Remote sensing approaches for evapotranspiration and yield estimations on irrigated corn fields

• Ano de defesa: 2019
• Autor(a) principal: Venancio, Luan Peroni
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa
• Palavras-chave em Português: Vegetação - Mapeamento; Milho - Rendimento; Evapotranspiração; Sensoriamento remoto; Irrigação e Drenagem
• Link de acesso: https://locus.ufv.br//handle/123456789/27367

Remote sensing data and applications have been experiencing a revolutionary advancement in various areas in the last fifteen years, including agriculture. These advancements are boosted by a large amount of the satellite sensors in orbit obtaining a large number of images of the Earth's surface every day in different temporal, spatial, spectral and radiometric resolutions. This dynamism is very useful to agriculture, since it is also a very dynamic system. Besides that, faced with the global problems of water and food shortages, climate change, environmental pollution, among others, high-efficiency agriculture will be increasingly required, which can be achieved more easily by means of the remote sensing data and applications. This thesis is divided into three chapters, making use of different satellites (Landsat 7, Landsat 8, Sentinel 2A and Sentinel 2B) with focus on the corn crop plantations irrigated by center pivot system in the western region of Bahia state, Brazil. The general focus was the estimation of evapotranspiration and yield along with vegetation spectral indices and spectral mixture analysis in irrigated corn fields using remote sensing approaches. The first study (chapter one) aimed to evaluate, calibrate and validate the SAFER algorithm for evapotranspiration estimation in irrigated corn fields. Meteorological and crop data were used to calculate corn evapotranspiration by means of the modified FAO method. In order to use SAFER algorithm, images of the sensors ETM+ and OLI/TIRS were acquired. SAFER algorithm with original regression coefficients has low accuracy for corn ET estimation, and after calibration with empirical data it showed a good performance, being a very useful tool for estimating water consumption by corn crop. Chapter two focused on corn yield estimation at farm level in Brazil using a new and simplified remote sensing approach, initially validated for North American corn production conditions. The formulation combines the methodology for biomass determination presented in the FAO-66 manual and a basal crop coefficient based on reflectance data adjusted by water and cold temperature stress. Data of 52 center pivots fields, collected during growing season of 2013 to 2016, were used. ETM+ and OLI surface reflectance images were used for the calculation of SAVI. The difference between predicted yield values and actual ones ranged between 12.2% and 18.8%, but with the majority of estimates between -10 and 10%, considering a single harvest index for all hybrids. After the reanalysis (grouping of similar hybrids and use of a specific HI) the performance of predictions increased, especially for Pioneer hybrids, with the majority of the differences, between predicted yield values and measured, remaining between -5 and 5%. Chapter three had as general objective to investigate the performance of vegetation indices for corn aboveground biomass estimation by means of their comparison with the fraction of photosynthetically active vegetation estimated from Spectral Mixture, defining the three best ones. Nine vegetation indices were calculated using the near infrared and visible bands of OLI sensor. Among the analyzed, VI, EVI, SAVI and OSAVI were considered the first, second and third best ones, respectively, for corn aboveground biomass estimation, based on their comparison with fraction of photosynthetically active vegetation. A second objective of this chapter was to find the best interval of VI accumulation (days) for corn grain yield estimation, using the three best classified in the general objective. For this purpose, field data of center pivots grown with irrigated corn during the season of 2018 were used along with Sentinel 2 images. The intervals that extended up to 120 days after sowing were the best. Finally, this work was a great challenge, mainly due to the use of data belonging to a commercial farm with a large number of cultivated corn hybrids. But, on the other hand, it brings very interesting results, showing the great potential of the remote sensing in agriculture. In turn, these results are useful for both the scientific community and farmers in Brazil, who are constantly being pressured by improvements in production processes, especially in water use.