On June 2, 2025, a major update has been implemented to our Crude Inventories methodology, designed to improve the accuracy and value of our Crude Inventories product. Consequently, there has been a revision of historical data for crude inventory figures.
What has been improved?
Three key improvements can be seen in this update:
Outlier removal, which has led to a significantly smoother and more consistent time series.
Improved fill level measurements resulting in greater delta precision.
Reduction in full tank overestimations, resulting in more accurate tank capacity measurements.
Examples of Old vs New time series on 2 installations: outliers have been removed, leading to smoother and more consistent time series
How did we improve the data?
We enhanced the accuracy and reliability of our crude oil inventory measurements by implementing key upgrades across our input data, algorithm, and processing pipeline— while maintaining our core methodology of using satellite imagery.
Input Data Enhancements:
We now use high-resolution precision images to measure tank radii, replacing older, less precise mapping methods. Tank heights are calculated from validated fixed points, improving consistency. Additionally, a clustering algorithm adjusts the heights of nearby tanks with similar dimensions, reducing outliers and refining overall accuracy.
Algorithm Improvements:
Our updated algorithm incorporates image registration as a post-processing step to correct geometrical distortions common in satellite imagery—such as misalignment, scaling issues, or skewing. This alignment to a shared reference frame significantly reduces positional inaccuracies.
Processing Pipeline Upgrades:
The pipeline is now more efficient, with streamlined image retrieval that reduces redundant steps. Calibration is applied selectively, avoiding unnecessary corrections and potential errors. Finally, data is stored in a more accessible format, allowing faster and more targeted data access.