Accuracy of satellite-based model such as SolarGIS can be characterized by following performance indicators:

 

1. Bias - characterizes systematic model deviation at a given site.

2. Root Mean Square Deviation (RMSD) and Mean Average Deviation (MAD) - indicate spread of error for instantaneous values.

3. Correlation coefficient

4. Kolmogorov-Smirnoff index (KSI) - characterizes representativeness of distribution of values

 

However for practical purposes and for easy understanding, statistical measures of accuracy are converted into uncertainty, which better characterizes probabilistic nature of a possible error of the model estimate.

 

Uncertainty of solar resource data is typically quantified for the P90 scenario (use of this confidence interval is considered as standard in solar resource assessment). For example, if we consider that uncertainty at P90 confidence interval is ±5%, it means that about 90% of the time, the true value will fall within ±5% range of the modelled best estimate.

 

If we consider normal distributions of deviation between modelled and measured values, standard deviation of bias of values represents 68% probability of occurrence. The uncertainty for the P90 confidence interval can be quantified using the following formula:

 

Uncertainty at P90 confidence interval = STDEV of biases x 1.2821

 

 

 

The uncertainty of SolarGIS model based on validation statistics at 189 sites globally is shown below for difference confidence intervals

 

 


 

 

Please note that above statistics represent uncertainty of the SolarGIS model and not user’s uncertainty. User’s uncertainty for SolarGIS GHI and DNI yearly summaries can be calculated as follows:

 



As SolarGIS data has been validated with only quality controlled data that were measured mostly with secondary standard instruments, we consider uncertainty of measurements to be ±2% and ±1% for GHI and DNI respectively. Based on this assumption, SolarGIS user’s uncertainty will be as follows: