# Assessment and Calculation | Grid and Regions methodology
Source: https://greenpixie.com/docs/methodology/grid/assessment-and-calculation

The calculation components and their outputs.

## 7.1 Source Data Collection [#71-source-data-collection]

For each period being calculated, the methodology retrieves each source's published data and retains every response as received.

On every update the methodology retrieves each source's data again for that source's revision window, being the period over which that source revises figures it has already published. Values inside that window can change between updates. Values older than the window are expected to be identical from one update to the next.

## 7.2 Grid Area Intensity Calculation [#72-grid-area-intensity-calculation]

For every grid area and hour, the methodology produces four figures, each in tonnes CO₂e per kWh. The total Grid Carbon Intensity is the sum of three components, being combustion, upstream and downstream, and transmission and distribution.

Where a source publishes generation by fuel, Greenpixie multiplies each fuel's share of the hour's generation by the emission factor for that fuel in that grid area, and sums the products. Each grid area's emission factor table holds a total value and a combustion value for every fuel. The upstream and downstream value is the difference between the two. Where a source publishes a carbon intensity directly, Greenpixie takes that figure as the grid area's intensity, as published. Where the source for a grid area is a national inventory figure, that figure covers every hour within the period its publisher states. A fixed published figure with no stated end date covers every hour after its publication until its publisher issues a replacement. Where a grid area's intensity is calculated from a yearly modelled dataset, the latest published year covers the hours after it. The year carried can be older than the year being calculated.

Where a source publishes electricity imported from another grid, that import takes the average intensity of the same grid area in the same hour. One grid area's imports take an intensity built from published fuel factors instead.

For each grid area, Greenpixie calculates the transmission and distribution losses by one of two methods, being a published percentage loss for the area or a loss component the source publishes itself. Under the first method, Greenpixie grosses the generated intensity up to a delivered basis, as the generated intensity divided by one minus the loss. The difference between the two figures is the loss component. Under the second method, used where the source publishes a loss component, as the Australian National Greenhouse Accounts Factors do, Greenpixie takes that component as the loss factor and grosses it up no further. A factor already published per kilowatt-hour delivered or sold accounts for losses. For every fixed published figure, Greenpixie records the point in the network at which the figure was measured.

## 7.3 Coverage of Principal Markets [#73-coverage-of-principal-markets]

Across AWS, Azure and Google Cloud, Greenpixie publishes factors for 269 cloud regions, built from 91 grid areas. Section 5 names every category of source those factors draw on. The paragraphs below describe the source and the approach Greenpixie uses in the United States, the United Kingdom, Europe and Australia, from the approaches Section 7.2 sets out. Grid areas outside those four markets are not described individually. Each source below publishes for more grid areas than are configured today. Section 6.1 sets out when a further area is added.

In the United States, Greenpixie produces a separate hourly intensity for each of the 33 balancing authorities it configures, covering areas such as Northern Virginia, Oregon, Northern California, Dallas and Chicago. Each is calculated from the hourly generation by fuel the Energy Information Administration publishes in Form EIA-930, using the subregion emission factors the Environmental Protection Agency publishes with eGRID. From publicly available information, Greenpixie resolves each cloud region to the county or counties it occupies. The hourly intensity is the weighted average of the balancing authorities serving those counties, which is a narrower basis than a state or national average. Greenpixie checks the configured subregion emission factors against those the Environmental Protection Agency publishes.

In the United Kingdom, Greenpixie takes the half-hourly generation mix the National Energy System Operator publishes for each distribution region and multiplies each fuel by an Intergovernmental Panel on Climate Change factor. Greenpixie calculates each European grid area the same way, multiplying its published generation mix by a factor for each fuel. Active cloud regions in London, Cardiff and the North West of England therefore carry distinct hourly figures.

Across Europe, Greenpixie produces an hourly intensity for each of the 14 ENTSO-E areas it configures, covering cloud regions such as Frankfurt, Dublin, Amsterdam, Stockholm, Milan, Madrid and Athens. An area is a whole country in most cases. Where a country is split into bidding zones, the area is the bidding zone. For Denmark, Greenpixie configures the DK2 bidding zone, covering the east of the country. Each is calculated from the actual generation per production type the ENTSO-E Transparency Platform publishes, at fifteen, thirty or sixty minute resolution depending on the country, averaged to the hour. The Transparency Platform separates around twenty production types, distinguishing three kinds of hydro, two of wind, and lignite from hard coal. Each country carries its own factor table.

In Australia, Greenpixie produces separate hourly intensities for active cloud regions such as Sydney, Melbourne, Canberra and Perth. Each state of the National Electricity Market is a grid area in its own right. Perth's South West Interconnected System is a separate grid area. Each is calculated from the National Greenhouse Accounts Factors for its state, adjusted to the daily or hourly market data the grid operator for that system publishes.

## 7.4 Cloud Region Mapping [#74-cloud-region-mapping]

For each cloud region, Greenpixie calculates the GCI as the weighted average of the grid areas supplying that cloud region, using the weights described in Section 6.3. Where no grid area is matched to a cloud region, that cloud region's GCI is the global mean described in Section 6.3. The total intensity and its three components are each averaged independently, as is the fuel mix. Those components sum to the cloud region's total as they do for a grid area. Where any contributing grid area has an unknown split between its combustion component and its upstream and downstream component, the cloud region's split is published as unknown.

Each cloud region's water stress category is the category recorded for the province or administrative region containing the city its data centres are located in. Where Aqueduct 4.0 records no value for a location, that absence is itself a category and ranks below every published category.

## 7.5 Gap Treatment and Provenance Labelling [#75-gap-treatment-and-provenance-labelling]

Against every published hour, Greenpixie keeps an internal record, the Provenance Label, of how that hour's value was produced. The labels distinguish measured source data, a published forecast, a figure carried forward from the latest period a source has published, a value filled from other published hours, and a fixed published figure applied across a period.

## 7.6 Cloud Provider Efficiency Factors [#76-cloud-provider-efficiency-factors]

Every supported cloud region carries a power usage effectiveness value, looked up against what its cloud provider publishes. Where a cloud provider publishes a figure covering the cloud region, Greenpixie takes that figure as the cloud region's power usage effectiveness, unchanged. AWS publishes a value for most of its cloud regions. Azure publishes a value for each of three large geographic groupings. Every Azure cloud region within a grouping takes that grouping's value. Google Cloud publishes by campus. A cloud region matching one campus takes that campus's published trailing twelve month value. A cloud region matching several campuses takes the mean of those campuses' published figures.

Where a cloud provider publishes nothing covering the cloud region, Greenpixie assigns a value from other published power usage effectiveness figures. Against every assignment, Greenpixie records which location the value was taken from, and why.

Reporting periods differ by cloud provider, being a calendar year for AWS, a fiscal year for Azure and a trailing twelve months for Google Cloud. Each is the latest annual figure that cloud provider publishes.