Interannual solar resource variability linked to ENSO and climate indices
Satellite climatologies from CERES and CM SAF SARAH-3 reveal how ENSO and the Indian Ocean Dipole drive year-to-year swings in solar irradiance across tropical and subtropical regions, letting project financiers build defensible P90 exceedance curves rather than guessing at tail risk.
Sensors
- CERES SYN1deg: Provides top-of-atmosphere and surface shortwave fluxes at 1-degree spatial resolution, daily and monthly aggregations, archived from March 2000 to present. Combines CERES broadband radiometry with MODIS-derived cloud properties to compute surface downwelling shortwave irradiance; absolute uncertainty in monthly GHI is typically cited at 5–10 W m⁻².
- CM SAF SARAH-3: Surface Solar Radiation Data Set from EUMETSAT's Climate Monitoring SAF, derived from Meteosat MVIRI and SEVIRI. Covers Europe, Africa and the Atlantic sector at 0.05-degree resolution with a daily timestep, archived from 1983 to near-present, giving roughly 40 years of record, the longest satellite-based surface irradiance climatology available for those regions.
- MODIS (Terra and Aqua): Cloud fraction and cloud optical depth products (MOD06/MYD06) at 1 km native resolution, aggregated to monthly means, used to attribute irradiance anomalies to specific cloud-cover changes driven by ENSO teleconnections. Two daily overpasses per satellite give four combined daily snapshots.
- ERA5 reanalysis: ECMWF's global reanalysis provides atmospheric state variables (total column water vapour, aerosol loading proxies, boundary-layer cloud fraction) at 0.25-degree resolution from 1940 to present. Used here as a dynamical bridge between sea-surface temperature indices and local cloud anomalies, not as a primary irradiance source.
Why a 20-year mean is not enough for a lender
Project finance for utility-scale solar typically requires a P90 yield estimate: the annual energy output that the plant will exceed in 9 years out of 10. Reaching that number demands knowing not just the long-run average global horizontal irradiance (GHI) but the shape of the year-to-year distribution around that average. In equatorial Africa, Southeast Asia, the Pacific and parts of Latin America, that distribution is not symmetric and not stationary. It is pulled and pushed by the El Niño–Southern Oscillation.
During El Niño years, the Walker circulation weakens, suppressing convection over the Maritime Continent and parts of the tropical Pacific while enhancing it elsewhere. The result can be GHI anomalies of 5–15% relative to the long-run mean, sustained over an entire wet season. A plant sized on the mean will underperform in La Niña years and overperform in El Niño years, but debt service is not symmetric: lenders care about the downside. Ignoring ENSO phasing in a bankable yield report is not conservative; it is incomplete.
What the satellite climatologies actually record
CERES SYN1deg derives surface shortwave irradiance by combining broadband radiance measurements from the CERES instruments on Terra, Aqua and SNPP with cloud properties retrieved by MODIS. The record runs from March 2000 to the present, giving roughly 24 years of monthly GHI fields at 1-degree resolution. That is enough to capture the 2002–03, 2009–10, 2015–16 and 2023–24 El Niño events and the intervening La Niña episodes, but it is short relative to the 20–30 year quasi-periodicity of multi-decadal Pacific variability. Tail estimates derived from it carry a residual uncertainty that honest assessments must state explicitly.
CM SAF SARAH-3 extends the picture for the Meteosat disk. Its 1983-to-present record covers roughly four complete ENSO cycles and includes the strong 1997–98 event, which CERES missed entirely. For sites in Africa, the Middle East and southern Europe, SARAH-3 is therefore the preferred primary climatology. The two datasets overlap from 2000 onwards and can be cross-validated; where they agree within a few percent, confidence in the anomaly signal is higher.
MODIS cloud-fraction time series serve a diagnostic role. When a GHI anomaly appears in CERES or SARAH-3, MODIS can confirm whether it is driven by changes in cloud cover (the dominant mechanism in ENSO teleconnection zones) or by aerosol loading, which requires separate treatment. The distinction matters for attribution and for deciding whether the anomaly is likely to recur with the same sign in future ENSO events of similar magnitude.
Connecting sea-surface temperatures to your site's irradiance
The Niño 3.4 index, the area-averaged sea-surface temperature anomaly in the central equatorial Pacific (5°N–5°S, 170°W–120°W), is the standard ENSO metric used in teleconnection studies. The Indian Ocean Dipole (IOD) index captures the east–west SST gradient across the Indian Ocean and has independent influence on irradiance over East Africa, South Asia and parts of Australia. Both indices are published monthly by NOAA and are available back to the mid-twentieth century from instrumental records.
The analytical step is to regress monthly GHI anomalies from the satellite climatology against concurrent Niño 3.4 and IOD values at each grid cell covering the project site. Where the regression is statistically significant (and in many tropical regions it is, explaining 20–40% of interannual GHI variance), it provides a physically grounded model for stratifying the historical irradiance distribution by ENSO phase. La Niña composites, El Niño composites and neutral-year composites each produce a different GHI distribution, and the P90 derived from the full mixed record is a weighted average of those three regimes. ERA5 reanalysis fields help explain the atmospheric pathway: SST anomalies drive circulation changes that alter boundary-layer cloud fraction, which is the proximate cause of the GHI shift.
