In this work solar erythemal ultraviolet radiation (EUV) which affects human skin and skin cancer was investigated at Songkhla province (7.2°N 100.6°E). EUV was converted to UV index then the data was used for model development and forecasting of hourly UV index using Artificial Neural Network (ANN). The ANN model has one input layer two hidden layers and one output layer. This input layer consists of extraterrestrial erythemal ultraviolet radiation solar zenith angle aerosol optical depth and cloud index which affect ultraviolet radiation and the output layer is hourly UV index. The results show that hourly UV index obtained from ANN and that from the measurement are in reasonable agreement with root mean square difference of 12.8% and mean bias difference of -2.4%. For forecasting of the hourly UV index the data for 7 days earlier was used for forecasting the next UV index for one day or nine-hour (08:00 am.-16:00 pm.). Multi-layer perceptron and back propagation algorithm were used in the model forecasting. The results show that the UV index from the model forecasting is reasonable agrees with UV index from measurement with root mean square difference of 17.0% and mean bias difference of 0.3%.

The objective of this study is to obtain a model that maximize production of γ-linolenic acid (GLA) by the moss Physcomitrella patens employing response surface methodology (RSM). The RSM with a three-variables (sucrose potassium nitrate and glutamate concentrations) and five-level (-2 -1 0 1 and 2) central composite design (CCD) including 20 experimental runs was employed to optimize the medium composition. Results showed that the experimental data could be appropriately fitted into a second-order polynomial model with a coefficient of determination (R2) more than 0 .9 5 for GLA production. Analysis of variance (ANOVA) revealed that the model was highly significant (p<0.0001) and the effects of the sucrose (20-100 g/L) and glutamate (0.5-2.5 g/L) concentrations on GLA production were significant (p<0.05). The GLA production with the optimized culture medium (sucrose concentration of 62.92 g/L potassium nitrate of 0.80 g/L and glutamate concentration of 1 .4 2 g/L) by RSM increased 4.61 folds when compared with the standard BCD medium. This experimental value (16.37 mg GLA/L) fit well with the predicted values (16.20 mg GLA/L).

This research developed a new model to determine the moisture diffusion coefficient for the thin-layer paddy drying with the spouted free-fall bed dryer. The design and construction of the dryer was done to experiment with the drying kinetics. The experiment has studied under conditions of hot air temperature of 60-150 °C the initial paddy mass of 5 - 15 kg and hot air velocity of 19 m/s. The results showed that the higher hot air temperature and the lower initial paddy mass resulted in a significant reduction in drying time. The experimental results were used to calibrate 12 of the most popular thin-layer drying models to find a suitable one considering based on the root mean square error (RMSE) chi-squared value (χ2) and the correlation coefficient (r). It was found that the popular thin layer drying model was able to accurately predict the effect of the moisture ratio considering both the influence of hot air temperature and the initial paddy mass in linear models. The highest accuracy was RMSE = 0.0279 χ2 = 0.00084 and r = 0.9938. In addition the experimental results have been used to synthesize the new diffusion moisture coefficient which is a function of initial paddy mass temperature and drying time. This developed model was based on an exponential function with only 7 model constants. The simulated results observed that the new model achieved the average error only 1.8 percent compared with the best model that uses up to 24 model constants.