Research method

How the SAPP grid atlas is built

Every source, every processing step, and which lines, substations and plants are confirmed by more than one source. Written so that someone else could check our work, or repeat it.

Atlas release 2026.11, 13 November 2026. Open the atlas · Download the data

  1. The question
  2. Coverage
  3. Sources and dates
  4. Processing steps
  5. Sources behind each feature
  6. How we check it
  7. Assumptions and limits
  8. Known gaps
  9. Updates and versions
  10. Data and citation

The question

Where is the high-voltage grid of the Southern African Power Pool today, what is promised next, and which public sources stand behind each piece of that picture? No single public map answers all three. Utility maps are out of date or not published, open maps do not say where each line comes from, and project lists rarely keep the dates that were promised before.

Coverage

  • Where: the 12 SAPP member countries: Angola, Botswana, DR Congo, Eswatini, Lesotho, Malawi, Mozambique, Namibia, South Africa, Tanzania, Zambia and Zimbabwe. Lines in five neighbors (Burundi, Congo, Kenya, Rwanda, Uganda) are drawn faintly for context and are not counted.
  • What: transmission lines of 66 kV and above, plus mapped power lines with no voltage tag; substations; power plants of 10 MW and above; transmission and generation projects in the pipeline; SAPP's draft master plan to 2045; SAPP's transfer limits.
  • When: the grid as mapped in November 2026, with plant and project dates from 2004 to 2045.
  • Not covered: distribution lines below 66 kV (see our grid-edge atlas), plants under 10 MW, line ratings, and the electrical state of any line (in service, de-energized, damaged).

Sources and dates

Public sources only. Commercial maps are never a source.

LayerSourceDate of dataLicense
Lines and substationsOpenStreetMap contributors, country extractsExtracted 4 November 2026ODbL 1.0
Lines missing from OpenStreetMapWorld Bank Africa electricity grid map; TANESCO power system; DR Congo and Mozambique transmission networks; SNEL network (World Bank DRC project)2015 to 2018CC BY 4.0
Power plantsGlobal Energy Monitor, Global Integrated Power Tracker, first; then the WRI Global Power Plant Database, OpenStreetMap and the World Bank AICD inventory where GEM has nothingGEM September 2026; WRI 2021CC BY 4.0; ODbL for OpenStreetMap
Project pipeline and promised datesOur own register, from SAPP documents, utilities, ministries, funders and dated news reports. Each promise keeps its link.2017 to November 2026CC BY 4.0 (ours)
Substation transformer capacityNTCSA Generation Connection Capacity Assessment 2025 (South Africa); a JICA power system study with ZESCO data (Zambia)2023; 2006Public documents, facts cited
Master plan, transfer limits, regional statisticsSAPP draft Generation and Transmission Master Plan 2025-2045 (public consultation), SAPP transfer limits, annual and monthly reports, Pool Plan 20172017 to 2026Public documents, facts cited
Towns and boundariesGeoNames; Natural Earth2026CC BY 4.0; public domain

The full source list, with links, is in the Sources tab of the atlas. Licenses are set out on our terms page.

Processing steps

  1. Extract. Take every power line and substation from the OpenStreetMap country extracts. Keep lines tagged 66 kV and above and lines with no voltage tag.
  2. Classify voltage. Round each tagged voltage to a standard class (66, 88, 110, 132, 220, 275, 330, 350, 400, 500 or 765 kV). A 120 kV line is shown as 132 kV. Untagged lines keep "voltage not stated".
  3. Remove duplicates. A cross-border line appears in two country extracts. We keep one copy and credit the whole line to the country that contains its midpoint. Duplicate substations at borders are dropped.
  4. Fill gaps. Where an older utility or World Bank map shows a line and OpenStreetMap has nothing nearby, we add the uncovered stretch, labeled with its source and year. Stretches under 8 km are ignored as noise.
  5. Plants. Start from our own register, then add Global Energy Monitor plants, then other databases only where no plant of the same kind is already nearby. A nearby plant counts as a duplicate only when the names share a real word, so that neighbors such as Medupi and Matimba both stay.
  6. Projects and promises. For each pipeline project we record every operating date that was promised, who said it, when, and the link. Old promises are never overwritten. "Slip" is the gap between the first promise we found and the current date.
  7. Simplify for the web. Line shapes are simplified and coordinates rounded to three decimals (about 100 m). The map is not a survey.
  8. Check sources. Every feature is compared with the other public maps and databases, and the ones that show it are recorded, as described next.

