Bibliographic citations
Estrada, A., (2021). Comunidades vegetales y estimación de biomasa con sensores multiespectrales y sistemas aéreos no tripulados en pastizales de Puna Seca [Universidad Nacional Agraria La Molina]. https://hdl.handle.net/20.500.12996/4788
Estrada, A., Comunidades vegetales y estimación de biomasa con sensores multiespectrales y sistemas aéreos no tripulados en pastizales de Puna Seca []. PE: Universidad Nacional Agraria La Molina; 2021. https://hdl.handle.net/20.500.12996/4788
@phdthesis{renati/242034,
title = "Comunidades vegetales y estimación de biomasa con sensores multiespectrales y sistemas aéreos no tripulados en pastizales de Puna Seca",
author = "Estrada Zúñiga, Andrés Corsino",
publisher = "Universidad Nacional Agraria La Molina",
year = "2021"
}
Remote sensing and geographic information systems are tools that have been used in the management of natural resources in the last decade. However, long-standing satellite images such as Landsat have presented limitations in spatial, spectral and temporal resolutions to develop precision agriculture and livestock. Faced with this, microsensors attached to unmanned aerial systems (UAS) appear as an alternative. The objective of the study was to identify plant communities and biomass estimation with multispectral sensors in UAS in plant communities of tolar and bofedal of Puna Seca. To determine the flight height and the study, images were collected with multispectral sensors at 25, 50.75 and 100 m flight height, at the same time that vegetation samples were collected in the fixed transects that were part of the control points. on land. The acquired images were processed in Pix 4D and Agisoft achieving high resolution RGB, NDVI orthophotos, which were used by the algorithms prepared for processing. Rapideye images Platform for the scaling to satellite image stage. The study has determined that the best flight height to identify the dry puna Tolar plant community and the segmented Tola shrub species is 25 meters, while for DIMU cushions it is 25 to 50 meters, observing the Tola and DIMU segmented species at 75 and 100 meters is diffuse. The Random Forest model with a training and test precision of 0.94, Kappa coefficient of 0.9071 and R2 = 0.482 predicted on average 3 g / pixel and 2 g / pixel of MV. of Lawn for rainy and dry season, the predicted biomass for the Tola component was 15 g / pixel of MV for both seasons. This same model with training and test precision of 0.8206, Kappa coefficient of 0.833 and R2 = 0.479 predicted an average of 2.5 g / pixel and 2 to g / pixel MV. DIMU for rainy and dry season. The scaling process was performed with an association level (p = 0.61 and p = 0.96) between orthophotography and Rapideye images, precision level 0.8253 training and 0.8306 test, Kappa index of 0.8091 and R2 = 46.20, the prediction of biomass with the Scaled images compared to the enhanced prediction for orthophotographs presented a difference that goes from 49% to 115.57%, therefore it is necessary to continue investigating algorithms and equations that decrease the difference found.
This item is licensed under a Creative Commons License