Building the P90/P50 exceedance curve
The standard approach is to fit a parametric distribution, commonly a normal or slightly skewed distribution, to the annual GHI totals extracted from the satellite record at the site grid cell. The P50 is the median annual GHI; the P90 is the value exceeded in 90% of years. The gap between P50 and P90 is the number a lender uses to stress-test revenue projections. In ENSO-sensitive regions, that gap is typically wider than in mid-latitude continental sites, often by 3–8 percentage points of annual GHI.
A more defensible construction stratifies the distribution by ENSO phase before fitting. If the current ENSO state at financial close can be identified, the conditional P90 for the following year is narrower and more accurate than the unconditional P90. Over a 25-year project life, however, the unconditional distribution is the correct one to use for debt sizing, since ENSO phase cannot be predicted more than roughly 9–12 months ahead with skill.
The honest caveat is unavoidable: 20–24 years of CERES data is a short sample for estimating the 10th percentile of a distribution that has multi-decadal structure. The 95% confidence interval on the P90 estimate can be several percent of GHI, wide enough to matter for project economics. SARAH-3's longer record reduces this uncertainty for Meteosat-covered sites. Where the two datasets overlap and agree, the combined record strengthens the estimate. Where they disagree by more than the stated instrument uncertainty, the disagreement itself is information about retrieval limitations that should appear in the yield report.
Practical limits and what they mean for due diligence
Satellite-derived GHI has a spatial resolution floor of 0.05 degrees for SARAH-3 and 1 degree for CERES. A 1-degree cell covers roughly 110 km × 110 km at the equator. Complex terrain, coastal gradients and urban heat islands introduce sub-grid variability that satellite climatologies cannot resolve. For flat, homogeneous sites in semi-arid regions, the 1-degree product is adequate for interannual variability analysis. For sites near coastlines or in hilly terrain, the climatology should be downscaled using local topographic and land-surface information before the exceedance curve is constructed.
Cloud cover is the dominant source of GHI uncertainty in tropical regions, and it is also the main ENSO signal carrier. Aerosol optical depth, which affects GHI independently of cloud, is not captured by CERES or SARAH-3 at the accuracy needed for high-dust environments such as the Sahel or Arabian Peninsula. For those sites, aerosol correction using MODIS or MERRA-2 aerosol products is a separate analytical step, covered in the sibling page on aerosol optical depth correction.
Satellize applies these methods to client sites using open satellite archives, with ERA5 as the dynamical backbone and Niño 3.4 and IOD indices from NOAA's published record. The output is a site-specific irradiance variability report formatted to meet the disclosure requirements of project finance due diligence. The Tonga crop-estimation programme demonstrated the same underlying approach of extracting interannual climate signals from multi-year satellite time series in a data-sparse Pacific context.
Typical figures
| Spatial resolution (CERES SYN1deg) | 1 degree (~110 km at equator); daily and monthly aggregations |
| Spatial resolution (CM SAF SARAH-3) | 0.05 degree (~5.5 km); daily and monthly aggregations |
| Temporal coverage | CERES: March 2000 to present (~24 years); SARAH-3: 1983 to near-present (~40 years) |
| GHI absolute uncertainty (monthly mean) | CERES: ~5–10 W m⁻² (published instrument + retrieval budget); SARAH-3: comparable order |
| ENSO index used | Niño 3.4 (5°N–5°S, 170°W–120°W SST anomaly); IOD (Dipole Mode Index) for Indian Ocean sites |
| Cloud fraction ancillary data | MODIS MOD06/MYD06 at 1 km native resolution, aggregated monthly |
| Reanalysis backbone | ERA5 at 0.25 degree, 1940 to present |
| Archive depth for exceedance statistics | 24 years (CERES alone); up to 40 years with SARAH-3 for Meteosat-covered sites |
| Deliverable formats | NetCDF time series, GeoTIFF anomaly composites, PDF bankable yield report with exceedance tables |
Analytics Satellize can run
| Interannual GHI anomaly composites by ENSO phase | Regression of monthly satellite GHI against Niño 3.4 and IOD indices; composite averaging by phase classification (El Niño / La Niña / neutral) | GeoTIFF anomaly maps and site-point time series, stratified by ENSO phase |
| P90/P50 annual GHI exceedance curve | Parametric distribution fit (normal or skewed) to annual GHI totals from CERES and/or SARAH-3 at site grid cell; bootstrap confidence intervals on P90 estimate | Exceedance probability table and curve, with stated confidence bounds, in PDF and CSV |
| ENSO teleconnection significance map for site region | Pearson and Spearman correlation of monthly GHI with Niño 3.4 at each grid cell; significance tested at p < 0.05 after autocorrelation correction | GeoTIFF correlation map with significance mask; site-specific correlation coefficient and explained variance |
| Cloud-fraction attribution time series | MODIS MOD06 monthly cloud fraction extracted at site; Pearson correlation with GHI anomaly and with Niño 3.4 to confirm cloud as the proximate driver | Time-series chart and correlation table in PDF report section |
| Multi-dataset cross-validation report | Comparison of CERES and SARAH-3 monthly GHI over overlapping 2000-to-present period at site; bias and RMSD statistics; ERA5 surface shortwave as third reference | Cross-validation table and bias-corrected best-estimate GHI time series in NetCDF and CSV |
| Conditional P90 under current ENSO state | Phase-stratified exceedance curve conditioned on current Niño 3.4 anomaly; applicable to near-term (1-year) revenue stress testing | Single-page conditional yield summary for lender technical adviser review |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.