GeneSE : um ambiente computacional para cálculo de balanço de energia da superfície utilizando imagens de satélites

• Ano de defesa: 2015
• Autor(a) principal: Gomes, Raphael de Souza Rosa
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Mato Grosso; Brasil; Instituto de Física (IF); UFMT CUC - Cuiabá; Programa de Pós-Graduação em Física Ambiental
• Palavras-chave em Português: SEB; Paralelismo; GPU; CUDA; OpenCL; CNPQ::CIENCIAS EXATAS E DA TERRA::FISICA; Paralelism
• Link de acesso: http://ri.ufmt.br/handle/1/2182

The knowledge about the forms of water use is becoming increasingly necessary, since this resource is also becoming scarce. Thus, using techniques that estimate the amount of water evapotranspired in a particular area is important for the management of hydrological processes, irrigation, industrial processes, among others. However, such techniques are relatively complexes and involve several steps for the final calculation. Thus, this study developed the GeneSE platform, which is an integrated computing platform that facilitates the use of these techniques for the determination of evapotranspiration using surface energy balance based on satellite images, such as SEBAL, SEBTA, SSEB and S-SEBI techniques. All the techniques involved in this work were implemented covering sequential and parallel processing, with CPU and GPU. The main facilities provided by the GeneSE platform are: (1) abstraction and encapsulation of the techniques; (2) ease of use of several techniques; (3) automation of various processing stages. To validate the Genesis platform was used a set of 85 satellite images from four different areas of the state of Mato Grosso. All these images were processed by 4 different GPU devices with 4 platforms developed in the computing environment (Java, OpenCL GPU, CPU OpenCL and CUDA). The tests showed that the environment spends an average of 150 milliseconds to create and compile the source code. The results also showed that the use of parallel platforms reduces the processing time up to 80 times with use the same amount of memory that sequentially platform does. After the performance tests, their validation was performed with the measured data from towers and three areas. The R² technique results presented values higher than 0.6, thus showing that the algorithms are significant for estimating evapotranspiration.