The Open Database of Greenhouses

The Open Database of Greenhouses (ODG) is a collection of digitized greenhouses and their locations across Canada and is made available under the Open Government Licence - Canada. The ODG is a derivative product of open data that leverages high and medium resolution earth observation imagery available from various sources, such as open data portals, partnerships, and agreements, which are controlled by municipal or provincial governments or were provided by companies that hold an existing National Standing Offer with the Federal Government, or International Space Agencies. This project uses leading-edge methods, data integration and advanced technologies in an effort to reduce the response burden on agricultural operators. As part of these efforts, the ODG is used as a tool for new technologies created to automate the collection of greenhouse data across Canada by Statistics Canada's Agriculture Division (Agriculture and food statistics) in collaboration with Statistics Canada's Data Science Division.

In the newest version of the ODG, some greenhouse footprints were automatically detected and digitized using leading-edge machine learning methods from Statistics Canada’s development of high-resolution through the application of super resolution methods  over open-source medium resolution imagery (Figure 1).

This research and accompanying data are some of the many ways Statistics Canada is endeavoring to reduce response burden on agricultural operators in Canada with its AgZero project. This database was released in collaboration with the Centre for Special Business Projects and is a component of the Linkable Open Data Environment (LODE).

Data sources and methodology

The ODG is a derived product of digitized footprints using a collection of earth observation imagery sources, which are available to the public through open government portals under various open data licences or obtained through partnership agreements with municipalities, provinces and companies which hold a National Standing Offer with the Federal Government.

In the first version of the ODG, the earth observation data was a collection of free and open high-resolution aerial and satellite imagery, with 7.5 cm-50 cm pixel resolution provided by the municipalities. For the second version of the ODG, manually digitized greenhouses were identified and confirmed using medium resolution Sentinel-2 imagery, and open-source imagery platforms for high resolution imagery for the 2023-2024 timeframe. The earth observation data and derived greenhouses are made available across 38 different municipalities within four provinces in Canada.

The first version of the database (version 1.0) contains approximately 2,400 records of greenhouses. Earth observation data as mentioned above is provided for the years 2017-2021.

The second version of the database (version 2.0) contains approximately 3,900 records of manually digitized greenhouses using a timeframe between 2023 and 2024. The machine-learning-based digitization component of the ODG was able to identify 1,006 greenhouses for the 2021 timeframe. For each earth observation dataset, visually identified greenhouses are digitized in relation to the image and then the dataset is further processed to create a collective dataset across Canada.

Moreover, Statistics Canada researched and applied super resolution methods to construct high resolution imagery from open-source medium resolution satellite imagery (ESA’s Sentinel-1 and Sentinel-2). The second version of the ODG includes leading edge methods for automatically detecting and digitizing greenhouses with a machine learning model applied solely to super-resolved images.

Greenhouses identified within each version of the datasets do not discriminate by greenhouse type, such as what is growing inside, and are not labelled differently based on any features that could help classify them. There is further possibility that the dataset includes buildings that look like greenhouses but are not defined as such by Statistics Canada's greenhouse definition, such as event spaces held in greenhouses, garden nurseries, 3-season greenhouses (tunnels) or greenhouses not yet active for vegetation growth. Data from the earth observation datasets is not copied into the ODG and instead original variables are created during processing of the dataset.

The machine learning results are experimental and thus may not represent a fully accurate and complete data of greenhouses on the ground. As will all modelling techniques, limitations exist: (i) The machine learning model was trained on Southern Ontario data and its ability to scale to other regions has not been tested, (ii) The machine learning model performance is affected by the resolution of the imagery and the size of greenhouses, specifically (like with human digitization) small greenhouses may be too small to detect.

To see the full list of variables and what information they include, in addition to data sources of earth observation imagery, please visit the metadata document that accompanies the ODG.

The variables included in the ODG version 2 are as follows:

• Unique ID
• Shape
• Province or Territory Unique Identifier
• Province or Territory Names
• Longitude
• Latitude
• Data Source
• Imagery Date
• Collection method (V2)

For more information on how these variables were compiled see the metadata document that accompanies the ODG.

IMDB: Surveys and statistical programs - The Open Database of Greenhouses (ODG)

Downloading the ODG

For ease of download, the ODG shapefile and accompanying metadata document is provided as a zipped file.

• ODG version 1.0
• ODG version 2.0

Visualizing the ODG

The ODG content is available for visualization on a map using the Linkable Open Data Environment Viewer.