The following measures have been defined to run statistical analysis of all the data.

• MAE (Mean absolute error) measures the average magnitude of the errors in a set of data, without considering their direction.
• MBE (Mean bias error) is an indicator of whether the model is over-predicting or under-predicting the measured values.
• RMSE (Root mean square error) is a quadratic scoring rule which measures the average magnitude of the error. This means the RMSE is most useful when large errors are particularly undesirable.

*Click on the picture to enlarge it

The 2 m air temperature is well modelled by the meteoblue Learning Multi-Model (mLM) with values of MAE < 1.2 K. Relative humidity shows a MAE<10% in most of the stations and the model tends to underestimate RH. The model uncertainty of the wind speed is 1.5 m s-1.

The model skill of daily precipitation events decreases with increasing precipitation intensity. In fact, MAE results to be less than 2 mm, but in some stations located in the tropics the variance becomes bigger. Here precipitation results unpredictable as it is only caused by thunderstorms. Simulations can thus predict well the trend like e.g. a drier or moister week but hard to tell the exact amount and when. Leaf wetness and soil temperature variables result to have discrepancy in most of the stations. As expected the accuracy of soil temperature is worse than the accuracy of the air temperature. MAE values are between 1.5-9K.

The Virtual Weather Station allows manual adjustments of local rain data if needed. The user can use a small rain gauge to measure actual rainfall and then correct data in FieldClimate. Manually adjusted rainfall data provides more accurate results on the water balance.