Global Renewables Watch, A First-of-Its Kind Living Atlas to Help Decisionmakers Facilitate a Much-Needed Transition to Renewable Energy
Overview
Mitigating the impacts of climate change is a massive undertaking that will require a rapid global buildout of renewable energy. We need at least a nine-fold increase in renewable energy production to meet the Paris Agreement goals and much more to achieve net-zero emissions by 2050.
The land-use footprint for this buildout will be enormous. Accelerating the transition to renewable energy requires a smart buildout that safeguards natural areas and supports goals for climate, conservation and communities.
The Global Renewables Watch is a first-of-its-kind living atlas intended to map and measure all utility-scale solar and wind installations on Earth using artificial intelligence (AI) and satellite imagery, allowing users to evaluate clean energy transition progress and track trends over time. In this exciting project, Microsoft is providing the AI and platform technology, Planet is contributing the underlying satellite imagery, and The Nature Conservancy is overlaying the subject matter expertise to analyze the output.
Q&A
What is the latest with Global Renewables Watch and when can we expect a global launch?
The team has been hard at work mapping all of the utility-scale wind and solar around the rest of the world. We are now validating the methodology through the scientific peer review process and we expect the atlas to launch in early 2024.
What is the problem the Global Renewables Watch is solving?
About three-quarters of global greenhouse gas emissions come from energy use, meaning a rapid, massive shift to clean energy is urgently needed for the world to get to net-zero emissions by 2050. The global development of renewable energy infrastructure will require a lot of land. This massive land footprint raises the potential for widespread land-use conflicts due to community and environmental concerns. These conflicts could slow the transition to clean energy and harm ecosystems that are important for both nature and people.
In order to optimize the transition to renewable energy, the world needs access to data that helps both researchers and policymakers understand current capacities and gaps and assists decisionmakers in planning and siting renewable energy development in ways that minimize disruption to ecosystems while promoting equal access to energy.
The dynamic Global Renewables Watch aims to provide that critical data, on a biannual basis, to allow users to evaluate clean energy transition progress and track trends over a period of time rather than as a moment in time.
At a more granular level, tracking every wind and solar energy project on Earth would be an immensely complex undertaking, and as such, there is currently no comprehensive, global system for independently assessing and monitoring the growth of renewable energy resources in real-time. Global Renewables Watch aims to address this by coupling the abilities of AI and satellite imagery.
Who is eligible to sign up for private preview access?
Global Renewables Watch is geared for academics, policymakers, and decisionmakers in the public and private sectors. Everyone is eligible to access the private preview for the Global Renewables Watch. In order to manage traffic while the tool is still in preview, Microsoft’s AI for Good Lab will approve requests on an ongoing basis. There may be a delay in approval depending on the volume of requests.
Can users who gain access to the private preview immediately begin using the data for their research, planning, etc.?
Users of the private preview will immediately be able to reference the data in its current form for their own research, but it’s important to note that downloadable data will not be available in private preview. We expect this functionality to be included in the full release of the tool in early 2024.
Why is Global Renewables Watch in private preview, is this data reliable for further analysis?
The Global Renewables Watch private preview has undergone initial validation for all 6 countries currently available. We are currently working on a peer-reviewed scientific publication of our methodology, along with the release of the full dataset, which will serve as a global inventory of solar and wind capacity. Until that time, expected early 2024, we do not recommend definitive conclusions are made using Global Renewables Watch.
Throughout the process, we’ve compared our estimated country level aggregates of wind and solar capacity to published numbers in BP’s Statistical Review of World Energy. To validate model outputs, we’ve used a combination of the Amazon Mechanical Turk platform and Microsoft’s own labeling tools. To ensure optimal accuracy of the data before publication, we’re contracting with a third-party validator.
As the Global Renewables Watch is designed to be a living atlas, we will continue to make updates over time. If you have comments on the mapping, we’d welcome your feedback using the form linked within the private preview experience.
Can users publish the insights derived from Global Renewables Watch data?
