Source
https://doi.org/10.1080/03036758.2020.1763409 — original source (opens in a new tab; the file is not redistributed)
Brent et al. (2020) — Solar Atlas of New Zealand from Satellite Imagery
Peer-reviewed short communication (Journal of the Royal Society of New Zealand) establishing a Solar Atlas of New Zealand from Solargis commercial satellite-modelled data (250 m raster, 2006–), developed by Victoria University of Wellington. Validated against NIWA Kelburn (Wellington) ground data for 2018: annual GHI rMBD 3.2% (summer 0.6%, winter 8.3%), within the atlas's stated ±4.0–5.5% GHI uncertainty. Delivers the peer-reviewed satellite-solar cross-check RT_101 sought.
⚠ Its tabulated figures are GHI in kWh/m² (horizontal plane) at point sites, not PVOUT specific yield (kWh/kWp) — the paper names PVOUT as an atlas layer but does not tabulate it. The load-bearing NI takeaways are (a) satellite-modelled NZ solar resource is accurate to ~±3% annually and (b) it systematically overestimates winter GHI (rMBD 8.3% in JJA, and up to 41.9% on an individual overcast winter day) — a direct caveat for off-grid winter sizing. Read verbatim via pdftotext →
data_quality: verified.
Summary
This short communication reports the development and validation of a Solar Atlas of New Zealand built from commercial satellite-modelled irradiance data. Victoria University of Wellington, building on the World Bank / ESMAP Global Solar Atlas platform, commissioned Solargis to model the NZ solar resource as GIS raster layers at ~250 m resolution using cloud-index imagery from the MTSAT-2 (2006–2015) and Himawari-8 (2016–) geostationary satellites plus atmospheric reanalysis inputs. The atlas produces GHI, DNI, GTI (tilted, at optimum angle), diffuse, optimum-angle, 2 m temperature and PVOUT (kWh/kWp) layers, with a stated GHI uncertainty of ±4.0–5.5%. The authors verify the modelled output against 4,335 hours of NIWA CliFlo ground measurements at the Kelburn (Wellington) station for 2018: the annual relative Mean Bias Deviation is 3.2%, comparable to the ±3.1% global-validation bias and the ±4% reported for NZ’s Lauder site — but strongly seasonal, dropping to 0.6% in summer and rising to 8.3% in winter (up to 41.9% on a single overcast July day), because the satellite model underestimates cloud effects and therefore overestimates winter GHI. Point GHI values are tabulated for eleven hydropower sites (Table 4). The paper positions the atlas as a screening/planning tool that augments NIWA’s SolarView but cannot replace on-site measurement for bankable utility-scale projects. It is the peer-reviewed primary that RT_101 was raised to retrieve as a cross-check on the PVGIS-ERA5 national solar layer (RD_013) and the Nelson–Tasman yield synthesis (CR_010).
Key claims
- claim: "The Solar Atlas of New Zealand was developed by Victoria University of Wellington with Solargis, building on the World Bank / ESMAP Global Solar Atlas platform. GHI and DNI are modelled by Solargis at ~250 m grid resolution, 2006 to date, at a 30-min original time step. Cloud-index inputs are from MTSAT-2 (07/2006–2015; 30 min; 4.6 × 7.1 km) and Himawari-8 (2016 to date; 10 min; 2.3 × 3.6 km) satellites (both JMA), with aerosol (MACC-II/CAMS), water-vapour (CFSR/GFS) and elevation (SRTM-3, 250 m) auxiliary layers."
source_location: "Section 'Establishing a Solar Atlas of New Zealand', p.4–5; Table 2, p.4"
- claim: "The atlas GIS raster layers comprise: GHI – Global Horizontal Irradiance (kWh/m2); DNI – Direct Normal Irradiance (kWh/m2); GTI – Global Tilted Irradiation at optimum angle (kWh/m2); DIF – Diffuse horizontal irradiation (kWh/m2); OPTA – Optimum Angle for GTI (°); TEMP – Temperature at 2 metres (°C); and PVOUT – Photovoltaic Electricity Output (kWh/kWp). [⚠ PVOUT is named as a layer but no PVOUT values are tabulated in this paper.]"
source_location: "Section 'Establishing a Solar Atlas of New Zealand', p.5"
- claim: "The uncertainty of the solar resource is estimated to be ±4.0 to ±5.5% (for GHI) and ±9.0 to ±13.0% (for DNI)."
source_location: "p.5 (paragraph after Table 2)"
- claim: "The modelled data were verified against measured data from the NIWA meteorological station at Kelburn, Wellington (eighteen years of solar data on record), using hourly CliFlo data for the 2018 year; the matched dataset comprises 4,335 hours (September–November points excluded due to instrument downtime). The difference between the cumulative annual measured and modelled data is less than 42.5 kWh/m2, and the relative Mean Bias Deviation (rMBD) is 3.2% for the annual GHI data — similar to the ±3.1% bias reported for 80% of 208 site validations across the globe and the ±4% reported for the Lauder measurement site in New Zealand."
