Our forecast data is rigorously validated

Hour-ahead (nowcast) accuracy

What we did

• Generated hour-ahead GHI forecasts using Solargis's internal Cloud Motion Vector (CMV) model, which tracks cloud movement from satellite imagery.
• Validated forecasts against satellite-based reference GHI for 153 globally distributed sites, covering all major climate zones.
• Compared accuracy at hourly and 15-minute temporal resolutions using data from 2025.
• Calculated Bias, MAD, and RMSD as standard accuracy indicators, both globally and per climate zone.

Results we obtained

GHI	Hourly	15-min
Average Bias	-0.18%	-0.18%
Average MAD	4.87%	5.39%
Average RMSD	8.58%	9.58%

Full results by climate zone ->

Conclusions

• Forecast accuracy is better for hourly than for 15-minute data, primarily due to temporal aggregation effects that smooth short-term variability.
• Arid regions achieve the best performance due to stable, clear-sky conditions.
• Temperate and cold climates show moderate errors driven by variable cloud cover and frontal systems.
• Tropical and polar regions exhibit the highest errors, caused by convective activity, rapid cloud formation, low sun angles, and seasonal extremes.

Day-ahead accuracy

What we did

• Generated day-ahead GHI forecasts (approximately 24–48 hours ahead) using Numerical Weather Prediction (NWP) models.
• Validated the same 153 global sites at hourly and 15-minute resolutions using 2025 data.
• Compared results with hour-ahead accuracy to quantify the effect of extended forecast horizon on accuracy.

Results we obtained

GHI	Hourly	15-min
Average Bias	0.56%	0.57%
Average MAD	6.15%	6.61%
Average RMSD	10.11%	10.89%

Full results by climate zone ->

Conclusions

• Accuracy is lower compared to hour-ahead forecasts due to the longer prediction horizon.
• Arid regions achieve the best performance due to stable conditions.
• Tropical regions perform worst due to convective cloud formation.
• Polar regions show lower accuracy, partly influenced by limited sample size.
• Temperate and cold climates show moderate errors.

Day-ahead accuracy

What we did

• Transformed day-ahead GHI forecasts into PVOUT forecasts using the same fixed PV system configuration.
• Validated accuracy for all 153 sites, comparing with hour-ahead results to quantify the horizon effect.
• Analyzed MAD min/max variability across sites to support realistic accuracy expectations for operational use.

Results we obtained

PVOUT	Hourly	15-min
Average Bias	1.08%	1.08%
Average MAD	6.27%	6.70%
Average RMSD	10.49%	11.24%

Full results by climate zone ->

Conclusions

• Day-ahead PVOUT accuracy is slightly worse than hour-ahead.
• The day-ahead horizon is the most critical for electricity market participation, grid planning, and battery storage optimization.
• Forecast accuracy can differ significantly by site — both the average MAD and the variability expressed by minimum and maximum values should be considered when evaluating expected performance.