The Global Renewables Watch website will make derivatives of the data (including the footprint, spatial data on project boundaries, growth, and estimated total available for download. The platform and algorithms used to derive these indicators will be released under open-source license for independent validation and testing purposes.
The underlying, raw satellite imagery is not being released, but researchers at credentialed research institutions may apply for access to sample data for testing and validation purposes under Planet’s Education and Research program.
How is solar and wind capacity measured? Why does Global Renewables Watch data not match up with publicly reported renewable energy capacity figures?
There have been many important efforts to map solar and wind capacity, and Global Renewables Watch is designed to complement them.
To convert wind turbine predictions to capacity, we currently use a standard factor of 4 megawatts (MW) per square kilometer of land that contains at least one wind turbine.
To convert solar farm predictions to capacity, we currently use a shifting factor of MW per square kilometer based on the year a solar farm was constructed:
| Year | Factor (MW) |
|---|---|
| 2018 | 51.5 |
| 2019 | 55.8 |
| 2020 | 60.09 |
| 2021 | 64.4 |
| 2022 | 68.69 |
These values scale linearly from a 2013 estimate of 30 MW/sq km (Ong et al. ) to a 2022 estimate of 659 MW/sq km (Bolinger and Bolinger ) using the following equation:
(69 – 30) / (2022 – 2013) = ~4.33
This method assumes that the first year a solar farm is detected equals the year of construction.
What are some of the early wins / challenges you’ve faced so far expanding the platform?
The main challenges of this project are:
- The scale at which we are working – we are working with many terabytes of satellite imagery and hundreds of thousands of solar and wind predictions per model
- Establishing proper validation pipelines so that we can iterate on the underlying machine learning methods used to create the dataset and release the most accurate dataset possible.
Are there any concerns of bad actors using the resulting mapping of energy sites for nefarious purposes?
Unfortunately, with any mapping project of this scale there is inherent risk. We are focused on democratizing access to this renewable energy data, and firmly believe that this transparency can drive accountability and the knowledge that everyone has access to this information can help deter bad actors.
Where does the data reside- in which data centers? Why?
All of the data and code associated with the Global Renewables Watch will be made open-source and available to all.
How is the Microsoft Planetary Computer used?
- The Planetary Computer hosts many open datasets (e.g. the Bing Maps global building footprints -- Microsoft Building Footprints) and allows users to search over and access these datasets programmatically. We are planning to host the datasets produced by the Global Renewables Watch on the Planetary Computer (specifically, the solar farm polygon and windmill points).
- Further, we currently leverage the datasets already hosted on the Planetary Computer in the Global Renewables Watch. We use the Sentinel 1 RCT dataset to find offshore wind farms and use the Sentinel 2 dataset to filter false positive predictions from our model.
What differentiates this tool from those other organizations have built?
There have been many important efforts to map solar and wind capacity, and Global Renewables Watch is designed to complement them. Succinctly, the Global Renewables Watch is unique in 3 ways:
- It is a time series, not a snapshot – the Global Renewables Watch monitors progress and trends over a period of time vs. a moment in time
- It is regularly updated – the Global Renewables Watch plans to release bi-annual updates of the atlas
- It documents development patterns, not just the development – the Global Renewables Watch tracks progress on renewable energy targets at a country-by-country basis, as well as accounts for underlying patterns where this development may cause conflicts
In more detail:
- The Global Renewables Watch is the first tool that will allow users to evaluate clean energy transition progress and track trends over a period of time rather than as a moment in time.
- With access to data as far back as 2018, and plans to update twice annually, the Global Renewables Watch will be able to help countries, regions, and ultimately the world, explain its renewable energy output, assist in understanding that output, and the potential impact of the renewable energy siting on the landscape over time.
- The Global Renewables Watch will be able to express the underlying patterns that are driving renewable energy development, highlight places where this development may cause conflicts, and assist in the development of alternative scenarios to reach renewable energy targets without conflict – See here for an example: SireRight.
