Estimation of microclimatic data in remote mountainous areas using an artificial neural network model-based approach

An artificial neural network (ANN) model-based approach was developed and applied to estimate
values of air temperature and relative humidity in remote mountainous areas. The application site
was the mountainous area of the Samaria National Forest canyon (Greece). Seven meteorological
stations were established in the area and ANNs were developed to predict air temperature and
relative humidity for the five most remote stations of the area using data only from two stations
located in the two more easily accessed sites. Measured and model-estimated data were compared
in terms of the determination coefficient (R2), the mean absolute error (MAE) and residuals
normality. Results showed that R2 values range from 0.7 to 0.9 for air temperature and from 0.7 to
0.8 for relative humidity whereas MAE values range from 0.9 to 1.8 oC and 5 to 9%, for air
temperature and relative humidity, respectively. In conclusion, the study demonstrated that ANNs,
when adequately trained, could have a high applicability in estimating meteorological data values in
remote mountainous areas with sparse network of meteorological stations, based on a series of
relatively limited number of data values from nearby and easily accessed meteorological stations.