source_location: "Section 'Verification of the Solar Atlas with NIWA data', p.5"
- claim: "The validation is strongly seasonal: for the summer months (December, January, February) the rMBD is 0.6%, but for the winter months (June, July, August) the rMBD is 8.3%. The modelling output indicates a clear underestimation of cloud effects (i.e. an overestimation of winter GHI); for locations with high cloud persistence and variability, ESMAP indicates a P90 uncertainty estimate of ±7%."
source_location: "Section 'Verification of the Solar Atlas with NIWA data', p.5"
- claim: "Table 3 (Kelburn 2018; average monthly GHI kWh/m2; CliFlo / Solar Atlas / difference / rMBD%): Annual 1269.7 / 1312.1 / 42.4 / 3.2. Winter is the weakest fit — June 37.8 / 43.8 / 6.0 / 13.8 (worst month); July 47.9 / 51.7 / 3.8 / 7.4; August 68.3 / 72.5 / 4.2 / 5.8 — vs summer January 189.6 / 188.6 / −0.9 / −0.5 and December 185.8 / 190.3 / 4.5 / 2.4."
source_location: "Table 3, p.8"
- claim: "Daily profile extremes at Kelburn 2018: maximum GHI at 13h00 recorded on 6 January (1.108 kWh/m2), minimum GHI at 13h00 on 1 July (0.039 kWh/m2). On a good-resource day (6 Jan) the Solar Atlas gives 6.285 kWh/m2 vs 6.481 kWh/m2 measured (rMBD −3.1%); on a poor-resource winter day (1 July) the Solar Atlas gives 0.665 kWh/m2 vs 0.386 kWh/m2 recorded (rMBD 41.9%), because cloud effects are not captured accurately in the modelled data."
source_location: "p.5 (13h00 max/min); p.7 (6 Jan comparison); p.8 (1 July comparison)"
- claim: "Table 4 — Solar Atlas average annual GHI (kWh/m2) at eleven hydropower sites (P90 uncertainty ±4% to ±8%): Lake Arapuni 1403.7; Maraetai Dam 1378.6; Lake Taupo 1478.0; Lake Tekapo 1503.2; Lake Pukaki 1425.2; Lake Benmore 1425.1; Lake Aviemore 1418.7; Clyde Dam 1375.4; Roxburgh Dam 1239.4; Lake Onslow 1261.4; Lake Manapouri 972.9. [⚠ These are GHI (horizontal-plane resource, kWh/m2), NOT PVOUT specific yield (kWh/kWp).]"
source_location: "Table 4, p.9; P90 range p.9"
- claim: "NZ solar context: in 2018 solar PV constituted less than 0.1% (0.72 PJ) of the total primary energy supply (890.68 PJ). Installed solar PV capacity increased 26% for the year ending September 2019, to over 100 MW. Transpower's Te Mauri Hiko addendum projects solar to generate between 36 and 115 PJ of electricity (of a total of 317 PJ) by 2050, of which at least half will be from distributed solar PV."
source_location: "Introduction, p.2–3"
- claim: "The solar resource in the Waikato region (above 1400 kWh/m2 from the Solar Atlas) is significantly higher than corresponding dairy-production regions in Germany (around 1150 kWh/m2), where thermal energy produced from solar is used for milk processing."
source_location: "Section 'Utilising the Solar Atlas', p.9"
- claim: "The Solar Atlas cannot replace on-site measurements, which are the industry standard and necessary for project financing: satellite data sources are acceptable where extensive historical data are not available, but ground-based station measurements are still required over a number of years to make utility-scale PV projects bankable. The atlas augments NIWA's SolarView tool, which is meant to be a guide and not for accurate predictions."
source_location: "Section 'Utilising the Solar Atlas', p.9; 'Previous and ongoing analyses', p.3"Neobiome Intelligence relevance
This is the peer-reviewed primary RT_101 was raised to retrieve: an NZ-authored, NZ-scope validation of a commercial satellite solar dataset against NIWA ground truth. Its value to NI is three-fold.
- Validates satellite-modelled solar resource for national NI coverage (D01). The atlas — built from Solargis on the Global Solar Atlas platform — reproduces the Kelburn ground record with an annual GHI rMBD of 3.2%, inside its stated ±4.0–5.5% GHI uncertainty and comparable to Solargis’s global ±3.1% and NZ-Lauder ±4% benchmarks. This is the independent, peer-reviewed evidence that satellite-derived solar resource (the same class of gridded product as the PVGIS-ERA5 layer in RD_013 and the Nelson–Tasman synthesis in CR_010) is accurate to ~±3% annually for NZ — i.e. adequate for the coordinate-first, screening-grade solar resource NI uses, without a ground station at every point.
- Quantifies the winter overestimate — a direct off-grid sizing caveat (D01). The fit is strongly seasonal: rMBD 0.6% in summer (DJF) but 8.3% in winter (JJA), and up to 41.9% on a single overcast July day, because the satellite model underestimates cloud. Winter is the binding constraint for off-grid autonomy, so any satellite/atlas solar figure fed to NI should carry a winter haircut — this paper puts a number on it and corroborates the winter-fog suppression risk already flagged for the Waimea/Moutere area in CR_010.