Sources behind each feature

We do not grade features. Each line, substation, plant and project lists the public sources that show it, and the map counts them. One source means the feature rests on that source alone. Two or more means separate sources confirm it. You see the list in every popup, the count in the "Number of sources" color view of the map, and both as columns in the downloads.

LayerWhat counts as a source
Lines from OpenStreetMapOpenStreetMap, plus each reference map that shows the same line: the World Bank Africa electricity grid map (2017) and the national network datasets the World Bank publishes on energydata.info: ZESCO (Zambia, 2022), the Zambia transmission network (2009), TANESCO (Tanzania, 2015) and Kenya Power (2014). The national "2017" files on that site are country extracts of the Africa map and are not counted a second time. A map shows the line when at least 60% of points along it, sampled every 1.5 km, lie within 4 km of a line in that map.
Lines missing from OpenStreetMap, and planned linesThe map they were taken from. One source, and the route is approximate.
SubstationsOpenStreetMap, plus the official document that lists the substation's transformer capacity, where one exists.
Power plantsThe database the record comes from; the WRI Global Power Plant Database when it shows a plant within 5 km of a Global Energy Monitor plant; and each different publisher linked in our own register.
Pipeline projectsEach different publisher linked in the project's record of promised dates.

The same rules apply to all four of our grid atlases (SAPP, EAPP, Madagascar and the combined map). The World Bank grid map is treated as a source in its own right. Where its record comes from a utility, as in Namibia, Botswana and Eswatini, the popup says "utility data".

Where things stand today

Of 163,501 km of line in the 12 countries, 72% is confirmed by two or more sources and 28% rests on one. Counts as of 13 November 2026.

LayerOne sourceTwoThree or moreShare
Lines, km46,298104,69912,504
Substations3,0581410
Operating plants3081260
Pipeline plants429333
Transmission projects1143

Which maps confirm the OpenStreetMap lines

Also shown inLine km
World Bank Africa grid map (2017)103,951
World Bank Africa grid map (2017, utility data)12,321
ZESCO network data (World Bank, 2022)7,892
Zambia transmission network (World Bank, 2009)4,807
TANESCO power system (World Bank, 2015)4,487

Lines by country, km

CountryOne sourceTwoThree or moreShare
Angola4,2182,8040
Botswana6,3101,47225
DR Congo3,1016,7450
Eswatini5945180
Lesotho1384980
Malawi7041,9960
Mozambique3,0886,6570
Namibia81210,7240
South Africa16,57965,8300
Tanzania4,2655704,371
Zambia3,7811,5948,102
Zimbabwe2,7085,2906

How to read this. A line with one source is not a doubtful line. Much of the region's grid has been traced by volunteers from satellite images and is usually in the right place; one source only means no other public map shows it yet. The count says nothing about voltage or whether a line is in service.

How we check it

  • Totals against Open Infrastructure Map. On 8 November 2026 we compared line length per country with openinframap.org, which draws on the same OpenStreetMap data. Totals agree within about 10% for eight of the 12 countries. The larger differences (Lesotho 67% of their figure, Angola 88%, Eswatini 115%, Botswana 117%) come from how cross-border lines are credited, from lines under 66 kV and from lines with no voltage tag. This checks our extraction, not the underlying map.
  • Plants against Global Energy Monitor. The same check found an error in our duplicate rule that had removed Medupi (4,764 MW) and other plants. It was fixed on 8 November 2026.
  • Official line lengths. Where a utility publishes its total line length, the country view of the atlas shows it next to our mapped total.
  • Not yet done. We have not checked line positions against ground surveys or utility as-built records, and we have not measured positional error. That is the next step for this method.