How do we know this approach works? What’s the underlying science? How does this effort improve upon prior efforts?
The Global Renewables Watch builds on two seminal pieces of work – a 2018 effort by Descartes Labs to map global solar capacity using 10-meter Sentinel 2 imagery and an earlier 2022 collaborative undertaking from Microsoft and The Nature Conservancy (TNC) that used satellite imagery to map installed solar capacity in India. The results of the latter were published in Nature Scientific Data (See Scientific Data volume 9, Article number: 497 (2022)).
As we continue to develop the tool, we will look to undertake peer review on outputs and share validation statistics on our website and any resulting peer-reviewed publications.
Are any utilities or government data used to locate solar/wind facilities or estimate capacity?
No, training data for the artificial intelligence model was sourced from publicly available sources. Specifically, we use datasets from Ortiz et al. (Nature Scientific Data, 2022), Kruitwagen et al. (Nature, 2021), and from OpenStreetMap.
What is the expected timeline of this project?
We have completed initial mapping of solar and wind installations for the United States, Germany, India, Brazil, Egypt, and Kenya. We are planning to have the first full global inventory completed in 2024, at which point the results will undergo both scientific and technical validation. The intent is that global updates to the data will be made available twice annually thereafter.
Our goal is to maintain this atlas over time so that it can be used as a resource for users including researchers, policymakers, and the public to track clean energy transition progress and trends over time in order to meet global emissions targets.
Is the underlying data publicly available and downloadable?
While this functionality is not available in private preview, the Global Renewables Watch website will make derivatives of the data (including the footprint, growth, and estimated total output) available for download alongside the full rollout in early 2024. The platform and algorithms used to derive these indicators will be released under open-source license for independent validation and testing purposes.
When does the timeseries begin (i.e. how far back does the data go)?
In order to maintain accuracy, the Global Renewables Watch relies on high spatial and temporal resolution data from Planet, and as such, mapping from as far back as 2018 will be made available for analysis.
Are there plans to move beyond wind and solar?
We are currently focused on delivering the first full global inventory for onshore wind and solar by early 2024. We’ve decided to start with onshore wind and solar as they are expected to account for over 75% of renewable energy growth by 2030 according to the International Energy Agency. Estimates for offshore wind are expected to also be available in 2024. We will evaluate additional data sources in the future, as we look to build the most complete picture of the world’s renewable energy resources. Going forward we may also use Global Renewables Watch data to predict future potential of renewable energy for any area of the world.
Are there plans to move beyond grid-scale renewable energy?
We are currently focused on delivering the first full global inventory of ground-mounted grid-scale wind and solar by 2024. We will evaluate additional sources in the future, as we look to build the most complete picture of the world’s renewable energy resources.
There are plenty of organizations that track renewable energy. Why should we believe your data over more traditional sources, and why, frankly, is this effort necessary?
The Global Renewables Watch is intended to complement, not compete with, existing sources of information about renewable energy sources.
Different entities report progress differently - using different methods, sources, and timeframes. Our effort aims to: use a globally consistent approach with standardized tools, satellite imagery and algorithms, update the atlas more frequently than most other sources; and address issues like the land-use impacts of renewable energy that are often overlooked in other efforts. As such, we hope it will be a helpful tool that can both inform and be informed by other sources, and support a more effective, environmentally beneficial, and timely transition to a more sustainable future.
When will the Global Renewables Watch have data online?
We have completed initial mapping of solar and wind for the United States, India, Brazil, Egypt, Kenya, and Germany. We are planning to have the first full global inventory completed by 2024, at which point the results will undergo both scientific and technical validation. We will have more to share at a later date.
Who is funding this effort and what does it cost?
The Microsoft AI for Good Research Lab is a philanthropic team of data scientists and researchers dedicated to using AI, Machine Learning and statistical modeling to tackle some of humanity’s greatest challenges. The Global Renewables Watch is the latest philanthropic project and we’re thrilled to be partnered with Planet and The Nature Conservancy.