Engine and Earth Blox
New and upcoming dataset releases sparked the same excitement in the Geo for Good community as opening a bag of chips does for seagulls at the seaside. Some that caught our attention were:
- MethaneSAT: will be available in the Google Earth Engine catalogue once the satellite starts producing data later this year. Full operations will start in Q1 2025. Meanwhile, the MethaneAIR datasets are already available.
- NEON Canopy Height: 1m resolution over North America NEON (US’s National Ecological Observatory Network). This dataset supplements the many global canopy height datasets available in Earth Blox already. The ability to substitute in best-available local data is one of the reasons that so many people use Earth Blox for their environmental analyses. Explore our dataset catalogue here.
- European Primary Forest Polygons: harmonizes 48 different, mostly field-based datasets of primary forests and contains 18,411 individual patches (41.1 Mha) spread across 33 countries, making it ideal for EUDR assessments.
Also announced were updates to Earth Engine Python SDK that will make it easier to work with familiar data science tools such as xarray and geemap. This merited promotion, at last, of the SDK to version 1.0.0, which feels like a symbolic milestone. Another exciting technical announcement was closer integration with BigQuery, including a read connector so that data can be read directly from BigQuery and processed in Earth Engine. This will be released soon.
Ever since we founded Earth Blox, Geo for Good has been a key fixture in our company calendar. We always look forward to the announcements, product launches, technical know-how and thought leadership that the Google team provides each year for the awesome GEE community.
For me, it was the first time I attended Geo for Good in person, and what really struck me was what a friendly and welcoming community it is. As well as great hospitality, the experts and technologists at Google are just as eager to collaborate, listen, and learn as they are teaching and presenting their own ideas. You do come away feeling you’re part of a global team working together to solve complex problems and that you share a common goal. That’s why Earth Blox is proud to be a Google Cloud Sustainability Partner, helping businesses harness the power of geospatial technology to ensure a sustainable future.
Ben Butchart
Ben Butchart is Head of Development at Earth Blox. Ben has over 20 years of software engineering and management experience at the Crown Office, JP Morgan, University College London, Dresdner Bank AG, and EDINA, where he led the delivery of geospatial services such as Digimap.