Estimativa de área agrícola a partir de sensoriamento remoto e banco de pixels amostrais / Estimates of crop area using remote sensing and database of sampling pixels

AUTOR(ES)
DATA DE PUBLICAÇÃO

2003

RESUMO

The methodology for national, state and municipal agriculture crop surveys must be efficient in terms of budget and estimate quality. This study has as main objective to create a methodology based on relational geographic database, in order to improve the identification of sampling points using orbital images to generate crop area estimates for the main cultures in the study area for three years. We used 22 images from TM/Landsat-5 and ETM+/Landsat-7 and field work as the main data source. A sampling frame of 300 points were randomly selected for unicipalities of Guará, Ipuã and São Joaquim da Barra and were monitored for three consecutive years. Multitemporal satellite images, used to identify the crops by visual analysis associated to a sampling frame, improved the crop surveys, reduced the fieldwork and provided objective estimates of the main cultures such as sugarcane and soybean. Furthermore, the estimate data recorded into the relational geographic database allowed the multitemporal study and the analysis of the dynamic of the agriculture landscape. The estimates generated by the direct expansion method of the sampling data seemed to provide good quality to the estimates when measured by coefficient of variation, mainly for crops with large areas, such as sugarcane and soybean. The crops of low occurrence, such as corn, cotton, bean and sorghum, presented high coefficient of variation for the estimates. The estimates calculated in these study were compared to the official estimates. For sugarcane the relative errors in relation to the official data were small for 2000 and 2002 - around 1% and 3%, while for 2001 it was -18%. For soybean, the relative errors were around 11%, 11% and 13% for 2000, 2001 and 2002, respectively. In general, the methodology applied to this study region proved to be useful for crop areas estimates, and could be considered as a tool for improving official estimates.

ASSUNTO(S)

soja geographic information systems (gis) milho cana-de-açucar amostragem landsat 5 agriculture corn sugar cane sistemas de informação geográfica imagens de satélite identificação de culturas agricultura landsat 5 satellite imagery agricultura soybeans crop identification sampling

ACESSO AO ARTIGO

Documentos Relacionados