- Spatial GHI spread + screening-tool status (D01). Table 4 gives point GHI across NZ (South Island alpine highest — Lake Tekapo 1503.2, Lake Taupo 1478.0; lowest Lake Manapouri 972.9 kWh/m²), a useful sense of resource dispersion; and the paper is explicit that atlas/satellite data is a screening and planning guide, not a bankable substitute for on-site measurement — which matches NI’s own posture (coordinate-first screening, not project-grade design).
⚠ Unit caveat (load-bearing — do NOT conflate): every tabulated number in this paper is GHI in kWh/m² on the horizontal plane, not PVOUT specific yield in kWh/kWp. Table 4’s ~1,240–1,503 kWh/m² values sit in a similar numeric range to the 1,350–1,380 kWh/kWp specific-yield figures in CR_010 / RD_013, but they are different quantities (GHI is the horizontal-plane resource; PVOUT is post-array energy per installed kWp). The atlas does produce a PVOUT layer, but this paper does not tabulate it — so this source is a methodological cross-check on the underlying dataset’s accuracy, not a direct kWh/kWp validation of the specific-yield cells. Do not enter any Table 4 figure into a specific-yield (kWh/kWp) NI cost/energy cell.
Research targets
Resolved
- RT_101 (RESOLVED → this page): Brent et al. (2020) “Solar Atlas of New Zealand from satellite imagery” retrieved, read verbatim and ingested. RT_101 is a
doctarget (retrieve-and-ingest) and the document is now held with its validation statistics captured — satisfying the “retrieve the peer-reviewed satellite-solar cross-check” purpose. ⚠ Scope of resolution: what it delivers is a methodological validation of the commercial-satellite dataset (annual GHI rMBD 3.2%; winter JJA 8.3%) plus point GHI values — it therefore corroborates the reliability of the satellite/gridded solar layers in RD_013 and CR_010 rather than directly cross-checking their 1,350–1,380 kWh/kWp specific-yield numbers (this paper tabulates GHI in kWh/m², not PVOUT in kWh/kWp).
Residual needs (no new RT raised)
- Direct PVOUT (kWh/kWp) specific-yield cross-check — not delivered by this paper. The underlying Solargis (2019) Solar resource raster GIS data, New Zealand report (nr. 990-26/2019; this paper’s ref “Solargis 2019b”) carries the PVOUT raster but is a restricted commercial document; the free Global Solar Atlas web viewer (globalsolaratlas.info) is the public proxy, and the existing PVGIS-vs-NIWA validation in CR_025 already serves this need — so no new retrieval RT is raised. → D01
- Generalised winter satellite-GHI overestimate derating — this paper gives rMBD 8.3% (JJA) and 41.9% on a worst-case overcast day at a single station (Kelburn); a usable NI winter-autonomy derating factor would need the monthly bias generalised beyond one station, but is partly covered by the winter-tilt discussion in RD_013 and the fog caveat in CR_010 — not raised as a formal RT. → D01
Notes
Primary peer-reviewed short communication (Journal of the Royal Society of New Zealand, DOI 10.1080/03036758.2020.1763409, published online 25 May 2020; received 15 Jan 2020, accepted 28 Apr 2020). Authors: Alan C. Brent, James (Jim) Hinkley, Daniel Burmester, Ramesh Rayudu — Sustainable Energy Systems, School of Engineering and Computer Science, Victoria University of Wellington. Read verbatim via pdftotext -layout → data_quality: verified; every quantitative claim is traceable to a stated page/table.
⚠ Author-citation note: RT_101 and the parent CR_010 cite this as “Brent A.C. (2020)” (first author only). The full author list is Brent, Hinkley, Burmester & Rayudu — captured here in author:. Not a duplicate; the RT citation was simply abbreviated.
⚠ GHI-vs-PVOUT unit trap (repeated because it is the single most likely mis-use): the paper’s tables report GHI (kWh/m², horizontal), not PVOUT specific yield (kWh/kWp). Do not treat Table 4 as a specific-yield table or use its values to validate/replace the kWh/kWp figures in CR_010 / RD_013.
⚠ No volume/issue/pages: the retrieved PDF is the online-first version; it carries only the DOI, not final volume/issue/page numbers. These fields are deliberately omitted (not fabricated). The in-text page numbers used in source_location: are the PDF’s own running page numbers (1–12).
Corpus fit: joins the D01 solar-resource evidence cluster — RD_013 (PVGIS-ERA5 national specific yield), CR_010 (Nelson–Tasman yield), CR_025 (PVGIS-vs-NIWA validation), OT_056 (utility-scale solar economics) — as the peer-reviewed, satellite-based accuracy anchor for the NZ solar resource that underlies all of them.
Connections
Links to
Referenced by
EDT domains (1): D01: Renewable Energy & Storage Systems