# A telescope in New Zealand: what it would add to a Chinese satellite-tracking network

On 13 August 2026 New Zealand's security service said that Purple Mountain Observatory, an institute of the Chinese Academy of Sciences with close links, the report says, to the Chinese government, had tried to install a satellite-tracking station in the country. To see what such a place adds, we computed one day of flight for 5,262 real objects in orbit and placed a station, one at a time, on every stretch of land.

- **99%**: of land sites add less than New Zealand to a Chinese network, for Starlink-type satellites
- **2.1**: extra minutes in every 100 of Starlink flight the network would see with a New Zealand station: from 8.1% to 10.2%
- **24%**: of Starlink passes over China are preceded or followed, within 30 minutes, by a pass over New Zealand

> The service's report says New Zealand can be the “first or last opportunity” to observe a satellite. The calculation tests the observing part and confirms it for low orbit. It does not make New Zealand an exception: Tasmania or Patagonia give as much, and for polar and high orbits other places do better. Geometry explains why such a place is of interest. It does not say what Purple Mountain Observatory was after.

## The calculation

### What we computed

We took the public orbits of the objects in the CelesTrak catalogue (16,436 objects, on 2 October 2026) and followed a full day, minute by minute, to see when each would be above the horizon from a given place.

Each coloured square on the map is a place on land. The colour shows how many extra minutes of flight a Chinese network would see if it had a station there. The starting network is the whole of China, from Hainan to Heilongjiang (the blue diamonds). The orange ring marks New Zealand.

### For Starlink, New Zealand is among the best places on land

Almost 78% of Starlink satellites fly orbits inclined at 43° or 53° to the equator, so they spend their time at middle latitudes, north and south. The Chinese network sees them 8.1% of the time.

A station in New Zealand, at 44° south, would add 2.1 minutes in every 100. The maximum anywhere is 2.1, and 99% of sites add less. Places in the same latitude band, such as Tasmania and Patagonia, give practically the same result.

### For polar orbits, the poles win

Weather and Earth-observation satellites, and the debris left by collisions and satellite-destruction tests (Fengyun-1C, Cosmos 2251, Iridium 33, Cosmos 1408), pass close to the poles: almost all the debris in those four groups is inclined at more than 73°.

For them high latitudes matter. The Antarctic coast would add 5.5 minutes in 100, Iceland 4.9. New Zealand adds 2.8, more than 75% of sites, but is not among the top ones.

### At 36,000 kilometres, New Zealand changes little

Geostationary satellites hang fixed in the sky, each above one longitude. China sees them 64.1% of the time; the rest sit above the Atlantic and the Americas.

For them the best places are in South America: up to 33.4 minutes in 100. New Zealand adds 3.3; 63% of sites add more.

### China already has stations in the southern hemisphere

Public sources describe two Chinese tracking stations in the south: at Neuquén in Argentina and at Swakopmund in Namibia. With them in the network, 98.4% of geostationary satellite time is already covered.

The Pacific remains. New Zealand covers this remainder better than 98% of sites, but the remainder adds up to only 1.6 minutes in 100.

### For an optical telescope the window is much narrower

A telescope sees a satellite only if it is night at the station (Sun at least 12° below the horizon) and the satellite is lit by the Sun. On that basis the Chinese network sees Starlink satellites 0.5% of the time, and a New Zealand station would add 0.1 minutes in 100.

The figures are small anywhere, but the order holds: New Zealand beats 99% of sites. Radio antennas do not have this limit.

## First or last opportunity: how often New Zealand sees the same satellite as China, shortly before or after

> New Zealand can be the first or last opportunity to observe and communicate with satellites transiting regions with low GBSI coverage.
> NZSIS, 2026 report, page 13 (GBSI is ground-based space infrastructure)

We checked the statement directly. A pass is a continuous interval in which an object is visible from somewhere in China. For each pass we asked whether the same object is visible from New Zealand in a time window before it or after it.

With a 30-minute window the answer is yes for 24% of Starlink passes, 21% of other low-orbit satellites and 22% of debris passes. With a ten-minute window it is yes for none (0%): New Zealand is too far away to see an object just before China does.

