Search

The paper describes two nonlinear regression models, applied to winter daily SO2concentration data and to the corresponding meteorological data from the metropolitan area of Vienna. The first model accounts for the role of wind speed and temperature (a proxy for emissions due to residential heating) on average SO2 concentration in the area. The second regression has an additional wind direction input and tries to identify the contribution by the industrial emissions (located primarily near the south-eastern border of the area) to concentration in the most polluted subarea. Both models offer a satisfactory fitting performance (e.g. correlations around 0.85 between observed and regression values). However, since model validation is a critical point for regressions, sensitivity tests of model fitting performance are carried out by using various data sets for the estimation of regression coefficients. One of such tests points out that there is an “optimal length” of the data set to be used, namely neither a too short set nor a set including “long past” data offer a satisfactory fitting quality.

The paper describes different mathematical models and real-time predictors of the short-term winter electric load in the biggest Italian district. First it is ascertained that weather variables suc...

The paper describes the application of three stochastic predictors of river flow-rate to a small basin (upper basin of Temo, in Sardinia ≃ 180 km 2 ). The first predictor (ARX, AutoRegressive with exogenous input) requires quantitative rainfall measurement, the second one (ARQX, AutoRegressive with Qualitative exogenous input) needs only a gross information about rainfall, the third one (AR, AutoRegressive) has no rainfall in put. The performance of the three predictors turns out satisfactory up to 3–4 hours, particularly by ARX. However, the gap of quality between the ARX and the other two predictors is counterbalanced by the cost of the rainfall telemetering system required by ARX.