# Accuracy

Differences between the various data sources and computational methods can be summarized
as follows:

## Quality of the Database

Global irradiance and air temperature datasets have undergone extensive validation. The
**RMSE** (root mean square error) for interpolated **yearly irradiance** values is
approximately **6%**, while for **temperature** it is **0.9°C**.

## Climatic Variability

The Meteonorm Climate database is based on the 20-year measurement period 2001–2020.
Comparisons with longer-term observations show that deviations in average global
irradiance caused by the selected time periods are typically **less than 1–3% (RMSE)**
across all weather stations.

## Model Accuracy

Meteonorm uses computational models to compute irradiance on tilted surfaces and to estimate
additional meteorological parameters.

The hourly model in Meteonorm tends to **slightly overestimate total irradiance on
inclined surfaces by 0–3%**. The discrepancy between modelled and measured values is
typically within **±10%** for individual months and within **±6%** for annual totals.

The modelled hourly global horizontal irradiance in Meteonorm
matches very well with measured hourly global horizontal irradiance. The
Kolmogorov-Smirnov test Integral (KSI) as described in Espinar et al. 2008 shows good
agreement with KSI OVER percent (KSI %) values between 6.4 and 27.7 for different
stations.

Direct normal irradiance also shows good agreement with measured values, with a slightly positive
bias. Yearly averages have an average relative difference of 3.4% (RMSE of 7.3%).

## General Remark

It is important for users to recognize that the underlying data and computational models
represent approximations of real-world conditions. Nevertheless, the **natural
year-to-year variability in measured global irradiance** is often greater than model
uncertainty, making Meteonorm a reliable and robust tool for climate-based simulations.