Banca de DEFESA: HUGO MACHADO RODRIGUES

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
DISCENTE : HUGO MACHADO RODRIGUES
DATA : 28/02/2020
HORA: 13:00
LOCAL: Depto de Solos, Sala 39
TÍTULO:

Processing of Electrical Conductivity Data from Sensor EM38-MK2 to Map Soil Variability.

PALAVRAS-CHAVES:

Proximal soil sensing, Sample design, Geostatistics, Krigagem, EM38-MK2

PÁGINAS: 68
GRANDE ÁREA: Ciências Agrárias
ÁREA: Agronomia
SUBÁREA: Ciência do Solo
ESPECIALIDADE: Física do Solo
RESUMO:

Establishing a sample design plan is one of the most important stages for detailed soil recognition and is a challenge for geostatistical interpolators, such as kriging, to be used in computerized support and decision support systems, due to the high processing time required. This study aimed to evaluate four levels of spacing (40, 80, 150 and 300m) between apparent electrical conductivity (CEa) data collection lines with the Geonics EM38-MK2 sensor, in an area of 72ha under irrigation pivot , located in the municipality of Itaí, São Paulo. To assess the effect of the distance between the lines in the uncertainties of the CEa maps produced by ordinary kriging, the root mean square error index (REQM) was used and the spacing levels were associated with seven spatial resolutions for mapping (100 , 80, 50, 30, 20, 10 and 5m), in order to identify the shortest machine time for the interpolation of each map via microbenchmark. The maps that used resolutions 100, 80 and 50m for the four spacing levels did not demonstrate compatibility with the extension of the study area, since they made the interpretation of the spatial variability of CEa unfeasible. The 5m resolution map with 40m spaced data took 17 minutes to execute and had the lowest REQM among all the evaluated maps, 0.47. The time associated with excess precision for precision agriculture operations has shown that 5m is an inadequate resolution. Thus, the remaining resolutions to choose from were 30, 20 and 10m, observing that the REQM values increase as the data lines are spaced and, as the resolution is refined, the REQM values decrease by spacing level, being 1.09 the lowest value found for 80m of spacing and 10m resolution. To reduce the time of 77s required for making the map, using the combination of the data set with 80m spacing and the resolution of 10m, 25, 50, 75 and 95% of the points were randomly removed. The REQM value for the map with 25% removal was 1.09, and the time for producing the map was reduced to 42s. The maps with the data removed in 50 and 75% presented REQM 1.21, and the times were 20 and 4.9 seconds, respectively. When producing the map with the data removed in 95% of the points, the time did not exceed 1s, however, the REQM of 1.54 presents greater uncertainties when using spacing between lines of 150m. Therefore, the sample design containing 512 points, referring to the set with 75% of data removed from the 80m spacing, produced a map with a 10m spatial resolution. In this way, it allowed the detailed recognition of the CEa variability with REQM 1.21 and processing time less than 10s

MEMBROS DA BANCA:
Presidente - 1220296 - MARCOS BACIS CEDDIA
Interna - 2223668 - HELENA SARAIVA KOENOW PINHEIRO
Externo à Instituição - RONALDO PEREIRA DE OLIVEIRA