The statement holds, but it is not specific to New Zealand. Iceland, Spain or Australia give values of the same order, sometimes higher. Chile, on the far side of the globe, sees no pass within 30 minutes of those over China.

| Place | Starlink | Other low orbits | Debris |
| --- | --- | --- | --- |
| New Zealand (Lake Tekapo) | 24.9% | 21.1% | 22% |
| New Zealand (Auckland) | 23.5% | 16.4% | 17% |
| Chatham Islands | 26.8% | 13.8% | 16.1% |
| Fiji | 16.1% | 10.5% | 11.7% |
| Tonga | 16.7% | 7.5% | 9.4% |
| Tahiti | 14.9% | 2.3% | 1% |
| Hawaii | 16.3% | 5.4% | 3.1% |
| Tarawa, Kiribati | 15.1% | 11.8% | 11.8% |
| Chile (Atacama) | 0% | 0% | 0% |
| Argentina (Neuquén) | 0% | 0% | 0% |
| Namibia (Gamsberg) | 16.7% | 3% | 3.1% |
| South Africa (Sutherland) | 23.3% | 5.6% | 6.3% |
| Western Australia (Yarragadee) | 13.5% | 37.1% | 38.1% |
| Eastern Australia (Canberra) | 20.4% | 29.8% | 27.2% |
| Brazil (Natal) | 15.4% | 20.2% | 20.9% |
| Spain (Madrid) | 22.8% | 14.1% | 15.6% |
| Réunion | 19.1% | 22.4% | 22.1% |
| Falkland Islands | 0% | 1.1% | 1.7% |
| Iceland (Reykjavik) | 7.7% | 51.5% | 55.7% |
| Kenya (Malindi) | 14.4% | 11.4% | 9.8% |

## Move the station and compare

Choose what to track, which network exists already and how it observes; then touch the map or use the arrow keys. The table below ranks the named places by what they add.

## What the public record says and what we do not know

The calculation above shows what geography can do. The real case is described in the NZSIS report (pages 13 and 14), which does not publish the evidence behind its account.

- **The attempt** (NZSIS report 2026): Purple Mountain Observatory, an organisation based in China with close links to its government, tried over the past year to install ground-based space infrastructure in New Zealand. NZSIS says it was not the first time the organisation had tried and that it is unlikely to be the last.
- **The local partner** (NZSIS report 2026): NZSIS says the local company was likely unaware that the equipment can collect intelligence of military value and would have had no idea who was receiving the data. NZSIS, working with other agencies, was able to disrupt the activity.
- **Who gets the data** (NZSIS assessment): The institute could be compelled to hand over data. In this case, the service says, it assesses the data would have been passed on willingly. NZSIS does not publish the evidence behind the assessment.
- **The earlier warning** (NZSIS report 2024): The 2024 report said a small number of entities had approached New Zealand organisations to develop space infrastructure. Such entities often claim civilian research purposes, but in each case it was later found that what was proposed could have assisted foreign military activity.

### What we know

- what NZSIS has said publicly about the attempt, the local partner and the end of the project
- that NZSIS describes the equipment as built to track satellites and space debris and also attractive for military purposes
- that the page of PMO's observation centre (an unverified official link) says it has seven stations; the six used in the calculation are in China
- that NZSIS says operators of ground-based space infrastructure are now regulated in New Zealand under the Outer Space and High Altitude Activities Act

### What we do not know

- the company, the proposed site and the type of equipment (optical, radio or other)
- which orbits the station would have tracked, and so which place in the ranking above it would have matched
- how the data would have travelled and who at Purple Mountain Observatory was involved
- what evidence lies behind the assessment that the data would have been passed on willingly

## Method, limits and how firm the figures are

**The data.** CelesTrak GP orbital elements, downloaded on 2 October 2026 (epochs between 2026-09-04 and 2026-10-04). Objects in the catalogue: 11,125 Starlink, 1,867 other active low-orbit satellites, 2,682 debris pieces from four fragmentations, 572 geostationary, 190 on medium or elongated orbits. From Starlink, debris and other low satellites we took a random sample of at most 1,500 per class (5,262 objects in all); the geostationary and medium-orbit objects are all included.

**The calculation.** Positions come from the SGP4 model, every 60 seconds for 24 hours from 2 October 2026, 00:00 UTC. An object is radio-visible if it is at least 10° above the horizon. Optically visible means: the Sun at least 12° below the station's horizon, the satellite lit by the Sun (Earth's shadow as a cylinder) and the satellite at least 20° above the horizon.

