|
Descripción
|
|
|---|---|
| ABSTRACT In the Yucatan Peninsula Mexico, the irrigated agriculture plays a key role in the crop production. In 2014, irrigated lands accounted for about 88,500 ha of the total cultivable area. The development of irrigation systems that make efficient and accurate use of water are essential for the sustainability of crop production systems. Thus, the reference evapotranspiration (ET0) is one of the most important factors related to irrigation system design, water management under irrigated and rainfed production. Also, the precise knowledge of daily global solar radiation (H) becomes very important in ET00 process. To address this challenge, the potential of various empirical models and soft computing (SC) techniques named Support vector machine (SVM), Artificial neural network (ANN), and Adaptive neuro fuzzy inference system (ANFIS) were evaluated for estimating ET0 and H under the Yucatan Peninsula environment. In the first part of this thesis, seven temperature based (TET) models and the standardize reference evapotranspiration equation for short canopies (ET0) method were compared. Using only temperature data, FAO-Penman Monteith Temperature (PMT) model was used to estimate daily values of ET0. Also, the ability and precision of SVM, ANFIS and ANN techniques were examined for estimating daily ET0 using measured meteorological variables. Three different combinations of minimum air temperature, maximum air temperature, rainfall, relative humidity and extraterrestrial radiation as input were investigated with air temperatures and extraterrestrial radiation as the base data set. In a second part, twelve existing empirical models on meteorological parameters ?based, four existing day of the year ?based (DYB) models, and three SC techniques (i.e., ANFIS, ANN, and SVM) were assessed for H predicting by using measured metrological variables. In addition, two new models were proposed for H modelling: (1) a model on meteorological parameters based and (2) a model on DYB. A qualitative analysis was performed on the database to find Incorrect or missing weather observations, mainly associated with the malfunction of measuring instruments, and to find weather observations affected by weather systems. The performance of the models in this thesis were evaluated using six different standard statistical measures: root mean squared error (RMSE), mean bias error (MBE), mean percentage error (MPE), mean absolute percentage error (MAPE), mean bias error (MAE) and coefficient of determination (R2). Results for ET0 modelling showed that the non-calibrated PMT expression using temperatures alone produced the best results. The others seven temperature-based models with and without calibration had poorer performance. The Hargreaves-Samani calibrated and Camargo calibrated models exhibited the best performance of the seven temperature-based models, but neither did as well as the PMT model. For the SC techniques approach, the results indicate that the SVM technique performed better than ANFIS and ANN approaches. Further, the influence of relative humidity and rainfall on the performance of models were investigated. The analysis revealed that the inclusion of the relative humidity data into the models significantly improves the accuracy of the ET0 estimates. As for the H modelling, the main findings were: According to the comparisons between empirical models, it was found that the newly developed empirical model which requires temperature, precipitation and relative humidity input variables obtained the best accuracy. However, if only temperature data are available, the Bristow and Campbell model can be used with good performance. Regarding comparisons between DYB models, it was found that the new proposed model estimates daily global solar radiation better than other DYB models. Furthermore, a seasonal analysis shows that the DYB model has good performance in all seasons, including in the rainy season. Finally, among the SC | |
|
Internacional
|
No |
|
ISBN
|
|
|
Tipo de Tesis
|
Doctoral |
|
Calificación
|
Sobresaliente cum laude |
|
Fecha
|
02/06/2017 |