Earth Engine in BigQuery: A New Geospatial SQL Analytics

Earth Engine in BigQuery

With the direct integration of Earth Engine in BigQuery, Google Cloud has shown a major extension of its geospatial analytics capabilities. This new tool, which was unveiled at Google Cloud Next ’25, allows the SQL community to access sophisticated analysis of geographical information obtained from satellite imagery by integrating potent raster analytics into the well-known BigQuery environment.

Google Cloud customers have favoured BigQuery for storing and querying vector data, which employs points, lines, or polygons to express spatial features like buildings and boundaries. However, Earth Engine in BigQuery is recommended for processing and storing raster data like satellite imagery, which encodes geographic information as a grid of pixels with values like temperature, height, and land cover.

“Earth Engine in BigQuery” combines the best of vector and raster analytics in one place. This integration has the potential to increase access to advanced raster analysis and make it easier to address a variety of real-world business problems.

This integration is being driven by the following key features:

• ST_RegionStats() is a new geography function in BigQuery. This function enables users to effectively extract statistics from raster data within designated geographic borders and is comparable to Earth Engine’s reduceRegion function. An Earth Engine-accessible raster picture and a geographic region (vector data) are used to calculate aggregate values (such as mean, min, max, total, or count) for the pixels that cross the geography.
• Previously known as Analytics Hub, BigQuery Sharing now offers new Earth Engine in BigQuery datasets. This makes it easier to find data and gain access to an expanding number of datasets, many of which are ready for analysis to derive statistics for a particular region of interest. These datasets may include information about risk prediction, elevation, or emissions.

There are usually five phases involved in performing raster analytics with this new capability:

• Locate the vector data that represents the areas of interest in a BigQuery table.
• Locate a raster dataset that came via Earth Engine in BigQuery image assets, Cloud GeoTiff, or BigQuery Sharing.
• To determine aggregate values on the intersecting data, use the ST_RegionStats() function with the raster ID, vector geometries, and optionally a band name.
• To gain understanding, examine the output from ST_RegionStats().
• Use tools like BigQuery Geo Viz to properly visualise the analysis findings on a map.

New opportunities for data-driven decision-making across a range of sustainability and geospatial use cases are made possible by this integration:

Climate, physical risk, and disaster response: Applying datasets on drought conditions, wildfire danger, or flood mapping to transportation, infrastructure, and urban planning applications. For example, utilizing the Wildfire danger to Communities dataset to evaluate the danger of wildfires or the Global River Flood Hazard dataset to comprehend the level of expected flood inundation.

Sustainable sourcing and agriculture: Examining land-use, elevation, and cover categories for agricultural evaluations and supply chain management. This involves determining if commodities are cultivated in non-deforested regions by utilizing data such as the JRC Global Forest Cover databases or the Forest Data Partnership maps.

Methane emissions monitoring: To support mitigation efforts, methane emission hotspots from tiny, scattered sources in regions like oil and gas basins can be found by analyzing datasets such as the MethaneSAT L4 Area Sources data.

Custom use cases: Enabling users to import their own raster datasets from Earth Engine in BigQuery image assets or Cloud Storage GeoTiffs.

ST_RegionStats()’s raster data sources are located in BigQuery Sharing, where the assets.image.href column usually contains the raster ID for each picture in an image table. By supplying their URI, Cloud Storage GeoTIFFs in the US or US-central1 regions may be utilised. Additionally supported are Earth Engine in BigQuery image asset locations such as ‘ee://IMAGE_PATH’.

Using the include argument in ST_RegionStats(), users can modify computations by giving pixel weights (ranging from 0 to 1), where 0 denotes an incorrect pixel that represents missing data. Pixels are weighted according to their position within the geometry if no weight is supplied. One important factor that influences calculation and output is the raster’s pixel size, sometimes known as its scale. Changing the scale (for example, by using options => JSON ‘{“scale”: 1000}’) can save query runtime and cost for prototyping, but because it affects results, it is not advised for production analysis.

Because Earth Engine does the calculation, using the ST_RegionStats() method is invoiced separately under the BigQuery Services SKU. The number of input rows, the particular raster picture and its quality, the size and complexity of the input geography, the number of crossing pixels, image projection, and formula usage are some of the variables that affect costs. By modifying a certain quota associated with Earth Engine in BigQuery slot time utilization from BigQuery, costs may be controlled.

At the moment, ST_RegionStats() queries need to be executed in the US, us-central1, or us-central2 areas.

This is a major advancement in Google Cloud’s geospatial analytics, opening up access to sophisticated raster capabilities and facilitating greater understanding of sustainability and other data-driven decision-making processes.

You can also read Column Granularity Indexing in BigQuery Alters Query Speed