**The starting network.** Three versions. Six PMO stations in China (Delingha, Ganyu, Xuyi, Honghe, Yao'an, Qingdao, at town coordinates, with an error of a few tenths of a degree). All of China, with fourteen points from Sansha and Hainan to Mohe and Kashgar: an upper bound for any sensor on Chinese soil. All of China plus Neuquén and Swakopmund.

**What “adds” means.** For a place we add up the minutes in which an object is seen from there and from no station in the starting network, and divide by the class's total minutes of flight. Read it as: of every 100 minutes of flight, this many are visible in addition.

**The places.** A four-degree grid; we keep the points within one and a half degrees of land, by the Natural Earth coastline: 1,181 points between 66° south and 66° north. “Better than X%” is the share of points that add strictly less.

**Limits.** One day, with mean orbits from the catalogue (typical error of the order of kilometres). We did not model weather, terrain, an instrument's sensitivity, integration time or the minimum length of an observation. “Visible” does not mean “useful”: the value of an observation depends on the sensor, the network and who analyses the data. We do not know what network Purple Mountain Observatory wanted to build; the three versions are hypotheses, not descriptions.

### Does the conclusion change if we change the calculation?

| Variant | Starlink | Other low orbits | Debris | Geostationary |
| --- | --- | --- | --- | --- |
| Main calculation (10°, main sample) | 2.1 (99%) | 2.8 (75%) | 2.7 (73%) | 3.3 (37%) |
| Horizon at 20° | 1.0 (97%) | 1.4 (73%) | 1.3 (71%) | 2.7 (43%) |
| Horizon at 5° | 3.1 (99%) | 4.0 (76%) | 3.9 (76%) | 4.3 (37%) |
| Another random sample | 2.1 (99%) | 2.8 (75%) | 2.7 (73%) | 3.3 (37%) |
| A third random sample | 2.1 (99%) | 2.8 (75%) | 2.7 (74%) | 3.3 (37%) |

Each cell: how many minutes in 100 New Zealand adds and, in brackets, the share of sites it beats. Starting network: all of China, radio.

## Sources

- [New Zealand's Security Threat Environment 2026, pages 13 and 14](https://www.nzsis.govt.nz/assets/NZSIS-Documents/New-Zealands-Security-Threat-Environment-2026.pdf), NZSIS (verified directly): The report text, read directly from the PDF: the “first or last opportunity” passage, the Purple Mountain case study, the assessment that the local company was unaware, the box on GBSI regulation.
- [New Zealand's Security Threat Environment 2024](https://www.nzsis.govt.nz/assets/NZSIS-Documents/New-Zealands-Security-Threat-Environment-2024.pdf), NZSIS (verified directly): The case study on foreign entities asking to build space infrastructure presented as civilian research.
- [GP orbital elements (JSON): the “active” group, Starlink and the four debris groups](https://celestrak.org/NORAD/elements/), CelesTrak (verified directly): The orbits of more than twenty thousand objects, downloaded on 2 October 2026. The data come from the United States tracking catalogue; CelesTrak publishes them freely.
- [Natural Earth, 1:50m land](https://www.naturalearthdata.com/downloads/50m-physical-vectors/50m-land/), Natural Earth (verified directly): The coastline for the map and for the list of land sites. Public domain.
- [Purple Mountain Observatory, Center for Space Object and Debris Research](https://english.pmo.cas.cn/research/researchdivisions/dacmsodr/csodr/), PMO (unverified official link): The centre dates from 1957 and, the page says, has seven observation stations. The Honghe station page places it in Heilongjiang province. The site's security certificate is self-signed, so the page counts as an unverified official link, and the station list (Delingha, Ganyu, Xuyi, Honghe, Qingdao, Yao'an) was confirmed from Wikipedia.
- [The Chinese station at Neuquén, Argentina](https://chinafocus.ucsd.edu/?p=7487), UC San Diego, China Focus (secondary): Describes the station, operated by a Chinese military unit and approved by Argentina's Congress in 2015. A secondary source; the Chinese station at Swakopmund, Namibia, is described on Wikipedia.
- [Revisiting Spacetrack Report #3 (AIAA 2006-6753)](https://celestrak.org/publications/AIAA/2006-6753/), Vallado et al. (verified directly): The SGP4 model used to turn orbital elements into positions. Typical error over one day is of the order of kilometres.

---

Text, calculation and map: Marius Comper. Published 13 August 2026, rebuilt on 2 October 2026. [All projects](https://mariuscomper.uk/).