Assumptions and limits

  • Routes are indicative. Simplified shapes can sit a few hundred meters from the real line, and gap-fill and planned routes can be several kilometers off.
  • A line with no voltage tag may be a distribution line. We keep it and label it.
  • The 4 km matching distance is wide, because the older reference maps are drawn roughly. A parallel line in the same corridor can therefore be counted as confirmed when it is not.
  • Sources can share origins. The World Bank map draws on OpenStreetMap in several countries, and plant databases often cite the same documents, so two sources are not always two independent surveys.
  • Shared plants on a border, such as Kariba, are counted under one country.
  • Promised dates are what was said in public. A date being quoted is not evidence that it will be met.
  • The atlas is a screening and learning tool. It is not for engineering design, system studies, navigation, safety or investment decisions. See our terms.

Known gaps

  • Angola, DR Congo and Tanzania have long stretches that are missing from OpenStreetMap and come from maps dated 2015 to 2018. They rest on that one map and their routes are approximate.
  • Only Zambia, Tanzania and Kenya have a national reference map besides the World Bank Africa map, so they are the only countries with many lines at three sources.
  • Transformer capacity is shown for South Africa and six substations in Zambia only.
  • Pipeline plants in Global Energy Monitor often have no expected year.
  • A small number of projects and places are not named on this site and are left out of the downloads.

If you hold a public dataset we could add as a source, or you see a line that is wrong, email [email protected].

Updates and versions

The atlas is checked on the 15th of every month against SAPP's reports, utilities, funders and project news. OpenStreetMap data is re-extracted every quarter. Sources are rechecked at every rebuild. Each change is listed under "What changed" on the atlas page.

ReleaseDateWhat changed
2026.1113 November 2026Sources and source counts on every feature, this method page, first data downloads. Corrected the same day: the DR Congo and Mozambique 2017 files were briefly counted as separate sources, although they are extracts of the World Bank Africa map.
Edition 28 to 11 November 2026Plants rebuilt on Global Energy Monitor (434 operating, 465 in the pipeline); voltage slider; source and age view; transformer capacity; cross-check against Open Infrastructure Map.
Edition 17 November 2026First publication: grid, 399 plants, pipeline, draft master plan to 2045.

From this release on, versions are numbered by year and month, and the downloads carry the same number.

Data and citation

The project register and the plant list, with their sources, are free downloads on the open data page under CC BY 4.0, archived on Zenodo at doi.org/10.5281/zenodo.24187977. Lines and substations are available on request.

Cite the atlas: Aloe Energy (2026). SAPP grid atlas, release 2026.11 [Interactive map]. https://aloeenergy.org/research/sapp-grid

Cite this method: Aloe Energy (2026). How the SAPP grid atlas is built: method and sources. https://aloeenergy.org/research/sapp-grid-method

Questions

Common questions

How accurate is the SAPP grid atlas?

It depends on the feature, which is why each one lists its sources. About seven in ten line kilometers are shown by two or more public maps. Lines with one source are usually well placed but not yet confirmed by another map, and lines taken from older maps are approximate. We have not yet measured positional error against survey data.

Where does each line on the map come from?

Click it. The popup lists every source that shows the line, such as OpenStreetMap and the World Bank grid map, and the "Number of sources" view colors the whole map by how many sources agree.

Is OpenStreetMap reliable for power lines?

For high-voltage lines it is often the best open source there is, since pylons are easy to trace from satellite images. Its weak points are voltage tags, lines under construction and whether a line is in service.

Can I reuse the method?

Yes, under CC BY 4.0 with credit to Aloe Energy.